close

Вход

Забыли?

вход по аккаунту

?

Evolution of the passive microwave signature of thin sea ice

код для вставкиСкачать
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted.
Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted.
Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back o f the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Aibor MI 48106-1346 USA
313/761-4700 800/521-0600
R ep ro d u ced with p erm ission of th e copyright ow ner. Further reproduction prohibited w ithout p erm ission .
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
Evolution of the
Passive Microwave Signature of
Thin Sea Ice
by
Mark R. Wensnahan
A dissertation submitted in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
University of Washington
1995
Approved by
L^
_________
/) . t
y l
(Chairpersons of Supervisory Committee)
Program Authorized
to Offer Degree _____ (Uv~-<v;p\A<?-< \c
Date
OcV trt n riS"_________
,
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
UMI Number: 9616691
UMI Microform 9616691
Copyright 1996, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Doctoral Thesis
In presenting this dissertation in partial fulfillment of the requirements for the
Doctoral degree at the University of Washington, I agree that the Library shall make its
copies freely available for inspection. I further agree that extensive copying of this
dissertation is allowable only for scholarly purposes, consistent with “fair use” as
prescribed by U.S. Copyright Law. Requests for copying or reproduction of this
dissertation may be referred to University Microfilms, 1490 Eisenhower Place, P.O. Box
975, Ann Arbor, MI 48106, to whom the author has granted “the right to reproduce and
sell (a) copies of the manuscript in microform and/or (b) printed copies of the manuscript
made from microform.”
Signature
Date
Oc-V
NT .
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
University o f W ashington
A bstract
E volution o f the Passive M icrow ave Signature o f Thin S e a Ice
by M ark W ensnahan
C hairpersons of the Supervisory C om m ittee:
P rofessor Thom as C. G renfell, D epartm ent o f A tm ospheric Sciences
Professor Gary A. M aykut, D epartm ent of A tm ospheric Sciences
Thin sea ice regulates heat exchange between the ocean and atm osphere, salt fluxes into th e
ocean and th e exchange o f shortw ave radiation at th e ocean surface. The degree o f regulation
depends critically on the ice thickness fo r ice less than 50 cm thick. It was fo u n d from a statistical
study o f S S M /I satellite d ata o f the B ering Sea that thin ice has its own unique passive
m icrow ave signature differentiable from open w ater and mature, thick ice. F urther, the H orizontal
polarization brightness tem perature at 19 and 37 G H z steadily increased dow nw ind o f land w ith
the signature changing o r evolving as the ice aged and thickened. Field observations indicate that
thin ice w ith a bare surface can produce only part o f this evolution and that th e signature o f
mature first-year ice ultim ately is the result o f frost flow er form ation or snow deposition. A
theoretical model o f em ission from sea was developed and sim ulations o f em ission corroborated
the observations and fu rth er indicated that the evolution of the thin ice signature can be explained
either by th e gradual d ry in g out o f a surface slush lay er or by the slow , steady form ation o f frost
flowers in com bination w ith brine w icking. Time sequences o f passive m icrow ave satellite d ata
were then com pared w ith m eteorological data. It w as determ ined that the thin ice signature near
land was, at tim es, sensitive to changes in w indspeed, air tem perature, and shortw ave radiation,
but, aw ay fro m land, th at sensitivity w as substantially reduced. S now also had a dram atic effect
on em ission from thin ice, increasing H polarization brightness tem perature by as m uch as 50K.
Finally, a theoretical m odel o f ice dynam ics was developed from w hich fields o f ice age and
thickness calculated. Relationships betw een brightness tem perature and ice age and thickness
were calculated from a regression analysis o f these fields and S S M /I data. It w as shown th at 37
GH z, H pol brightness tem perature provides a reasonable estim ate o f ice age and thickness up to
as much as 12 days old and 30 cm in thickness.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table
of
C
ontents
L ist O f F ig u r e s .................................................................................................................................................... ii
L ist O f T a b l e s .................................................................................................................................................... iv
1. In t r o d u c tio n ....................................................................................................................................................1
2. S atellite C a se S t u d y : B ering S e a , S pring 1 9 8 8 ........................................................................ 11
2.1 Principal component a n alysis.............................................................................................................. 11
2.2 Independent evidence o f thin i c e ......................................................................................................... 13
2.3 From thin ice to FY i c e .........................................................................................................................14
3. F u n d a m e n t a l s O f M icrow ave E m issio n F rom S e a Ic e ............................................................24
4. F ield O b ser v a tio n s O f E mission F rom T hin S ea Ic e ................................................................. 27
4.1 CEAREX, 1989...................................................................................................................................... 28
4.2 CRRELEX, 1 9 9 0 ...................................................................................................................................29
4.3 LEA D EX , 1993.......................................................................................................................................31
4.4 CRRELEX 1 9 9 4 .................................................................................................................................... 33
4.5 Sum m ary...................................................................................................................................................37
5. T heoretical S tudies O f E m ission ....................................................................................................... 5 1
5.1 Multilayer Fresnel form ulation........................................................................................................... 51
5.2 Permittivity mixing form ulae...............................................................................................................53
5.2.1 S ea ic e .............................................................................................................................................. 58
5.2.2 S n o w a nd fr o s t flo w e r s ................................................................................................................ 59
5.3 Sim ulations............................................................................................................................................... 60
5.3.1 Ice ancl b r in e ..................................................................................................................................61
5.3.2 S n o w ..................................................................................................................................................62
5.3.3 F ro st flo w e r s ..................................................................................................................................64
6. M icrow ave S atellite D ata A n d E n v ir on m ental C o n d it io n s ............................................ 79
6.1 Meteorological d a ta ...............................................................................................................................80
6.2 Passive microwave satellite data........................................................................................................ 80
6.3 The effect o f currents............................................................................................................................ 82
6.4 Time se r ie s............................................................................................................................................... 83
6.5 Correlation studies..................................................................................................................................85
6.5.1 W ind ..................................................................................................................................................86
6.5.2 A ir tem perature a nd shortw ave radiation............................................................................. 87
6.5.3 S n o w ..................................................................................................................................................90
6.6 Sum m ary...................................................................................................................................................91
7. D rift M o d el C o m p a r iso n ..................................................................................................................... 106
7.1 Theory...................................................................................................................................................... 106
7.2 Tb, ice age, and ice thickn ess............................................................................................................109
8. D isc u ssio n .....................................................................................................................................................125
9. R e f e r e n c e s .................................................................................................................................................. 133
10. A ppe n d ix ...................................................................................................................................................... 138
10.1 Methods o f Principal Component A n a ly sis................................................................................ 138
10.2 Details o f the ice growth m odel...................................................................................................... 140
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
L
is t o f
F
ig u r e s
1-1: Polarization ratio as a function of ice ty p e............................................................................................. 6
1-2: Polarization ratios at tw o frequencies as a function o f ice type........................................................ 7
1-3: B rightness tem perature record for the ice sheet o f C R R E L E X 1990..............................................8
1-4: Grow th sequence data fro m CR R E L EX 1990 saline tank
ice........................................................9
1-5: C luster plots o f the C R R E L E X 1990 in principal com ponent sp ace............................................ 10
2-1: SSM /I land m ask m ap o f th e Bering S ea case study are a.................................................................16
2-2: N A SA T eam algorithm cluster plot o f 25 km SSM /I d a ta ............................................................... 17
2-3: E ndm em ber spectra used in the PCA o f the Bering S ea case study...............................................18
2-4: E igenvector coefficients o f a supervised PC A of the S SM /I Bering S ea data.............................19
2-5: C luster p lo t o f the B ering S ea data in P C space.................................................................................20
2-6: C oncentration maps for the B ering S ea on April 2, 1988................................................................ 21
2-7: A V H R R visible image o f the Bering S ea study area.........................................................................22
2-8: C hanges in T b( 19) and P R (19) dow nw ind o f St. L aw rence Island................................................23
4-1: M icrow ave observations m ade during the C EA R EX 1989 cruise.................................................39
4-2: F rost flow ers on 5-10 cm thick rafted nilas observed on M arch 18, 1989..................................40
4-3: O bservations before and after snowfall on thin artificial sea ice....................................................41
4-4 Passive m icrow ave m easurem ents o f L ead 4 during L E A D E X , 1992.......................................... 42
4-5: T im e sequence o f em ission during ice grow th at Lead 3 o f L E A D E X
1992........................... 43
4-6: F rost flow ers on the surface thin ice at L ead 3 o f L E A D E X , 1992.............................................. 44
4-7: O bservations o f bare and snow covered 30 cm . thick ice, C R R E L E X
1994........................... 45
4-8: T he effect o f artificial surface roughness on bare thick ice o f C R R E L E X 1994..................... 46
4-9: Physical properties o f the second ice sh eet grown during C R R E L E X 1994.............................. 47
4-10: Snow salinities from the second ice sheet at C R R E L E X 1994.................................................... 48
4-11: Tim e sequences o f Tb(H ) from the grow th o f sheet 2 at C R R E L E X 1994.............................. 49
4-12: The effect o f snowfall and brine w icking from sheet 2, C R R E L E X 1994................................50
5-1: Theoretical model sim ulations of em ission from optically thick ice.............................................. 69
5-2: Sim ulations o f a .7 m m brine slick over optically thick ic e ............................................................. 70
5-3: Fresnel m odel output fo r optically thick ice with no b rin e............................................................... 71
ii
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
5-4: Theoretical calculations o f the m icrow ave signature o f snow over sea ice................................. 72
5-5: C alculations o f the m icrow ave signature o f a very th in snow over sea ice................................. 73
5-6: Brine distribution used in the m icrow ave sim ulations o f frost flo w ers.........................................74
5-7: Sim ulations o f frost flow ers over sea ice for a m inim um ice volum e o f 5 % ...............................75
5-8: Possible scenario o f thin ice signature evolution.................................................................................76
5-9: Sim ulations o f frost flow ers over sea ic e .............................................................................................. 77
5-10: Second possible scenario o f thin ice signature ev o lu tio n ............................................................... 78
6-1: M eteorological data fo r th e Bering S ea during the spring o f 1988................................................ 93
6-2: Regional m aps o f Tb(19, H ) for the B ering Sea from 1988........................................................... 94
6-3: Bering S ea current field as modeled by O verland and Roach (1987).......................................... 96
6-4: M ap o f T b(19, H) from A pril 2 ,1 9 8 8 w ith the location o f the three sites...................................97
6-5: Tim e series o f Tb(19) at three sites dow nw ind o f St. L aw rence Islan d ........................................ 98
6-6: Tim e series o f PR(19) fo r three sites dow nw ind o f St. Law rence Islan d .....................................99
6-7: Tim e series o f Tb(37) at three sites dow nw ind of St. L aw rence Islan d ...................................... 100
6-8: Data used in correlation studies o f em ission vs. environm ental co n d itio n s...............................101
6-9: M aps o f T b(19, H) from m orning and afternoon on successive d ay s..........................................103
7-1: W ater d rag coefficient as a function o f depth in the w ater colum n..............................................114
7-2: C om parison o f the drift m odel with buoy data from M uench at al.( 1988)................................115
7-3: M ethod o f producing fields o f ice age and thickness using the drift m o d e l...............................116
7-4: Ice age as a function o f T b(19,H) betw een M arch 11 and April 10, 1988................................ 117
7-5: Ice thickness as a function o f Tb(19,H ) betw een M arch 11 and A pril 10, 1988......................119
7-6: Ice age vs. T b(19, H) for ice parcels originating south o f St. L aw rence Island........................120
7-7: Ice thickness vs. Tb( 19, H) for ice parcels originating south o f St. L aw rence Islan d.............121
7-8: Ice age vs. T b(37, H) fo r ice parcels originating south o f St. L aw rence Island........................122
7-9: Ice thickness vs. Tb(37, H ) for ice parcels originating south o f St. L aw rence Islan d.............123
7-10: R egression analysis o f ag e and thickness vs. three passive m icrow ave param eters............. 124
8-1: G eophysical param eters predicted from passive m icrow ave satellite d a ta ................................ 132
iii
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
L
is t o f
Tables
6-1: M ultilayer Fresnel m odel sim ulations o f diurnal variations in em ission ..................................102
7-1: R egression equations fo r ice age and thickness vs. passive m icrow ave p a ra m e te rs..............118
9-1: Shortw ave radiation param eters used in the therm odynam ic m o d e l...........................................143
iv
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
A
cknow ledgm ents
I w ish to thank T om G renfell and G ary M aykut fo r their w illingness to oversee m y graduate
career. I thank Tom fo r his patience and help in teaching me experim ental m ethods in the
observation o f m icrow ave emission fro m sea ice. M ore importantly, he was a source o f calm and
reassurance during the frantic and not alw ays happy tim es that go w ith field research. Gary has
proved an unfailing source o f encouragem ent, trust, and friendship. O f the many things he has
taught m e, I will never forget to alw ays consider the broader im plications of my w ork.
I also w ish to thank D ale W inebrenner. Dale has contributed m any useful insights into
m icrow ave theory and has always im pressed me w ith his willingness to don the h at o f
experim entalist. N orbert U ntersteiner continues to b e a source laughter. His opinions are
outspoken, often unconventional, and alm ost always correct.
Paul B oynton gave m e my initial lessons in the art and craft o f science. His enthusiasm ,
tem pered w ith clear and critical thinking continue to be an exam ple fo r my own w ork. If it hadn’t
been for Paul, I probably w ouldn’t be a scientist.
Finally, I wish to than k the most im portant person in my life. R egan W ensnahan has never
faltered in her support o f m e or my w ork. She has given m e insights into what it m ean s to learn
and, by d o in g so, has im proved my ow n education. She has been w illing to stand by m e in both
good and bad times. F o r this, she deserves my deepest thanks.
This w o rk was supported by the O ffice o f N aval Research and N A S A under contract
N 00014 -8 9 -J-1 140, by the Office o f N aval Research u nder contract N 00014-90-J-1075, and by
N A SA u n d er contracts N A G W -2574 and N G T-50723.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
1. I n t r o d u c t io n
Sea ice is an im portant com ponent o f the Earth's clim ate system. It affects not only global
albedo, but also energy exchange and boundary layer structure in the p o lar atm osphere and
oceans. Interactions betw een the atm osphere and polar oceans are strongly influenced by the
thickness o f the intervening ice. For exam ple, ice o f less than 50 cm has turbulent heat fluxes, ice
production rates and salt fluxes to the ocean that can be 1 to 2 orders o f m agnitude g reater than
that of thick first-year (FY ) ice or m ultiyear (MY) ice (M a yku t, 1978). T hickness-dependent
differences in albedo also affect solar energy input to th e upper ocean (G renfell, 1983) and the
oceanic heat flux at the underside o f the ice (M aykut a n d Perovich, 1987; M aykut a n d M cP hee,
1995). This strong thickness dependence m eans that relatively small areas o f open w ater (OW ),
nilas and young ice could dom inate the regional heat and m ass balance o f the ice pack (M a yku t,
1982). T his is especially true in areas such as the southern polar seas, w here large am ounts of
nilas and young ice are present ( W orby a n d A lliso n, 1992). M onitoring exchanges betw een the
ice, ocean, and atm osphere therefore requires the ability to determ ine n ot only ice extent, but also
the distribution o f ice thickness - especially ice less than 50 cm thick.
Sea ice m onitoring currently depends largely on data from passive m icrow ave satellites which
provide all-w eather coverage on a global scale. The m ost useful instrum ents have been the
Scanning M ultichannel M icrow ave R adiom eter (SM M R ) and the Special Sensor M icrow ave
Im ager (SSM /I) which m easure m icrow ave em ission from the surface at various frequencies
between 6.7 and 85 GH z, at both vertical and horizontal polarizations (V p o l and H p o l,
respectively). These data have been utilized in a num ber o f algorithm s to predict not only ice
extent and concentration, but also the presence of two thick ice classes, F Y and M Y ice
(C avalieri et al., 1984; S w ift e ta l., 1985; Svendsen et al., 1987; W alters e ta l., 1987; Thom as
a n d Rothroch, 1989).
Indications from surface and aircraft based m easurem ents are that nilas and young ice can
have m icrow ave signatures significantly different from that o f O W and thick ice. (C a m p b ell et
al., 1975\ C avalieri et al., 1986; G renfell, 1986; C om iso et al., 1984, 1989; T ucker e t a l., 1991,
G renfell e t al., 1992). Fig. 1-1 dem onstrates these differences in terms o f the m icrow ave spectral
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
characteristic o f polarization ratio (PR). The PR is a m easure o f th e difference in brightness
tem perature (Tb) at H and V pols and is defined as
T . ( v , V ) - T (v ,H )
P R (v ) = — ---------------5--------Tb ( v , V ) + T b (v ,H )
( 1- 1)
where v is a particular frequency. T he thin ice observed in these studies has a P R betw een that o f
O W and thick FY ice. Further, the P R is sm aller for older, thicker form s o f ice.
Based on these results it might seem like a simple m atter to detect thin ice in the passive
m icrow ave data. H ow ever, studies o f satellite data by Steffen a n d M a slanik (1988),
C o m iso (l9 8 6 ) and R othrock et al. (1988) have suggested that nilas and young ice are not readily
distinguishable from m ixtures o f O W , F Y and M Y ice. In Fig. 1-2, for example, the thin ice
points lie along the line betw een O W and FY ice and hence would be indistinguishable from
mixtures o f w ater and thick ice. By the sam e token, algorithm s that solve only fo r O W and thick
ice can have errors o f o v er 25% in areas o f nilas and young ice (C a va lieri et al., 1986). The
whole question o f the nature o f the signature of nilas and young ice rem ains largely unanswered;
m ost o f o ur present know ledge about the m icrowave spectra o f nilas and young ice has been
obtained from general field surveys w here little inform ation is available regarding either the
evolution o f the spectra w ith tim e or its relationship to specific environm ental conditions during
growth.
W ensnahan et al (1993a) have provided some insight into the evolution o f the thin ice
signature. They presented tim e sequences o f passive m icrowave em ission taken o v er ice o f less
than 10 c m in thickness. T hree o f these sequences w ere from artificial sea ice grow n at the Cold
Regions Research and Engineering Laboratory (C R R E L) in H anover, N ew H am pshire during the
years 1988, 1989 and 1990, while a fourth was m ade o f a refreezing lead in the G reenland Sea
during the C om bined E ast A rctic R esearch E xperim ent (CEA REX ) in 1989.
These four sets o f experim ental results share a num ber of com m on features w hich appear to
be characteristic o f m icrow ave em ission from very thin saline ice. T h e first is a rapid initial rise
in brightness tem perature (T b) from the relatively low values o f open w ater to the higher values
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
of thick ice (Fig. 1-3). T h e thickness at w hich T b reached m axim um o r near m axim um values
varied w ith frequency; at 90 GHz it occurred when the ice was no m o re than 1 cm thick while at
6.7
G H z the ice was over 6 cm thick. Sim ilar behavior has been reported by G renfell and
Com iso (1986) and Sw ift et al. (1986). T he second com m on feature w as a sudden increase in
surface tem perature w hen the ice was between 1 and 2 cm thick. B e fo re this occurred the surface
tem perature had slow ly fallen as the ice increasingly insulated the surface from the underlying
warm w ater. The sudden w arm ing o f the surface could not be attributed to environm ental causes
and appears to be due to the transport o f w arm brine from within the ice onto the surface. The
third feature was a decline follow ing a m axim um in T b at 19, 37 and 9 0 G H z but not at 6.7 or 10
G H z. T he decrease in T b(37 GHz) occurred ju st after the surface tem perature increase, while
T b(90) began to decrease at the same tim e as or slightly before T b(37). D uring C R R E L E X 1988
and 1989 the 19 G H z signal didn’t drop until the ice w as around 3 cm thick. A t C R R E L E X
1990, T b(37) began to drop ju st before a dusting o f snow fell on the ice when it w as 2.7 cm thick.
W hile initially dry, the snow quickly m etam orphosed, turning to liquid and refreezing onto the
surface. B y the tim e the ice reached 4 cm in thickness, no trace o f the snow crystals remained.
The m ajor effect o f this snow was to increase the w etness o f the ice surface for several hours.
O nce th e surface becam e w et, Tb(19), T b(37) and T b(90) dropped sim ultaneously. In all of the
C R R EL experim ents the decreases in T b persisted to a thickness o f at least 6 cm an d were larger
at H pol than at V pol (about 30-50 K vs. 15-20 K). T h e largest drop occurred during
C R R E L E X 1990, suggesting that the increased surface w etness during snow m etam orphosis may
have enhanced the decrease in Tb. N um erical m odeling ( W ensnahan e t al, 1993a) has shown that
even a sm all am ount o f brine on the surface of sea ice can dram atically change its m icrowave
properties. Brine and high salinity layers at the surface o f the ice are a com m on observation for
nilas and young ice in the A rctic as w ell as the ice at C R R E L (M a rtin , 1979; G renfell et al.,
1988).
Taken together these com m on features suggest a pattern in the evolution o f the microwave
spectrum o f sea ice grown in relatively quiescent conditions. As the ice initially form s, emission
at all frequencies respond to changes in thickness. O nce the ice is 1 to 2 cm thick, upw ard brine
transport due to expulsion begins to carry warm brine from the low er reaches o f the ice up to the
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
surface. A s a consequence, the surface tem perature increases and T b decreases at frequencies
greater than 19 GHz. A s long as the surface remains w et, Tb rem ains depressed.
The observations from these experim ents allowed W ensnahan e t al. (1993b) to address two
questions concerning the rem ote sensing o f sea ice: (1) how does the presence o f nilas and young
ice affect the com m only used algorithm s; and (2) are the spectra o f nilas and young ice
sufficiently distinct from that o f open w ater and older ice such that it w ould be possible to detect
them from space?
They began their analysis by plotting the field data in terms o f a com m only used satellite
algorithm fo r the interpretation of m ultichannel m icrow ave data from the polar oceans, namely
the N A SA Team algorithm ( C avalieri et al., 1984; G loerson and C avalieri, 1986). T he
developers o f this algorithm argued that only the three dom inant surface types (com m only
referred to as the tiepoints o r endm em bers) o f OW, FY , and M Y ice could be readily
distinguished in the A rctic. To reduce the direct effect o f surface tem perature on T b, the N A SA
Team algorithm uses ratios to characterize differences in the endm em ber spectra. O ne o f these is
the PR at 19 G H z and the other is the gradient ratio (G R ) which is a m easure o f the slope of the
spectrum fo r two frequencies. For the N A S A Team algorithm it is defined as
GR(I9V. 37V) = A g .V )-T ,(1 9 .V )
Tb(37,V) + Tb(19,V)
where 19 and 37 refer to the m icrow ave frequencies (in GHz).
The O W , FY and M Y ice endm em bers form a triangle within the space defined by PR and GR
(Fig. 1-4). D ata whose signature falls w ithin that triangle can be interpreted as m ixtures o f those
three endm em bers. The C R R E L EX 1990 data (also plotted in Fig. 1-4) started w ith O W and, as
the ice grew , the data described a path from O W to F Y ice. Once a significant am ount o f liquid
had accum ulated on the surface, the P R increased and the data settle into a cluster in the middle
o f the triangle, with a P R o f around 0.8 and a G R near zero. D espite representing only thin ice of
uniform thickness, this cluster o f points w ould be interpreted by the N A SA algorithm as around
40% OW , 40% FY and 20% MY.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
T o avoid this kind of m isinterpretation o f satellite data, algorithm s w hich take into account the
presence of thin ice m ust be developed. W ensnahan et al (1993b) conducted a series o f
supervised principal com ponent analyses (PC A ) on the field data from C E A R E X and the CRREL
experim ents to determ ine w hether it would be possible to detect thin ice using passive m icrowave
data (for an explanation o f this technique see the Appendix: M ethods o f P rincipal C om ponent
A nalysis). It was determ ined that tw o distinct classes o f young ice (referred to as new ly-form ed
ice and thin ice) w ere detectable using the full suite of m icrow ave frequencies from 7 to 90 GHz.
H ow ever, using only 19, 37 and 90 G H z (the channels available for the current SSM /I satellite)
only thin ice could be accurately differentiated from OW , FY and MY ice (Fig. 1-5).
T he distinguishing features o f the spectra o f these four endm em bers w ere the m agnitude o f the
6 channels of T b, the separation between H and V pol and the slope o f the spectra betw een
different frequencies. The first principal com ponent score (Pi) was a m easure o f the overall
m agnitude of em ission and as such separated O W with low levels o f em ission from thin and FY
ice with much higher levels. M Y ice lay betw een the two extrem es due to reduced em ission at 37
and 9 0 G H z. T he second score, P2, responded to differences in the slope o f the spectra and
differentiated the negative slope o f M Y ice from that o f the other scene types. Finally, thin ice
w as distinguished from FY ice by P3 which indicated the degree of polarization by contrasting
T h(V ) at 19 and 37 G H z with T b(H ) at 37 an d 90 GHz.
It is apparent from this and previous work that thin sea ice with a brine ladened surface has a
passive m icrow ave signature different from that o f w ater or thick ice. Further, current algorithm s
designed to solve only for concentrations o f thick ice types are subject to substantial error in
regions o f thin ice. T h e PCA o f C R R ELEX d ata indicated that, at least in theory, it should be
possible to detect thin ice as a separate and unique class o f ice in satellite data. H ow ever, it
rem ained to be seen w hether this potential could be exploited in actual satellite data.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
6
0 .3
w a te r
o
0.2
w a ter
w ater
w ater
w ater
new
CO
c
o
d a rk n ila s
CO
N
thin(bare)
CO
I
nilas
O 0.1
nilas
pancakes
Q.
' g re y
grey w hite
s h o re fa s t
grey
^ " 9
-F
white
first y e a r
first y e a r
thick(snow )
0.0
%
%
e,
y,
S'
%
Oo
%
Fig. 1-1: Polarization ratio as a function o f ice type.
All values have been scaled to be the same for w ater. D ata are at 18-19 G Hz except
C am pbell et. al. (1975) which is at 11 GHz. (Inform ation taken from a table in Cavalieri
(1993)).
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7
0 .1 6 -
Bering Sea, 1986
0 Cavalieri et al. (airplane)
A Grenfell (surface)
V , ..
“ W a te r
0 .1 2 0 New
0 .0 8 -
^ B a r e nilas
0 .0 4 -
O Y oung
O FY
^ T h ic k + s n o w
i
i
i
i
Steffen a n d U aslanik
0 . 1 6 - (Baffin Bay, .1988, SMMR satellite)
a.
i
i
i
i
W a te r
0 .1 2 -
0 .0 8 -
1 Nilas
0 .0 4 ^ H G re y
0 . 0 0 -^
0 .0 0
W hite
]
■ 'i.........
0 .0 4
!
i
0.08
!------------ ,i
i
0 .1 2
b.
—
0 .1 6
P o la riz atio n ratio(37 GHz)
Fig. 1-2: P olarization ratios at tw o frequencies as a function o f ice type.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Brightness temperature (K)
8
250
-2 5 0
200
-200
150
150
6.7 GHz
100
a . H pol
0
2
3
4
Ice th ic k n e ss (cm)
5
6
b. V pol
0
1
2
3
4
5
-100
6
Ice thick n ess (cm)
F ig. 1-3: B rightness tem perature record fo r the ice sheet o f C R R E L E X 1990.
Both H pol (a) and V pol (b) data are show n. The thickness at w hich a significant rise in
surface tem perature occurred is noted on th e H pol plot. A dditionally, the point at w hich a
trace o f snow fell on the ice is denoted by a line m arked (1). M elting o f the snow cau sed an
increase in surface w etness fo r an extended period and is noted by the shaded region. A fter
several hours the snow had m elted and refrozen onto the surface such that by the tim e o f the
line m arked (2) there w as n o trace left, (from W ensnahan e t al., 1993b)
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
9
OW
0 .0 5 -
>
>-
O)
2.4
6.2
0 .0 0 -
4.3
FY
cc
C3
3.0
- 0 .0 5 -
(3E)
A v e E rro r
MY
- 0 .1 0 4 —
0.00
0 .0 5
0.10
0 .1 5
0.20
0.2 5
PR(19)
Fig. 1-4: Grow th sequence data from C R R E L E X 1990 saline tank ice.
D ata are plotted in the context o f the PR and G R used by the NASA T eam sea ice
algorithm. Included in the plots are the triangles formed by O W -FY -M Y mixtures as w ell as
the average uncertainty o f P R and G R for the data. The ice thickness (in cm ) o f selected
points indicated by triangles are also given, (from W ensnahan et al., 1993b)
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
10
175 Thin
Newlyform ed
CL
FY
MY
75 -
1
1
40 Thin
MY
20
-
Q.
OW
Newlyform ed
-* •
FY
350
450
550
Pf
Fig. 1-5: C luster plots o f the C R R E L E X 1990 in principal com ponent space.
A xes are the first three eigenvectors o f a supervised principal com ponent analysis using the
equivalent o f the 6 channels o f SSM /I m icrow ave data from 19-90 G H z at both
polarizations. T he dotted line shows the volum e defined by possible m ixtures of the
endm em bers. (from W ensnahan et al., 1993b)
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
2 . S a t e l l i t e c a s e s t u d y : B e r in g S e a , S p r in g 1988
A s an initial test o f the potential o f space based detection u f thin ice, a satellite case study was
conducted using S S M /I data to attem pt to answ er three questions: does a signature with spectral
characteristics like the thin ice from the CR REL experim ents exist in the satellite data? Is the
signature unique such that an algorithm can be form ed to solve for mixtures o f it and OW , FY
and M Y ice? Finally, is the new scene type actually thin ice o f som e kind?
A prom ising region for this kind o f work is th e Bering Sea (Fig. 2-1) w hich is noted for the
frequent form ation o f wind driven polynyas (S m ith et al., 1990). U sing 25 km gridded S SM /I
data from April 2, 1988, PR(19) and GR(19V , 37V ) were calculated for this region (Fig. 2-2).
M ost o f the points in the cluster p lot lie along the O W -FY line but there is a distinct cluster o f
points extending from FY to the m iddle of the triangle. The spectra o f the circled points at the end
o f this cluster are characterized by high brightness tem peratures, a large P R and a V pol G R near
zero (Fig. 2-3) - m uch like the thin ice of C R R E L E X 1990. T he exact position o f the end o f the
cluster does, how ever, differ som ew hat from th at o f the field data. T he cluster extends further
into the triangle for the Bering S ea data, resulting in a higher P R than that observed at C R R E L
and a G R that is slightly positive rather than slightly negative. It is possible th at the apparent thin
ice in the Bering S ea represents a different stage in the signature evolution, o r that the ice has
distinct surface characteristics caused by differences in environm ental conditions.
2.1 Principal component analysis
A supervised P C A was carried out using the endm em ber spectra for OW , F Y , M Y, and the
apparent thin ice (Fig. 2-3). The O W spectrum is ju st the average o f a large num ber o f pixels in
the southern part o f the study region which fall near the OW p o in t in Fig. 2-2. T he FY ice
spectrum is the average o f the points near the F Y ice endm em ber in the PR -G R cluster plot o f
Fig. 2-2. The M Y ice spectrum is the SSM /I tiepoint used in the N A SA T eam algorithm. Finally,
the spectrum o f w hat w e assume to be thin ice is the average o f spectra from points at the right
end o f the cluster in Fig. 2-2.
T hree eigenvectors were found (Fig. 2-4), all spanning variances above the 9 K 2 noise level
(H o llin g er et. al., 1990). As with th e C R R E L E X data, the positive coefficients o f the first
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout p erm ission .
eigenvector em phasize changes in the magnitude o f T b. However, in this case the second vector
prim arily contrasts T b(19H ) and T b(37V ) and the th ird com pares T b(19V ) and T b(37H ). C luster
plots in P C space show good separation between endm em bers (Fig. 2-5) with m uch o f the data
falling in a region defined by m ixtures o f OW, FY, and the possible thin ice.
U sing these three eigenvectors, w e formed an algorithm for all four scene types and then
m apped the concentrations in the study area (Fig. 2-6). Both the N A S A Team algorithm and the
PCA algorithm predict the same location for the ice edge. In both cases the ice zone in the Bering
Sea is dom inated by FY ice, with predictions o f M Y in the C huckchi S ea and in isolated pixels on
the A laska coast; the latter are m ost likely either fast ice or contam ination o f the data by land.
The study area also includes two signatures that are, to some extent, anom alous com pared to
OW , FY , M Y and the possible thin ice. Pixels along the eastern portion o f the ice edge are
predicted by both algorithm s to include negative M Y ice concentrations, as well as negative
concentrations o f thin ice in the PC A case. W e suspect that this signature is the result o f either
m elting o r flooding o f the ice. There are also estim ates o f positive and negative concentrations o f
ice in th e O W south o f the ice edge. T h is is probably due to w eather effects such as w ind
roughening which cause the O W spectrum to be highly variable (note the sm earing o f the O W in Fig. 2-2 and Fig. 2-5). It is ju st such variability in the O W signature that inspired the
developm ent o f a w eather correction filter for the N A SA Team algorithm (G loerson and
C avalieri, 1986).
The striking difference between the tw o algorithm s is in the interpretation o f th e areas
dow nw ind o f land. W hile both predict a large area o f O W east o f N univak Island, the
interpretations o f the areas dow nw ind o f St. Law rence Island, the Sew ard Peninsula, and the
Siberian coast are very different. T he N A SA algorithm predicts 50-60% O W and 20-30% M Y
ice w hile the PCA algorithm calculates O W as no m ore than 20-30% (and that only near land)
with M Y ice concentration near 0% throughout. Instead, these regions are predicted to be heavy
concentrations o f thin ice extending dow nw ind o f the land edges, eventually tapering into 100%
FY ice.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
13
2.2 Independent evidence of thin ice
H aving show n that a signature like the thin ice o f CR R E L E X 1990 exists in the satellite data
and that it is unique com pared to O W , F Y and M Y ice, we have yet to dem onstrate that this new
scene type is in fact thin ice.
C onsider first the environm ental conditions. In late M arch o f 1988, a high pressure center
over Siberia and a low pressure center o ver the G ulf o f A laska resulted in strong w inds from the
north. U nder these conditions polynyas are formed as the ice is driven away from the coasts
{Pease, 1987). Frazil crystals form ed in the polynya are advected southw ard, eventually
consolidating into a solid ice sheet within a few tens o f kilometers. T he new ly-form ed ice
continues to thicken as it is driven southw ard at a rate o f up to 3% o f the w ind speed {Pease a n d
S a lo , 1987; Pease an d O verland, 1984). A s long as offshore w inds persist, new ice is constantly
generated in the polynya and distributed downwind.
In late M arch and early A pril o f 1988 weather stations in the region reported w inds from the
north at 8 to 10 m/s. A ir tem peratures w ere relatively cold, ranging from -10° to -24°C , with very
little snow fall reported. F or these conditions, the O W polynya width should have been about 10
km {Pease, 1987). W ith a m axim um free ice drift rate o f 20 to 30 km /day, ice w ithin 100 km or
so o f land should have been relatively thin. In fact, D rinkw ater et. al. (1991) observed w hat they
called thin F Y ice during aircraft overflights dow nw ind o f St. L aw rence Island shortly after the
winds started blow ing out o f the north.
Considering the physical situation, the N A SA algorithm prediction o f thousands o f km 2 o f
high concentrations o f O W and M Y ice not likely to be correct. A t air tem peratures o f -10° to
-24°C, T he O W polynyas w ould have been restricted to narrow zones near land. Further, there
should be virtually no M Y ice in this seasonal ice zone since southw ard transport o f M Y ice
through B ering and A nadyr Straits is extrem ely limited.
Visible im agery from the A dvanced V ery High Resolution R adiom eter (A V H RR ) indicate that
large areas dow nw ind o f land had a low albedo, indicating either O W o r nilas and young ice (Fig.
2-7). Again, given the cold air tem peratures, these areas were m ost likely not O W and therefore
should have been som e form o f thin. ice. H eavy concentrations (>70% ) o f possible thin ice
predicted by the PCA algorithm correspond with the regions o f very low albedo in the visible
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
14
image. South o f St. L aw rence Island thin ice concentrations drop sharply beyond 150 to 200 km ,
going fro m 90% to 20% in a span o f 50 to 100 km. T his corresponds with the region in Fig. 2-7
where the surface changes from gray to white.
As a final note, M a sso m and C om iso (1992) analyzed corresponding A V H R R therm al
im agery to derive surface tem perature, achieving agreem ent o f ab o u t ± 1-2°C with ground based
m easurem ents. T heir w ork indicates th a t the surface tem peratures in the areas o f heavy thin ice
concentrations were around -6°, w ell below the freezing point o f O W . They used near infrared
and therm al infrared im agery to identify two types o f recently form ed ice in the B ering Sea during
this tim e; w hat they refer to as “new ice” and “young ice”. Their new ice occurs m ostly just
dow nw ind o f the O W in the Bering S ea polynyas, but their young ice extends far dow nw ind of
the polynya regions. T h e distribution o f the young ice they identify closely m atches both the dark
and light gray regions o f Fig. 2-7, as w ell as the area covered by concentrations o f >70% thin ice
predicted by the PC A algorithm .
T aken together, the w inds from the north, the cold air tem peratures, the predicted ice
production and transport in response to these conditions, the in-situ observation o f the initial
form ation o f the polynya and the A V H R R imagery leave little doubt that the predicted regions o f
heavy thin ice concentration are som e form o f nilas o r young ice. T hus, we were able to
distinguish thin ice using PCA on only four channels o f m icrowave data. Rather than the O W and
M Y ice predicted by the N A SA algorithm , the PCA algorithm yields a realistic distribution of
thin ice w hile still distinguishing O W F Y and M Y ice in other portions o f the study area.
2.3 From thin ice to FY ice
O ne intriguing feature o f this exam ple is that the P C A algorithm predicts a transition from
100% thin ice to 100% FY dow nw ind o f St. L aw rence Island (Fig. 2-6). This is a reflection o f a
progressive change in em ission (Fig. 2-8). Values o f T b( 19, H) w ere around 170K near land but
increased steadily to 23 0 K for ice 4 0 0 km dow nw ind o f land. S im ilar b ut sm aller changes in
em ission also occurred at V pol (from 230 to 250K) b ut these w ere prim arily restricted to within
100 km o f land. As a result of the differing responses at the two polarizations, P R ( 19)
progressively decreased dow nw ind o f land.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Ice further from land is older and thicker. Therefore, data from a single day represent a tim e
history o f the physical and microwave evolution o f the ice suggesting that changes in em ission are
som ehow related to changes in the age o r thickness o f the ice. A pparently, em ission at H pol
steadily increases (causing PR(19) to decrease) as the ice ages.
C avalieri (1993) used the apparent relationship betw een ice age and PR(19) to devise an
algorithm for interpreting SSM /I data in term s of ice categories. He exam ined data from the
Bering S ea in the Spring o f 1988 as well. His algorithm involved using the N A SA T eam
param eters o f PR(19) and GR(19V , 37V ) to solve for concentrations o f OW , FY and tw o classes
o f thin ice, new and young ice. Essentially the cluster o f points in Fig. 2-2 extending from the FY
tiepoint to the middle o f the PR-GR triangle was divided into categories o f new, young and FY
ice based on PR. M aps o f ice class derived from SSM /I were then com pared with sim ilar maps
derived by M assoni a n d C om iso (1992) using A V H R R imagery. S ubstantial agreem ent was
found betw een the two interpretations.
A ssum ing that there is a correspondence between ice age and PR (19), we are left with
som ething o f a mystery. W hat are the physical m echanism s that cause the thin ice spectrum to
change to that of FY ice? A re the physical m echanism s, and therefore the m icrowave evolution,
dependent on the environm ental conditions during ice grow th? Can the m icrowave data be used to
produce m aps of actual ice thickness rather than som ew hat vague classes o f ice? W e will
exam ine these questions in the next few chapters.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
16
Chukchi Sea
Siberia
Alaska
Bering Sea
Siberia
North
/
G ulf o f
Anadyr
AnadyrStrait
/
S t. Law rence Is.
St. M atth ew Is.
Bering Sea
Alaska
N u nivak Is.
SATCASCorg
IM-BS
Fig. 2-1: SSM /I land m ask m ap o f the Bering S ea case study area.
Pixels are 25 km by 25 km.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
17
0.08
apjsejR*-: ow
0.04
m
>
o.oo
A pparent thin ice
FY'
ON
MY'
0.0
0.1
0.2
0.3
Fig. 2-2: N A SA T eam algorithm cluster plot o f 25 km S SM /I data.
Date are from the B ering S ea on April 2, 1988. The triangle is defined by the S SM /I
tiepoints used in the T eam algorithm. T h e end of a clu ster o f apparent thin ice is also
indicated
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Brightness Temperature (K)
18
b . A p p a re n t T h in Ic e
Frequency (GHz)
SAtCASiag
Fig. 2-3; E ndm em ber spectra used in the PCA o f the Bering S ea case study.
Including (a) O W south o f the Bering Sea, (b) apparent thin ice from Fig. 2-2; (c) F Y ice
from the B ering Sea and (d) the M Y ice SSM /I tiepoint from the central Arctic. Frequencies
are 19 and 37 G H z with V pol indicated by circles and H pol by squares.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
19
1.0
a
0.5
OJ
e
§
EZUl
,1
0.0
1
o
iS
1.0
________
19h
19v
P^:q2= 48K2
P7: q 2 = 586 K2
P : o 2 = 4441 K2
-
37h
37v
19h
19v
37h
37v
19h
19v
37h
37v
C hannel
.urcun»r<
hum
Fig. 2-4: E igenvector coefficients o f a supervised PCA o f the SSM /I Bering Sea data.
The variance spanned by each vector is also given.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
20
Apparent Thin
Apparent Thin
"#n
Apparent Thin
iV uji 1
UtTASa,
V )tM
Fig. 2-5: C luster plot o f th e Bering S ea data in PC space.
A xes are principal com ponent scores defined by the first three eigenvectors o f a PCA using
the endm em ber spectra of Fig. 2-3.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
21
NASA team algorithm
Principal component algorithm
WTrAM-Xn-MMri
Fig. 2-6: C oncentration m aps fo r the B ering Sea on A pril 2, 1988.
Pixel size is 25 km by 25 km . High concentrations are show n in red and low concentrations
in blue. T he top three panels show predictions o f the N A SA Team algorithm in term s of
OW , F Y and M Y ice. T h e bottom fo u r panels show results from the P C A algorithm for
OW , F Y , M Y and thin ice. C oncentrations o f g reater than 100% o r less than 0% represent
spectra w hich cannot b e entirely explained as sim ple m ixtures o f th e endm em bers.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
22
Siberia
Alaska
St. Law rence Is.
St. M atliew Is.
Fig. 2-7: A V IIR R visible image o f the Bering Sea study area.
Im age is for th e m ost part cloud free. The dark and gray areas south o f St. Law rence Island
and the Siberian and Alaskan coastlines have been interpreted by M assom a n d Comiso
(1992) as consisting o f young ice.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
23
260
e
-0 .1 6
240
-
0.12
.220
-0 .0 8
200
PR(19)
£
St. Law rence Is.
-0 .0 4
180
400
600
8 00
400
600
800
Distance along line (km)
Fig. 2-8: C hanges in Tb(19) and PR(19) dow nw ind o f St. Law rence Island.
a) m ap o f T b(19, H ) fro m April 2, 1988. b) Tb(19, H ) and T b( 19, V ) as function o f distance
along the line indicated on the map. c) PR(19) as function of distance along the line.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
3. F u n d a m e n t a l s
o f m ic r o w a v e e m is s i o n f r o m s e a i c e
The am ount of radiation em itted by thin sea ice is directly related to the dielectric properties of
the ice. T he dielectric properties are in turn determ ined by the physical properties o f the ice such
as its tem perature, salinity and crystal structure. To understand the evolution o f the thin ice
signature or spectrum, w e will need to relate changes in the physical properties o f the ice as it
thickens and ages to em ission from the ice. This chapter describes the rudim ents o f m icrowave
theory pertaining to thin sea ice; an in depth treatm ent is given in C hapter 5.
For ice that is isotherm al at tem perature T , the brightness tem perature (Tb) m easured by
ground-based instrum ents is a com bination o f em ission from the ice plus reflection o f radiation
em itted by the atmosphere:
T. = e*T. + R*T
b
i
a
(3-1)
v
J
where e is the em issivity o f the ice, R is reflectivity o f the ice and T a is the brightness tem perature
o f the atm osphere. A ll o f the quantities in this equation are functions o f frequency, except for Tj.
In addition, both e, R and T b are functions the polarization and the angle betw een the propagation
vector o f the em itted radiation and a vector norm al to the ice surface, know n as the nadir angle 9 .'
Tn the sam e way that atm ospheric radiation is partially reflected by the ice surface,
electrom agnetic waves originating w ithin the ice are partially reflected back into the ice. Only the
radiation that avoids reflection contributes to em ission from the surface. A further reduction in
em ission occurs if upw ard propagating radiation is scattered by air o r brine inclusions w ithin the
ice. The em issivity can be written as the sum o f these processes
e = 1- R -a
(3-2)
where a describes the effect o f scattering by the ice. T he scattering term is frequency dependent
and accounts for both volum e and surface scattering. If the scattering elem ents w ithin the ice or
at its surface are much sm aller that th e w avelength o f the radiation then a is negligibly small. As
the size o f the elem ents increases, Tb decreases with the largest loss occurring at the highest
frequencies. The dom inant scatterers in sea ice are heavily m etam orphosed snow as well as ice
that has undergone significant m elting and drainage. T he scattering term is particularly im portant
for M Y ice. Since loss is greater at 37 than 19 GH z, scattering causes the large negative
GR( 19,37) typical o f M Y ice. Scattering is generally negligible for young ice. Even for snow
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
25
covered FY ice, the effect o f scattering is apparently small enough at 37 G H z that it causes only
a slight depression o f G R.
The reflectivity o f a surface is prim arily a function o f the difference in the com plex relative
perm itivities (e) of the tw o media. T he larger the contrast in perm itivities, the larger the reflection
coefficient and the sm aller the em issivity and brightness tem perature. The reflectivity at H pol is
more sensitive than V pol to changes in the perm ittivity contrast. T herefore, as the contrast
increases, Tb(H) decreases more than Tb(V).
At m icrow ave frequencies, the bulk perm ittivities o f sea ice and snow are approxim ately
volume w eighted sum s o f the perm ittivities of air, brine and pure ice. (This is a crude
approxim ation, more accurate approxim ations will be derived in C hapter 5). T h e perm ittivity o f
air is essentially 1, w hile that o f pure ice is relatively small at around 3.2+.00H . B oth are
essentially independent o f frequency. T he perm ittivity o f brine (eh) is large at all frequencies w ith
the highest values at the low est frequencies. For exam ple, £b(7 G H z) is approxim ately 55+40i
while £b(90) is 7+9i. E m ission from polar surfaces, especially at lo w er frequencies, is therefore a
strong function o f the am ount and distribution o f brine or liquid w ater within the ice or snow.
W e can now begin to explain the physical causes for m icrowave signatures observed in the
Arctic. T here is, for exam ple, has a high perm ittivity contrast betw een the ocean w ater and the
atm osphere. This m eans the reflectivity o f the ocean is high and T b values are low . Further, Tb(H)
is much low er than Tb(V), resulting in a large PR. S now and air have a small perm ittivity
contrast so that Tb(H) and Tb(V) are high and PR is small.
If the dielectric properties of a m edium vary with depth then the em issivity w ill depend on
more than ju st the reflectivity o f its surface. For exam ple, m icrow ave radiation m easured over
very thin ice originates both within the ice and the underlying w ater (G renfell a n d C om iso ,
1986). H ow ever, radiation em itted by the water is at least partially absorbed in th e ice. A s ice
thickness increases, absorption increases and the contribution o f the w ater to surface em ission
decreases to near zero.
The decrease in the am plitude (E) o f an electrom agnetic wave as it propagates through a
horizontally hom ogeneous absorbing layer is given by
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
26
E = e -kvZ /cose
( 3 _3 )
where k " is the absorption coefficient at a particular frequency v, z is the thickness o f the layer
and 0 is the nadir angle. T he absorption coefficient is a function o f the perm ittivity o f the medium
and is defined as
k" = I m p f^ /e ) '
(3-4)
where c is the speed o f light in free space.
The quantity in the exponent o f (3-3) is known as the optical depth and the value o f z at which
the optical depth equals one is known as the penetration depth. A t an optical depth o f 5 alm ost all
o f the radiation incident on the layer w ill be absorbed, at which point the layer is said to be
optically thick. At m icrow ave frequencies, the absorption coefficient increases with frequency.
Even so, it is sm all enough for pure ice that it w ould take at least 2 m eters o f m aterial for ice to
be optically thick. The absorption coefficient of brine is very large; an optically thick layer of
brine is .5 m m thick at 7 G H z and .04 m m thick at 90 GHz. A bsorption in sea ice and snow is a
strong function o f the fractional brine content, just like the reflection coefficient.
From K irch h o ff s law , it is known that a material th at is a good ab so rb er o f radiation at a
particular frequency is also a good em itter at that frequency. O ptically thin layers ab so rb very
little radiation and, consequently, em it very little radiation. In the exam ple o f thin ice o ver water,
very little radiation is em itted by the optically thin ice com pared w ith the underlying water.
A dditionally, there is a strong perm ittivity contrast betw een the ice and water. U ltim ately, little
radiation is em itted from very thin ice and most o f the radiation em itted by the w ater is reflected
dow nw ard at the ice/w ater interface. A s the ice thickens, progressively m ore radiation is emitted
from the ice and, due to the relatively low permittivity contrast betw een the air and the ice, much
of that radiation is em itted causing the Tb to increase.
The intricacies o f perm ittivity contrasts, their effect on reflectivity, the role of optical depth
and vertical variations in perm ittivity can m ake it difficult to com prehend the physical factors
controlling the signature o f sea ice. Owe w ay to understand the physical causes of em ission from
thin sea ice is by applying these fundam entals to the interpretation o f observations, w hich we will
do in the n ext chapter.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
4 . F ie l d
o b s e r v a t io n s o f e m i s s i o n f r o m t h i n s e a i c e .
There are a num ber o f obvious differences in the physical properties o f thin and FY ice that
may affect m icrow ave em ission. For exam ple, thicker FY ice has a colder and less saline surface
and therefore a low er surface brine volume. G iven that a surface brine slick is an im portant factor
in producing the thin ice spectrum , it seems likely that reductions in surface brine will play a role
in the evolution o f the passive signature from thin to FY ice.
A nother difference betw een newly form ed ice and thick ice is the presence o f some form of
low density surface layer. O ften within a day or two o f initial growth, vapor deposition will
cause delicate tree-like structures to form on the surface o f thin ice. O v er tim e these fr o s t flo w ers
can populate the surface so densely that the ice is blanketed by them. C apillary action w icks brine
up onto the flow ers; both C ro cke rf1984) and Perovich a n d R ichter-M enge (1994) observed
salinities o f ov er 100 ppt on frost flow ers in the Arctic. D rinkw ater a n d C rocker (1988) used a
sim ple theoretical model to dem onstrate that the brine enriched flowers could significantly
influence the radar backscatter o f thin FY ice. It is likely that the sam e is true o f m icrow ave
em ission, although no studies have so far been conducted to test this.
In addition to frost flow ers, FY ice eventually becomes covered with snow , either from
snow fall or from snow blow n o ff the surrounding ice. M ost studies o f snow on sea ice have
concentrated on thick FY ice and M Y ice. U nlike snow on thick ice, snow on thin ice will
generally have large am ounts o f brine incorporated into it due to the w icking up o f liquid from the
surface o f the ice. T he liquid w ater content o f the snow may be further augm ented by an increase
in ice surface tem perature due to the insulating effect o f the snow and a reduction in freeboard
w hich may lead to flooding. T he net result is the form ation o f a slush layer at the snow /ice
interface. T he effect on m icrow ave em ission can vary radically from situation to situation. For
exam ple, G renfell a n d C om iso (1986) observed a 4.5 cm snowfall (the bottom 2 cm o f w hich
w as visibly w et) on 9 cm thick ice and m easured essentially no change in em ission at 10 GHz.
L ohanick (1993) observed a 2 0 cm snow fall on 25 cm ice and, over the course of tw o days,
m easured a 100 K drop at 10 G H z as a slush layer form ed at the base o f the snow. T hereafter,
the slush refroze and T b(10) rose by m ore than 100K and ended with values higher than before
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
28
the snowfall. O bviously snow is important to em ission but it is difficult to form system atic
conclusions from such diverse results.
To determ ine the effect o f physical changes in surface properties on em ission we have carried
out m icrowave observations o f thin sea ice, both in an outdoor laboratory setting and in the
Arctic. These field experim ents have allowed concurrent m easurem ents o f both em ission from the
ice and the surface properties o f the ice under a variety of environm ental conditions.
U nfortunately, w e normally cannot control environm ental conditions such as air tem perature,
dow nw elling solar radiation and snowfall, and it is these conditions that determ ine the physical
properties of the ice. Only over the years have w e accum ulated enough observations to piece
together a few general conclusions about the signature o f thin sea ice grow n in quiescent
conditions. In this chapter w e provide an overview o f previously unpublished m easurem ents.
From this collection o f m easurem ents, we then draw conclusions about the im pact surface brine,
snow , frost flow ers and surface roughness h av e on the m icrow ave signature o f thin sea ice and its
evolution as the ice thickens.
The observations reported below were all done with the sam e set o f portable m icrowave
radiom eters operating at the frequencies 6.7, 10, 19, 37 and 9 0 GHz. In m ost cases data w ere
obtained at both H and V pol. T h e m easurem ents were calibrated using both high and low
tem perature references (G renfell and L ohanick, 1985) and, unless indicated otherw ise, individual
m easurem ents are accurate to ±1 K. Physical properties o f the ice were generally obtained in
sim ilar ways at each experim ent. M easurem ents o f ice thickness were m ade by boring a hole in
the ice, inserting a metal probe in the shape o f an “L ” and draw ing up the horizontal leg o f the
“L ” against the bottom o f the ice; these m easurem ents were accurate to ±3-5 mm. Snow d epths
w ere measured w ith a ruler and w ere accurate to about ±2 m m . Salinity w as obtained from a
handheld optical salinom eter w ith an accuracy o f ±2 ppt.
4.1 CEAREX, 1989
In 1989, during the spring phase of the C om bined East A rctic Research E xperim ent
(CEA REX ), ship-based m icrow ave and physical property observations w ere m ade o f a variety of
ice types in the G reenland Sea. A short growth sequence from this cruise w as reported as p art o f
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
29
the thin ice w ork o f W ensnahan et al. (1993a, 1993b) and details o f the observation techniques
were given W ensnahan e t al. (1993a).
A sum m ary of station sites which include concurrent H and V pol m easurem ents is given in
Fig. 4-1. In PR-GR space the observations define the typical O W -FY -M Y triangle. T he O W
exhibits the sm earing o f PR typically caused by variable surface roughness. The M Y ice covers a
large range o f negative G R ’s due to strong but highly variable scattering o f 37 G H z radiation
typical o f old ice (G renfell, 1992). The FY ice data form a sm all clu ster with a PR o f around .02
and a slightly positive G R.
Several large fields o f nilas were encountered during the cruise and dual polarization
m easurem ents were m ade at tw o such sites. In both cases the ice was 5 -1 0 cm thick w ith rafting
prevalent throughout the fields. Both sites had a dense co v er o f frost flow ers around 2 cm thick
(Fig. 4-2). In both cases the air tem perature was around -18°C. The surface tem peratures were
m easured at locations near the ship as -10°C in one case and -17°C in the other. It isn’t clear why
the two sites had such different surface tem peratures, although it may be du e to differences in the
am ount o f rafting at the locations where the temperature was measured. N o m easurem ents were
made o f the salinity o f the flowers or the underlying ice. surface rior was it noted w hether the
surface o f th e ice was wet.
Even though the ice w as very thin at these sites, the T b at frequencies o f 19 G H z o r greater
were sim ilar to those o f F Y ice. At 7 and 10 GHz, the thin ice had low er T b values than those of
FY ice, perhaps due to the higher bulk brine volume o f the thin ice or sim ply because th e ice was
optically thin at those frequencies. The P R ’s is the thin ice was much sm aller than either those
observed at C R R E L or the thin endm em ber spectrum o f the SSM /I data. T his low PR w as likely
the result o f either a relatively dry ice surface, the presence o f frost flow ers or perhaps both, but
by them selves the observations provide no conclusive evidence of the physical cause.
4.2 CRRELEX, 1990
The C old Regions R esearch and E ngineering Laboratory has been th e site o f num erous
experim ents investigating the m icrowave properties o f thin sea ice (A rco n e e ta l., 1986; G renfell
a n d Com iso, 1986; S w ift et al., 1986\ G ren fell e ta l., 1988, W ensnahan e t al., 1993a). T he
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
30
facility includes a large outdoor tank o f saline w ater where observations o f ice form ation, grow th
and subsequent aging can be carried out.
O bservations m ade in 1990 during the first 6.2 cm of growth w ere sum m arized in the
Introduction. These data w ere all taken o f bare ice w ith essentially no snow o r frost flowers.
D uring the early hours o f January 15 a small am ount o f snow fell on this ice. E m ission was
m easured starting shortly after the snow fall and continuing for the next 6 hours (Fig. 4-3). T he
snow w as .5 cm thick at the tim e o f our m easurem ents. The top o f the snow appeared w hite and
relatively dry and had a salinity near 0 ppt. The snow at the snow /ice interface appeared gray and
slushy and had a salinity o f 25 ppt. From the C ox a n d Weeks (1983) equations for sea ice w e can
calculate an approxim ate brine volum e (vb) for the snow. With a snow /ice interface tem perature
o f -4° and a density o f 300 kg/m 3, the brine volum e o f the bottom o f the slushy layer would have
been around 8%.
T he addition o f snow caused a m arked change in the signature o f the ice. B oth PR and G R
increased sharply w ith the snow data lying well above the O W -FY line defined by the SSM /I
endm em bers. The response was a strong function o f frequency. T here was a large decrease in T b
at low frequencies, especially at H pol. A t high frequencies, T b(V) remained largely unchanged
but T b(H) increased dram atically. A s a result, T b(19) values for the snow covered ice were close
to those o f the thin ice endm em ber from the SSM /I data, but em ission at 37 G H z was essentially
the sam e as that o f F Y ice.
T his behavior is apparently a direct consequence o f the vertical brine distribution within the
snow and frequency dependent differences in the optical depth o f the snow. Since high
frequencies have a relatively shallow penetration depth, em ission at 37 and 90 G H z probably
originated alm ost entirely in the upper, relatively dry portions o f the snow. R elatively dry, but
optically thick snow , results in high T b(H) values and a low PR at these frequencies. At low er
frequencies the penetration depth is larger and em ission may have included significant
contributions from the w et slush layer at the base o f the snow. T he high liquid w ater content o f
the slush apparently drove dow n the T b, especially at H pol, w hich increased PR.
From this exam ple we conclude that dry, bare ice has a PR interm ediate betw een that o f FY
and thin ice. Further it w ould seem that the FY signature might be the result o f the presence o f
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
31
relatively dry snow on ice. In any event, the signature o f snow covered ice appears to be strongly
dependent on the am ount and distribution o f brine in the snowpack.
4.3 LEADEX, 1992
Sea ice forms a veneer over the A rctic Ocean. W ind and ocean stress can cause this veneer to
crack and deform, creating small areas o f w ater called leads. In cold w eather the w ater quickly
freezes form ing areas o f thin ice. D uring the Spring o f 1992, the L E A D E x p erim en t (hence
LEA D EX ) brought together scientists from many disciplines to study lead ice form ation in the
A rctic Basin. Investigators were stationed at a base cam p in the B eaufort Sea and helicopters
were used to locate leads suitable for study. W hen an appropriate lead was found, equipm ent and
living quarters w ere transported to the site. In this w ay w e were able to obtain data on A rctic thin
ice evolution beginning shortly after form ation and continuing for up to 3 days thereafter.
W e obtained data from three different leads (officially Leads 2, 3 an d 4) and several results
stand out as typical. U nfortunately, one o f the typical results was that the helicopters invariably
blew snow from nearby thick ice out onto the lead, i.e. onto our study area. Brine w icked up from
the surface o f the ice quickly turned the entire 1 to 3 c m thick snow layer to slush. T h e bulk
salinity o f the snow m easured at tw o leads ranged from 75 to 100 ppt with snow /ice interface
tem peratures o f -8° to -15°. A ssum ing a w et snow density o f 300 k g /m 3, the brine volum e o f the
slush varied from 10 to 22% .
Lead 4 is a good exam ple o f m icrow ave em ission fro m this brine ladened snow (Fig. 4-4).
Initially, th e spectrum w as sim ilar to th e SSM /I thin ice endm em ber and higher than that o f the
bare ice o f C R R E L E X 1990. Tb(6.7) values were very low, apparently due to the high w ater
content o f the snow (Fig. 4-4b). The spectrum had changed significantly by the next day;
T b(19, H ) rose about 2 0 K resulting in a decrease in PR (19) from approxim ately .12 to .07.
A dditionally, T[,(6.7) increased by 60K at H pol and 4 0 K at V pol. S im ilar behavior w as
observed at Lead 3 w ith an overnight decrease in PR ( 19) from .11 to .04.
The m ost likely reason for the overnight change in em ission was a decrease in the brine
volume o f the slush. O nly a few m easurem ents o f the bulk salinity o f the slush w ere m ade but
concurrent m easurem ents indicate a great deal o f horizontal variability and it is difficult to draw
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
32
any conclusions concerning the tim e history of vb. H ow ever, the surface slush layer at both leads
w as described as “ hardening gradually” or “firm ing up” with tim e. A t the same tim e, T b(6.7)
increased overnight at both leads. Since the penetration depth is large at 6.7 G H z, much o f the
radiation at this frequency should have originated at the wet snow /ice interface. C onsequently, the
increase in em ission at 6.7 G H z also suggests that vb o f the slush decreased overnight. This
suggests that snow w hich is wet throughout can have a GR near 0 but with PR( 19) varying in
direct proportion to the vbof the snow .
O n the second day (April 12) at lead 4, the ice in our study area rafted, form ing a com bined
sheet around 10 cm thick. The surface properties w ere those o f ice from the m iddle o f the lead: a
dam p surface, a surface tem perature o f approxim ately -12°C and a dense cover o f 1 to 2 cm
high frost flowers. T h e resulting signature was sim ilar to that o f FY ice (Fig. 4-4), although in
m ost cases with a PR (19) slightly less than that o f FY ice.
Frost flowers w ere also present at L ead 3 w here a 2 meter by 2 m eter section o f ice was
rem oved from the lead and a sequence o f m easurem ents was then m ade o f grow ing ice. T he air
tem perature was betw een -16° and -25°C , resulting in over 9 cm o f growth in less than 20 hours.
T he first few m easurem ents in the growth sequence correspond to frazil crystals suspended in
O W (Fig. 4-5). T hese data and the data for ice less than 3.4 cm in the thickness all lie well above
the O W -FY line in P R -G R space. Sim ilar behavior for very thin ice was reported in W ensnahan
et al. (1 993b) and is the result o f the shallow er penetration depth at 37 G H z com pared to 19
G H z. F o r ice greater than 3.4 cm thick, the data lie along the O W -FY line and have relatively low
values o f PR. T here is no obvious thin ice signature like that o f W ensnahan e ta l. (1993b) or that
observed in the S SM /I data, ju st a steady progression towards FY ice.
A gain the m icrow ave spectrum appears to be a direct result o f the surface properties o f the
ice. A t LEAD EX , surface brine layers and frost flow ers formed on all o f the leads studied,
suggesting that they m ay be a com m on feature o f thin ice in the A rctic Basin. Perovich and
Richter^M enge (1994) reported that very thin ice at L E A D EX had a salinity-enhanced surface
layer approxim ately 1 m m thick w ith a salinity o f greater than 100 ppt and a vb o f upw ards of
50% . W ithin a day o r tw o o f initial form ation, a 1 to 3 cm thick layer of frost flow ers form ed on
the ice. T he frost flow ers had salinities o f 40 to 80 ppt but, due to their low density, a vb o f only 1
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
33
or 2%. T hey postulated that the flow ers form ed as a resu lt o f the surface brine layer saturating
the air near the surface w ith vapor. V apor is then deposited on ice projecting above the brine
layer, with preferential deposition occurring on already formed flowers. The delicate structures
quickly w ick up surface brine by capillary action resulting in the high observed salinities.
In the case o f the ice from the Lead 3 growth sequence, a wet surface with a m easured salinity
o f betw een 45 and 100 ppt formed on the ice shortly after it began to grow . A few frost flowers
had started to form by the tim e the ice w as 20 mm thick . Over tim e m ore and m ore appeared,
eventually creating a m oderately dense cover on the ice (Fig. 4-6). By the end o f the sequence the
ice was still only 9 cm thick and had th e usual wet surface. The only obvious explanation for the
low PR(19) is the frost flow ers. It appears from these results, as w ell as the Lead 4 and
C E A R E X data, that frost flow er-covered thin ice can have a signature sim ilar to that o f FY ice at
19 and 37 G H z, even w ith a wet, salinity-enhanced surface layer,.
4.4 CRRELEX 94
W e grew tw o ice sheets at CR REL in 1994. The first sheet was used to study the m icrow ave
properties o f thick ice w ith and w ithout a snow cover, as well as w ith artificial surface roughness.- •
Grow th started in D ecem ber and by the tim e o f the observations the sheet was over 20 days old
and 30 cm thick. The ice surface appeared dry and had a salinity o f only around 5 to 7 ppt. H alf
o f the ice sheet was kept free o f snow throughout this portion o f the experim ent (Jan. 8 to Jan.
16). Bare ice m easurem ents were made during this tim e with air tem peratures ranging from -2° to
-16°C. T he G R o f m ost m easurem ents o f the bare ice w as slightly negative at around -.015, while
PR( 19) w as approxim ately .95 (Fig. 4-7). T hree m easurem ents had positive G R and higher
PR( 19) values but, in at least tw o o f those cases, the surface of the sites m easured had been
disturbed by people w alking on the ice. Just as in C R R E L E X 1990, the low est PR observed for
the bare ice w as not that o f FY ice but rather midway betw een the S S M /I values o f F Y and thin
ice.
On January 4, snow w as allow ed to fall on part o f the sheet. M easurem ents o f the snow covered ice began on January 8 when the snow was 5.5 cm thick with a slush layer in the lowest
1.5 cm. Initially, the spectrum exhibited very low Tb(6.7) values, a positive GR and a PR of
around .06 (Fig. 4-7). Snow grain size increased over the next three days and brine w as wicked
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
34
up into the snow . T he m icrowave signature changed continuously during this tim e until, by
January 11, the spectrum had stabilized. The spectrum had h ig h er Tb(6.7) values than on January
8, with both PR and G R having decreased to approxim ately FY values (Fig. 4-7e). The addition
o f another 4 cm o f fresh snow on January 15 caused only m inor changes in the spectrum. T he
m ain effect o f the additional snow was to decrease Tb(37) by ab o u t 5 K and T b(90) by about 2 0
K, indicating an apparent increase in scattering w ithin the snow (Fig. 4-7f).
T his sequence o f m easurem ents shares at least two sim ilarities with observations from
previous experim ents. As at C R R E L E X 1990, snow over sea ice resulted in a signature that, at
least briefly, lay above the O W -FY line. As at LEA D EX , changes in the vertical distribution o f
b rine within the snow pack led to relatively rapid changes in T b and, eventually, to a PR and G R
like that o f FY ice.
T he bare portion o f the thick ice sheet was also used to exam ine the effect o f surface
roughness on the m icrow ave properties o f thin ice. O n January 13, approxim ately 1 mm ice
crystals were spread over a section o f the bare ice to a depth o f a few m illim eters. In response,
T b(6.7) rem ained virtually unchanged while Tb at frequencies o f 19 G H z and ab o v e rose by 5 to
5 0 K (Fig. 4-8). T h e increase w as larger at H pol than at V pol, causing a d ecrease in PR to
aro u n d .06.
T he air tem perature rose above freezing shortly after this set o f m easurem ents. They rem ained
at o r above freezing until the night o f January 14, w hen they dropped to -14°C. D uring the w arm
w eather, the layer o f crystals partially melted and then refroze, solidly bonding the crystals to the
surface o f the ice by the m orning o f January 15. In response, P R and G R reverted to values
sim ilar to the surrounding bare ice. However, T b(37, H) and T b(90, H) rem ained 15 to 20 K
higher than the surrounding unm odified bare ice, indicating p ossible surface scattering at these
frequencies.
O n January 16, a 1 cm thick layer o f 1 cm diam eter ice pellets was laid o ver this same site.
T his new layer slightly depressed the T b(37) and T b(90) at both polarizations b ut it’s major effect
w as to increase T b(19, H) and once again decrease PR to around .06.
W hat can we conclude from this set o f observations? Does surface roughening cause PR( 19)
to decrease? N ot necessarily. T he problem is th at th e addition o f both o f these layers did more
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
35
than sim ply increase the surface roughness. Both were, at least initially, loose aggregates o f ice
crystals w ith a low er density than the underlying ice. T hat in it itself could have a m ajor im pact
on m icrow ave emission. T he fact that PR(19) reverted to higher values once the 1 mm size
crystals had bonded to the surface suggests that 19 G H z is insensitive to such small scale
roughness. It further suggests that the original decrease in PR was due to the low er density o f the
surface layer.
Two sim ilar experim ents were conducted at CR REL in 1985 and 1988. In 1985, the top o f a 7
cm thick ice sheet was gouged with ice picks. G renfell a n d Com iso (1986) m easured increases in
T b(H) resulting in significant decreases in PR(19) and PR(37). Unfortunately, snow fell on the
sheet during this series o f m easurem ents and the results are difficult to interpret. In 1988, 2 x 5
cm ice chunks were added to the surface o f a 15 cm thick sheet. G renfell et. al. (1988) m easured
large decreases in T b(V) w hich again resulted in a decrease in PR(19) and PR(37). H ow ever, it is
not clear w hether these chunks contributed only to surface roughness or also changed the density
o f the surface. T he results from these three experim ents are inconclusive; in all cases PR
decreased b ut w e cannot be certain o f the exact cause.
After the roughness experim ent, the ice w as removed from the C R R E L tank and a new sheet
was grown starting on January 19. The average w ater tem perature was ab o u t 1°C above the
freezing point at the start o f growth. This, com bined with bright sunshine, caused the ice
thickness to increase very slow ly for the first 12 hours (Fig. 4-9) and, consequently, very little
brine was trapped in the u pper portions o f the ice. The surface o f the bare ice appeared dry with a
salinity o f only around 7 ppt.
A few frost flowers form ed on one part o f the ice late on January 19. T hey continued to form
and grow ov er the next 30 hours, eventually covering about 60 to 70% o f this portion o f the ice.
The flowers w ere generally 1 to 5 cm across and about 1 cm high. They appeared dry although
the ice at their bases appeared loose, coarse, and damp. T he salinity o f the top half o f the flow ers
varied from 16 to 35 ppt and their bases w ere around 36 ppt. Subsequently, 6 sites w ere
designated on the ice for the purpose of m icrow ave m easurem ents. Sites 1 through 4 had about
10% flow er coverage w hile Sites 5 and 6 had a dense cover o f flowers o v er 70% of the surface.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
36
A round m idday on January 22, about 2-4 m m o f snow fell on the ice. O n January 23, an o th er
1.5
to 2 cm. o f snow fell. By early on January 24, Sites 1 through 4 had approxim ately 2.5 cm of
snow and Sites 5 and 6 had a com bined frost flow er and snow layer of around 3.5 cm. T he upper
portion of the surface layer consisted o f dry fine grain snow w ith a salinity o f 0 ppt at all sites
(Fig. 4-10). T he bottom 3 mm o r so had salinities ranging from 36 to 42 p p t and looked “d am p
but not slushy” . By January 26, brine w icking had given the snow a 3 tiered structure with the
top third consisting o f dry fine grain crystals, th e middle third w as slightly m oist with som ew hat
coarser grains o f 1-2 mm diam eter and the bottom third dam p with 2 mm d iam eter grains.
T he Tb record had the typical rapid rise during initial grow th. Beyond that, however, things
w ere less than typical (Fig. 4-11). There w as virtually no difference in em ission from the b are ice
and the frost flow ers covered sites before the snow fall on January 22, other th an slightly higher
T b(37, H) values. W ith the addition o f snow, T b(6.7, H) and T b(19, H) decreased dram atically,
w hile T b(37, H) rem ained m ore o r less unchanged and Tb(90, H) actually increased. O nly in the
case o f T b( 19, H) w as there significant variability from site to site. Tb(19, H) decreased 30 to
60K at sites w ithout frost flow ers, while decreasing by no m ore than 10K at the sites with frost
flow ers. W hen air tem perature w arm ed to -1°C on January 24, it signaled the beginning o f a
sharp increase in Tb(6.7, H), Tb(19, H) and Tb(37, H) at all sites.
Many o f the m icrow ave observations may be explained by a com bination o f 3 factors: the
thickness o f the low density surface layer, the distribution o f brine within that layer and the
penetration depth at different frequencies. F or exam ple, the freshly fallen snow was dry ex cep t at
th e bottom . If this 2-3 cm thick upper portion w as transparent at 6.7 GHz, then m ost o f the
em ission at that frequency w ould com e from the newly dam pened surface w hich would cause
T b(6.7) to decrease. In a sim ilar w ay, the thinner cover at sites 1 to 4 seem s to have been optically
thin at 19 G H z causing T b( 19) to drop in response to increased surface brine volume. The snow
and frost flow er layers at Sites 5 and 6 may have been thick enough (physically and optically)
that m ore o f the radiation originated from the d ry er layers above the ice surface. By this
reasoning, only the middle and upperm ost layers contributed to em ission at 37 and 90 G H z and
we w ould expect T b to increase a t both frequencies. But this only happened a t 90 GH z. It is not
clea r why T b(37) rem ained essentially unchanged with the addition o f snow.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
37
As air tem perature rose starting on January 24, brine w as wicked up into the snow pack. This
brine increased the optical depth of the m iddle and upper layers, causing m ore radiation to
originate from slightly dam p snow. This apparently caused the dram atic increase in T b at all
frequencies.
One thing that rem ains unexplained is that, before any snow fell, the frost flow er covered sites
had spectra nearly identical to those o f the bare sites. It is possible that th e frost flow ers w ere so
dry that they were essentially transparent at all frequencies. The salinity o f the flow ers was
measured as 16 to 35 ppt. T he tem perature o f the flow ers would have been between that o f the
surface (ab o u t -7°C) and that o f the air (usually much less than -12°C). If we assum e a flow er
tem perature o f -10° and use the density o f 25 kg/m 3 estim ated by P erovich and R ichter-M enge
( 1994) then the brine volum e would be betw een .2 to .4%. T his is significantly low er than the vb
o f 1% for frost flowers at LE A D EX . N onetheless, it is difficult to understand why the flow ers at
Sites 5 and 6 had no effect w hatsoever on em issivity.
All o f the changes in T b w ere, o f course, reflected in P R -G R space (Fig. 4-12). S hortly after
consolidation the signature had a high PR o f about .14 and a G R som ew hat less than 0. PR
steadily decreased as the ice thickened and, by the tim e th e ice was 9 cm . thick, PR w as .08 to .09
at all sites. T his is the sam e lim iting value o f PR for bare ice observed at previous experim ents.
The effect o f snowfall on PR and G R varied from site to site. Both increased dram atically at
Site 1 due to the large decrease in 19 G H z T b at polarizations. Site 6 exhibited very little response
in Tb( 19, H) but a sm all decrease in Tb(19, V). The result w as an increase in GR and a decrease
in PR. H ere, as in previous exam ples, the signature of snow covered ice varies dram atically
depending on how brine w as distributed w ithin the snow. A t all sites, as brine was redistributed
upward starting on January 24, PR and G R decreased, eventually stabilizing with a G R near 0
and a PR actually low er than that o f FY.
4.5 Summary
There are a several general conclusions that can be draw n from the observations presented
above.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
38
B are unroughened ice: T he C R R E L E X 1994 data show n in Fig. 4-12 suggests that b are ice
has a high PR like that of thin ice but that, as the ice ages, the PR decreases to about .08. This
appears to be the low est PR possible for bare unroughened ice. This result appears to hold for
thick ice as w ell as thin. It was true o f 6 cm thick ice from C R R E L E X 1990 and it was also true
o f 30 cm thick ice from C R R E L E X 1994. T he thick ice from the 1994 experim ent had low bulk
and surface salinities sim ilar to that o f m ature FY ice and y et the PR was still no low er than .08.
It w ould seem that further surface m odification is required to achieve a P R low er than this.
Roughened ice: T he experim ents on artificially roughened ice are inconclusive. T he P R was
decreased in all cases, but w hether this was du e to roughness or som e other factor is not clear.
Frost flo w ers: T he low PR o f FY ice was only observed fo r ice having som e form o f low
density surface layer. In both arctic field experim ents, thin ice covered w ith frost flow ers had
spectra like that o f FY ice for frequencies o f 19 G H z and higher. Further, frost flow er form ation
apparently caused a steady transition from the O W to FY ice spectrum with no intervening thin
ice signature d uring ice growth at L ead 3 o f L E A D E X . T he exception w as C R R E L E X 1994,
w hen frost flow er covered sites had relatively high PR s o f .08 to .09, perhaps due to the low
salinity o f these flow ers com pared to those observed in the Arctic.
Snow: S now -covered thin ice can also produce a FY ice signature. H ow ever, the signature of
snow -covered ice is a strong function o f the snow thickness and the distribution o f brine w ithin
the snow. If the snow pack is thin and the brine is lim ited to a slush layer near the snow /ice
interface, then th e signature can have a relatively high PR but lie above the O W -FY line in PRG R space. If a sm all am ount .of brine is distributed upw ard w ithin the snow , then the PR can be
very low - even below that o f FY ice. If the entire snow pack is saturated, as at LE A D E X , then
the signature lies along the FY -thin ice line, w ith PR varying in direct proportion to brine volume.
Finally, the satellite case study from the previous chapter im plies that a steady decrease in PR
occurs as thin ice ages and thickens. In none o f the experim ental results w as this progression of
PR observed. H ow ever, it now seem s clear that the signature evolution m ust at some point
involve snow or frost flowers w ith accom panying brine.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
39
«- —
‘
-
1
_ »
■
I
.
-- i
. .i
0.08 -
°o
0 0
1
■
o
o
oo
o
u.
0.04 *
p
o.oo -
O>N
os
3 -0.04
X
X
-
X
X
-0.08 -
X
o
OW
A
N ilas (5 -1 0 c m )
*
FY ice
X
M Y ice
.n n .
0.00
0.08
0.16
0.24
PR (19)’*
250
- 250
O 200
150
-
b. Nilas. M arch 18.
10
19
c. N ilas. M arch 19.
100 10
19
- 150
d . FY ice
100
10
200
19
100
F requency (G H z)
t T W I W .R i « i ..H /IW l
Fig. 4-1: M icrow ave observations made d u rin g the C E A R E X 1989 cm ise.
a) PR vs. G R plot o f a variety o f ice types, b) spectrum o f 5-10 cm thick rafted nilas
observed on M arch 18. c) spectrum o f 5-10 cm thick rafted nilas observed on M arch 19. d)
average spectrum o f all FY ice station data. Error bars in this and all subsequent p lots of
spectra are standard deviation.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
40
Fig 4-2: Frosl flowers on 5-10 cm thick rafted nilas observed on March 18, 1989.
Taken in the G reenland Sea during CEA REX
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
41
0 .0 8 -
Snow
B are
04
O
-0 .0 4 -
-0 .0 8 -
0.00
0.16
0.0 8
0.24
PR (19)
250
200
150
100
b. Before snow
10
After snow
100
10
100
Freq (GHz)
Fig. 4-3: O bservations before and after snow fall on thin artificial sea ice.
M easurem ents were made during C R R E L E X 1990 o f 6 cm . thick nilas. T he snow was 5
mm thick and had a substantial slush layer, a) PR vs G R plot of the data along w ith the
O W -FY -M Y triangle defined by the SSM /I N A SA T eam algorithm endm em bers. b) average
spectrum o f bare ice before snow fall, c) average spectrum after snowfall.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
42
0.05 -
£
cn
>
On
0.00
-
Apr 11
F ro st'"
flowers
Apr 12
-0.05 -
- 0 .1 0 -
0.00
0.08
0.16
0.24
PR (19)
Fig. 4-4 Passive m icrow ave m easurem ents o f Lead 4 during LEA D EX , 1992.
T hree separate groups o f thin ice d ata are indicated and their average spectra are plotted.
E m ission from the slushy snow at the edge of the lead was m easured on April 11 (fig. b) and
again on April 12 (fig. c). Frost flo w er covered thin ice that had rafted over itself w as also
observed on A pril 12 (fig. d)
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
43
2.0
- 250
-
3.4
0.00
200
- 150
Tb (K)
GR (19v, 37v)
0.05
-0.05
b. 9.4 cm
100
-
10
0.10
19
Freq (GHz)
0.00
0.08
0.16
0.24
PR (19)
LnXJO HO .orR .... N /J 1 M
F ig. 4-5: Tim e sequence o f em ission during ice growth at L ead 3 o f L E A D E X 1992.
The ice w as initially bare but frost flow ers form ed on the ice rapidly once the sheet had
consolidated. There is no thin ice signature to speak of, ju st a steady progression to FY ice.
a) Time sequence in term s o f PR vs. G R. T he numbers associated w ith solid circles are ice
thickness in cm . b) The spectrum of the 9.4 cm thick ice.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
44
Fig 4-6: Frost flowers 011 the surface thin ice at Lead 3 o f LEADEX, 1992.
lee w as grown in a 2 m by 2 m hole and w as 9 cm thick at the time o f the photo.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
45
<**!
>
O
N
0.00
Om
o
-0.04
0.00
0.04
0.08
0.12
250
- 250
200
-
150
b. Bare ice
c. Bare ice outliers
o
e
■
□
a
m
Bare ice
Bare ice outliers
Snow with slush
Aged snow
New snow on aged snow
200
- 150
250
- 250
200
-
150
d. Snow w/slush
10
e. Aged snow
100
10
19
f. New snow over aged snow
100
10
19
200
- 150
100
Frequency (G H z)
tlL M ntht
Fig. 4-7: O bservations o f bare and snow covered 30 cm . thick ice, C R R E L E X 1994.
a) C luster plot in P R -G R space o f all th e data. Bare ice data are eith er unfilled circles with
crosses. Squares represent a sequence o f m easurem ents o f snow covered ice. D ifferent types
of circles and squares indicate the points used in calculating the av erag e spectra in plots b
through f. b) Average spectrum o f m ost o f the bare ice data (indicated by unfilled circles in
the P R -G R plot), c) A verage spectrum o f outliers from the bare ice d ata m easured on Jan. 9,
11 and 13 (circles with crosses), d) S pectrum of ice w ith a four day o ld snow co v er from
Jan. 8. (the single solid square), e) A verage spectrum o f the snow covered ice after three
more days o f aging and m etam orphosis (squares with x ’s). f) A verage spectrum after an
additional 4 cm o f new , dry snow had fallen (squares w ith vertical lines).
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
46
0.04 -
>
Jan 13
0.00
>
-
o\
Jan 15
0i
O
Jan 16
-0.04 -
0.00
0.08
PR (19)
C.
250 -
- 250
200 -
-
150 4 -
b. Jan 13.
10
c .Jan 15
19
100
10
d. Jan 16
19
100
10
200
150
19
100
F ig. 4-8: The effect o f artificial surface roughness on bare thick ice o f C R R E L E X 1994.
All m easurem ents are o f the same site on the bare ice b u t with various form s o f surface
m odification, a) The four data points from the roughness experim ent are plotted as triangles.
The data from Fig. 4-7 are plotted in gray fo r reference, b) Spectrum obtained on January
13, after a layer o f 1 m m ice crystals had been freshly laid over the bare ice (also show n in
gray is the average spectrum o f bare ice given in Fig. 4-7b); c) Spectrum from January 15,
after the 1 m m ice crystals had melted and refrozen onto the surface; d) Spectrum from
January 16, after a layer o f 1 cm diam eter ice crystals had been laid o v er the site as w ell.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
47
Jan 19
Jan 2 0
Ja n 21
Jan 2 2
Jan 23
Jan 2 4
Ja n 25
Jan 2 6
Jan 27
J a n 28
4 -
Ice th ick n ess
• 16 -
-1 0
Sn o w th ic k n e ss
-
20)
Cl
£
<D
E-
-30 -
—o — A ir tem p
* " V - S urface tem p (K T 1 9 )
—
Interface tem p (th erm o m eter)
-35
Jan 19
J a n 20
Ja n 21
Jan 22
Jan 23
Jan 24
Ja n 25
J a n 26
Jan 27
J a n 28
D a te in 1994
IWMMtTtri klUd
Fig. 4-9: Physical properties o f the second ice sheet grow n during C R R E L E X 1994.
a) average ice thickness and snow thickness at Site 1 plotted as function o f time, b) air
tem perature, surface tem perature and snow /ice interface tem perature plotted as function of
time. T he surface tem perature was m easured with a K T-19 infrared radiom eter w hile the
interface tem perature w as measured w ith a therm ometer.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
48
a.
Jan 23, 19:00
A v e sites 1 and 6
b.
Jan 24, 17:00
c.
Jan 25, 16:00
Site 3
e. Jan 26, 9:30
Site 1
Site I
to p o f snow
£
o
1)
t
.a
2
20
0
20
0
Salinity (ppt)
F ig. 4-10: Snow salinities from the second ice sheet at C R R E L E X 1994.
The salinity as a function o f height above the ice surface is plotted. T he top of the snow
pack is given by a dotted line in each case. T he majority o f the snow fell midday on January
23. S ignificant upw ard brine m ovem ent occurred within a few days o f snowfall.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
49
B rightness tem perature
(K e lv in )
19
20
21
22
23
24
25
26
27
19
20
21
22
23
24
25
26
27
240-
-2 4 0
210
-210
180
180
150
150
a. 7 GH z
b. 19 GHz
240
240
210
210
180
180
Site number
150-
- 150
d. 90 G H z
c. 37 GHz
19
20
21
22
23
24
25
26
27
19
20
21
22
23
24
25
26
27
D a y in J a n u a ry o f 1994
Fig. 4-11: T im e sequences o f T b(H ) from the grow th o f sheet 2 a t C R R E L E X 1994.
In each graph a single frequency is plotted fo r Sites 1 through 6. V ertical dotted lines
indicate three snow fall events with approxim ate depths o f 2-4 m m on January 2 2 ,2 cm on
January 23, and 1 cm on January 25. F requencies plotted are a) 7 G H z, b) 19 G H z, c) 37
G H z, and d) 9 0 GHz.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
50
0.0 8
— o— Site I
Jan. 24
S ite 6
Ja n . 24
0 .0 4 -
Jan. 23
O
Jan. 23
Jan. 22
0 .0 0
-
-0 .0 4
0.00
0 .0 4
0 .0 8
0.12
0 .1 6
P R (1 9 G H z)
Site i
250 -
200
-
150 c. New snow
b. Before snow
d. Aged snow
Site 6
250 -
20 0
-
150 e. B efore snow
10
19
f. New snow
100
10
19
g. Aged snow
100
10
100
F requency (G H z)
C R L 9 4 B .n r g , 9/1 S /9 5
Fig. 4-12: T he effect o f snow fall and brine w icking from sheet 2, C R R E L E X 1994.
a) sequences from sites 1 and 6 plotted in P R -G R space. L abeled points indicate the first
observation m ade on that day. b) average spectrum from Site 1 of data taken the day before
snow fall, c) spectrum of the first observation m ade on January 24 at Site 1 d) spectrum o f
all d ata taken on o r after January 25 at Site 1. e) spectrum o f data taken th e day before
snow fall at Site 6. f) spectrum o f the first observation m ade on January 24 at Site 6 g)
spectrum o f all d ata taken on o r a fter January 25 at Site 6.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
5. Theoretical studies of emission
The field d ata give clear indications o f som e of the relationships betw een m icrow ave
em ission and the physical properties of ice and snow. H ow ever, it is difficult to determ ine the
exact nature o f these relationships given the limited range o f properties encountered an d the
difficulty o f m easuring m inute changes in those properties. Only with a theoretical m odel can we
fully explore the effects o f the brine slick, snow and frost flowers on em ission.
In this chapter we develop a simple m odel to calculate microwave em ission from sea ice.
T he model uses a m ultilayer form ulation o f the Fresnel equations to p redict the propagation and
absorption o f radiation w ithin the ice. This approach requires the specification of the bulk
perm ittivities o f sea ice, snow and frost flow ers, which are calculated from the Polder-van Santen
perm ittivity m ixing formula. W e present sensitivity studies o f em ission from bare ice and ice with
a snow or frost flow er cover. T he bare ice sim ulations include calculations o f em ission from ice
w ith variable bulk brine and air volumes as w ell as ice w ith a surface brine slick. T he studies of
snow and frost flow ers on sea ice consider th e effects o f surface layer b rine volume, brine
distribution, and ice volume. U sing this m odel, we are able to quantitatively verify m any o f the
qualitative argum ents used to explain the experim ental observations o f the last chapter, and to
extend the results over a w ide range o f physical properties.
5.1 Multilayer Fresnel formulation
A num ber o f different m odels have been developed to predict both m icrow ave em ission and
radar backscatter front sea ice (see W inebrenner et. al„ 1992 for a review ). M ost of th e current
generation o f m odels concern themselves w ith the com plicated processes o f surface and volume
scattering and, hence, are based on relatively com plex theories. However, scattering is generally
considered negligible for young ice. For exam ple, model com parisons w ith CR RELEX 1988
observations o f a 9 cm. thick ice sheet indicate that em ission at frequencies below 37 G H z was
largely controlled by the reflectivity o f the ice, not by surface or volum e scattering {W inebrenner
et. al„ 1992). W e have chosen, therefore, to model em ission from thin ice using a relatively
sim ple, nonscattering model. T he ice and any attendant low density cover are m odeled as a stack
o f horizontally hom ogeneous layers to account for vertical variations in perm ittivity. B ecause the
snow or frost flow er cover on thin ice is often optically thin, the radiation field is treated as
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
coherent, w ith absorption occurring within the layers and reflection and transm ission occurring at
layer interfaces.
Reflection and transm ission at a boundary between tw o media with different perm ittivities
can be calculated using the Fresnel equations. For a w ave traveling in m edium 1 and reflected at
the boundary w ith medium 2, the am plitude reflection coefficient (r) o f the reflected w ave is
given by the Fresnel equations as
(5-1)
w here the subscript H and V refer to horizontal and vertical polarization, e is the relative
perm ittivity o f each medium and 0Ois the direction o f propagation w ithin m edium 0. T he angle of
propagation relative to nadir w ithin m edium 1 (0i) can be obtained from Snell's law. T he
transm ission coefficient (t) is given by tH = l+ r H and tv = ( l+ r v)(cos 0o/cos 0i).
Each interface in a layered m edium will give rise to a reflected and transm itted w ave. The
effect o f all o f the reflections, transm issions and interactions between the different w aves can be
accounted for by a series o f m atrix m ultiplications (see e.g. Collin, 1991). The upw ard and
dow nw ard propagating w aves at the top o f a m edium consisting of n layers can be related to
those at the bottom o f the nth layer by
w here a and b are the am plitudes o f the dow nw ard and upw ard propagating waves respectively.
T he first m atrix on the right hand side (along with the factor 1/t) accounts for reflection and
transm ission at the boundary betw een the i and i+1 layers. The second m atrix on the right
accounts for propagation through the i+1 layer. The quantity P is som etim es referred to as the
electric path length where
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
53
P i (v ) = k 0>/ e i (v )
z-,
vCOS0i (V)/
and k0 is the wave num ber in free space, z is the vertical depth and v is frequency.
In the case o f sea ice, w e assume that the bottom layer, the ocean, extends to infinity such
that there is no upw ard propagating w ave from below (b n= 0). D ividing (5-2) by ao, w e obtain
1
n T
;_n
v
l
i.i+l
e -iPu,
‘ i.i+I
1
0
0
dui.l
(5-3)
which can be solved for r for the entire medium. The pow er reflection coefficient is R = rr* (r
tim es its com plex conjugate). The em issivity of the m edium can then be calculated from (3-2) by
assum ing that scattering is negligible (This assum ption will be exam ined in m ore detail in the
next section). Finally, T b can be obtained from (3-1) by assuming th at T a is equal to the
frequency in G H z ( W ensnahan et al., 1993a).
5.2 Permittivity mixing formulae
The solution for r(v) requires specification o f the incident nadir angle and the thickness and
perm ittivity o f each layer in the system . R ather than specifying the perm ittivities directly, they are
calculated from the physical properties o f each layer using a m ixing form ula. S ea ice, snow and
frost flow ers are assum ed to consist o f three basic constituents: ice, a ir and brine. In most m ixing
formulae the ensem ble o f constituents is treated as a host medium containing inclusions o f the
other constituents (som etim es referred to as scatterers). The effective perm ittivity (Eeff) of the
medium is then calculated as a com bination of the perm ittivities o f the host and the inclusions.
H allikainen and W inebrenner (1992) review various mixing m odels for snow and sea ice.
For our w ork we have chosen one of the best known o f these, the Polder-van Santen equation.
The Polder-van Santen equation has been used to successfully model dry snow (H allikainen et.
al., 1986), w et snow (H allikainen et. a l., 1986; C olbeck, 1980) and sea ice (Sihvola a n d Kong,
1988). Further, in the low frequency lim it, the m ixing form ula developed by S togryn (1985,
1987) for his strong fluctuation theory m odel of sea ice and snow reduces to the Polder-van
Santen equation.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
54
Ice, brine and air are all linear isotropic media and their individual perm ittivities are all
scalars. H ow ever, nonspherical inclusions result in an effective perm ittivity tensor e defined by
the equation
D = e • Ea
(5-4)
where D is the electric displacem ent field resulting from an applied field, E a . The displacem ent
field can also be written as the sum o f E a interacting with a host m edium (with perm ittivity e h )
and the dipole m om ent/volum e ( P ) o f the scatterers as induced by the applied field.
D = 8 hE 7 + P
(5-5)
Equations (5-4) and (5-5)can be com bined to yield
P = ( e - e hI).E7
(5-6)
W here I is the identity m atrix. Equation (5-6) relates P to the applied field, but w e can also
think o f P as the result o f the localelectric field ( E, ) near the scatterers.
P = Na«E^
(5-7)
w here N is the num ber o f scatterers per unit volume and a is the polarizability o f a single
inclusion. T he local electric field is not the same as the applied field, rather it is the sum o f the
applied field and the electric field due to P itself ( E p ), i.e.
17 = 17 + 1 ;
(5-8)
where E p is given by
i ; = (ehi) ' . L . P
(5-9)
and L is referred to as th e depolarization dyadic. U sing this system o f equations the effective
perm ittivity can be solved by com bining (5-6), (5-7) and (5-9) to elim inate the fields in (5-8).
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
55
The depolarization dyadic quantifies how the shape of the scatterer affects the electric field
produced by P . It is m ost often assum ed that the scatterers are spheroids (ellipsoids with two
equal axes) in w hich case L is diagonal and
L i i +
L22 +
L33 =
1
with
L 22 = L 33
T he value o f L M depends on the ratio o f the axes lengths. For prolate (cigar shaped) spheroids
Lll=i z # ( ± J h * U
"
E2
U
e
U -
e
J
w here E is the eccentricity. For oblate (pancake shaped) spheroids
L" ' i f r d
" arc M e)) ' i "
where
1- E
g:
E2
F o r exam ple,
Li j -
0
needles
1/3
spheres
1
discs
As an exam ple o f calculating an effective perm ittivity, consider the sim ple case o f spherical
scatterers em bedded in an isotropic host; a , L and e reduce to scalars with L equal to 1/3.
Equations (5-6) to (5-9) can be com bined to yield
1+
2N a
-
3e
e
1_ N a
3eh
w hich can be rearranged to
a
^~
—
N
E
+ 2eh
'h
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
56
T he first equation is known as the M axw ell-G am ett m ixing formula w hile the second is the
C laussius-M ossotti or Lorentz-Lorenz form ula.
The M axw ell-G arnet form ula in the exam ple above is only valid fo r situations w here the
scatterers occupy a small fraction of the m edium . As the volume fraction o f scatterers increases,
the fields due to individual scatterers interact. The Polder-van Santen form ula was developed to
address this lim itation. Sihvola a nd K ong (1988) dem onstrated that m any well know n formulae,
including Polder-van Santen, can be reproduced by replacing e h I in (5-9) with w hat they refer to
as an apparent perm ittivity e a . M aking this substitution (5-6) to (5-9) reduce to
T h e perm ittivity o f m edia containing m ore than one species (s) o f scatterer is calculated by
substituting fsys = N s0 Cs and (L s)h = L,j, and then sum m ing these new term s over all species.
M aking this substitution (5-6) to (5-9) then reduce to
(5-10)
w here
v o lu m e o f an in c lu sio n
(5-11)
and fs is the fraction of the total volume occupied by a particular species o f scatterer. These
equations apply to an electric field in line w ith the axes o f the ellipsoidal inclusions. W e will
assum e that the inclusions are random ly and isotropically distributed w ith a scalar effective
perm ittivity and an apparent perm ittivity ten so r which is diagonal (S ih vo la a n d K ong, 1988)
A veraging (5-10) o v era ll angles w e obtain
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
57
(5-12)
Equations (5-11) and (5-12) are generalized equations. The Polder-van Santen equation is
obtained by substituting
(5-13)
into (5-11) and (5-12).
T he derivation above is generally know n as the quasi-static form ulation and is applicable to
radiation w ith a w avelength much greater than the diam eter of the inclusions involved. A t shorter
w avelengths, the actual size o f the inclusion becomes increasingly important, volum e scattering
increases, and the quasi-static mixing form ula becom es less and less accurate. V ant et. al. (1978)
argue that this type o f form ulation should be suitable fo r FY ice at frequencies o f up to 24 GHz,
while H allikainen et. al. (1986) successfully used a quasi-static m ixing form ula to m odel the
perm ittivity o f wet snow w ith grain sizes from .5 to 1.5 mm at frequencies up to 37 GHz. There
is evidence that the quasi-static assum ption can begin to break dow n at around 37 G H z. For
exam ple, m ature FY ice as observed from SSM /I has som ew hat-depressed values o f T b(37, V)
relative to T b( 19, V) caused by volum e scattering at 37 GHz. H ow ever, the difference between
the two is sm all (on the order of a few Kelvin) indicating that scattering at 37 G H z is relatively
weak. W e expect, therefore, that the quasi-static m ixing formula w ill w ork well at 19 G H z but
may slightly underpredict the imaginary part of the perm ittivity at 37 G H z (and hence slightly
overpredict em ission) fo r th e thin and F Y cases considered here.
To calculate the perm ittivity o f sea ice and snow using (5-11), (5-12) and (5-13) requires
specification of a host m edium , the shape o f the scatterers, and the perm ittivities o f the host and
the scatterers. Stogryn a n d D esargent (1985) m easured the perm ittivity o f brine in sea ice for the
tem perature range -2.8 to -25°C. They found that th eir data could b e closely approxim ated by the
Debye equation (D ebye [1929]):
£br
-
£ oo +
e s - £~
1 - i27tvx
+
1o
27I£0 v
where e 0 is the perm ittivity o f free space. The other param eters are given by em pirical formulae
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
58
es = (939.66 - I9.068VC10.737 - T)
e« = (82.79 + 8.19 T 2)/( 15.58 + T 2)
2nx = .1099+. 13603 x 10'2 T + .20894 x 10° T 2 + .28167 x 10'5 T3
and
o = -T e x p (.5 1 9 3 + .08755 T)
-T exp( 1 .0 3 3 4 + .1 1 T )
T > -22.9
T<-22.9
where T is the brine tem perature in °C. T he perm ittivity o f pure ice is nearly independent o f
tem perature and frequency at m icrow ave frequencies. W e assum e that it is a constant 3.2 + .0 0 1i .
5.2.1
Sea ice
W e assum e that sea ice is com posed o f a background of ice with inclusions o f air and brine.
T he air inclusions are treated as spheres (G renfell, 1983). Brine pockets, on the other hand, are in
general not spherical. V ery little inform ation is available on brine pocket geom etry. F or
congelation ice, the pockets are taller than they are w ide but a typical value for the ratio o f height
to width is not clear. Vant et. al. (1978) quote values fo r brine pockets in FY ice o f 5 mm tall and
.025 mm w ide (a ratio o f 200; 1) but there is no indication where these values com e from . In the
sam e paper, they model th e perm ittivity o f sea ice using a value o f 20:1 and achieve reasonable
results. A rco n e et. al. (1986) exam ined saline laboratory ice and described the brine pockets as
generally ellipsoidal with a vertical length o f 1 to 2 m m but a horizontal width that varied from
.01 to 1.7 m m . A ty p ical ratio o f the axial lengths w as 10:3:1. U sing this ratio, and assum ing that
the pockets are spheroids, w e assum e a length to width ratio o f 5:1, corresponding to L n ~ .06.
A ssum ing spherical air bubbles and ellipsoidal brine pockets, (5-11), (5-12) and (5-13)
com bine to produce
^ ^ a i r 0 ~~ E i c e )
2e + 1
£ f b r ( e br — E ic e )
e (1 —L
n ) + Li ] Ebr
e (1 +
L | 1) + £ br (l —L n )
T his form ula w as com pared with m easurem ents by a num ber o f investigators as sum m arized by
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
59
H allikainen and W inebrenner (1992). Substantial agreem ent was obtained at 4 and 10 G H z o ver
a range o f salinities from 2 to 10 ppt and ice tem peratures from -3 to -20°C.
5.2.2 S n o w and fr o s t flo w e rs
Dry snow was assum ed to consist o f air with inclusions o f ice. Sihvola and K o n g (1988)
com pared the M axw ell-G arnett m ixing form ula w ith data at .85 to 12.6 GHz from T iu ri et. al.
(1984). T hey dem onstrated that the predicted perm ittivity was relatively insensitive to the shape
of the inclusions but th a t the best agreem ent was fo r discs. They fu rth er stated that sim ilar results
occurred using the Polder-van Santen equation. Follow ing their exam ple, we assum e that the ice
in dry snow can be m odeled as discs, hence the m ixing formula is
0
„ _ , , ^ ic c
ice —
£ — I 1 --------------------------------
3
W hile it is a relatively straightforw ard matter to model the perm ittivity of dry snow , wet
snow is a m uch more com plicated situation. The problem is that the ice and water in w et snow
form an intim ate m ixture rather than discrete inclusions em bedded in a host. S to g r y n f 1985)
noted that the Polder-van Santen equation fails to properly model m oist mixtures, particularly the
im aginary part of the perm ittivity. A ttem pts have been made to rectify the problem . F or example,
H allikainen et. al. (1986) obtained reasonable agreem ent between the Polder-van S anten equation
and experim ental data w hen the liquid inclusions were allowed to be asym m etrical and their shape
was varied as a function o f liquid w ater content. H ow ever, Stogryn( 1985) argues th at a more
reasonable approach is to model the ice grains as being coated by som e percentage o f the total
liquid w ater volume. B ased on this idea he developed an equation fo r w et snow w hich agreed
extrem ely w ell with the experim ental d ata o f L in lo r{ 1980). Follow ing Stogryn’s w ork, we will
assume th a t w et snow consists o f w ater inclusions and water coated ice inclusions w ith air as the
host m edium .
In general, wet snow grains rapidly m etam orphose until they are approxim ately spherical
{Colbeck, 1982). B ohren a n d H uffm an (1983) derive an equation fo r polarizability o f a coated
spherical scatterer w hich, w hen applied to a brine coated snow grain, yields
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
60
_
8f
K ^br
~
£ ice ) ( 2 £ b r
+ l) + (2Ebr + £jc(. )(l —£ br)
f ( £ br _ £ i c e ) ( 2 e b r _ 2 ) + ( 2 £ br + £ ice ) ( 2 + £ b r )
w here f is the fraction of total snow grain volum e occupied by ice. U sing this equation for the
snow grains and (5-11) for the w ater not coating the grains, (5-12) becomes
e - 1+
e -1
3fgrYgr 1 + -2e + 1
£ ^br (£ br — 0
e ( l ~ ( L b r ) t, i ) + ( L b r ) U e br
e ( I + ( L b r ) i , 1) + E b r ( J ~ ( L b r ) U )
w here fg, is the fraction o f the total volume occupied by either the coated grains and ft* is the
fraction occupied by water not coating the grain.
Stogryn{ 1985) determ ined that the best m atch to the L inlor data w as achieved when 25% o f
the w ater was assum ed to coat the ice grains. U sing this value and assum ing that the w ater
inclusions are spherical ( (L br )
= 1/3), the last equation alm ost exactly m atches the sim ulation
o f S to g ryn (\9 8 5 ) and the data o f L inlor (1980).
F rost flow ers are a low density layer consisting of air as the host and ice and brine as
inclusions. The structure o f the flowers varies from delicate fan shapes to am orphous clum ps
depending on air tem perature (P erovich a nd R ichter-M enge (1994)). T he clum ps occur in warm
air tem peratures and have sm all crystals equivalent in size to snow grains. G iven the w arm air
tem peratures in the Bering Sea, w e assume th a t frost flow ers w ould form in clum ps w hich can be
m odeled as a low density snow cover.
5.3 Simulations
The SSM /I data suggest that there is a steady evolution o f the thin ice signature to that of
m ature FY ice. A prom inent indicator of this evolution is an apparent continuous decrease in PR
as the ice ages. T he field data indicate som e o f the ways in w hich em ission from thin ice m ay be
affected by changes in the physical state o f the ice. However, as noted earlier, it is difficult to
isolate the effect o f any single physical attribute using field data alone. U sing the m ultilayer
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
61
Fresnel model we will attem pt to quantify the relationship between changes in em ission and
changes in specific physical properties typical o f aging young ice.
5.3.1 Ice an d brine
As thin ice ages it becom es colder and less saline, leading to an overall reduction in brine
volum e. At C R R E L E X 1994 bare ice without a noticeable brine slick had a P R that decreased
steadily as the ice aged. Is it possible to reproduce the signature evolution sim ply through
reductions in bulk brine volume (vb)? To answ er this question, a series o f sim ulations were d one
o f optically thick ice with uniform properties. T he ice was m odeled as containing no air but
varying am ounts o f brine. The m odel output w as T b( 19) and T b(37) from w hich PR( 19) and
G R (19V , 37V ) w ere calculated (Fig. 5-1). The m odel predicts that ice containing no air but large
am ounts o f brine has a very high PR . A s vb decreases, so too d o es PR. This is a direct result o f a
decrease in the effective perm ittivity o f the ice. T he permittivity contrast with the air decreases as
vb decreases. A s a result, both Tb(H ) and Tb(V) go up, the increase being m ore pronounced at
T b(H), thus causing P R to decrease.
T he predicted signature coincides with the clu ster of thin ice points from the Bering S ea
when vb decreases from 20% dow n to 0%. Only very thin, new ly-form ed ice has a bulk vb as high
as 20% ; generally, vb is on the order o f 5 to 10%. N onetheless, the sim ulations suggest that
decreases in bulk brine volume may contribute to th e signature evolution o f thin ice. In the lim it
that the ice contains no brine, the perm ittivity is ju st that of pure ice and the signature has a PR
o f .08. T his is much higher than the P R o f FY ice b u t precisely the same lim iting value m easured
for bare ice at C R R E L .
W hile a very high bulk vb may not be typical o f young ice, high vb is characteristic o f the
surface brine slick. C om parable calculations were done with a brine slick laid o v er optically thick
ice (Fig. 5-2). T he slick was a .7 m m thick m ixture o f ice and brin e but with no air. O bservations
indicate that ice even a few cm in thickness is optically thick at 19 and 37 G H z. T herefore, the
underlying ice was m odeled as an optically thick hom ogeneous layer containing 6% brine and 3%
air (this ice base w as used in all the sim ulations o f surface layers w hich follow ). T he results are
R ep ro d u ced with p erm ission of th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
62
sim ilar to the first set: for a slick w ith vb from 0 to 20% , the signature falls in the range o f the
thin ice cluster but the low est PR is again only .08.
5.3,2 Snow
W hile the initial evolution of the thin ice signature may be explained by decreases in the
brine volum e of either the surface or the ice as a w hole, it is obvious that som ething more is
necessary to com plete the transition to a F T ice signature. This p o in t is made even clearer if we
consider ice with no brine but increasing am ounts o f air (Fig. 5-3). W ith no air, the signature
starts w here the previous sim ulations left off. In all cases the G R is above that o f F Y ice, m ost
likely d u e to volume scattering at 37 G H z which is n o t accounted fo r in the quasi-static m ixing
formula. A s pointed out above, volum e scattering at 19 GHz should be negligible and should not
affect the calculated values o f Tb(19) o r PR(19). W e see that PR (19) decreases as the air volum e
increases. The PR is sim ilar to that o f FY ice w hen the ice is 60% air and 40% pure ice. This
confirm s the conclusion that a low density layer is needed to produce a FY ice signature. T he P R
decreases below that o f FY ice, m uch like observations at L E A D E X and C R R E L E X 1994^ for
air volum es greater than 60%.
H ere again, m icrow ave em ission can be understood in term s o f the perm ittivity contrast
betw een air and the ice. A s the air volum e in the ice increases, the perm ittivity o f the ice
decreases and the contrast is reduced. T b at both polarizations increases but the increase is larger
at H pol, ultim ately causing the PR to decrease. In the limit that vair —» 100%, the perm ittivity
contrast approaches 0 an d PR also approaches 0.
T he sim ulations o f Fig. 5-3 confirm the relationship between low density layers and low PR
values. H ow ever, it does not explain the signature evolution o f bare ice to m ature F Y ice. W hile
brine drainage may increase the air volum e of ice o ver time, the total air volume o f FY ice is
norm ally no more than a few percent (Tucker e t al., 1991). A low density layer m ust be added to
the ice either in the form o f snow o r frost flowers to achieve air volum es of 50% o r more. If an
optically thick layer o f snow is sim ply added to the ice, then, from F ig. 5-3, the PR should ju m p
from that o f bare ice to a value close to that o f FY ice. Further, the field observations indicate
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
63
that even a sm all am ount o f snow can cause a PR sim ilar to FY ice. H ow is it possible to add
snow to the ice surface and obtain the gradual evolution of the signature observed in SSM /I?
The field data suggest that em ission from snow covered ice is extrem ely sensitive to the
liquid content o f the snow . At LEA D EX , an overnight decrease in PR was observed as the overall
brine volum e o f a slush layer appeared to decrease. In general, if snow falls on very thin ice,
slush form s. The surface tem perature decreases as the ice ages and thickens. At th e sam e time,
brine drains into the underlying ice. T he net result is a decrease in the brine volum e o f the snow.
W e sim ulated this situation using a hom ogeneous snow layer over the sam e ice base used for the
brine slick (Fig. 5-4). T w o different ice volum es (hereafter Vj) w ere used for the snow with vb
varied from 0 to 24%. T he snow is optically thin at low brine volum es which can cause coherent
interference o f the radiation em itted from different layers within the model. To m oderate the
effects o f interference, snow depth was varied from 1 to 3 cm in 1 m m increm ents and the output
was averaged to produce a single spectrum at a given vb.
The calculated PR spans most o f the range observed in the thin ice regions o f the Bering
Sea. For vb = 24% , the m odel predictions lie near the thin ice end o f the thin-FY ice cluster. The
PR o f . 12 fo r a vb of 20-24% also agrees reasonably w ell with the observations o f slush at
LEAD EX . A s vb decreases, PR also decreases and approaches that o f FY ice w hen.vb reaches 2
to 6%. T he snow is still reasonably close to optically thick at a vb o f 2%. If vb is decreased
further, the 1-3 cm snow layer becom es transparent to m icrow ave radiation and P R increases
because em ission is prim arily from the underlying ice.
This sim ulation dem onstrates a physically reasonable way for the thin ice spectrum to
progressively change into that of FY ice. However, it is not likely that it explains the Bering Sea
evolution. T here was no snow fall in the region during late M arch and early April except for one
day when a trace o f snow was recorded at Nome. W hile it is possible that airborne ice crystals
originating from the polynya fell on the ice downw ind, this would in all likelihood result in only a
very thin snow cover. To test the effect o f a very thin snow cover, the previous sim ulations were
redone w ith a 0.1 to 1 cm thick layer o f snow (Fig. 5-5). W ith high vb even a thin cover is
optically thick, resulting in high PR. A gain, the data co v er much o f the range o f observed PR as
vb is reduced. T he low est values o f PR are around .045, indicating that further surface
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
64
m odification would be needed to com pletely produce a FY ice signature. N onetheless, a thin
slushy snow cover may explain much o f the Bering S ea thin ice signature evolution.
5.3.3 F rost flo w ers
In the absence o f snowfall, the m ost likely way for a low density layer to have form ed on the
Bering S ea thin ice is by frost flow er growth. It has been observed th at the vertical brine
distribution on frost flow ers can vary dram atically, w ith the bases having a Vb o f nearly 50% and
the tops only a fraction o f a percent (P erovich and R ichter-M enge, 1994). H ow ever, the precise
distribution o f brine as a function o f height has never been m easured. W e, therefore, conducted a
series o f sensitivity studies where vb w as varied as a function o f heig h t such that
r
vb(z) = vb(z
) + vb(0)
z
\p
(5-14)
V Z , °P )
w here z is the height above the ice surface and z,op is total height o f the frost flow ers. The brine
volume is vb(0) at the base o f the flow ers and vb(ztop) at the top. T h e param eter p w as varied fro m ,
10 to 400, w ith higher pow ers indicating that the brine was increasingly confined n ear the surface
o f the ice (Fig. 5-6).
As the liquid w ater content o f snow increases, the snow rapidly undergoes m etam orphosis
with the grains enlarging and the pack as a whole collapsing (C olbeck, 1982). P erovich and
R ichter-M enge (1994) observed a sim ilar m etam orphosis on frost flow ers as they w ere w armed
by sunlight. They also observed that the frost flow ers w ere destroyed by flooding o f the ice
surface. H ence, we expect that frost flow ers will not accom m odate large brine volum es relative to
the ice volum e. For the sim ulations that follow, we set v, to some m inim um value w ith the
adjustm ent that Vi m ust alw ays be equal to at least 2 .5 * v b(z). This w as done up to the point that
no air rem ained in the ice, beyond w hich
V j( z )
= 1- vb(z). The frost flow er layer w as again
assumed to be 1 to 3 cm thick for all o f the sim ulations. The practical effect o f this adjustm ent o f
Vi was that, in m ost cases, the lowest part o f the layer (up to around .5 cm) has increasing vb and
Vj, much as one might expect for a m ature field o f flow ers of varying heights.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
65
In our first set of sim ulations, the m inim um Vj w as set to 5%, vb(ztop ) = 0% and vb(0) was
varied from 2 to 50% (Fig. 5-7). Both the am ount o f brine and the w ay in which the b rin e is
distributed had a major effect on the predicted signature. For exam ple, changes in th e surface
brine volum e had little effect on em ission when the brine was restricted to a very thin surface
layer (p=400). But, if the brine was distributed upw ard by even a sm all am ount (p= 100), then PR
varied by a factor of tw o in direct proportion to changes in vb(0). T he sensitivity o f P R to vb(0)
and PR decreased and eventually reversed as brine w as distributed still further up o nto the
flowers. S im ilar results hold for GR as w ell. W hen the brine was near the surface, G R did not
vary with vb but, as the brine was distributed higher on the flower, G R becam e a strong function
o f vb. H ow ever, this relationship again w eakened for p > 20.
M ost o f these effects are directly related to the optical thickness o f the different layers within
the frost flo w er cover. W hen p = 400, the brine was isolated in a layer little more than . 1 mm
thick such that the layer w as optically thin even for large vb and contributed little to em ission. If
the surface brine was reduced to 2%, the only layers contributing to em ission were the dry frost
flowers and the underlying ice. In that case, the PR w as around .08. T his is the asym ptotic limit
o f PR we obtained for bare ice when the brine volum e w as reduced to 0% . It is also the same
value obtained at bare and frost flow er covered sites during C R R ELEX 1994 and m ay explain
why there w as no difference in PR betw een the two sites. In fact, a 1 to 3 cm. layer o f dry frost
flowers is optically thin and therefore has alm ost no effect on em ission. F o r all o f the brine
distributions tested, this sam e lim iting value o f PR was obtained w hen vb(0) was 2% . W e
conclude that dry frost flow ers by them selves contribute essentially nothing to em ission from sea
ice.
W hen p = 100, the brine occupies a layer about .5 m m thick, essentially a brine slick at the
base of the flow ers. The portion of the flow ers above this level is dry an d contribute little to
em ission. E ssentially, the p= 100 case reproduces the b rin e slick sim ulation of Fig. 5-2, but for a
slightly d ifferent thickness and brine distribution. The range of PR from thin to bare F Y ice was
again reproduced by decreasing vb, the only difference being the range o f brine volum es involved.
At low er values o f p, m ore brine is distributed upw ards and the relationship betw een PR and
vb depends on a delicate balance between the optical thickness and perm ittivity o f the frost
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
66
flowers. For p=30, the 19 G H z radiation originates at the brine-soaked surface w here increasing
vb sim ply increases the perm ittivity contrast between the surface and the air and therefore PR.
W hen p=10, enough brine exists on the flow ers that they are on the verge o f being optically thick.
A s vb increases, more radiation originates in the flowers, w hich have a low perm ittivity contrast
with the air, and therefore P R decreases. A balance betw een these two effects is nearly achieved
w hen p=20.
A sim ilar balance of com peting effects determ ines G R, but with the difference in penetration
depth at 19 and 37 G H z playing an im portant role. For p= 100 and p=400, the frost flo w er layer
is optically thin at both 19 and 37 G H z and em ission at both frequencies originates at the brine
soaked surface. A s a result, T b(19, V) and T b(37, V) respond in a sim ilar way to changes in ice
properties, m aking GR essentially independent o f changes in vb. A s m ore brine is distributed up
onto the flow ers, either by decreasing p or increasing vb(0), radiation at 37 G H z is increasingly
em itted by low density layers above the surface. Em ission at 19 G H z continues to include
significant am ounts o f radiation from the brine soaked low er layers because o f the larger
penetration depth. A s a consequence, em ission at 37 G H z increases and so does GR. W hen p<20,
greater am ounts o f 19 G H z radiation originate from the m oist frost flow ers and G R decreases.
From these results w e can construct a scenario to explain the thin ice signature evolution.
C onsider ice that has ju st form ed and has a surface brine slick which is com posed o f 20% liquid
(Fig. 5-8). If, over tim e, the brine volum e o f the slick decreases due to surface cooling or
drainage, then the P R would begin to decrease. W ithin a few days, frost flow ers might form on
the ice and begin to w ick up brine. As the brine moves upw ard, PR continues to decrease,
eventually reaching a value o f less than .04. In this way, PR could evolve from the thin ice values
to nearly those o f FY ice.
This scenario is supported by two pieces o f evidence. First, significant decreases in the
surface brine volum e have been observed w ithin the first few days after initial ice form ation on
leads in the A rctic (P erovich a n d Richter-M enge, 1994). Second, the ice directly dow nw ind of
land in the B ering Sea had a very low albedo but, beyond the first 100 kilom eters or so
dow nw ind, the ice albedo increased steadily. This suggests that the ice near land was both thin
and bare, and that frost flow er form ation did not begin until the ice was at least several days old.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
- .
67
O n the other hand, the points shown in Fig. 5-8 indicate that GR should increase with decreasing
PR, while the reverse is true o f the satellite data. This m ay indicate a problem with th e scenario
or simply that scattering losses at 37 G H z m ust be included in the m odel to achieve full
agreement. A lso, it is not clear whether a slow but steady m ovem ent o f brine up o nto the flowers
is realistic. A t LEA D EX , P erovich a nd R ichter-M enge (1994) m easured salinities at the top of
frost flow ers o f between 4 0 and 80 ppt w ithin a day o f formation.
To test w hether small am ounts o f brine distributed throughout th e frost flow er layer would
affect the signature we varied vb(top) from 0 to 2% (Fig. 5-9). In addition, V; was varied from 5 to
25% to sim ulate frost flow er and snow covers o f different densities. E ach plot in Fig. 5-9 has the
sam e pattern o f results as F ig. 5-7. T he different values o f p form a series o f lines radiating away
from a com m on point defined by low brine volume. T h e intricate relationships betw een em ission
and brine content and distribution discussed with regard to the previous plot also apply to these
plots.
From this set o f sim ulations we can draw a few general conclusions. As vb(top) increases,
both the P R an d the .dynam ic range of P R as a function o f vb(0) decreases. This is tru e for all
three values o f Vj tested. In essence, as vb(top) increases, the optical depth o f the low density
cover increases m asking th e underlying brine-ladened ice surface. F or p> 10, PR decreases as Vi
increases . F o r exam ple, w ith vb(top) = 0 an d p = 100, P R ranges all the way up to . 15 when v* =
5% , but rem ains less than . 10 when v, = 25% . Here again, as Vi increases it masks em ission from
the surface m aking surface brine volume changes less important.
The only case with high the PR o f thin ice occurred when a thin w et surface layer was
overlaid w ith an optically transparent, dry cover o f low density (the p = 100 line in F ig. 5-9a). On
the other hand, the low P R o f FY ice can occur for a variety o f situations, usually as the result o f
brine redistribution onto th e flow ers or as a consequence o f an increase in the frost flow er
density. A sm ooth transition to FY ice can also occur in a num ber o f different ways. For
exam ple, the full range o f P R can be traversed if Vj rem ains constant but, due to brine drainage
and wicking, vb(0) decreases and at the sam e tim e vb(top) increases.
Perhaps the m ost likely scenario is a m odification o f the scenario proposed in F ig. 5-8.
C onsider ice w ith a surface brine slick w hose vb decreases with time (F ig. 5-10). A fter a few
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
days, a light frost flow er cover forms and im m ediately wicks up brine such that the top o f each
flow er has a small but essentially fixed am ount of brine on it. Then, o ver time, m ore flowers
grow, thus increasing both the ice and brine volume o f the frost flow er layer. By the end of this
process, PR has covered the full range o f signatures from thin to F Y ice.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
69
ow
0 .0 5
..5 0 %
•40
P 0.00 -
-0 .0 5 -
-
0 .1 0
my
-
0.00
0 .0 5
0.10
0 .1 5
0.20
0.2 5
0 .3 0
P R (19)
Fig. 5-1: Theoretical model sim ulations of em ission from optically thick ice.
T he ice contains no air and the bulk brine volum e is varied as indicated. A lso show n are the
SSM /I O W -FY -M Y triangle. D ata from the thin ice region in the Bering Sea is plotted as
sm all open gray circles.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
70
ow
0.05
50%
>
r-»
cn
0.00
,20
-0.05
-0.10 0.00
my
0.05
0.10
0.15
0.20
0.25
0.30
PR (19)
Fig. 5-2: Sim ulations o f a .7 m m brine slick over optically thick ice.
The slick contains no a ir and the brine volume is varied as indicated. The underlying ice is
91% ice, 6% brine and 3%air.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
71
.... ow
90 70 50
30
HF
0%
fy \
my
0.00
0.05
0.10
0.15
0.20
0.25
0.30
PR (19)
5KlMor|
VlbM
F ig. 5-3: Fresnel model output for optically thick ice w ith no brine.
Air volum e was varied as indicated.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
72
0.00
0 .0 4
0 .0 8
0.12
0.16
0.12
0.16
a. v. = 2 0 %
0.03 -
0.00
-
-0.03 -
b. v. = 4 0 %
>
Ov
0.03 -
10. • • " '1 4
0.00
-
-0.03 -
0.04
0.08
P R (1 9 )
W ETSNOW2.iifg
Fig. 5-4: Theoretical calculations o f the m icrowave signature o f snow over sea ice.
Data are the average o f model output from 1 to 3 cm snow thickness. T w o different ice
volum e fractions w ere used for the snow: a) 20% ice volume and b) 40% . B rine volum e was
varied from 0 to 26% as indicated by the num bered dots.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
73
0.00
0.04
0.08
0.12
0.16
0.08
0.12
0.16
a. v. = 20%
0.03
0.00
-
-0.03 -
a
u
0.03 -
10-
0.00
-
-0.03 -
0.04
P R (1 9 )
W ETSNOW 2.org
F ig. 5-5: T heoretical calculations o f the m icrow ave signature o f a very thin snow over sea ice.
Data are the average o f m odel output from 0.1 to 1 cm snow thickness. Tw o different ice
volume fractions w ere used for the snow : a) 20% ice volume and b) 40% . Brine volum e was
varied as indicated by th e num bered points.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
74
0.6
o
u_
O
c
o
100
400
0.0
0.2
0.4
0.6
0.8
1.0
F rac tio n o f v b at th e b a s e o f the fro st flo w e rs
wwi
Fig. 5-6: B rine distribution used in th e m icrowave sim ulations o f fro st flowers.
Each curve represents a different value o f p in equation (5-14).
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
75
b . p = 100
0.02
-
0.00
-
2 -1 0
2-10
d. p = 20
0.04 - c ' P " 30
0.02
-
0.00
-
2-10
0.04 £
r<i
>
o
0.04
0.08
0.12
0.16
0.02 0.00
-
0.00
0.04
0.08
0.12
0.16
PR(19 GHz)
Fig. 5-7: Simulations of frost flowers over sea ice for a minimum ice volume of 5%.
In all cases, vb(zlop) was set to 0%. Each plot is for a different value of p in equation (5-14).
Each point represents a different value of vb(0) which ranges from 2 to 50% as indicated (in
all cases down triangles are 2% and up triangles are 50%). Also shown the NASA Team
algorithm triangle and the thin ice points from the Bering Sea.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
76
0 .0 4
>
0.02
p = 10
0.00
0 .0 4
s lic k
p = 2 0 p=IOO
0 .0 8
vb='0
0 .1 2
s lic k ^
v.h = 2 0
0 .1 6
P R (1 9 G H z)
n o w o j .tn g , A/H A)s. ••
Fig. 5-8: Possible scenario o f thin ice signature evolution.
Scenario includes an initial brine slick, the form ation o f frost flowers and brine w icking onto
the flowers. T he triangles are sim ulations o f a brine slick on thick ice as show n in Fig. 5-2.
T he brine volum e of the slick w as assum ed to decrease from 20% to 10% during the initial
signature evolution. The circles represent sim ulations o f a 1 to 3 cm thick frost flow er layer
as shown in F ig. 5-7. The ice volume o f the frost flowers layer was fixed at 5% and th e top
and bottom brine volumes w ere fixed at 0 an d 10% respectively. The distribution o f brine
w as varied using (5-14) by varying p from 100 down to 10 as indicated.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
77
viKP) =5%
15%
25%
0,04
0.02
....
...
0.00
o
iy
^ .
0.04
0.04
0.02
......
0.00
: oi..:
0.04 -
0.02
-
0.00
-
0.00
N)
$
....
0.04
0.08
0.12
0.16
0.04
0.08
0.12
P R (1 9 G H z )
0.16
0.04
0.08
0.12
0.16
- p = 10
-20
-3 0
- 100
-4 0 0
Fig. 5-9: Sim ulations o f frost flowers ov er sea ice.
E ach colum n o f plots is for a constant m inim um value o f v, and each row is a particular
value o f V b(z,op) . W ithin each plot the lines are different values o f p and are indicated by the
sam e color schem e as Fig. 5-7. V alues o f vb(0) indicated by dow n triangles are 2% and up
triangles are 50% . A lso shown is th e N A SA T eam algorithm triangle and the thin ice points
from the Bering Sea.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
78
0.04 -
t— 0.02 -
O 0.00 heavy ff
0.00
0.04
light ff
0.08
0.12
0.16
P R (1 9 G H z )
n o m : K s .o r t, h/ u n s ••
Fig. 5-10: Second possible scenario o f thin ice signature evolution
T his scenario includes an initial brine slick and the eventual grow th o f frost flowers. T he
triangles are sim ulations o f a brine slick as described in Fig. 5-2. The brine volume o f the
slick w as assum ed to decrease from 20% to 10% during the initial signature evolution. T he
circles represent sim ulations o f frost flowers as described in F ig. 5-9. T he “ light f f ’ point is
from a sim ulation o f a 1 to 3 cm thick frost flow er layer with Vj = 5% , vb(top) = .5% and
vb(btm ) = 10%. T he “ heavy f f ’ point is from a sim ulation o f a 1 to 3 cm th ick frost flow er
layer w ith Vj = 15%, vb(top) = 2% and vb(btm ) = 6% . The brine was distributed on the
flow ers using (5-14) with p = 100 in both cases.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
6. M ic ro w a ve sa t e l l it e da ta a n d e n v ir o n m e n t a l c o n d it io n s
From the theoretical calculations discussed in chapter 5, we have established how surface
physical properties affect m icrowave em ission from thin sea ice. W ith this know ledge we have
been able to construct plausible scenarios for how the passive m icrow ave signature o f thin ice
evolves into that o f FY ice. These scenarios involve brine drainage and frost flo w er growth.
Unfortunately, neither o f these m echanism s is w ell understood. T here are few detailed
observations o f these processes and the observations are limited in scope. C ertainly there is no
theory fo r how either occurs. As a result, the m icrow ave model cannot be used to quantify the
relationship between ice age and em ission. N or can it be used to quantify the effect environm ental
conditions will have on these processes and, therefore, on em ission. O ne possible way to gain
greater insights into these relationships is through an exam ination o f the satellite data itself. T he
broad expanse o f thin ice apparent in the April 2 SSM /I data from the Bering S ea took several
weeks to form. By com paring the tim e series o f m eteorological d ata and satellite imagery, it m ay
be possible to infer not only the relationship betw een ice age and signature, but perhaps also how
that relationship is affected by environm ental conditions.
In this chapter, w e begin by com paring local m eteorological d ata from the B ering Sea in the
spring o f 1988 with S S M /I data. F rom this com parison, we will see that the thin ice signature
was prim arily associated with periods o f no snow fall and that the regional distribution o f thin ice
was strongly dependent on the current Field. Tim e sequences o f Tb from 3 sites dow nw ind o f St.
Law rence Island are then used to obtain an initial estim ate of the relationship betw een ice age and
em ission. The Tb sequences are correlated with w ind, air tem perature, dow nw elling shortw ave
radiation, snowfall and other m eteorological param eters. In general, the correlations are relatively
w eak but certain patterns can be discerned. U ltim ately, emission from a given p iece o f ice
depends on the surface properties as influenced by the m eteorological conditions w hile em ission
from a given pixel depends also on the nature of the ice advected through the pixel. Isolating the
effects o f ice dynam ics is the subject o f the next chapter.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
6.1 Meteorological data
Fig. 6-1 sum m arizes the m eteorological conditions in the Bering S ea during M arch and early
April o f 1988. A ir tem peratures and snow fall were observed at the N om e station at the north edge
of the polynya region in N orton Sound. W ind data recorded at N om e include local orographic
effects and are, in general, not indicative o f mean wind patterns for the region (C. Pease, pers.
comm.). The gradient w ind was, therefore, calculated from the regional pressure fields near St.
Law rence Island as reported by the N ational M eteorological C enter (N M C). The m eteorological
data can be divided into three distinct periods. D uring the first period, from late F ebruary to early
M arch, w inds were usually from the north or northeast at 5 to 12 m /s w ith air tem perature
varying from slightly above freezing to nearly -30°C. T h e distinguishing feature o f this period,
how ever, w as the regular daily accum ulations of 3 or m ore centim eters o f snow. D uring the
second period, from M arch 13 to April 10, winds were consistently from the northeast at 15 to 25
m/s and air tem peratures ranged from -6 to -30°C with a strong diurnal cycle o f betw een 10 and
20°C. U nlike the previous several weeks, the N om e station recorded virtually no snow during this
period with the exception o f trace am ounts on M arch 25 and April 8. C onditions changed rapidly
starting on A pril 10 when a 2 cm snow fall signaled the beginning o f spring melt w ith air
tem peratures at or above freezing, winds from the east-southeast, and consistent snow fall.
The m eteorological data suggest that large am ounts thin ice would have been generated in the
Bering S ea from the very beginning o f M arch. H ow ever, it is likely that the passive signature o f
the ice w ould be m asked by snowfall until M arch 14 or 15. A fter that, polynyas dow nw ind of
land would rem ain sources o f thin bare ice that could be advected southw ard until the second
w eek in A pril. Snowfall starting on A pril 8 would again alter the signature, as w ould melting
shortly thereafter.
6.2 Passive microwave satellite data
Passive m icrow ave data from the B ering Sea were obtained anyw here from 1 to 6 times a day
depending on the satellites orbital path relative to the Earth. Norm ally a series o f passes were
m ade during the morning between 2 and 5 A M local tim e and another set during the afternoon
betw een 4 and 7 PM local tim e. The data exam ined here com e from data tapes provided by
Rem ote S ensing System s in Santa Rosa, California and w ere calibrated and geolocated in the
R ep ro d u ced with p erm ission of th e copyright ow ner. Further reproduction prohibited w ithout p erm ission .
normal m anner with one exception. D uring processing, a correction fo r th e along-track
geolocation o f individual footprints must be applied. In essence, the satellite incorrectly reports
the location o f an individual footprint as being slightly further along the orbital path than it
actually is. T he standard correction is 15 km , how ever tests conducted for this particular region
and time suggested that a slightly larger correction was needed, and all o f the data have been
shifted by 19 km along track instead1.
After calibration, the d ata w ere translated onto the po lar stereographic grid o f 25 km by 25
km pixels com m only used fo r sea ice research (Com iso a n d Z w a lly, 1989). This gridding process
often results in a scattering o f pixels w hich contain no data. Such pixels w ere assigned values
equal to the average o f adjacent pixels to avoid visual clutter. This process does not en h an ce the
inform ation content of the d ata but it does m ake the m aps easier to look at.
The SSM /I footprint at 19 G H z is on the order o f 50 km , twice the size o f a grid pixel. The
effect of the large footprint is evident in regions where the T b field varies rapidly over short
distances. D ow nw ind o f land, a 50 km footprint may include emission fro m thin ice, the open
w ater in the polynya and the land itself. Consequently, the sharp contrast in em ission from land to
thin ice is sm eared out over both the land and the ice (e.g. Fig. 6-4). T he low est T h values are
thus often found in regions one pixel rem oved from the land, where em ission is prim arily from
very thin ice and the polynya. This makes it difficult to determ ine the ex act signature o f newly
form ed ice in the Bering Sea. However, the transition from thin to FY ice south o f St. Law rence
Island occurs gradually over a distances 4 0 0 km or more, m aking it resolvable at the 19 G H z
resolution.
W e begin ou r exam ination o f the satellite data with regional maps o f T b( 19, H) (Fig. 6-2). The
d ata are daily averages o f all satellite passes betw een 12Z and 12Z G M T. A typical m ap shows
ice covering m uch of the B ering Sea with open water along the southw est (left) edge o f each map.
Since Tb is only plotted for values greater than 165K the w ater appears w hite with the ice edge
1 M uch of the work that follow s was originally done using calibrated d ata distributed by the
National Snow and Ice D ata Center on C D -R O M . T he d ata had apparently not been corrected
fo r along track error and m uch o f the original work had to be redone using the original data
tapes.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
82
sharply defined by a line o f m ixed green, yellow and red pixels marking the transition from ice
w ith Tb values o f more than 200K to w ater values o f around 100K.
T he three d istinct periods in the m eteorological data correspond to three phases in the T b
record. W ith heavy snowfall and lack o f consistent offshore w inds before M arch 15, areas o f low
T b were restricted to regions adjacent to land, with the low est T b being above 200K. O nce winds
turned consistently northerly and snowfall stopped, regions o f low T b spread dow nw ind o f
S ew ard Peninsula, Siberia, St. Law rence Island, and even St. M atthew Island. The spread of the
thin ice reached its peak on A pril 2 with average daily T b( 19, H) values as low as 170K next to
land. A s winds died and snow again fell, the low T b characteristic o f thin ice all but disappeared.
T hereafter, w inds from the south and air tem peratures above freezing led to inconsistent thin ice
form ation and eventually to the annual spring breakup of the ice pack.
6.3 The effect of currents
T he satellite d ata suggest that each o f the thin ice regions represents a different regim e o f ice
dynam ics. Ice m otion is the com bined result o f wind stress, ocean current stress, internal ice
stress and the C oriolis force. O cean currents in the Bering S ea are induced by an overall height
difference betw een the Pacific and Arctic O ceans as well as local height differences due to wind.
O verland a nd R o ach (1987) m odeled ocean currents in the B ering Sea u sin g a tw o dim ensional
barotropic model that included ocean bottom topography. A typical w intertim e case w ith surface
w inds o f 8 m/s from the northeast causes an overall water transport from th e Pacific to the Arctic
prim arily by flow through the G u lf o f A nadyr to A nadyr S trait and on to B ering Strait (Fig. 6-3).
S econdary flow traverses the southern edge o f St. Law rence Island, contributing to a return
current into the P acific along th e A laskan coast. Flow in N orton Sound is severely restricted and
therefore weak.
South of St. Law rence Island, ice motion induced by northerly winds is largely unhindered by
internal ice stress and the ice is norm ally assum ed to be in a state o f free drift. The currents are
also southbound, except near land, and the resulting motion can be quite rapid. D uring the second
tim e period, the com bination o f w ind and current led to a broad expanse o f lowered Tb up to 400
k m w ide. Sim ilarly, thin ice produced dow nw ind of St. M atthew Island w as also in a state o f free
drift. In the satellite data, this ice appears as if it were an extension of the thin ice region
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
-
83
dow nw ind o f St. Law rence Island. T o the west, in the G ulf o f A nadyr, northbound currents
substantially reduce the rate o f southw ard ice advection. This region o f low Tb w as slow er to
form and w as never w ider than 200 km o r so. There is even evidence o f ice having been carried
northw ard through A nadyr Strait and into Bering S trait during periods of calm o r winds from the
south. For exam ple, from April 6 to 8, the region o f T b < 200K (w hich had been restricted to the
southern portion o f A nadyr Strait) apparently m igrated north o f St. Law rence Island (Fig. 6-2).
Southw ard ice m otion in Norton Sound, south o f Seward Peninsula, is im peded by the
southern coast o f the Sound. Along the w estern edge o f Seward Peninsula, the ice is free to m ove
southw ard betw een St. Law rence Island and the coast o f Alaska, leading to a region o f the
lowered T b that extended from land fo r several hundred kilom eters dow nw ind d u rin g early A pril.
H ow ever, along this route the thin ice also mixes w ith mature FY and M Y ice originating in the
Chukchi Sea. In general, satellite-observed variations in emission from this area are not purely
due to thin ice evolution.
6.4 Time series
To illustrate the variability o f em ission in M arch and April, tim e series o f T b from selected
locations in the Bering S ea were produced. Three points south o f St. Law rence Island were
chosen for study (Fig. 6-4). The average T b values from a 2 pixel by 2 pixel (50 km by 50 km )
area were calculated for each orbital pass where d ata w ere available.
On M arch 9, Tb(19) values were above 230K at all o f the St. L aw rence sites with
correspondingly low PR values o f around .04 (Fig. 6-5 and Fig. 6-6). A t site S L - 1, this rapidly
changed on the M arch 10 as Tb(19, H ) dropped nearly 30K, stabilized for 5 days, and then
dropped by another 30K over the course o f 3 days. It rem ained at around 180K throughout the
next three w eeks. Sm all scale changes in emission during this three w eek period define three
distinct segm ents in the tim e series. F rom M arch 17 to 22, Tb(19, H ) averaged 180K; from
M arch 23 to 29, Tb w as about 10K higher (except fo r a dip on M arch 26); after M arch 29,
em ission reached its low est point and then slowly rose. Emission also varied substantially from
day to night during the periods of low est T b. Finally, on April 8, T b(19,H ) values at SL-1
suddenly jum ped up by about 50K to values close to those at the start of the sequence. Sim ilar,
though less dram atic, behavior was evident at the o th er two St. L aw rence sites. In each case, T b
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
84
decreased on M arch 10, stabilized and then eventually decreased again. Each site registered
elevated Tb from M arch 23 to 29, after which Tb at all sites decreased to their low est values. On
April 8, sites SL-1 and SL-2 had dram atic increases in em ission w hile Tb(19,.H ) at SL-3 actu ally
decreased by about 15K.
S im ilar patterns are present in the Tb(37, H) d ata (Fig. 6-7). T h e timing o f decreases and
increases was the sam e as that o f the 19 G H z data, so much so that the data at these two
frequencies have a linear correlation coefficient o f between .90 and .94. The o nly major
difference is that 37 G H z had a slightly sm aller dynam ic range. Both had starting values o f
235K, b u t the low est values o f Tb(37, H) observed at SL-1 w ere slightly higher, around 185 to
190K, com pared w ith 175 to 180K at 19 GHz. O th er than that, m easurem ents at the two
frequencies were nearly indistinguishable.
M uch o f this variability in em ission can be directly tied to environm ental conditions. For
exam ple, the decrease in Tb on M arch 10 was part o f a regionw ide decrease in em ission that
roughly coincided w ith a period o f heavy snowfall. T he decrease w as largest near land, perhaps
due to snow interacting with brine on very thin ice o r possibly due to increased thin ice
production as winds began to pick u p from the north.
A nother 3 0 K d ro p in T b( 19, H) occurred at SL-1 on M arch 15 (Fig. 6-5). S im ilar decreases
occurred at the other sites but w ere delayed by ab o u t 3 days at S L -2 and another 2 or 3 days at
SL-3. T h e decrease in T b and the delay dow nw ind are clear indicators of the production and
advection o f thin, bare ice. By M arch 17, average T b values at SL-1 had bottom ed out. W e can
assum e from then on th a t the polynya was in a steady state o f ice production and southw ard
advection. Likew ise, after M arch 20, the flux of thin ice through SL-2 appears constant. T b at
SL-3 reached a m inim um on M arch 22, although it began to rise shortly thereafter. We can
obtain a crude estim ate o f the relationship between T b and ice age from this portion o f the tim e
sequence. It is likely th at bare ice produced in and around S L -1 w as advected southw ard through
SL-2 and SL-3. The decrease in T b indicates that old er snow -covered ice was advected out o f
each site and replaced by the bare thin ice produced upwind. A ssum ing this is true, the delay in
the decrease in Tb dow nw ind indicates the speed o f the ice. Since it took approxim ately 5 to
6
days for the thin ice to travel from SL-1 to SL-3, the ice drift rate w as about 30 km /day. On
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
85
M arch 2 1, Tb( 19, H) at SL-3 reached its low est value of 21 OK; at the sam e tim e em ission was
on the order o f 19 5 K at SL-2, and 180K at S L - 1. This would seem to indicate that em ission
from the advected ice was steadily increasing with tim e, with Tb rising by about 5K p er day.
Thus, it would have taken about 12 days for the signature to change from that of thin ice (with
T b( 19, H) ~ 180K) to that o f FY ice (T b(19,H ) - 240K).
T he same reasoning applies to at 37 GH z. On M arch 21, Tb(37, H) w as 1 9 0 K at S L -1, 205K
at S L -2, and 215 at SL-3. This suggests an increase in em ission o f betw een 4 and 5 K per day.
T he full range o f T b, from 190K for thin ice to 240K for FY ice, w ould again have taken on the
order o f 12 days.
6.5 Correlation studies
O nce established, the thin ice zone was m aintained by northerly w inds for the next 3 weeks.
T he three sites exhibited a num ber o f small scale changes in em ission during this tim e that were
undoubtedly related to variations in environm ental conditions such as w ind, air tem perature and
solar radiation, as w ell as occasional snow fall. It is difficult to unravel th e effect each o f these
had on em ission since they often varied sim ultaneously. A correlation study o f daily m axim um
and m inim um values o f Tb(19, H ), T b(37, H), and several environm ental param eters was
conducted on d ata from M arch 16 to April 7 (Fig. 6-8). Environm ental param eters considered
included: (i) m axim um and m inim um daily values o f T ^ recorded at N om e, (ii) daily values of
the u (east-w est) and v (north-south) com ponents o f the gradient wind o ver St. Law rence Island,
and (iii) down w elling shortw ave radiation estim ated using the equations o f Shine (1984) with
cloud fraction determ ined from ISC C P satellite data for a 2.5° latitude by 2.5° longitude box
centered on the B ering S ea (Steffen and Schw eiger, 1992). T he correlation between T b( 19, H) or
T b(37, H) and environm ental conditions was w eak to nonexistent in alm ost all cases, w ith
correlation coefficients o f 0.2 o r less. Correlation coefficients were highest at SL-1, but the
largest o f these (fo r the v com ponent o f the w ind) was still only 0.29 at 19 G H z and 0 .24 at 37
G H z. M ultiple linear regression o f T b at this site versus all o f the environm ental conditions
together yielded a correlation coefficient o f around .63 at 19 G H z and .46 at 37 GH z. Daily
averages o f T b( 19, H ), T b(37, H), and environm ental data w ere also com pared. A gain, the largest
correlation (0.46 at 19 G H z and 0.34 at 37 G H z) was betw een v and T b at S L - 1.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
86
Form ally, a correlation coefficient o f .63 or less is not generally considered statistically
significant. H ow ever, the lack o f statistically significant correlation between em ission and
environm ental param eters does not mean that no correlation exists; the physical processes are
com plex, the record is relatively short and the environm ental conditions are not always
independent o f each other. There are discernible patterns in the data, patterns w e will exam ine in
the next few sections.
6.5.1
W ind
W e have already discussed the general relationship between wind, currents and the
distribution o f thin ice dow nw ind o f land. The data from SL-1 used in the correlation analysis
em phasizes this relationship (Fig. 6-8). The Tb values were low est when the w inds were strong
from the north, as from M arch 17 to 22 and again from M arch 30 to April 2. W hen winds w ere
w eaker, as from M arch 23 to 29 and after April 2, em ission increased. The response to changes
in w ind could be both sensitive and imm ediate at this site. On M arch 26, w hen w ind speed briefly
■
increased by 6 m/s, T b(H) decreased by 20K at 19 G H z and 15K at 37 GHz. A sim ilar sensitivity
was evident from A pril 3 to 7, w hen wind steadily declined fro m . 17 m/s to 3 m /s while average
T b( 19, H ) progressively increased from 173K to 190K (with com parable resp o n se at 37 G H z).
S im ilar but w eaker correspondence exists betw een wind and em ission at SL -2 and SL-3.
M uch o f the correlation at these tw o sites appears to simply be d u e to the age o f the ice as
influenced by the w ind. Low T b values were m easured sometim e after the o nset o f episodes o f
strong southerly w inds. Since ice is advected m ost rapidly during periods o f peak wind, ice at the
dow nw ind sites w ould have been relatively young and would have had a low T b. The advection
rate apparently decreased when the v com ponent o f the wind slackened on M arch 22. This w ould
have caused the ice to be older by the time it reached SL-2 and SL -3 and thus explains the
increase in Tb.
A t least some o f the sensitivity to wind at site SL-1 may have been the resu lt o f changes in the
w idth o f the polynya and not sim ply a reflection o f the age o f the ice near land. Polynya w idth
depends on a balance between ice production and ice advection, both o f w hich are increasing
functions o f wind speed. Pease (1987) constructed a simple m odel for polynya width and
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
-
87
calculated that w idth varies rapidly with wind strength for surface speeds fro m 0 to 5 m/s.
H ow ever, as speed increases beyond 5 m/s, enhanced ice production keeps p ace with advection
and the calculated polynya width rem ains constant. The surface wind speed is approxim ately
60% o f the gradient wind speed in the Arctic (O verland and D a vid so n, 1992). In the case o f data
from April 3 to 7, estim ated surface wind speed w ould be from 2 to 10 m/s covering a range
som ew hat beyond th e calculated dom ain o f strong sensitivity. O f course, the calculated sensitivity
m ay also be underestim ated. A possible source o f inaccuracy in the model is the assum ption that
the depth to which frazil accum ulates before consolidating rem ains constant w ith wind speed. It is
reasonable to think that the accum ulation depth actually increases with wind speed due to
enhanced turbulence in the w ater colum n. This w ould enhance the sensitivity o f polynya w idth to
w ind strength in general and m ight explain the observed sensitivity of em ission at speeds abo v e 5
m/s.
6.5.2 A ir tem perature and shortw ave radiation
O f course, polynya width doesn’t simply d epend on wind speed; it also varies with therm al
forcing. Surprisingly, there is very little overall correlation betw een the m icrow ave data and the
air tem perature o r shortw ave radiation (Fig. 6-8). F or exam ple, from M arch 2 0 to 22, Tb
rem ained essentially unchanged at S L -1 even though the daily average T ^ decreased from -1 0 ° to
-20°C. D uring this tim e, Tb(19, H ) and Tb(37, H ) were at or n ear their low est values. Such low
T b values were m easured again on M arch 29, even though Tair w as 14°C higher and the m axim um
daily shortw ave radiation had increased by 150 W /m 2. The lack o f correlation is surprising. F or
the range o f tem peratures recorded at Nome, theory suggests that polynya w idth should have
increased by at least a factor o f 2, from 20 to 4 0 km . Why this is not reflected in the m icrow ave
d ata is not clear.
T here are tw o periods when the influence o f air tem perature and shortw ave radiation are
noticeable in the m icrow ave data. F rom M arch 17 to 22, and again from M arch 30 to A pril 5,
there were large diu m al cycles in em ission at SL-1 where Tb(19, H) was as m uch as 20K g reater
during local early m orning com pared to the afternoon. The oscillation was sm aller at 37 G H z,
only around 10K, T he oscillation is also noticeable to a lesser extent at SL-2 and SL-3 during
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
-
early April when m orning em ission was about 5K larger at both frequencies (Fig. 6-5 and Fig. 6 7). T he diurnal variability o f em ission is also evident in the regional m aps o f d ata from m orning
and afternoon (Fig. 6-9). A s the thin ice region expanded, the diurnal oscillation generally becam e
m ore pronounced. T here are obvious 50 km wide regions where Tb(19, H) varied from 180K in
the m orning to less than 160K during the afternoon. The dium al cycle essentially disappeared as
the strength o f the w ind decreased in both mid-M arch and early April.
T he oscillation in Tb was out o f phase with air tem perature and, presum ably, therefore ice
tem perature. This m eans that it w as not the result o f an oscillation in the em itting tem perature of
the ice. A t sites SL-2 and SL-3, w here only ice was present, the diurnal cycle w as probably the
result o f changes in surface brine volum e as the surface warmed and cooled o ver the course o f a
day. In the case o f SL-1 , possible causes include both changes in surface brine volum e and
changes in polynya width. F or 10 m/s winds and an air tem perature o f -14°C, Pease (1987)
calculated that polynya width changes slowly, requiring nearly 6 0 hours from the tim e o f initial
form ation to achieve 95% o f the full opening w idth o f 30 km. H ow ever, we know that T b at S L -1
responded very rapidly to changes in wind speed at times. It seem s unlikely that dium al changes
in air tem perature and shortw ave radiation could produce a significant oscillation in polynya
width over the course o f a day.
W ind-driven polynyas form ed in cold tem peratures generally consist o f a sm all area o f open
w ater extending no m ore than a few hundred m eters dow nw ind o f land. The w ater beyond that
contains frazil with the concentration o f ice crystals in the water increasing dow nw ind o f land.
E ventually, the frazil-ladened w ater forms slushy m ats which finally consolidate into ice floes.
The polynya width usually refers to the region from land out to the consolidated mats. The 30
km. w idth estim ated from the form ulae o f Pease (1987) is about as large as a 37 G H z footprint
and a large portion o f a 19 G Hz footprint. Is it possible that tem perature-induced changes in the
em issive properties o f the mats and new ly consolidated ice could produce the observed
oscillation?
T o explore this idea, w e used a simple linear tem perature profile therm odynam ic model based
on M aykut (1978) to predict the surface tem perature o f the mats. W e then calculated vb from the
surface tem perature and em ission using the m icrowave model. T h e mats and floes were m odeled
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
89
as 10 cm thick ice with a therm al conductivity o f 2 (corresponding to a vb o f 20% ) and an albedo
o f .35. The model requires specification of a variety of m eteorological param eters as input.
D uring early A pril, gradient w inds were around 17 m/s, giving an estim ated surface w ind o f 10
m /s. A ir tem peratures ranged from -8°C during the day to -18°C at night. A ssum ing a cloud
cover o f 30%, dow nw elling longw ave radiation calculated from the equations o f M aykut a n d
C hurch (1973) w as estim ated to be 222 W /m 2 at -8°C and 190 W /m 2 at -18°C . D ow nw elling
shortw ave radiation at 1700 local tim e was calculated to be 125 W /m 2 (the equations for both
long and shortw ave radiation are given in A ppendix: D etails o f the ice g ro w th m odel) U sin g this
set o f param eters, the ice surface tem perature w as calculated to be about -6 °C during the
afternoon and -14°C at night.
Since we cannot be certain o f the physical structure o f ice grow n in these conditions, w e chose
to m odel the physical properties o f the mats in several different ways (T able 6-1). First th e m ats
w ere assumed to consist o f ice w ith no air and a fixed salinity o f 24 ppt. T h e salinity was selected
to give the best agreem ent betw een the predicted PR and the daytim e observations o f A pril 1 at
SL-1 (chosen as representative o f low Tb and high PR with large dium al variability). A t a salinity
o f 24 ppt and a tem perature o f -6°C , vb o f the sim ulated ice w as calculated to be 20% , resulting
in a P R o f .146. A t the colder tem perature o f - 14°C, vb decreased to 10% and consequently PR
decreased to .110. T hese PR values cover a som ew hat larger range than the observed values o f
. 146 to . 123 for A pril 1. H ow ever, the calculated T b values at both 19 and 37 G H z are h ig h er
than observed by as much as 25K at H pol an d 30K at V pol. T he sim ulations were redone w ith
salinity adjusted to produce a reasonable m atch to observed daytim e T b(19, H ) but, even then,
T b( 19, V) and T b(37, V) were too large by upw ards o f 15K.
T he second set o f physical param eters considered were those o f a surface brine slick on the
ice. W e assum ed the ice included a .7 mm thick brine slick on top o f an optically thick ice base
com posed o f 20% brine, 3% air and 77% ice. T he slick was m odeled as having no air and a
salinity o f 42 ppt. T hus, vb of the slick was 35% at -6°C and 20% at - 14°C. T his set of
sim ulations m atched the observed range o f T b(19, H), but here again, T b(V) remained high by
around 15K. As a consequence, the PR values were higher than observed by .02-.03.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
90
B oth these and the last set o f calculations indicate that lower frequencies are m ore sensitive to
changes in surface brine volume. F or example, H pol emission is predicted to vary by 23K at 19
GHz but only 1 IK at 37 GHz. T his trend is also true o f the S SM /I observations in w hich Tb(H)
varied by 17K at 19 GHz but only 10K at 37 GHz. T he difference in response is a direct
consequence of differences in the perm ittivity o f brine at these tw o frequencies. T he higher
perm ittivity at 19 GHz com pared to 37 GHz results in greater sensitivity to changes in the brine
volume.
Ice grown in turbulent conditions is likely to have at least som e sm all-scale surface roughness
that will result in brine coated ice crystals projecting slightly above the surface. W e sim ulated
this condition using a surface layer 1.5 to 4.5 m m thick over the sam e ice base used in the last
sim ulation. The properties o f the surface layer w ere calculated u sin g the sam e m ethod of
specifying vertical distribution o f brine volume an d ice volume used for the frost flow ers. T he
brine volum e was assum ed to be 0 at the top o f the surface cover and was varied using (5-14)
w ith p = 15. Again, the ice volum e w as assumed to be a m inim um o f 5% or 2 .5*vb . R easonable
agreem ent with the daytim e H pol values was achieved for a surface brine volum e o f 60%
corresponding to a salinity o f 42 p p t at -6°C. In this case, the discrepancy betw een m odeled an d observed daytime T b(V) was reduced to only 9K at 19 GHz and 12K at 37 G H z. A t least som e o f
the rem aining discrepancy at T b(37, H) may be du e to scattering losses not accounted for in the
m odel. T he brine volum e varied by a factor o f tw o o ver the expected range o f surface
tem perature. As a result, the calculated dynam ic range o f Tb(H) is actually larger than the diurnal
variability at SL-1.
U ndoubtedly, still better agreem ent could be achieved with continued refinem ent o f the
sim ulations. H ow ever, it is apparent from these calculations that a large diurnal oscillation in
em ission can be caused by tem perature induced changes in brine volume.
6.5.3 Snow
T races o f snow w ere recorded at N om e on M arch 25 and on A pril 8 and both occasions
coincide with observed changes in T bat the SL sites. Around M arch 23, there w as an increase in
H pol em ission at all three sites (Fig. 6-5 and Fig. 6-7), The increase was part o f a w idespread
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
pattern affecting all o f the m ajor areas o f thin ice production (Fig. 6-2). The increase in em ission
happened a day or so before both a light snow fall and calm er w inds were reported at Nome.
H ow ever, this particular storm system traveled from west to east and it is likely that these
changes occurred a day or so earlier near St. L aw rence Island. B oth snowfall and decreased ice
advection would have affected em ission. There is evidence o f decreased southw ard advection,
particularly in Norton Sound and along the Siberian coast, w here the regions o f low T b narrow ed
with tim e. There is also some hint o f advection into A nadyr Strait. U nfortunately, since these
changes occurred at the same tim e as the snowfall, it is im possible to determ ine the effect snow ,
by itself, had on em ission.
At the beginning o f A pril, snow fall and a shift in the wind again marked a rapid change in T b.
The effect o f flagging w inds on ice dynam ics is apparent in the regional m aps (Fig. 6-2).
D ecreasing wind speeds on April 5 to 7 allowed currents in the G u lf o f A nadyr to play a larger
role in ice dynam ics and, as a result, the region o f thin ice narrow ed progressively. There is again
evidence in the satellite data of ice advection into A nadyr Strait. A t about the sam e time,
T b(19, H) and T b(37, H ) ju st south o f St. Law rence Island actually increased, perhaps due to
shearing and ice deform ation brought on by the east flow ing cu rren t in this area. D uring A pril 7 and 8, T b(19, H) values throughout the region rose above 195K. T h e increase w as nothing short
o f dram atic. A t S L -1, T b( 19, H) jum p ed nearly 5 0 K w hile Tb(37, H ) w ent up by 40K . A t SL-2,
the increase was 40 K at 19 G Hz and 30K at 37 G H z. (Fig. 6-5 and Fig. 6-n). C uriously, the
outline o f the thin ice region rem ained, delineated by T b < 215K.
C learly, ice dynam ics played a role in increasing emission before April 7. H ow ever, the
changes from April 7 to 8 were rapid and affected an area over 100 km wide south o f St.
L aw rence Island. It is unlikely that ice dynam ics alone could have caused such a rapid, large
scale change in em ission. It is much more likely that snow was the primary cause.
6.6 Summary
In th is chapter, data from sites dow nw ind o f St. Law rence Island have been used to estim ate
the rate at w hich ice w as sw ept aw ay from land. From that estim ate, the general increase in T b
dow nw ind was interpreted in term s o f a relationship between T b and ice age, with em ission
increasing from thin ice values to FY ice values over the course o f 12 days. T he steady increase
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
92
in Tb w ith age holds true as long as the ice remained snow free. E m ission was, how ever,
som ew hat sensitive to changes in environm ental conditions. Increased wind speed produced a
decrease in em ission apparently due to changes in the polynya w idth near land and increased rates
of ice advection rate further dow nw ind. Diurnal changes in air tem perature and shortw ave
radiation caused an oscillation in surface tem perature and, therefore, in surface brine volume and
em ission. Snow appears capable o f causing sudden and dramatic increases in em ission to FY ice
values. T o som e extent, T b( 19) was m ore sensitive to m eteorological conditions than T b(37). F or
exam ple, the diurnal oscillation caused a 20K fluctuation in T b( 19, H ) at SL-1 b u t only 10K in
T b(37, H ).
T here are some phenom ena in the T b record w hich had no obvious correlation w ith the
m eteorological data. T he lack o f sensitivity of T b to a long term change in air tem perature and
dow nw elling solar radiation was one exam ple. A nother was the regional change in T b(19,H) from
the afternoon of A pril 2 to the afternoon o f April 3 (Fig. 6-9). In 2 4 hours, em ission increased by
as m uch as 20K over a large area, even though the environm ental param eters rem ained
essentially unchanged. W hile the cause o f these phenom ena is unknow n, one thing is clear: over
the range o f conditions considered in this chapter, m uch o f the variability in em ission, including
this last exam ple, occurred near land. T h e diurnal oscillation, for exam ple, caused variations o f
upw ards o f 20K in T b(19) at SL-1 b u t no more than ab o u t 5K at S L -2 and SL-3. A w ay from
land, em ission was m ore stable, perhaps indicating a relatively stable relationship betw een T b and
ice age o v er a large portion of the region.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
-
93
26
Mar 2
2
7
12
Mar 17
17
22
27
Apr 1
1
6
11
Apr 16
16
21
10m/s
O.
a.
-10
-
Snowfall (cm)
10
-20
- -30
26
2
M ar2
7
12
17
Mar 17
22
27
6
Apr 1
I1
jl L
16
Apr 16
21
Date in 1988
Fig. 6-1: M eteorological d a ta for the B ering Sea during the spring of 1988.
G eostrophic winds w ere calculated fro m NM C pressure fields in th e vicinity o f St.
Law rence Island. M axim um and m inim um air tem peratures are N o m e station observations;
the solid line is the average o f the tw o. Snowfall is the daily accum ulation recorded at
N om e. Reports of trace am ounts o f snow are plotted as .5 cm.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
94
T B M A P IR .o rg , 6/1 /9 5 , version 3 .5 2
Fig. 6-2: R egional m aps o f T b(19, H ) for the Bering S ea from 1988.
D ata are daily averages (from 12 Z to 12 Z G M T) o f all satellite p asses, w ith dates
indicated fo r each map. T he SSM /I landm ask is show n in gray. R egions o f T b < 165 are
prim arily open w ater and are show n in white.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
95
TB>4AP2R.nrg, 6///9S. v tntnn 3.32
F ig. 6-2 (cont.): Regional m aps o f T b(19, H ) fo r the B ering Sea from 1988.
D ata are daily averages (from 12 Z to 12 Z G M T ) o f all satellite passes, w ith d ates
indicated fo r each m ap. T he SSM /I landm ask is show n in gray. R egions o f Tb < 165 are
prim arily open w ater an d are shown in white.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
96
Siberia
Alaska
.5 m /sec
Fig. 6-3: B ering Sea current field as m odeled by O verland a n d R oach (1987).
The sim ulation was fo r typical w intertim e conditions of 8 m/s surface winds from the
northeast. The SSM /I 25 km landm ask has been laid over their data.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
S T U W .o rg , S A W S . version 1 3 2
Fig. 6-4: M ap o f Tb(19, H ) from A pril 2 , 1988 w ith the location o f the th ree sites (SL-1 to SL-3).
T im e series from these sites are exam ined an d correlated w ith m eteorological data.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
98
M ar 9
12
22
27
Apr 1
22
27
A pr 1
260
240
220
200
180
1 60
N
X
o
Si
240
220
to
t/3
<u
JZ
euo
200
5
£ 180
b. SL-2
160
240
220
200
180
160
M ar 9
12
17
6
D ate in 1988
02 574
STlAW.org. I /I M . J
Fig. 6-5: T im e series of Tb(19) at three sites.dow nw ind o f St. Law rence Island.
The locations of the three sites are m arked in Fig. 6-4. In each p lo t the triangles are H pol
and the circles are V pol. The x axis show s the local date, rather than GMT.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
99
M ar 9
12
17
22
27
Apr 1
6
11
0 .0 4
0 .0 8 -
0. 12-
a. SL0 .0 4 -
0 .0 8 s
C\
0 .12-
b. SL-2
0 .0 4 -
0 .0 8 -
0 .12-
c. SL-3
M ar 9
12
17
22
27
A pr 1
6
Date in 1988
STlAW.ott. I&IW* MM
Fig. 6-6: T im e series of P R (19) for three sites dow nw ind o f St. Law rence Island.
The locations o f the three sites w hose locations are marked in Fig. 6-4. The x axis show s the
local d ate rather than G M T. The y axis has been reversed, with P R decreasing upw ards, to
aid com parison with the T b record.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
100
Mar 9
260
12
17
22
27
11
A pr 1
240
220
200
a. SLc
j>
13
*
240
3
u
220
1
200
a.
E
0
u
e
180
b. SL-2
160
240-
220-
200 -
180-
c. SL-3
160
M ar 9
12
17
22
27
A pr 1
6
Date in 1988
STU W .tt. I&ltw* i 5)4
Fig. 6-7: Tim e series o f T b(37) at three sites dow nw ind of St. Law rence Island.
T he three sites are marked in Fig. 6-4. In each plot the triangles are FI pol and the circles are
V pol. The x axis represents the local date, rather than G M T .
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
101
M ar 17
M ar 22
M ar 27
A pr 1
A pr 6
210
195-
180 -
165 -
-5 -
■a
B
-1 5 -
-20
-
— A— -1 * total
....p.... y component
-
10-
-20
-
550
^
500 -
o
450 -
400 c/i
350
M a r 17
M a r 22
M ar 27
A pr 1
A pr 6
D a te in 1 9 8 8
Fig. 6-8: Data used in correlation studies o f em ission vs. environm ental conditions.
Shown are T b(19, H) and T b(37, H) at SL-1, gradient w ind (both the total wind speed and
the N-S or v com ponent), air tem perature from the N om e Station (w ith daily average
tem perature also given) and incom ing shortw ave radiation (with th e daily m axim um also
indicated).
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
102
Table 6-1: M ultilayer Fresnel model sim ulations o f diurnal variations in em ission.
C hanges in the bulk brine volum e w ere considered (the no slick case) as w ell as several
different types o f surface brine slick. In each case the salinity w as fixed and brine volum e
was varied as a function o f ice tem perature. O utput was Tb( 19) from w hich PR(19) was
calculated. A lso given are the observation from SL-1 on A pril 2.
(°C)
Vb
{%)
-6°
-10°
-14°
20
14
10
-6°
-10°
-14°
35
25
T su rf
Bulk brine volume
S = 24 ppt
Surface slick
S = 4 2 ppt
“R ough” slick
S = 0 to 65 ppt
V i(z ,op) = 5 %
“R ough” slick
S = 4 0 to 83. ppt
Vj(z,0p) = 5%
O bserved
S L-1, A pril 1, 1988
20
60
Tb
Tb
(19, H ) (19, V)
Tb
Tb
(37, H) (37, V)
187
251
202
195
202
253
252
207
210
161
177
183
237
244
245
185
195
196
256
256
254
PR
(19)
.146
.129
.110
249
251
251
.191
.159
238
246
246
.173
.142
.126
.145
40
30
163
181
231
241
187
241
200
204
-6 °
3-80
164
221
202
234
.148
-10°
-14°
2-50
1-40
179
184
236
238
208
210
243
244
.128
Afternoon
M orning
164
181
222
232
181
191
228
234
.147
.123
-6°
-10°
-14°
185
.137
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
103
Moming
Afternoon
S ib eria
Alaska
Fig. 6-9: M aps o f T b( 19, H) from m om ing and afternoon on successive days.
M om ing data are the average o f all passes from 2 to 5 A M local tim e; afternoon d a ta are
the average o f all passes from 4 to 7 PM local time. M issing data and lan d are indicated in
gray.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
104
Morning
Afternoon
Siberia >
A la s k a
F ig. 6-9 (cont.): M ap s o f Tb(19, H ) from m o m in g and afternoon on successive days.
M om ing d a ta are the average o f all passes from 2 to 5 A M local tim e; afternoon d a ta are
the average o f all passes fro m 4 to 7 P M local time. M issin g data an d land are indicated in
gray.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
105
M orning
Afternoon
S ib e r ia ,
Alaska
lUMUXJar*
«*»IJM
F ig. 6-9 (cont.): M aps o f T b(19, H) from m om ing and afternoon on successive days.
M orning d ata are the average o f all passes from 2 to 5 A M local tim e; afternoon data are
the average o f all passes from 4 to 7 P M local tim e. M issing data and land are indicated in
gray.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
7. D r if t m o d e l c o m p a r iso n
It is clear that em ission increases as the ice ages and thickens in the B ering S ea case study.
W e have been able to determ ine that the signature evolution from thin to FY should take ab o u t 12
days. W e know that environm ental conditions can affect em ission from the ice, b ut our
know ledge is incomplete. The estim ate o f the rate o f signature evolution is from a single short
piece o f th e time record. O ur understanding o f the effect of environm ental conditions is based
largely on qualitative arguments. T o determ ine w hether the relationship betw een T b and age or
thickness is robust requires a direct com parison o f these quantities under a variety of
m eteorological conditions. In this chapter, we d evelop simple m odels of ice drift and growth for
estim ating fields o f ice age and thickness in the B ering Sea. These fields are then com pared w ith
the SSM /I data and the viability o f determ ining the properties o f thin ice from space are assessed.
7.1 Theory
Free drift ice m otion is driven by w inds and currents. If we assum e that inertial influences are
minimal ( Thorndike a n d Colony, 1982), then the balance of forces acting on an ice floe involves
the w ater surface stress (xw), the air surface stress (xa) and the'C oriolis force p er unit area (C ):
xw + x a +C = 0
(7-1)
We adopt the standard convention o f representing each of these quantities as a com plex num ber
such that
x = x x + i - x y = A - e 1'1*’
The surface stresses are generally param eterized as quadratics in term s of the ice, air and w ater
velocities (Vi, va and v w) with
^ = P aC a | va - v i | ( va - v i > i'V
(7-2)
“tw = P w C w| v w —V i|(v w - V i ) ; 1'5
(7-3)
and
where p a = 1.3 kg/m 3 is the air density and pw = 1026 kg/m 3 is th e w ater density. T he air and
water drag coefficients are given by C a and Cw. T h e exponential term , \j/, is the turning angle
between the far field w ind and the actual surface air stress; £ is the angle betw een the current and
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
107
the w ater stress. The C oriolis force is expressed as
C = p i H i f v i • e - i 't/2
(7-4)
where Hi is the ice thickness, p, = 925 kg/m 3 is the ice density, f = 1.458 ■10-4 • sin(tp) is the
Coriolis param eter, and (p is the latitude. By substituting (7-2), (7-3) and (7-4) into (7-1) we
obtain
P a C a | va - V j | ( v a - V i > i v + p w C w | v w - V j | ( v w - V i j e 1'^ -t - pj Hj f ■v f - e -1' ^ 2 = 0
which can be solved for the ice velocity induced by given wind and current velocities.
To solve this equation all that rem ains is the specification of ice thickness, drag coefficients
and turning angles. T he m odel is relatively insensitive to variations in ice thickness o f less than 1
meter (P ea se an d O verland, 1984) and w e assume Hi = 0.5 m. The d rag coefficients and turning
angles depend on the reference height and depth at w hich the wind and current are specified. The
air drag coefficient is typically m easured at a 10 m reference height an d varies with the roughness
o f the ice and with atm ospheric stability. F or air tem peratures below freezing, typical values of
C a range from 0.0015 fo r sm ooth ice to 0.0037 for rough ice in the in n er marginal ice zone
(O verland, 1985). V alues fo r marginal ice zones such as the Bering S ea generally range from
0.0027 to 0.0030. For th ese calculations, w e will use 0.0028.
To obtain the 10 m w ind speed, the gradient wind at 500 mb was calculated from N M C
pressure m aps o f the region and then scaled and turned. A n overview o f the relationship between
gradient w ind and near surface wind height is given in O verland (1985) with additional data
available from O verland (1992) and A lb rig h t { 1980). T h e average relationship (and its standard
deviation) is
vw(10m)
v w(5 0 0 m b )
= 60 ± .0 9 .
Little or no turning o f th e w ind is observed from 10 m dow n to the surface but the average value
o f \j/ betw een 500 mb height and 10 m is around 27°±6° in the Arctic.
Sim ilarly, the water drag coefficient depends on the depth of the specified current. Using
geostrophic currents in the A rctic Basin, M cPhee (1980) calculated th a t C w = 0.0055, while
Pease et al. (1983) m easured Cw = 0.0180 fo r currents at a depth o f 2 m in the B ering Sea.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
108
E stim ation o f C w is further com plicated in our case by the fact that the current Field o f O verland
a n d R oach ( 1987)(Fig. 6-3), used as m odel input, is not for a specific depth but is the mean
current averaged over all depths.
The depth at which the current achieves its average value can be estim ated from the
sim ulations o f O verland et al. (1984) w ho calculated current velocity as a function o f depth for
shallow seas such as the Bering. W ater depth is on the order o f 50 to 90 m for the region south o f
St. L aw rence Island and Siberia. For an 80 m deep ocean covered w ith ice and subject to a 10 m
w ind o f 15 m/s, the average current speed occurs at a depth of approxim ately 30 m; for a
w indspeed o f 5 m/s the average occurs at 23 m. Shallow er seas achieve the average speed at
som ew hat shallow er depths. Hence, the current field o f O verland a n d R oach (1987) should apply
at a depth on the order o f 20 to 25 m. T he standard logarithm ic profile o f current results in a drag
coefficient given by
f
Cw( Z)
k
y
(jn (z /z w ) y
where k = 0.04 is von K arm en’s constant, z is the distance below the ice and zw is the roughness
length. T he m easurem ents o f Pease et al.{ 1983) at 2 m below the w ater line correspond to zw =
0.102 m, Fig. 7-1 show s the drag coefficient as a function o f depth corresponding to this value.
A t a depth o f 23 m, the drag coefficient is 0.0054 - nearly identical to the value derived by
M cPhee (1980) for the central Arctic. W e will, therefore, adopt the values derived by M cPhee
(1980) for both the drag coefficient and the turning angle, nam ely C w = 0.0055 and £ = 23°.
C urrents in the B ering S ea are, at least in part, the result o f the w ind. It has been observed
that currents in both B ering and A nadyr Straits can at tim es actually reverse direction w ithin a
m atter o f 1 to 4 days after th e onset o f particularly strong winds from the north (P ease a n d Salo,
1987; C oachm an a nd A agaard, 1988; Coachm an, 1993). The sim ulated currents o f O verland
a n d Roach (1987) represent conditions sim ilar to average conditions during much o f the
observation period in 1988. However, the wind conditions did vary so the current field could be
som ew hat inaccurate. On the other hand, the current is m ost sensitive w here flow is restricted, in
the straits and near land. O v er m ost o f the Bering Sea south of St. L aw rence Island flow is not
restricted and the sim ulations o f O verland and Roach (1987) suggest that the currents are
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
109
relatively insensitive to changes in w ind strength and direction. H ence, w e expect that the current
field will produce a satisfactory description o f ice m otion south o f St. Law rence Island and
Siberia.
W e can dem onstrate the effectiveness o f the drift m odel using buoy data reported by M uench
et al. (1988). T he drift occurred during February and M arch of 1985 w hen strong w inds blew
from the northeast for an extended period (Fig. 7-2). T h e predicted d rift was for an ice parcel
driven by gradient winds and the O verland and R oach (1987) currents. The agreem ent between
the m odel and observed d rift is surprisingly good. T he model accurately predicts the rapid
m ovem ent southward as w ell as the retrograde motion when the w ind blew from the south in early
M arch.
7.2 Brightness temperature, ice age, and ice thickness
Given th e apparent ability to predict the motion o f an ice parcel in the Bering Sea, how do we
predict regional fields o f ice age and thickness? Fig. 7-3 sum m arizes th e process. F o r every day
from F ebruary 1 to A pril 12, a set o f m odel ice parcels w ere released along the co ast o f St.
Law rence Island and S ib eria and the displacem ent calculated (Fig. 7-3a). The parcels were
allowed to drift until A pril 13 or until they intersected either land o r predefined boundaries. The
boundaries w ere chosen to avoid: (i) areas that include older ice in the ice stream to the east, (ii)
landfast ice to the w est o r (iii) open ocean to the south. F or a given day, the locations o f all ice
parcels w ere segregated into bins corresponding to S S M /I grid pixels and the average age o f all
parcels in each pixel w as then calculated (Fig. 7-3b).
At any point in tim e, the thickness o f ice represented by a parcel w as estim ated u sin g a
modified version of the M a yk u t and U ntersteiner (1971) model for ice growth. T he M a yku t and
U ntersteiner (1971) m odel uses the surface energy balance plus internal heating due to shortw ave
radiation to predict the tem perature profile o f the ice. F rom the tem perature profile, the am ount o f
heat conducted away from the ice/w ater interface is calculated which, in turn, is used to predict
ice growth. D ue to frazil accum ulation, the drifting ice w as assum ed to be 5 cm thick after the
first 12 hours o f drift but all subsequent grow th was assum ed to be du e only to heat conduction
as predicted by the therm odynam ic m odel. M odel input includes surface winds calculated from
the gradient w ind field, air tem perature and snowfall from the N om e station data and cloud cover
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
i 10
estim ated from ISC C P (further details of the model are given in the A pp en d ix: D etails o f the ice
grow th m odel). B y calculating ice growth for each parcel, it is possible to obtain a regional map
o f thickness in the same way as ice age (Fig. 7-3c). Both the ice age and thickness for each pixel
can be com pared with the S S M /I T b values (Fig. 7-3d). Fig. 7-3 is typical o f the results from the
com bined drift and therm odynam ic model (hereafter, sim ply referred to as the drift m odel). In
both the model and in the S S M /I data, the area o f thin ice south o f land can clearly be seen with
age, thickness and Tb(19, H) all increasing downw ind.
Fig. 7-4 show s daily averaged SSM /I T b( 19, H) observations plotted against corresponding
drift model predictions of ice age. Early in M arch, the ice w as snow covered for the m ost part and
T b was on the o rd er of 235K for ice greater than 5 days old. Em ission varied with age only near
land, where the ice was youngest. A s the snow -covered ice w as sw ept aw ay and replaced by bare
ice, a roughly linear correlation between age and emission becom es evident. O ver the course o f a
w eek, the region o f thin ice broadened to larger than the 50 km footprint at 19 GHz, and the Tb o f
th e youngest ice progressively decreased. B y M arch 23, the pattern was w ell established.
The result o f a linear regression is indicated in each plot in Fig. 7-4. T h e data used in the
regression included all data from March 23 to A pril 6 for m odel ice parcels covered by less than
0.5 cm o f snow (the filled circles in Fig. 7-4). Regression coefficients are given in T able 7-1.
B ased on the regression, a FY ice T b(19, H) o f 240K would be reached in about 17 days, nearly
50% longer than the 12 days w e estim ated in the last chapter. A sim ilar linear correlation
between T b and the thickness o f model ice is apparent in Fig. 7-5 with FY ice values reached
w hen the ice is around 34 cm thick.
A com parison w as also conducted at 37 G H z (Table 7-1) using the nonsnow -covered parcels.
A s with the tim e sequences o f the last chapter, T b(37, H) and T b(19, H) are highly correlated,
resulting in behavior much like that of Fig. 7-4 and Fig. 7-5. H ere again, T b is predicted to reach
FY ice values in about 16 or 17 days, when the ice is 33 cm thick.
The relationship between T b and ice age o r thickness is approxim ate at best, with r2 < 0.60 in
all cases. The d ata exhibit a g reat deal o f spread away from the regression lines. Some o f th is is
m ost likely the result o f environm ental conditions, as pointed out in the last chapter. For exam ple,
em ission from th e thinnest ice near land varies by 10-20K, at least in part, in response to the
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
I ll
strength o f the southw ard com ponent o f the wind. Also, the trace o f snow on M arch 25 coincides
w ith the appearance of a spur o f points w ith ages less than 10 days b ut uncharacteristically high
Tb. The snow fall around A pril 10 coincides with a general breakdown in the relationship between
age and em ission. More puzzling, however, is evidence o f bias as a function of tim e - the
estim ated age is consistently below the regression line in late March but shifts to consistently
above it during early April.
At least som e o f the spread in the data appears to be caused by the specified currents in and
around A nadyr Strait. The current is strong enough that m ost sim ulated ice parcels starting in the
S trait were carried northw ard, w hile parcels south o f S iberia were carried eastw ard tow ard, but
not through, the strait (Fig. 7-3a). As a result, the ice south o f A nadyr S trait is estim ated to be
o ld er and thicker than ice ju st to the east and west (Fig. 7-3b and c). W hile there is som e evidence
in the Tb data o f m ovem ent northw ard in the strait, it generally occurred only during periods when
northerly w inds were w eak (e.g. March 25 in Fig. 6-2). It is possible th at the specified current
alo n g the coast o f Siberia and through A nad y r Strait is to o strong.
If we lim it the drift model to only those parcels that originated south o f St. Law rence Island
(Fig. 7-3), we again find that it took tim e fo r the pattern to develop (Fig. 7-6 to Fig. 7-9). Once
established, Tb(19, H) and T b(37, H) are again predicted to increase nearly linearly w ith ice age
and thickness. H ow ever, the predicted increase is more rapid than in than in the previous
sim ulation. FY values are reached when the ice is about 11 or 12 days old, by w hich tim e the ice
is around 29 cm thick. This is the same age vs. T b relationship estim ated from observing the
advance o f the thin ice front through the tim e series sites in the last chapter.
U sing only the St. Law rence Island parcels, the linear relationships betw een age an
thickness
and Tb are im proved, with r2 values increasing by betw een 0.03 and 0 .12 (Table 7-1). T he highest
correlation is for T b(37, H) vs. thickness w ith r2 = 0.72. In all cases, there is an obvious trend of
increasing em ission with age or thickness. However, significant spread o f the data ab o u t the
regression lines rem ains (Fig. 7-10). A m ajor source o f disagreem ent is associated w ith the oldest
ice parcels w here the sensitivity o f em ission to age dim inishes and the d ata trail upw ards in the
age vs. T b plots. Since growth slows as the ice thickens, these data form a tight cluster in the
thickness vs. T b plot and r2 is higher by 0.08 at 19 G Hz and 0.11 at 37 G H z com pared with the
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
112
T b vs. age regression. A nother source o f disagreem ent is ice less than 5 cm thick (Fig. 7-10d).
These d ata are from pixels near land w here contam ination o f the signal by em ission from the land
apparently causes the T b to be consistently higher than that predicted by the regression. These
points also exhibit a great deal of spread due to variability in em ission caused by changes in
environm ental conditions.
The higher correlations at 37 G Hz com pared with 19 G Hz are som ew hat surprising. Em ission
data at 19 and 37 GHz are strongly correlated relative to each other an d a linear regression of
T b(19, H ) vs. Tb(37, H ) h as an r2 of 0.94. However, th e small differences between the tw o are
enough to affect the com parisons of Tb, age, and thickness. The differences are m ost likely due to
at least tw o factors: the sm aller footprint at the higher frequency and, to a lesser ex ten t, lower
sensitivity to changes in environm ental conditions. T h e effect o f both is m ost apparent near land,
for pixels w ith an average thickness o f less than 0.5 cm . The sm aller footprint leads to less
contam ination o f the 37 G H z due to the land. N ear land, Tb(19, H) w as very sensitive to changes
in m eteorological conditions. The net result is that Tb(19, H) values are higher than expected and
vary by nearly 45K (Fig. 7- 10b) while Tb(37, H) values are closer to the regression line and vary
by only ab o u t 30K (Fig. 7-10d). F ootprint size also m akes a difference aw ay from land. In the 19G H z data, there is a cluster o f points w ith T b values from 220 to 240K b ut with relatively low
ages (5 to 8 days) and thicknesses (15 to 22 cm). This cluster is m uch less pronounced in the 37
G H z data. M any of the points in this cluster are from near the east boundary of the drift model, a
region o f old ice advected from Bering Strait. The older ice has a high T b of around 240K . The
larger footprint at 19 G H z m ay include som e o f this o ld er ice and thus cause the high values
com pared w ith 37 GHz.
Perform ance was significantly worse using PR(19) w hich had correlation coefficients as much
as 0.07 less than that o f T b(19, H) and 0.1 5 less than th a t of T b(37, H ) (Table 7-1, F ig. 7 -1 0 ean d
0 . This is true even though PR is m eant to improve perform ance by u sin g a ratio o f brightness
tem peratures to reduce the effect physical tem perature has on em ission. W hy perform ance should
decrease in this case is not totally clear, b u t it appears th at some o f the inform ation content o f the
data is lost by using ratios.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
113
M ultiple regressions were conducted using various com binations o f m icrow ave channels. The
im provem ent over using T b(37, H) by itself w as relatively sm all in all cases. T he best
perform ance w as achieved by using T b(37, H) and T b(37, V) together, yielding an increase in r2of
around 0.02. R egression analyses were also conducted using four channels o f m icrow ave data
plus m eteorological data on winds, air tem perature and dow nw elling radiation but there was little
im provem ent in the correlations.
The actual accuracy o f ice age and thickness calculated from the form ulae o f T able 7-1 is
difficult to assess. The equations them selves can be used to estim ate the possible effect o f the
variability o f the passive m icrow ave signature o f the ice. For exam ple, T|,(H) was observed to
vary by 5K at sites a few pixels removed from land. A 5 K change in T b(37, H) due to a diurnal
oscillation w ould lead to variations o f 1 day in predicted age and 2 cm in thickness. T h e cluster
plots o f Fig. 7-10, how ever, exhibit much larger variability. For a given value o f age o r thickness
calculated from the drift m odel, T b can vary by 40K. F o r a given value o f T b(37, H), ice age can
vary by 8 days and ice thickness can vary by 12 cm. H ow ever, both th e drift model an d SSM /I
data have uncertainty associated with them. T he regressions represent the am ount o f variance in
the drift m odel output that can be explained by the satellite data. B ut th e accuracy o f the drift
model, beyond a single exam ple from 1985, remains uncertain. N onetheless, the results are
encouraging.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
114
30-
0.005
0.010
0.015
T K SW B .uit. 7/1/Vt I <’
Fig. 7-1: W ater drag coefficient as a function o f depth in the w ater colum n.
A ssum es a bottom roughness o f 0 . 102m.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Feb 2
Feb 12
Feb 22
M ar 4
M a r 14
M ar 24
A pr 3
Date in 1985
S ib eria
2 -2 0 -8 5
3 -1 0
3 -2 0
St. L a w ren ce Is
157. M a th e w s Is.
— A — B uoy d a ta (M u e n c h et u l ., 1988)
—O — D rift m odel
A la ska
Fig. 7-2: Com parison o f the drift m odel w ith buoy d ata from M uench a t al.( 1988).
a) the gradient w ind data used to drive the drift m odel, b) observed and model drift tracks.
The triangles and circles are on the same dates fo r observed an d m odeled tracks.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
116
c .T h ic k n e s s (cm )
d . T b(1 9 , H ) ( K °
M A P S * r g .7 /2 /) S ..U .U
Fig. 7-3: M ethod o f producing fields o f ice age and thickness using the drift model.
a) D rift tracks for a set o f ice parcels buoys released on M arch 23, 1988; b) estim ated age
field on A pril 2, 1988; c) estim ated thickness field fo r A pril 2; d) SSM /I observations o f
Tb(19, H ) from April 2. T he outlined area in b, c, an d d is drift ou tp u t from only those
parcels released south o f St. Law rence Island.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
160
ISO
Ml
220
24(1
180
2011
220
240
ISO
200
220
240
180
2(0
220
240
Tb(19,H )(K )
Fig. 7-4: Ice age as a function o f T b(19,H) betw een March 11 and April 10, 1988.
A ge was predicted by the drift model using trajectories o f all ice parcels. T b is from S SM /I.
Filled points are ice with < 0.5 cm snow cover, unfilled have snow covers o f > 0.5 cm . T he
regression line is a least squares fit o f all filled points show n from M arch 23 to A pril 6.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 7-1: Regression equations for ice age and thickness vs. passive m icrow ave param eters.
R egressions were perform ed using all o f the parcels as well as only those parcels that
originated south of St. Law rence Island. Data were lim ited to non-snow covered parcels
from M arch 23 to A pril 6, 1988. T he regression coefficient and fo rm u la are listed.
Ice age
Ice thickness
r2
formula
r2
formula
19H
.53
.2640 * 1 9 H - 46.0
.56
.0041* 19H - 0.65
37H
.57
.3174 * 37H - 60.2
.60
.0051* 37H - 0.895
All parcels
St Lawrence parcels only
19H
.56
0.1756* 1 9 H - 30.40
.64
0.0034 * 19H - .55
37H
.61
0.2134 * 37H - 40.28
.72
0.0043 * 37H - .75
37V
and 37H
.64
.2104 * 37V +
. 1605 * 37H - 80.62
.74
.0035 * 37V +
.0034 * 37H -1.43
PR(19)
.49
-88.523 * PR + 13.66
.57
-1.754 * PR + .3174
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
119
w
0.5
-1-11
0.4
0.3
3'15
> 6
Se°
°o
o° °°°
3-17
„
■«,. j
° 8
° ■ v 0c°
y
\ s °
y
\
0.1
0.0
0.4
0.3
3-19
o 3-2!
°°8
° o’
.*
y ^ \
0 °c£0o 3-23
0
o
°00 3-25
O0
°
**1
0.2
Ice thickness (days)
• \
• •: Q\
0.2
0.1
/•••
•t
().()
3-29
0.4
0.3
•••
8
0.2
■
J
1
• x y y ,*
■ .• U ’ J . ' .
*
*
0.1
0.0
» . « '. .
4-4
CD 4*
0.4
0.3
o° 4-8
•o
y y * *n * *' i»
M
0
\'J £ s n
z « V . •
4-10
0.2
0.1
y
h
0.0
Itt)
180 Zt)
220 240
180 200 220 240
* c ' :
180 21)0 220 240
y/
Jl
180 2CX) 220 240
Tb(19,H )(K )
Fig. 7-5: Ice thickness as a function o f T b(19,H) betw een M arch 11 and A pril 10, 1988.
T hickness was predicted by the drift model using trajectories o f all ice parcels. T b is from
SSM /I. Filled points are ice w ith < 0.5 cm snow cover, unfilled have snow covers o f > 0.5
cm . T he regression line is a least squares fit o f all filled points show n from M arch 23 to
A pril 6.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
120
3-tl
10
3-13
-
3-2!
30 20 -
Ice age (days)
10
3-15
q,
-
3-29
30 20
-
10
-
30 - 4-4
3-31
4-2
4-8
4-10
V.
180
200
220
240
220
240
220
240
200
220
240
Tb(I9 ,H )(K )
Fig. 7-6: Ice age vs. T b(19, H) for ice parcels originating south o f St. Law rence Island.
A ge was predicted by the drift m odel; Tb is from SSM /I. Filled points are ice w ith < 0.5 cm
snow cover, unfilled have snow covers o f > 0.5 cm . The solid regression line is a least
squares fit o f all filled points show n from M arch 23 to April 6. T he dotted line is the
regression obtained previously u sing all ice parcels, as given in Fig. 7-4.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
121
3-11
3-13
3-15
3-17
,Y3
321
325
323
Ice thickness (m )
o°0
V
327
4-2
329
•%
4-8
44
4-10
1— '
180
240
180
220
240
180
220
240
180
200
220
240
Tb(19,H )(K )
Fig. 7-7: Ice thickness vs. T b(19, H) for ice parcels originating south o f St. Law rence Island.
Thickness w as predicted by the drift m odel; T b is from SSM /I. Filled points are ice w ith <
0.5 cm snow cover, unfilled have snow covers of > 0.5 cm . The solid regression line is a
least squares fit o f ail filled points show n from M arch 23 to April 6. T h e dotted line is the
regression obtained previously using all ice parcels, as given in Fig. 7-5.
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
Ice age (days)
122
Fig. 7-8: Ice age vs. T b(37, H) for ice parcels originating south o f St. Law rence Island.
A ge was predicted by the drift model; T b is from SSM /I. Filled points are ice with < 0.5 cm
snow cover, unfilled have snow covers o f > 0.5 cm. The solid regression line is a least
squares fit o f all filled points show n from M arch 23 to A pril 6. The dotted line is the
regression obtained previously using all ice parcels.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
123
0,5
3-13
3-15
3-17
3-19
3-21
3-23
3-25
3-27
3-29
3-31
4-2
0.4 0.3 -
0.2
-
0.5
0.4 -
Ice thickness (m)
0.3 0.2
-
0.4 0.3 0.2
-
* ••
0.5
4-10
0.4 0.3 0.2
-
0.0
160
180
200
220
240
180
200
220
240
180
200
220
240
180
200
220
240
T (3 7 ,H )(K )
b
Fig. 7-9: Ice thickness vs. T b(37, H) fo r ice parcels originating south o f St. Law rence Island.
T hickness was predicted by the d rift model; Tb is from SSM/I. F illed points are ice with <
0.5 cm snow cover, unfilled have snow covers o f > 0.5 cm. T he solid regression line is a
least squares fit o f all filled points show n from M arch 23 to A pril 6. The dotted line is the
regression obtained previously using all ice parcels.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
124
P R (I9 )
Fig. 7-10: Regression analysis o f age and thickness vs. three passive m icrow ave parameters.
A ge and thickness are drift m odel predictions for all dates from M arch 23 to April 6. O nly
parcels originating south o f St. Lawrence Island were used in the model. T h e three
param eters are Tb(19, H), Tb(37, H), and PR(19). The regression equations and correlation
coefficients are given in T able 7-1
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
8. D
is c u s s io n
U ntil recently, it w as expected that passive m icrow ave data could only be used to detect open
water and two form s o f thick sea ice. It was know n that thin sea ice could have a som ew hat
different signature from that o f thick FY ice, but the precise nature o f the signature was not
known. T he presence o f thin ice was, if anything, treated as a source o f noise in ice concentration
algorithm s. However, in the research presented here, we have dem onstrated th at thin ice is
detectable. W e have derived an algorithm that distinguishes thin ice even in the presence o f OW ,
FY and M Y ice. M ore importantly, w e have show n that, in som e circum stances, the ice thickness
distribution can be estim ated directly from the satellite data.
The interpretation o f satellite im agery can be a difficult undertaking. W hile the images
invariably make for pretty pictures, it is often difficult to determ ine the underlying physics that
produced the observations or'how to interpret the data in term s o f geophysically relevant
param eters. In this dissertation, w e have used a variety o f approaches to understand and interpret
em ission from sea ice. Principal com ponent analysis, field observations, theoretical m odeling o f
em ission, correlation studies o f em ission and environm ental conditions, and ice dynam ics
m odeling each bring a different perspective to the problem . B y com bining these perspectives, we
w ere able to form a m ore com plete picture o f em ission from thin ice.
F rom principal com ponent analysis o f satellite data from the Bering Sea, w e found that thin
ice has its own unique signature. F urther exam ination o f the data revealed that H pol em ission
steadily increased dow nw ind o f land, suggesting a link between Tb and ice age or thickness.
Field observations and theoretical sim ulations o f microwave em ission indicated that the
evolution o f the signature o f thin ice is almost entirely a result o f changes in the surface
properties o f the ice. Bare ice with a highly saline surface brine slick can produce the thin ice
spectrum . A s the ice thickens and the surface cools, the brine slick freezes and em ission
increases. H ow ever, even if all the brine freezes, bare ice has a P R no low er than 0.08, only about
halfway along the thin-FY cluster in the SSM/I data. A FY ice signature is only achieved w ith the
addition o f frost flow ers or snow to the surface. T he sim ulations further indicated that the steady
increase in em ission from thin ice to FY ice values w as possible as the result o f either the
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
126
progressive d ry in g out o f a surface slush layer or the slow, steady form ation o f frost flow ers
w hich wick brine o ff the surface.
A com parison o f time sequences o f satellite data with m eteorological d ata indicated th at the
signature near land was, at tim es, sensitive to changes in air tem perature, shortw ave radiation,
snow fall and w indspeed. This signature variability can be explained in term s o f the physical
properties o f the ice. Dium al oscillations o f T b observed near land were likely due to changes in
th e surface brine volum e that w ere caused by changes in air tem perature an d shortw ave radiation.
Snow fall added a low density surface layer to the ice and thus caused a substantial increase in
em ission. The sensitivity to w ind speed appears to be the result o f changes in the rate o f ice
advection and changes in polynya width near land. In all cases, the sensitivity was greatest near
land. A way from land the signature was m ore stable.
Tim e sequences o f Tb allow ed us to m ake an initial estim ate o f the relationship between
em ission and ice age. This estim ate was confirm ed using a theoretical m odel o f ice dynam ics and
therm odynam ics, allow ing us to quantify the relationship betw een em ission and ice thickness.
T here is a strong linear correlation between m odel thickness and both Tb(19, H) and Tb(37, H).
T h e correlation is som ew hat higher at 37 G H z primarily due to the sm aller satellite footprint at
that channel. T he low est correlation was for PR (19), suggesting that the m agnitude o f T b is an
im portant part o f the inform ation content o f th e spectrum.
Field observations, theoretical m odeling and satellite data analysis have m ade it possible to
determ ine the cause o f the thin ice signature evolution and how best to interpret that evolution in
term s o f basic ice properties. T he regression equations for em ission vs. thickness derived using
the ice drift m odel m ake it possible to construct regional m aps o f ice age and thickness for the
B ering Sea from passive m icrow ave satellite d ata (Fig. 8-1). T h e equation using both H and V
pol data at 37 G H z was used. T h e maps o f age and thickness both indicate the now fam iliar
pattern o f very thin ice near land and older thicker ice dow nw ind. The actual range o f age and
thickness detectable using the passive data is prim arily lim ited by the T b at w hich thin ice
becom es indistinguishable from F Y ice. The A V H R R im agery and the S SM /I data suggest that
thick snow covered ice in the G u lf o f A nadyr can have a Tb(37, H) as low as 230K . A s a
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
127
consequence, ice greater than 10 days old or 24 cm thick can be difficult to distinguish from
much older ice (and all such pixels have been shaded dark blue).
T he regional m aps o f ice thickness can be used to predict a host o f geophysically significant
param eters. Surface tem perature, surface heat fluxes, ice growth rates and salt fluxes all depend
on the ice thickness and environm ental conditions and can be calculated using the thickness
estim ates and local m eteorological d ata as input in a sim ple therm odynam ic m odel o f sea ice. Fig.
8-1 c, d, and e are exam ples o f such calculations m ade using the therm odynam ic model of
M aykut (1978). The m eteorological data include an air tem perature o f-1 2 °C m easured at N om e,
shortw ave radiation o f 200 W /m2 estim ated from ISC C P data, a calculated L W radiation flux o f
205 W /m 2, and a surface w indspeed o f 10 m/s derived from the gradient wind. T he albedo w as
assum ed to increase w ith ice thickness following the equations o f W eller et. al. (1972) and the ice
salinity w as fixed at 10 ppt for the calculation o f salt flux.
T he influence of ice thickness on surface tem perature, heat flux, and salt flux can be seen in
Fig. 8-1. A ll are high fo r areas o f ice less than 10 cm thick but d ecrease rapidly as ice thickness
increases. It is apparent from this figure that, by com bining the ice thickness estim ate with
meteorological data and a therm odynam ic model, the passive m icrow ave data take us well beyond
sim ply describing the state o f the ice pack. The d ata provide a tool for studying the impact o f the
ice on the ocean and atm osphere.
The range o f conditions under w hich the thin ice signature and evolution o ccur have yet to be
evaluated. Q uestions rem ain as to how universal and uniform the thin ice spectrum is. A re there
significant regional differences in the spectrum ? O v er what range o f environm ental conditions
does it exist? D o the relationships betw een age or thickness and passive m icrow ave data
established in the springtim e Bering S ea hold in other seasons?
The answ ers to these questions hinge on the physics producing the signature. W e have been
able to construct tw o likely scenarios to explain the transition from the thin to FY ice signature.
In the first scenario, the ice is initially bare and sm ooth but with a surface brine slick. As the ice
thickens, frost flow ers slow ly populate the surface, w ick up brine and eventually produce the FY
ice signature. This scenario appears to be supported by observations m ade during the Bering S ea
E xperim ent in 1973 (R am seier et. al., 1974). They reported that gray ice generally had “various
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
128
degrees o f hoarfrost form ed on the surface; usually the thinner the ice, the sm aller the am ount o f
hoarfrost” . Further, the A V H R R visible imagery taken in early A pril o f 1988 indicates that the
ice had a very low albedo just dow nw ind of land (M assom and C om iso, 1994), as one might
expect from smooth, bare thin ice. T he albedo increased dow nw ind o f land, w hich is consistent
with frost flow ers increasingly populating the surface o f the ice. H ow ever, the slow accum ulation
o f frost flow ers on the surface w ould require that flow er growth o ccu r steadily o v er the course o f
as m uch as 12 days. F rost flow er grow th is only possible if a source o f vapor is present. O bvious
sources o f vapor are the polynya and surface brine on the ice. B ut fro st flow er grow th would
have to continue well dow nw ind o f the polynya and after much o f the surface brine has been
frozen or w icked up onto existing flow ers. W hile this scenario is not inconceivable, it is difficult
to im agine this process happening over a broad region for weeks on end.
T he other scenario is that o f ice covered with a thin, <1 cm, lay er o f slush. O v er tim e, the
slush w ould dry and em ission would increase. O f course, slush im plies snow and little snowfall
occurred (at least at N om e) during the tim e period exam ined here. A irborne ice crystals
originating from the polynya probably w ere deposited o n the ice ju s t dow nw ind o f land but the
am ount is difficult to estim ate. H ow ever, there is evidence that a slushy layer m ay play a role in producing the thin ice signature in the B ering Sea. Photographs taken during aircraft overflights
from three separate years show that thin ice 10 to 15 km dow nw ind o f St. Law rence Island had
moderately high albedos {Pease, 1987). In all three cases, the ice had what looked like either a
snow or slush cover. In at least tw o o f these cases, no snowfall w as reported at N om e just prior
to, or shortly after, the dates o f the photos.
W hether frost flow er growth or the drying of a slushy surface occurred in the B ering Sea is
unclear but the two regim es would result in very different albedos ju s t dow nw ind o f land. The
low albedo o f the thin ice regions presented in M assom and C om iso (1994) suggests that the ice
was bare. H ow ever, visible satellite im agery o f the B ering Sea from M arch 22, 1988 (M artin et.
al., 1992) shows albedos in the thin ice regions that are much higher than those o f M assom and
C om iso (1994). T he higher albedo suggests the presence of snow o r slush. F urther analysis
utilizing visible im agery would help clarify this issue.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
129
Both slow fro st flower grow th and steady freezing of brine in a surface slush layer im ply that
the thin ice signature may occu r primarily in w arm air tem peratures. D uring LEAD EX , frost
flow er growth w as extrem ely rapid at air tem peratures o f around -35°C, producing a FY ice
signature in less than a day. R ecent experim ents growing artificial frost flow ers in a cold room
suggest that frost flow er grow th is substantially slower at tem peratures above -20°C (S. M artin,
pers. com m .). T he rate at w hich brine freezes o ut of a slush lay er depends on how rapidly the
surface cools as it grows. T he surface would cool slowly in w arm air tem peratures, causing the
signature to chan g e slowly. It seem s likely that the existence o f the thin ice signature and the rate
at which the signature changes will depend on the thermal forcing.
The actual surface processes occurring in the Bering Sea are undoubtedly m ore com plicated
than the idealized sim ulations o f frost flow ers o r slush. A dditionally, it is likely that the ice pack
includes a variety o f surface types all mixed together in any given 25 km by 25 km pixel rath er
than the individual surface conditions sim ulated with the m icrow ave model. A reas of frost
flow ers, slush and com binations o f the. two m ay all be present in a given pixel. Natural horizontal •
variations in surface brine volum e occur over short distances and would affect the m icrow ave
signature. Ice rafting has been observed in-the B ering Sea. R afting would effectively increase the •
thickness o f th e ice, affecting the surface tem perature and ultim ately the surface brine volum e.
T h e slow, steady change in em ission over the broad areas in th e Bering S ea suggests that a single
m echanism may be responsible for much of th e thin ice signature evolution b u t subpixel
variability will continue to be a source o f uncertainty in estim ating ice thickness using passive
m icrow ave data alone.
Many o f the questions concerning the physical causes o f the thin ice evolution, the effect o f
environm ental conditions and the degree o f subpixel variability can be addressed through
continued research. Field m easurem ents of em ission and ice physical properties would provide
direct confirm ation o f the physics involved. A cruise through the Bering S ea thin ice zone w ould
clea r up m any o f the questions about the surface properties causing the signature evolution.
M icrow ave m easurem ents should include em ission at 19 and 37 G H z for both H and V pol.
Surface characterization will be crucial with the m ost obvious observation sim ply being w hether
frost flowers o r slush dom inate the surface o f the thin ice. Salinity and surface tem perature d ata
are needed to determ ine the surface brine volum e over a range o f ice thicknesses. Horizontal
R ep ro d u ced with p erm ission o f the copyright ow ner. Further reproduction prohibited w ithout perm ission.
130
surface variability and its effect on em ission should also be observed. Finally, Ice thickness
should be m easured and com pared directly w ith the m icrow ave signature.
M icrowave em ission m odeling has proved invaluable for studying em ission from thin ice but
a few questions rem ain. The C R R E L observations suggest that surface scattering may play a role
in producing the m icrowave signature of thin ice, although the results are inconclusive. It should
be relatively straightforw ard to include surface scattering in the current m odel and assess the
im pact of various degrees o f roughness on the signature. T he model used here treats frost flow ers
as a tenuous snow cover. Field m easurem ents should help to determ ine the validity o f the frost
flow er sim ulations as well as guiding any im provem ents in the theoretical treatm ent o f them.
The SSM /I d ata should be com pared with data from other satellite sensors. Albedos from
visible imagery can be used to clarify w hether the ice just dow nw ind of land is bare. Infrared
im agery can be used to determ ine the surface tem perature o f the ice which can then be com pared
w ith tem peratures derived from the passive m icrow ave data. V isible and infrared imagery m ay be
used to detect the presence o f fresh snow. If the ice is initially bare, snow fall will cause a
detectable increase in visible albedo. Snow fall acts as an insulator and w ould cause a decrease
the surface tem perature in the infrared imagery.
The insulating effect o f the snow also causes a sharp decrease in heat flux from the high
values o f thin ice to much low er values close to those o f thick FY ice. At the sam e time, em ission
from the ice w ould sharply increase from that o f thin ice to som ething close to that o f FY ice.
H ence, em ission m ay be a good indicator o f the therm al regim e o f the ice even with the addition
o f snow. This possibility can be tested by com paring infrared and passive m icrow ave imagery
from before and after a snowfall.
Tucker et. al. (1992) dem onstrated that passive m icrow ave and radar m easurem ents
com plim ent each other and can be used together to discrim inate young ice types. Thin ice regions
such as the B ering S ea are an ideal place to test the relationship between the two. Both the
passive m icrow ave and radar signatures of thin ice depend on the surface properties of the ice.
T h e backscatter from thin ice is very low com pared to that o f FY ice or w ind roughened water. Is
it possible that backscatter increases steadily dow nw ind ju st as em ission does? A re there
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
131
differences in the active and passive m icrow ave response to snowfall that would help to
distinguish snowfall events?
One o f the advantages o f radar is the sm all footprint size (on the o rd er o f tens o f m eters rather
than tens o f kilom eters). R adar might be used to determ ine the extent o f the polynya and thus
quantify the role of polynya width in producing the passive signature n ear land. H eat and salt
fluxes are largest near land. Radar m ay allow us to estim ate the fluxes n ear land w here the
passive m icrow ave data is contam inated.
O f course, any future w ork would need to go beyond the confines o f springtim e in th e Bering
Sea. A prelim inary survey o f SSM /I d ata indicates that the thin ice signature also show s up in
m any o f the other m arginal seas, apparently as part o f polynya building events. In addition, clear
indications o f thin ice can occasionally be seen at the ice edge during freeze up. The range o f
conditions that produce a thin ice signature can be tested in case studies from different regions
and different environm ental conditions. F o r example, the record should b e searched for cases
w hen air tem peratures w ere much low er and little or no dow nw elling shortw ave radiation was
present. R egional differences can be exam ined using the SSM /I data in conjunction w ith local
m eteorological data and d rift model sim ulations. In this way, we can test w hether techniques
developed in the Bering fo r associating regional heat and mass fluxes w ith passive m icrow ave
and m eteorological data apply in general.
Satellite rem ote sensing is a pow erful tool for m onitoring large-scale geophysical processes
involving the E arth’s oceans and atm osphere. Passive m icrow ave sensors have proven capable of
m apping the extent o f thick ice in the A rctic and A ntarctic. W e have show n that the
therm odynam ically im portant thinner ice can also be m apped. The potential exists to use the
techniques developed here over a variety o f regions and seasons. The potential exists to expand
our know ledge and understanding o f the E arth ’s clim ate system and oceans.
R ep ro d u ced with p erm ission of th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
e. Salt F lux'
4 0 W/m
I’R w ic r .o r g . w n m . x s 2
Fig. 8-1: G eophysical param eters predicted from passive m icrowave satellite data.
F ields are a) ice age, b) ice thickness, c) surface tem perature, d) sensible heat flux, and salt
flux. T he age and thickness w ere calculated directly from SSM /I d a ta using Tb(37, V ) and
Tb(37, H) from A pril 2, 1988. T h e surface tem perature, heat flux, an d salt flux w ere
calculated using a sim ple therm odynam ic model w ith the predicted ice thickness an d
m eteorological d ata from N om e as input.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
9. R e f e r e n c e s
Arcone, S.A., A.J. G ow and S. M cG rew , Structure and dielectric properties at 4.8 and 9.5
G H z o f saline ice, J. Geophys. Res., 91(C12), 14281-14303, 1986.
Bohren, C.F., and D .R . Huffman, A bsorption and Scattering o f Light by Sm all Particles.
W iley and Sons, N ew York, 1983.
Campbell, W.J., R.O . Ramseier, R.J. W eaver, and W.F. W eeks, Skvlab Floating Ice
E xperim ent - Final Report. U SG S R eport E76-10161, N A SA -C -47446, 67 pp., 1975.
C avalieri, D.J., P. G loerson and W .J. Campbell, Determ ination o f sea ice param eters with the
N IM B U S 7 SM M R , J. G eophys. Res., 89, 5355-5369, 1984.
Cavalieri, D.J., P. G loerson and T .L . W ilheit, A ircraft and satellite passive m icrow ave
observations o f the Bering S ea cover during M IZ EX West, IE E E Trans. G eosci. Rem.
Sens., G E-24(3), 368-377, 1986.
Cavalieri, D.J., A passive m icrow ave technique fo r mapping new, young and first-year sea ice
types in seasonal sea ice zones, J. G eophys. Res., 99, 12561-12572, 1994.
Coachm an, L.K., and K. Aagaard, T ransports through Bering Strait: A nnual and interannual
variability, / G eophys. Res., 93, 15,535-15,539, 1988.
Coachm an, L.K., O n the flow field in the Chirikov Basin, Cont. S h e lf Res., 13(5/6), 481-508,
1993.
Colbeck, S.C., L iquid distribution and the dielectric constant o f w et snow, N ASA Conf. Pub.
2153, A. R ango (ed.), 21-38, 1980.
Colbeck, S.C., An overview o f seasonal snow metamorphism, Rev. G eophys. Spa ce Sci.,
20(1), 45-61, 1982.
Collin, R. E„ Field T heory o f G uided W aves. IE E E press, New York, 1991.
Com iso, J.C., S.F. A ckley and A .L. Gordon, A ntarctic sea ice m icrow ave signatures and their
correlation with in-situ observations, J. Geophys. Res., 89, 662-672, 1984.
Comiso, J.C., Characteristics o f arctic w inter sea ice from satellite m ultispectral m icrow ave
observations, J. G eophys. Res., 91, 975-994, 1986.
Com iso, J.C., and H .J. Zwally, P olar M icrow ave Brightness T em perature From Nim bus-7
S M M R . N A SA R ef. Publ. 1223, 82 ppg., 1989.
Cox, G .F.N ., and W .F. W eeks, E quations for D eterm ining the G as and Brine V olum es in Sea
Ice Samples. C R R E L Rept. 82-30, llp p , 1982.
Cox, G .F.N ., and W .F. W eeks, N um erical sim ulations o f the profile properties o f undeform ed
first-year sea ice during the grow th season, J. G eophys. Res., 93, 12449-12460, 1988.
Crocker, G.B, The Physical Properties o f S now cover on Sea Ice in the Central H igh Arctic.
M .Sc. Thesis, M cG ill University, 1984.
Debye, P.. Polar M olecules. Dover, N ew York, 1929.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
134
Drinkwater, M .R., and G .B . Crocker, M odeling changes in the dielectric and scattering
properties o f young snow -covered sea ice at GHz frequencies, J. G lac., 34(118), 274-282,
1988.
Drinkwater, M .R., R. Kwok, D.P. W inebrenner and E. Rignot, M ultifrequency polarim etric
synthetic aperature radar observations o f sea ice, J. G eophys. Res., 96„ 20679-20698,
1991.
Ebert, E.E. and J.A. Curry, A n interm ediate one-dim ensional therm odynam ic sea ice m odel
for investigating ice-atm osphere interactions, J. G eophys. Res., 98, 10,085-10,109, 1993..
Fily, M. and J.P. Benoist, Large-scale statistical study o f the Scanning M ultichannel
M icrow ave R adiom eter (SM M R) d ata over A ntarctica, J. Glac., 37, 129-138, 1991.
Gloerson, P., and D.J. Cavalieri, Reduction o f weather effects in the calculation of sea ice
concentration from m icrow ave radiances, J. Geophys. Res., 91, 3913-3919, 1986.
G re n fe ll, T .C ., T h e e ffe c ts o f ice th ic k n e s s o n th e e x c h a n g e o f s o la r ra d ia tio n o v e r the
p o la r o c e a n s , J. G la c io l., 2 2 , 3 0 5 -3 2 0 , 1979.
Grenfell, T .C ., A theoretical model o f the optical properties o f sea ice in the visible and near
infrared, J. Geophys. R es., 88, 9723-9735, 1983.
Grenfell, T .C ., Surface based brightness tem peratures o f sea ice in the B ering and G reenland
Seas, IE E E Trans. G eosci. Rem. S e n s ., GE-24, 378-382, 1986.
Grenfell, T .C ., Surface-based passive m icrow ave studies o f m ultiyear se a ice., J. G eophys.
Res., 97, 3485-3501, 1992.
Grenfell, T .C . and D.K. P erovich, Spectral albedos o f sea ice and incident solar irradiance in
the S outhern Beaufort S ea, J. Geophys. Res., 89, 3573-3580,1984.
Grenfell, T .C ., and A.W . Lohanick, T em poral variations o f the m icrow ave signatures o f sea
ice during the late S pring and early S u m m er near M ould Bay NW T, J. Geophys. R es., 90,
5063-5074, 1985.
Grenfell, T .C ., and J.C. C om iso, M ultifrequency passive microwave observations of first-year
ice grow n in a tank, IE E E Trans. G eosci. Rem. Sens., GE-24, 826-831, 1986.
Grenfell, T .C ., D.L. Bell, A .W . Lohanick, C.T. Swift and K. St. G erm ain, M ultifrequency
passive m icrow ave observation o f saline ice grown in a tank, Proc. IG A R R S 1988 Sym p.,
ESA SP-284, 1687-1690, 1988.
Grenfell, T .C ., D.J. C avalieri, J.C. Com iso, M.R. D rinkw ater, R.G. O nstott, I. Rubinstein, K.
Steffen and D.P. W inebrenner, Considerations for m icrow ave rem ote sensing o f thin sea
ice, in R em ote Sensing o f Sea Ice. F. C arsey (ed.), AGU, Geophys. M ono. 68, 291-301,
1992.
Hallikainen, M . T., F. T. U laby and M. A bdelrazik, D ielectric properties o f snow in the 3 to
37 G H z range, IE E E Trans. Ant. Prop., A P -3 4 (ll), 1329-1339, 1986.
Hallikainen, M . T., and D. P. W inebrenner, The physical basis for sea ice remote sensing, in
M icrow ave Remote S ensing o f Sea Ice. F. Carsey (ed.), Amer. G eophys. Union, 29-46,
1992.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
135
H ollinger, J. P., J. L. P eirce and G. A. P oe, SSM /I instrum ent evaluation, IE E E Trans.
Geosci. Rem. Sens., 28, 781-790, 1990.
Johnson, P.E., M.O. Sm ith, and J.B. A dam s, Q uantitative analysis o f planetary reflectance
spectra with principal com ponent analysis, Proc. L u n a r Sci. Conf. 15: J. Geophys. Res.
S u p p .,9 0 ,C 805-C 810, 1985.
K rzanow ski, W.J., Principles of M ultivariate Analysis: A U ser's P erspective. New York,
O xford U niversity P ress, 1988.
Linlor, W „ Perm ittivity o f w et snow betw een 4 and 12 G H z., J. Appl. P hys., 2811-2816,
1980.
Lohanick, A.W ., M icrow ave brightness tem perature o f laboratory-grow n undeform ed firstyear ice with an evolving snow cover, J. Geophys. Res., 98, 4667-4674, 1993.
Massom, R.A., and J.C. Com iso, The classification o f arctic sea ice types and the
determ ination o f surface tem perature using advanced very high resolution radiom eter data,
J. G eophys. Res., 99, 5201-5218, 1994.
Maykut, G.A., Energy exchange over young sea ice in the Central Arctic, J. Geophys. Res.,
83, 3646-3658, 1978.
Maykut, G.A., Large-scale heat exchange and ice production in the C entral Arctic, J.
G eophys. Res., 87, 7971-7984, 1982.
Maykut, G .A „ N. Untersteiner, Some results from a time dependent therm odynam ic m odel o f
sea ice, 7. Geophys. Res., 76, 1550-1575, 1971.
Maykut, G. A., and P.E. Church, R adiation climate of Barrow , Alaska, 1962-1966, J. A p p l.
M eteorol., 12, 620-628, 1973.
M aykut, G .A ., and M .G. M cPhee, S olar heating of the A rctic mixed layer, J. Geophys. Res.,
in press, 1995.
Maykut, G .A ., and D.K. Perovich, The role o f shortwave radiation in the sum m er decay o f a
sea ice cover, J. G eophys. Res., 92, 7032-7044, 1987.
McPhee, M .G ., An analysis o f pack ice drift in summer, in Sea Ice P rocesses and M odels.
R.S. P ritchard (ed.), U niv. of W ashington Press, Seattle, 1980.
Muench, R .D ., J.D. S chum acher and S.A . Salo, W inter currents and hydrographic conditions
on the N orthern Central Bering Sea shelf, J. Geophys. Res., 93, 516-526, 1988.
Overland, J.E., H.O. M ofjeld and C.H. Pease, W ind driven ice drift in a shallow sea, J.
G eophys. Res., 89, 6525-6531, 1984.
Overland, J.E., A tm ospheric boundary layer structure and drag coefficients over sea ice, J.
G eophys. Res., 90, 9029-9049, 1985.
Overland, J.E., and A.T. Roach, N orthw ard flow in the B ering and C hukchi Seas, J. G eophys.
Res., 92, 7097-7105, 1987.
Overland, J.E ., and K.L. Davidson, G eostrophic drag coefficients over sea ice, Tellus, 44a,
54-66, 1992.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
136
Pease, C.H., S.A . Salo and J.E. Overland, D rag m easurem ents for first-year sea ice over a
shallow sea, J. Geophys. Res., 88, 2853-2862, 1983.
Pease, C.H., and J.E. Overland, An atm ospherically driven sea-ice drift m odel for the B ering
Sea, Ann. G lac., 5, 111-114, 1984.
Pease, C.H., T he size o f w ind-driven coastal polynyas, J. G eophys. Res., 92, 7049-7059,
1987.
Pease, C.H., and S.A. Salo, S ea ice drift near B ering Strait d u rin g 1982, J. G eophys. Res.,
92, 7107-7126, 1987.
Perovich, D.K. and J. R ichter-M enge, Surface characteristics o f lead ice, J. G eophys. Res.,
99, 16,341-16,350, 1994.
R a m se ie r, R .O ., P. G lo e rso n , W .J. C a m p b e ll, an d T .C . C h a n g , M e s o s c a le d e s c rip tio n
fo r the p rin c ip a l B e rin g S e a ice e x p e rim e n t, in R e su lts o f th e U .S . C o n trib u tio n to
th e J o in t U .S ./U .S .S .R . B e rin g S e a E x p e rim e n t. N A S A T M X -7 0 6 4 8 , 3 1 -7 5 ,
1974.
Rothrock, D.A., D.R. T hom as and A.S. Thorndike, Principal com ponent analysis o f satellite
passive m icrow ave data o ver sea ice, J. G eophys. Res., 93, 2321-2332, 1988.
Schweiger, A.J., and J.R. Key, A rctic cloudiness: com parison o f ISC C P-C 2 and N im bus-7
satellite-derived cloud products with a surface-based cloud clim atology, J. Clim ate, 5( 12),
1514-1527, 1992.
Shine, K.P., Param eterization o f the shortw ave flux over high albedo surfaces as a function ofcloud thickness and surface albedo, Quart. J. R. Met. Soc., 1 10, 747-764, 1984.
Sihvola, A.H., and J. A. Kong, Effective perm ittivity o f dielectric mixtures, IE E E Trans. Geo.
Rem. Sens., 26, 420-429, 1988.
Sm ith, M .O., P.E. Johnson and J.B. A dam s, Q uantitative determ ination o f m ineral types and
abundances from reflectance spectra using principal com ponent analysis, Proc. L u n a r Sci.
Conf. 15: J. G eophys. Res. Supp., 90, C 797-C 804, 1985.
Smith, S.D., R.D. M uench and C.H . Pease, Polynyas and leads: An overview o f physical
processes and environm ent, J. Geophys. Res., 95, 9461-9479, 1990.
Steffen, K„ and J.A . M aslanik, Com parison o f NIM BUS 7 scanning m ultichannel m icrow ave
radiom eter radiance and derived sea ice concentrations with L A N D SA T imagery o f the
North W ater area o f Baffin Bay, J. G eophys. Res., 93, 10769-10781, 1988.
Steffen, K. and A. Schw eiger, N A SA team algorithm for sea ice concentration retrieval from
Defense M eteorological Satellite Program Special Sensor M icrow ave Im ager: Com parison
with L andsat satellite im agery, J. Geophys. Res., 96, 21,971-21,987, 1991.
Stogryn, A., S trong fluctuation theory for m oist granular media., IE E E Trans. Geo. Rem.
Sens., G E-23, 78-83, 1985.
Stogryn, A., and G .J. Desargent, T he dielectric properties o f brine in sea ice at m icrow ave
frequencies, IE E E Trans. A nt. Prop., A P-33, 523-532, 1985.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
137
Stogryn, A., An analysis o f the tensor dielectric constant o f sea ice at m icrow ave frequencies,
IE E E Trans. Geo. R em . Sens., GE-25, 147-158, 1987.
Svendsen, E„ C. M atzler and T.C. G renfell, A model for retrieving total sea ice concentration
from a spacebom e dual polarized passive m icrowave instrum ent operating near 90 G H z,
Int. J. Rem . Sens., 8, 1479-1487, 1987.
Swift, C.T., L.S. Fedor and R.O. Ram seier, An algorithm to measure sea ice concentration
with m icrow ave radiom eters, J. G eophys. Res., 90, 1087-1099, 1985.
Swift, C .T ., D.C. D ehority, A.B. T anner and R.E. M cIntosh, Passive m icrow ave spectral
em ission from saline ice at C-band during the growth phase, IE E E Trans. Geosci. Rem .
Sens., G E-24, 840-848, 1986.
Thomas, D .R ., and D.A. Rothrock, B lending sequential scanning m ultichannel m icrow ave
radiom eter and buoy data into a sea ice model, J. G eophys. Res., 94, 10,907-10,920, 1989.
Thorndike, A .S., D.A. R othrock, G.A. M aykut and R. C olony, The thickness distribution of
sea ice, J. Geophys. R es., 8 0 ,4501-4513, 1975.
Thorndike, A.S., and R. Colony, Sea ice m otion in response to geostrophic winds, J. G eophys.
Res., 87, 5845-5852, 1982.
Tiuri, M .E ., A.H. Sihvola, E.G . Nyfors and M. T. H allikainen, The com plex dielectric
constant o f snow at m icrow ave frequencies, IE E E J . O ceanic Eng., O E-9, 377-382, 1984.
Tucker, W .B ., T.C. G renfell, R.G. O nstott, D.K. Perovich, A.J. Gow, R .A . Shuchm an and
L.L. Sutherland, M icrow ave and physical properties o f sea ice in the w inter m arginal ice
zone, J. Geophys. Res., 96, 4573-4587, 1991.
Vant, M .R ., R.O. R am seier and V. M akios, The com plex-dielectric co n stan t of sea ice at
frequencies in the range 0.1-40 GHz, J. Appl. Phys., 49, 1264-1257, 1978.
Walters, J.M ., C. R u f and C .T. Swift, A m icrowave radiom eter w eather-correcting sea ice
algorithm, J. G eophys. Res., 92, 6521-6534, 1987.
W eller, G., R adiation flux investigation, A ID JE X Bull., 14, 28-30, 1972.
W ensnahan, M .R., T.C. G renfell, D.P. W inebrenner and G .A . Maykut, O bservations and
theoretical studies o f m icrow ave em ission from thin saline ice, J. G eophys. Res., 98, 85318545, 1993a.
W ensnahan, M .R., G.A. M aykut, T.C. G renfell and D.P. W inebrenner, Passive m icrow ave
rem ote sensing o f thin sea ice using principal com ponent analysis, J. G eophys. Res., 98,
12,453-12,468, 1993b.
Worby, A .P., and I. Allison, O cean-atm osphere energy exchange over thin Antarctic pack-ice
o f variable concentration, Ann. Glac., 15, 1992.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
10. A p p e n d ix
10.1 Methods of Principal Component Analysis
PCA is a statistical technique for linearly transform ing m ultidim ensional data into a new
orthonorm ai coordinate system (K rzanow ski, 1988). The first axis defined by this transform ation
is directed along the line o f m axim um variance in the original coordinate system . Each successive
axis (or vector) points in the direction w hich encom passes the largest am ount o f variance not
accounted fo r by the previous vectors subject to the constraint that the axes form an orthonorm al
basis set. PC A is com m only used to determ ine m ajor forms o f variance in d ata ( R othrock et al.,
1988; Fily a n d Benoist, 1991) and can also be used as the basis for m ixture algorithm s (Jo hnson
e ta l., 1985; S m ith et al., 1985).
The original data vector is com posed o f the N channels o f brightness tem perature data such
that
T =
( 10-1)
W here the subscripts 1 to N designate the different frequencies and polarizations. T he linear
transform ation to the principal com ponent coordinate system can be expressed as
p
=
vtt
( 10-2)
where V is a transform ation m atrix whose colum ns are unit length vectors w hose coefficients are
the direction cosines for the axes in principal com ponent space. T he vector P is the data vector
transform ed into principal com ponent space. T h e individual elem ents o f P are often referred to as
principal com ponent scores. F o r com m ensurate data, V is calculated from the characteristic
eigenvalue equation for the covariance m atrix . T he eigenvalues are the variance spanned by each
o f the colum ns, o r eigenvectors, o f V.
PCA o f m icrow ave data can be done using either unsupervised or su p e n d sed methods.
U nsupervised analysis uses all pertinent raw d ata in the determ ination o f the eigenvectors (e.g. all
data from the A rctic on a particular day) w hile the supervised m ethod uses only spectra for
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
139
chosen endm em bers (e.g. average spectra from OW , FY and M Y ice). T he unsupervised method
has the advantage o f describing all the variance in the data; not only are the differences between
the spectra o f potential endm em bers taken into account but also the variance in the signature of
the endm em bers. T he m ajor disadvantage o f this type o f analysis is that it favors endm em bers
with a great many m easurem ents and may not adequately represent endm em bers for w hich there
are only a few observations.
The advantage o f supervised PCA is that weighted sam ples of endm em ber spectra can be used
so that the eigenvectors will not be skew ed toward the more prevalent scene types. A t the same
tim e, the variance o f the endm em ber spectra can be accounted for by including an equal num ber
o f spectra fo r each m em ber such that the spectra cover the range and distribution observed in
nature. The disadvantage o f this approach is that if the endm em bers are not well chosen, or if not
all the endm em bers that are present in the d ata are accounted for in the analysis, then the
eigenvectors may not include important inform ation about the data. A com m on technique for
choosing endm em bers is to look for obvious groups and outliers in clu ster plots o f the data.
O nce the endm em bers have been selected, a PCA is carried out and the result is N
eigenvectors. In m ost cases only the first few eigenvectors will Span variance significantly above
the noise level o f the data. T hese eigenvectors represent the independent dim ensions defined by
the data. F or m eigenvectors which are above the noise level we can solve for relative
concentrations of m+1 endm em bers if w e assum e
£ c : = 1
(10-3)
where i stands for the ith endm em ber and C stands for the concentration fraction of an
endm em ber. Since PCA is a linear transform ation from the original data space, it then follow s
from (10-2) that
(10-4)
where j is the jth com ponent o f P. U sing the first m com ponents o f P w e obtain m equations
which can be inverted to solve for m+1 concentrations.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
140
R othrock et a l., (1988) used unsupervised PCA to study the inform ation content o f the
satellite passive m icrow ave data from SM M R for northern polar oceans during the m onths of
A pril, A ugust and D ecem ber, 1979. T h ey determ ined that in all three m onths the first 2
eigenvectors contained 99% o f the variance in the data. The first vector described the magnitude
o f the brightness tem peratures in the spectra while the second vector responded to differences in
the spectral gradient. T h e third eigenvector explained less than 21 K 2 o f the variance for
D ecem ber, and even less during the oth er months. T his third com ponent was dom inated by a
heavy w eighting o f the 22 GHz, H Pol channel w hich is located at the center o f a w ater
absorption band in th e m icrow ave and so responds to atm ospheric liquid water and w ater vapor,
i.e. w eather effects. T he authors concluded that only tw o significant eigenvectors w ere associated
with surface scene types and hence only the three endm em bers OW , F Y and M Y ice could be
determ ined from the data.
T his result is likely due to the lim ited areal coverage o f thin ice in the Arctic. It has been
estim ated that thin ice represents no m ore than a few percent o f the total w inter ice co v er
{Thorndike et al., 1975). A dditionally, thin ice exists primarily as leads a few hundred meters
wide - m uch sm aller than a 25 km sa tellite footprint. A n unsupervised PC A o f 25 km data could easily overlook w hatever thin ice w as present. •
10.2 Details of the ice growth model
T he ice growth m odel used in conjunction with the drift model is based on the therm odynam ic
model o f M aykut a n d U ntersteiner (1971). It predicts ice growth as a result o f heat conduction
due to both the surface energy balance as well as the internal heating from shortw ave radiation.
The m odel has been m odified from the original in a num ber of ways. T he largest ch an g e has been
the inclusion of the C o x a n d Weeks (1988) em pirical equations for predicting brine entrapm ent
and distribution w ithin the ice. As a result, the model predicts realistic profiles o f salinity, thermal
conductivity and heat capacity. The equations are im plem ented at the end of each tim e step when
the salinity o f the new ly added ice is calculated and averaged into the salinity o f the bottom layer
o f the ice. At the sam e tim e, equations fo r brine drainage and expulsion are applied to all o f the
internal layers. Since no adequate theory exists for internal transport, w e follow the original Cox
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
141
an d W eeks (1988) m odel and brine lost internally is sim ply removed from the ice rather than
transported from layer to layer.
T he other major m odification o f the M aykut a n d U ntersteiner (1971) model is in the treatm ent
o f internal absorption o f shortw ave radiation. A bsorption is calculated from B eer’s Law but,
rather than a single bulk formula, the shortw ave spectrum is broken up into w avelength bands as
sum m arized in T able 10-1. The total dow nw elling radiation is calculated from the equations of
Shine (1984). U nder clea r sky conditions the incident shortwave radiation flux (Fcir) is
c„
c o s 2 0 ,s_____________
_ _c __________________
clr
° 1 .2 -c o s0 s + e a 10“3(1 + cos 0 S) + 0.046
. - i I
w here S0 = 1353 W /m 2 is the solar constant, 0S is the solar zenith angle and ea is the atm ospheric
vapor pressure in m illibars (assum ed to be ~7 m illibars). The incident flux under cloudy
conditions (FC|dy) is
_ 1 - 0 .9 9 6 a x
cldy “
1-ax
Vc o s 0 s ( s l + s 2 c o s 0 s)
1+ 0 .1 3 9 (1 -0 .9 3 5 -a x K
where a* is the albedo o f the ice for a given w avelength band and t c is the cloud optical thickness
(assum ed to be 7.5). T h e param eter Si = 53.5 W /m 2 an d S2 = 1274.5 W /m 2. and C loud cover was
estim ated from ISC C P (Schw eiger a n d Key, 1992) and the total dow nw elling radiation was then
calculated as a w eighted sum of these tw o equations. T he incident radiation is broken up into
spectral bands using the weights in T able 10-1. The w eights were derived from typical spectra
given in G renfell a n d P erovich (1984) for clear and cloudy conditions in the Arctic. T he albedos
and extinction coefficients were obtained from G renfell (1979). A lbedo varies with ice thickness
as observed by W eller (1972). Based on a regression o f those observations, the ice albedos were
scaled as a function o f thickness using
a x (z) = a x H
• (l-,2 7 - e _z/037 -.4 2 • e ’ ^ 361)
where ax(°°) is the m axim um value as given in T able 10-1. It was further assum ed th a t once any
am ount o f snow fell on the ice the albedos would im m ediately jum p to those o f snow .
A few other, m inor modifications to the model w ere made. D ow nw elling longw ave radiation
LWjn w as com puted using the equations o f M aykut a n d Church (1973) where
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout p erm ission .
142
LW in = 0.7855^1 + 0.2232 • C 2’75) • a • Ta4
w here C is the cloud fraction, a is the Stefan-Boltzm ann constant (5.67* 10s) and Ta is the air
tem perature at the N om e m eteorological station. Sensible heat transfer (SH) w as calculated using
the equations o f E bert a n d C urry (1993). T hey em ploy the usual equation
SH = p a ca C Sh va (Ta -T s)
w here ca = 1006 J/kgK is the specific heat o f the ice, va is the wind speed and T s is the ice surface
tem perature. T he sensible heat transfer coefficient C Sh is defined by E bert a n d Curry (1993) in
term s o f the R ichardson num ber (Rj):
r
. _
' §(T a ~ T s )
Ta v 2 (10m)
w here Az = 10 m is the wind reference height, va(10m) is the 10 m w indspeed, and g = 9.8 m /sec2
is the acceleration due to gravity. If the Richardson num ber is negative (i.e. th e air is co ld er than
the surface) then
C SH = 1.3-10~3 1~ 2 0 _ g_
S
l + 5 lV -R
If the Richardson num ber is positive (i.e. the air is w arm er than the surface) then
1.3 ■10-3
C cu
SH “—7
FT
(l H- 20 • R )
O ther model input param eters included relative humidity w hich w as assum ed to be .9 and th e
oceanic heat flux w hich was fixed at 2 W /m 2. T he tim e step used was 1800 seconds and the layer
thickness was 2 cm .
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Table 10-1 Shortw ave radiation param eters used in the therm odynam ic model.
D ow nw elling shortw ave radiation (S W in) weights, albedo and extinction coefficient as a
function o f w avelength used in the therm odynam ic model.
Band
(nm)
40 0 -6 0 0
60 0 -7 5 0
750-925
925-1125
> 1125
sw inweights
Albedo
clear
cloudy
ice
snow
.29
.23
.18
.13
.17
.34
.28
.21
.13
.05
.78
.70
.55
.35
.05
.92
.92
.87
.80
.15
Extinction coeff.
(1/m)
ice
snow
2
4
10
80
1000
11
30
90
110
1000
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
CURRICULUM VITAE
Mark R. Wensahan
A ddress
D ept, o f A tm ospheric Sciences, A K -40
U niversity o f W ashington
Seattle, W ashington 98195
(206) 543-0628
thinice@ atm os.w ashington.edu
E ducation
1988, B.S., U niversity o f W ashington, Seattle, W A
1991, M .S., A tm ospheric Sciences, U niversity o f W ashington, Seattle, W A
1995, Ph.D ., A tm ospheric Sciences, U niversity o f W ashington, Seattle, W A
Selected P ublications
W ensnahan, M. R., T . C. G renfell, D. P. W inebrenner and G. A. M aykut, 1993: O bservations
and theoretical studies o f m icrowave em ission from thin saline ice, J. G eophys. R es., 98, 8531 8545.
W ensnahan, M. R., G . A. M aykut, T. C. G renfell and D. P. W inebrenner, 1993: Passive
m icrow ave rem ote sensing o f thin sea ice using principal com ponent analysis, J. G eophys. Res.,
98, 12,453-12,468.
G renfell, T. C., D. P. W inebrenner, M. R. W ensnahan, Passive m icrow ave signatures of
sim ulated pancake ice and young pressure ridges, IG A R SS '92, H ouston, Texas, M ay 26-29,
1992, pg. 1253-1255.
W inebrenner, D .P., T.C . G renfell, and M .R .W ensnahan, 1993:M odeling the T em poral
E volution o f L-Band Polarim etric S A R O bservations o f G row ing S ea Ice in A rctic Leads,
A bstract in PIERS 1993 Proceedings, 478.
G renfell, T. C., J. C. Com iso, M. A . Lange, H. E icken, M. R. W ensnahan, P assive
m icrow ave observations o f the W eddell S ea during austral w inter and early spring, J. G eophys.
Res., 99, 9,995-10,010, 1994.
G renfell, T.C ., M .R . W ensnahan, and D.P. W inebrenner, Passive m icrow ave signatures o f
sim ulated pancake ice and young pressure ridges, R em o te Sensing R ev., V ol. 9, pp. 51-64, 1994.
G renfell, T.C ., M .R . W ensnahan, and D.P. W inebrenner, 1994: M easurem ents o f M icrow ave
Em ission from N ew and Y oung S aline Ice during the 1993 C R R EL P ond Experim ent, IGARSS
'94 P roceedings, 605-607.
G renfell, T.C ., M .R . W ensnahan, D .P. W inebrenner and L. Z urk, 1995:Passive M icrow ave
and Infrared O bservations o f N ew and Y oung Sea Ice - The 1994 C R R E L Pond E xperim ent,
A bstract in PIERS 1995 Proceedings, 66.
R ep ro d u ced with p erm ission o f th e copyright ow ner. Further reproduction prohibited w ithout perm ission.
Документ
Категория
Без категории
Просмотров
0
Размер файла
6 032 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа