вход по аккаунту



код для вставкиСкачать
Atmos. Sci. Let. 10: 164–169 (2009)
Published online 2 June 2009 in Wiley InterScience
( DOI: 10.1002/asl.226
Comparison of ozone fluxes over grassland by gradient
and eddy covariance technique
Jennifer B. A. Muller,a * Mhairi Coyle,b David Fowler,a Martin W. Gallagher,a Eiko G. Nemitz,a and Carl J. Percivala
a School of Earth, Atmospheric and Environmental Sciences, The University of Manchester,
b Centre for Ecology & Hydrology, Bush Estate, Penicuick, Midlothian, EH26 0QB, UK
*Correspondence to:
Jennifer B. A. Muller, School of
Earth, Atmospheric and
Environmental Sciences, The
University of Manchester, Simon
Building, Brunswick Street,
Manchester, M13 9PL, UK.
E-mail: jennifer.muller@
Received: 23 January 2009
Revised: 2 April 2009
Accepted: 21 April 2009
Simon Building, Brunswick Street, Manchester, UK
Ozone flux measurements over vegetation are important to estimate surface losses and
ozone uptake by plants. The gradient and eddy covariance flux technique were used for
measurements over grassland at a flux-monitoring site in southern Scotland during August
2007. The comparison of the two methods shows that the aerodynamic flux-gradient method
provides very similar long-term average fluxes of ozone as the eddy covariance method. The
eddy covariance technique is better at capturing diurnal cycles and short-term changes, but
the comparison of two fast analysers illustrate that there can be considerable measurement
uncertainty. Copyright  2009 Royal Meteorological Society
ozone; eddy covariance; aerodynamic gradient; deposition velocity; grassland
1. Introduction
The atmospheric ozone budget is crucially important
in determining the Earth’s atmospheric composition,
radiative balance and future climatic changes. Ozone
deposition to the surface is an important part of the
total ozone budget and constitutes one of the major
sinks in the lower troposphere. Elevated ozone concentrations near the ground are associated with poor
air quality and thus negative impacts on human health,
materials and vegetation. Ozone is considered the most
phytotoxic air pollutant in rural areas (Ashmore, 2005)
and the potential of ozone to impede plant growth and
reduce crop yield in agriculture is of ecological and
economic importance. To assess impacts on plants, a
UNECE critical-level approach has been used and the
exceedance of the critical level corresponds to a negative response in plants. For ozone, an accumulated
exposure concept was originally conceived, which is
based on exposure–response relationships obtained in
chamber studies. AOT40, short for the “accumulated
ozone concentration above the threshold of 40 ppb”
for daytime conditions during the growing season,
was used to set a critical level for different vegetation
groups. One of the advantages of the AOT40 criticallevel concept is that it is relatively straightforward,
easy to implement and monitor across Europe. Many
ozone concentration measurements are made operationally across the continent, allowing direct mapping
of exposures and assessing exceedances. However,
the AOT40 approach has several uncertainties and
conceptual problems. One limitation of the AOT40
concept is that accumulated ambient ozone concentrations do not necessarily correspond to effects in
Copyright  2009 Royal Meteorological Society
non-optimal growth conditions, and so, potentially
overestimating damage (Gruenhage and Jaeger, 1994).
Thus, a more mechanistic approach was sought and a
flux-based metric subsequently developed. The accumulated stomatal flux, AFst, is more directly linked
to the ozone taken up through the stomata of plants,
allowing the inclusion of response processes such
as stomatal closure in hot, dry conditions. A fluxbased effects model has been developed and tested
by Emberson et al. (2000) and now forms the basis
of some of the UNECE vegetation risk assessments.
High-quality ozone flux measurements are required to
help validate such flux-based effects models. These
can be measured using the aerodynamic gradient or
eddy covariance method. A comparison of analysers
and techniques was previously carried out by Keronen
et al. (2003) and Mikkelsen et al. (2000) for tall forest vegetation, and by Droppo (1985) and Pilegaard
et al. (1998) for short vegetation. This letter presents
a unique dataset of ozone fluxes over grassland measured by the gradient technique and two dry chemiluminescence analysers for the eddy covariance method.
2. Micrometeorological flux measurement
2.1. Aerodynamic gradient method
Analogous to Fick’s law of molecular diffusion, a trace
gas flux (F χ ) can be calculated by multiplying the
vertical gradient in trace gas concentration with the
eddy diffusivity coefficient K χ , viz
Fχ = −Kχ .
Ozone fluxes by gradient and eddy covariance technique
where z is the height above the surface. The vertical
gradient of the trace gas can be obtained relatively
easily by measuring average concentrations at several
heights above the surface. The eddy diffusivity coefficient K χ can be assumed to approximate the turbulent
diffusivity of momentum Km in neutral conditions. Km
can be found from the equation defining momentum
flux τ
τ = ρ.Km .
where ρ is air density and u is average horizontal
wind speed. For non-neutral conditions, the MoninObukhov Similarity theory allows extending the gradient transport relation by applying stability corrections.
2.2. Eddy covariance technique
v =v +v
w = w + w
χ =χ +χ
The left hand side of equations 3–6 is the instantaneous value (u, v and w for wind, χ for another
variable, e.g. trace gas) which is made of the sum of
the mean part (u,v ,w & χ ) and the fluctuating part
(u ,v ,w & χ ) on the right hand side of the equation.
Based on the correlation of two variables, the average of the product of the two fluctuating parts–or
covariance–is the turbulent flux or transport. For
the vertical turbulent flux of horizontal momentum
(momentum flux) τ = u w and the vertical flux of
a trace gas χ is hence Fχ = χ w . The peak of the
power spectrum of eddy size depends on the measurement height, with eddies increasing in scale with
height, and eddies become greater with increased vegetation/surface roughness height and increased wind
Thus, a trace gas flux can be directly measured
by eddy covariance provided the instrumentation is
sufficiently fast-response (∼10–20 Hz) to capture the
dominant range of eddy sizes contributing to the flux.
3. The Easter Bush long-term flux
measurement site
Ozone flux measurements using the gradient and eddy
covariance technique were made at a grassland site
in Southern Scotland from 17 August to 3 September 2007. The field site is situated on the University
of Edinburgh experimental farm (Easter Bush), about
Copyright  2009 Royal Meteorological Society
10 km south of Edinburgh. The site is maintained by
the Centre for Ecology and Hydrology and has been
used extensively for biosphere–atmosphere exchange
research and monitoring for many years. It is located
in the foothills of the Pentland Hills at 190 m above
sea level (Grid Ref NT245641) and is surrounded by
agricultural fields and farms. The fields surrounding
the Easter Bush site are grasslands with predominantly perennial rye-grass (Lolium perenne). During
the measurement period the two adjacent fields were
both grazed by sheep, and the canopy height reached
10 cm. A comprehensive description and characterisation of the site as well as analysis of long-term ozone
flux measurements can be found in Coyle (2005).
4. Instrumentation and data quality control
The less empirical and more direct method to measure
fluxes is by eddy covariance. Eddy covariance is based
on the statistical analysis of turbulence and flow and
makes use of the Reynolds de-composition of the flow
variables being sums of a mean and a fluctuating part:
u = u + u
Ozone fluxes were measured using the aerodynamic
gradient and the eddy covariance technique. For the
gradient approach, ozone concentrations were sampled
sequentially at four heights above the canopy (1.72,
1.24, 0.89 and 0.64 m) using a UV-absorption analyser (Thermo Scientific TECO 49C) and the ozone
concentration gradient based on all four heights was
used for the flux calculations. This photometric analyser has a detection limit of 1 ppb and a response
time of 10 s. Eddy covariance ozone measurements
were made using two dry chemiluminescence analysers: GFAS ozone sonde (Guesten et al., 1992) and
Rapid Ozone Flux Instrument (ROFI), whose design is
based on that of the GFAS instrument (Coyle, 2005).
The inlet tubes were located side by side below the
sonic anemometer at 2.1 m. The chemiluminescent target disks are silica gel disks with a coumarin/gallic
acid coating and were obtained by Bagus Consulting
(Speyer, Germany). Each disk has a limited lifetime
and disks were renewed in both analysers on 20, 24
and 28 August 2007. As the GFAS and ROFI only
provide a relative measure of ozone concentrations,
the absolute ozone concentrations from the top height
of the gradient mast was used to calculate absolute
ozone fluxes by eddy covariance. Turbulence measurements, used for both the gradient flux and eddy
covariance calculations, were made using a Metek
USA-1 sonic anemometer mounted at 2.5 m height,
sampling at 20.8 Hz. Data logging, online flux analysis and post-logging re-analysis of the data was done
using a LabVIEW programme.
The following standard corrections and filters were
applied to the gradient data, as fully detailed in
Coyle (2005): correction for sequential sampling,
stability corrections, correction for storage effects and
filters for wind direction (flow distortion by near
by monitoring cabin), insufficient fetch (based on
Cumulative Normalised Flux CNF), friction velocity,
u∗ , below threshold of 0.08 ms−1 , extreme stability
using Monin-Obukhov Length |L| above or equal
to the value of 2 and instationarity (using integral
turbulence characteristics) (Foken and Wichura, 1996).
Atmos. Sci. Let. 10: 164–169 (2009)
DOI: 10.1002/asl
Additionally, data were filtered for the quality of the
4-point gradient, and only periods with a regression
coefficient R2 above 0.5 accepted. Application of all
filters resulted in a final data coverage of 25%.
Eddy covariance data series were de-spiked, wind
vectors rotated according to a best plane-of-fit correction, filtered for stationarity and corrected for fluctuations by temperature and water vapour (Webb et al.,
1980). Data for up to three hours immediately after a
disk change were discarded to avoid spurious fluxes.
Other quality control measures included using regressions of relative and absolute ozone concentrations
on a disk-by-disk basis. If the regression coefficient
R2 was below 0.5, outliers and end-of-the-disk-period
concentrations were removed to achieve correlations
above that threshold. This is justified on the basis that
sensitivity of a disk can decrease towards the end of a
disk period, and those periods were removed from the
time series. Final data capture was 68% for the GFAS
and 48% for the ROFI analyser.
Absolute ozone fluxes were obtained using two
separate methods. The Ratio method calculates absolute fluxes every half hour by multiplying the halfhourly absolute ozone concentration with the deposition velocity as obtained by the ratio of the raw voltage
flux to the mean raw voltage output. The Disk Calibration method uses a disk specific calibration factor
as obtained by regressing the absolute ozone concentrations against the mean relative concentrations for
each disk period. The calibration factor (ppb/mV) is
then multiplied with the raw voltage flux.
5. Results
Ozone concentrations at Easter Bush ranged from 3 to
36 ppb during the measurement period, with low concentrations occurring during anticyclonic conditions
and advection of NOx from local sources which led
to chemical losses of ozone in the boundary layer on
some days. Ozone levels during August 2007 were
typical for the time of the year and similar in magnitude to previous years. After quality control of the
data, the final dataset consists of 194 half-hourly flux
Jennifer B. A. Muller et al.
Table I. Mean and median statistics on gradient and eddy
covariance ozone fluxes. Values in ngm−2 s−1 .
values for the gradient method and 588/402 values for
the GFAS/ROFI eddy covariance method.
The mean ozone flux for the measurement period is
largest for the gradient method. The eddy covariance
fluxes are smaller than the gradient fluxes and the
two calculation methods can result in considerably
different mean values (Table 1). Providing a clearer
picture on the distribution of fluxes, Figure 1 shows a
box-and-whisker plot for ozone fluxes and deposition
velocities in the left and right panels respectively.
The box-and-whisker plot reveals that the gradient method produces the largest spread of values and
the most positive flux values. These positive values
are often considered erroneous as ozone is not emitted from the surface, but only deposited onto the
ground/plant canopy/surface. The gradient and GFASRatio fluxes agree reasonably well as the inter-quartile
ranges mostly overlap. However the fluxes from the
ROFI analyser are clearly smaller than the gradient
fluxes. The absolute difference between the techniques
(gradient versus eddy covariance) is dependent on the
chosen method for calculating absolute fluxes from the
fast-response analysers. The Ratio method results in a
better agreement with the gradient technique, but also
produces a larger discrepancy between the two effectively identical analysers GFAS and ROFI. The Disk
Calibration method gives a better agreement between
the two eddy covariance measurements, but the disk
calibration fluxes are clearly smaller than the gradient
The different measurements can be compared in
some more detail using scatter plots. Figure 2 compares the two eddy covariance analysers (left panel)
and the gradient with the eddy covariance fluxes (right
panel). (The Ratio method is used for the eddy covariance fluxes in this instance). It is obvious that GFAS
Figure 1. Box-and-Whisker Plot for ozone fluxes (left) and deposition velocities (right). (Whiskers indicate the 5th and 95th
percentile) Grad refers to Gradient, GF stands for GFAS, RO is ROFI and D denotes Disk Calibration method and R represents
Ratio method.
Copyright  2009 Royal Meteorological Society
Atmos. Sci. Let. 10: 164–169 (2009)
DOI: 10.1002/asl
Ozone fluxes by gradient and eddy covariance technique
Figure 2. Scatter plots of ozone fluxes from the two eddy covariance measurements (left panel) and comparison of gradient
fluxes with eddy covariance fluxes (solid symbol is Gradient–ROFI comparison, open symbol is Gradient–GFAS comparison).
fluxes are generally larger than those from the ROFI
and the scatter is greater for larger deposition fluxes.
The slope of the relationship is 0.58 with a small intercept (−3.6 ngm−2 s−1 ). The R2 is good with value of
0.78. The comparison of the eddy covariance fluxes
with the gradient fluxes (Figure 2, right panel) shows
that for the GFAS the scatter is very large, with a
slope of 1.19 and an intercept of −30 ngm−2 s−1 . The
correlation is poor with a coefficient R2 of 0.15. The
relationship of the gradient with the ROFI fluxes is not
significantly better, and the gradient fluxes are generally larger than those from the ROFI. The slope is 1.78
and the intercept is −47 ngm−2 s−1 . The correlation is
also poor with a coefficient R2 of 0.14.
6. Discussion
As detailed in the results section, the gradient method
produced fewer half-hourly points for the measurement period than the eddy covariance technique. This
means that on average less information is available
on the temporal distribution of ozone fluxes by the
gradient method. The various filters for the gradient
method are required to remove periods when the aerodynamic approach might lead to erroneous fluxes, e.g.
extreme stability, very low friction velocity, instationarity or insufficient quality of the ozone gradient, but
the process of filtering for specific conditions might
also introduce a bias in fluxes. For instance, fluxes
can be expected to be smaller for low friction speeds,
the filtered dataset will be biased towards larger fluxes.
This could be one reason why the gradient fluxes are
larger than the eddy covariance fluxes. However, it is
important to appreciate that even with the eddy covariance method a range of flux values is observed, especially when the different methods of calculating fluxes
is taken into account. This suggests that although eddy
covariance is the more sophisticated micrometeorological method flux measurements, there is a considerable
source of uncertainty in measurement of ozone fluctuations using a dry chemiluminescence analyser. To
exclude differences in instrument performance as the
source of uncertainty, spectral analysis was carried out
Copyright  2009 Royal Meteorological Society
and no significant difference in the power and cospectra was found (data not shown).
Figure 1 reveals that the gradient method produces
positive fluxes for a substantial amount of time and
it has the largest range of values. Since the eddy
covariance statistics do not include positive fluxes for
more than 5% of the time, this seems to suggest that
indeed eddy covariance performs better at measuring
these small fluxes during the measurement campaign
in summer 2007. However, as with the mean values,
the box-and-whisker plot shows that the ROFI and
GFAS differ somewhat from each other, with the
ROFI giving a more peaked distribution than the
The comparison for all half-hourly fluxes (Figure 2)
illustrates that the general agreement in mean or
median values does not provide any useful indication of good agreement on any one day. The diurnal
cycles and changes throughout the day as measured
by eddy covariance are captured more clearly than
for the gradient method. This point is highlighted by
Figure 3 which shows the distribution of deposition
velocity values throughout the diurnal cycle for the
gradient system (A), eddy covariance ROFI (B), and
eddy covariance GFAS (C). The gradient approach
produces a less distinct diurnal profile and negative
deposition velocities occur frequently. In contrast, the
eddy covariance measurements show distributions of
vd that are non-zero during the day and some negative
values occurring during night time. If hourly means
are considered (Figure 3 D), the gradient does show
a diurnal profile that is not too dissimilar from the
eddy covariance measurements made by the GFAS.
The mean hourly gradient deposition velocities are
rather variable though and it can be assumed that
the distribution of vd by the gradient improves with
longer averaging periods when more data points are
binned for each hour. The gradient and GFAS-Ratio
method both produce larger deposition velocities than
the ROFI analyser and the GFAS-DiskCal method,
differing by about 2 mms−1 around midday. This difference can be in part explained by the potential
bias towards larger fluxes in the gradient method and
the uncertainty in the eddy covariance measurements.
Atmos. Sci. Let. 10: 164–169 (2009)
DOI: 10.1002/asl
Jennifer B. A. Muller et al.
Figure 3. Panels A, B and C: Box-and-whisker plots for hourly deposition velocities (vd ) from the Gradient (A), ROFI-Ratio
(B) and GFAS-Ratio (C) system. Whiskers indicate the 5th and 95th percentile. D: Mean hourly deposition velocities (vd ) by eddy
covariance compared to gradient. GFAS blue, and ROFI green, with solid lines for the Ratio method and dotted lines for the Disk
Calibration method. Black solid squares for all gradient data and black open square symbol for positive gradient values only.
Additionally, there might be some difference due to
the difference in inlet height of the eddy covariance
and gradient measurements. Although the turbulence
data from the eddy covariance system is used for the
gradient calculations and the top height ozone concentrations from the gradient is used for the absolute
ozone fluxes from the eddy covariance system, there
could be cases of flux divergence due to reactions of
NO with ozone. As the gradient data are filtered for
insufficient quality of the ozone gradient, many periods when flux divergence might be significant (during periods of high NO concentrations) are removed
from the dataset. To assess the impact on fluxes and
flux divergence during periods of low NO concentrations, modelling of ozone fluxes at several heights
would be required, which is beyond the scope of this
Overall, the results presented here show that the
gradient method provides reasonable ozone flux and
deposition velocity estimates when longer-term averages are considered. This validates the long-term flux
measurements made at the Easter Bush site using the
gradient method. However, these results also highlight
the difficulty of using the gradient data for shortterm process studies where changes of timescales of
a few hours need to be fully resolved and captured.
For this, clearly eddy covariance measurements are
better suited, as even small changes in fluxes are measured well in difficult micrometeorological conditions
Copyright  2009 Royal Meteorological Society
and the number of erroneous fluxes (positive values)
is small. Nonetheless the analysis presented in this
letter also reveals an uncertainty in the measurement of
eddy covariance fluxes using dry chemiluminescence
analysers such as the GFAS and the ROFI and the
magnitude of the difference can depend on the choice
of calculation method. The reason for the discrepancy arising from the calculation methods requires further investigation. Sources of uncertainties associated
with eddy covariance measurements in general have
been well described in the literature (e.g. Foken and
Wichura, 1996), but the uncertainty associated with
fast dry chemiluminescence ozone analysers specifically has not been reported so far.
7. Conclusion
The eddy covariance flux technique is considered
superior to the gradient method as fewer assumptions
are needed to be made about the structure of the atmosphere and data coverage is larger for the eddy covariance measurements. Results presented here show that
the gradient method provides reasonably reliable measurements when longer-term averages are considered.
In case of diurnal cycles and short-term changes, the
eddy covariance method has been found to give somewhat more reliable results.
Atmos. Sci. Let. 10: 164–169 (2009)
DOI: 10.1002/asl
Ozone fluxes by gradient and eddy covariance technique
J.B.A.M would like to acknowledge the Natural Environment Research Council (NERC) for studentship funding
(NER/S/A/2006/14037) and the Centre for Ecology and
Hydrology (CEH) for a CASE-studentship.
Ashmore MR. 2005. Assessing the future global impacts of ozone on
vegetation. Plant, Cell and Environment 28: 949–964.
Coyle M. 2005. The Gaseous Exchange of Ozone at Terrestrial
Surfaces: Non-stomatal Deposition to Grassland. PhD thesis,
University of Edinburgh, UK.
Droppo J. 1985. Concurrent measurements of ozone dry deposition
using eddy covariance and profile flux methods. Journal of
Geophysical Research 90: 2111–2118.
Emberson LD, Wieser G, Ashmore MR. 2000. Modelling of stomatal
conductance and ozone flux of Norway spruce: Comparison with
field data. Environmental Pollution 109: 393–402.
Foken T, Wichura B. 1996. Tools for quality assessment of surfacebased flux measurements. Agricultural and Forest Meteorology 78:
Copyright  2009 Royal Meteorological Society
Gruenhage L, Jaeger H-J. 1994. Influence of atmospheric conductivity
on the ozone exposure of plants under ambient conditions:
Considerations for establishing ozone standards to protect
vegetation. Environmental Pollution 85: 125–129.
Guesten H, Heinrich G, Schmidt RWH, Schurath U. 1992. A novel
ozone sensor for direct eddy flux measurements. Journal of
Atmospheric Chemistry 14: 73–84.
Keronen P, Reissell A, Rannik U, Pohja T, Siivola E, Hiltunen V,
Hari P, Kulmala M, Vesala T. 2003. Ozone flux measurements over
a Scots pine forest using eddy covariance method: Performance
evaluation and comparison with flux-profile method. Boreal
Environment Research 8: 425–443.
Mikkelsen TN, Ro-Poulsen H, Pilegaard K, Hovmand MF, Jensen
NO, Christensen CS, Hummelshoej P. 2000. Ozone uptake by
an evergreen forest canopy : temporal variation and possible
mechanisms. Environmental Pollution 109: 423–429.
Pilegaard K, Hummelshoj P, Jensen NO. 1998. Fluxes of ozone and
nitrogen dioxide measured by eddy correlation over a harvested
wheat field. Atmospheric Environment 32: 1167–1177.
Webb EK, Pearman GI, Leuning R. 1980. Corrections of flux
measurements for density effects due to heat and water vapour
transfer. Quarterly Journal of the Royal Meteorological Society 106:
Atmos. Sci. Let. 10: 164–169 (2009)
DOI: 10.1002/asl
Без категории
Размер файла
180 Кб
asl, 226
Пожаловаться на содержимое документа