6.2011-6385

AIAA 2011-6385
AIAA Guidance, Navigation, and Control Conference
08 - 11 August 2011, Portland, Oregon
Comparison of the Impacts of Airport Terminal/Surface
Weather Hazards
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Shubh Krishna * , Rafal Kicinger, Ph.D. † , Girishkumar Sabhnani, Ph.D., ‡ and Jimmy Krozel, Ph.D. §
Metron Aviation, Inc., Dulles, VA, 20166
This study analyzes the Air Traffic Management (ATM) impacts of surface/terminal
weather constraints at the 35 busiest airports in the National Airspace System (NAS). The
degradation in airport departure and arrival rate performance is investigated during time
periods where ceiling, visibility, surface winds, freezing precipitation, and thunderstorms
prevent clear weather throughput. Using cumulative distribution functions (CDFs), we
compare airport performance during time periods when weather hazards are present with
clear weather baseline performance. Comparing against the clear weather baseline as well as
across weather hazard types allows us to rank the effects of various weather hazards.
Weighting such results by the frequency of occurrence of the weather event allows us to
identify the most costly weather hazards in the NAS for terminal area operations. Such
information allows us to focus future ATM investments on those technologies that can best
mitigate the most significant weather hazards affecting the NAS. All comparisons are
performed with 2009 operational ATM data.
Abbreviations
AAR
Airport Arrival Rate
ADR
Airport Departure Rate
ASPM Aviation System Performance Metrics
ATM
Air Traffic Management
C&V
Ceiling and Visibility
CDF
Cumulative Distribution Function
FAA
Federal Aviation Administration
GDP
Ground Delay Program
kts
knots
ft
m
METAR
NAS
nmi
OEP
SM
TFMS
WITI
I.
Introduction
feet
meters
Meteorological Aviation Report
National Airspace System
nautical mile
Operational Evolution Plan
Statute Mile
Traffic Flow Management System
Weather Induced Traffic Index
W
eather is a major limiting factor in the National Airspace System (NAS) today. The average weather-related
delays in the NAS are twice the average non-weather delays [KH03, KC03]. The Aviation Capacity
Enhancement Plan [FAA03] lists weather as the leading cause (65% to 75%) of delays greater than 15 minutes, with
terminal volume as the second leading cause (12% to 22% of delays). In prior work, we have investigated how to
categorize a wide variety of aviation weather impacts [KM07, KMP08]. A review of the scientific analysis
approaches that formulate Air Traffic Management (ATM) impact models is given by Krozel [K10], including
surface/terminal, en route, as well as oceanic models. In this paper, we only focus only on surface/terminal impacts.
In order to study airport capacity impacts, we compare surface/terminal area ATM impacts for a variety of
weather phenomena. In the terminal/surface domain, weather constraints are caused by a large spectrum of weather
phenomena: fog, haze, smoke, ceiling and visibility (C&V), thunderstorms, lightning, hail, heavy rain, in-flight icing
(typically a terminal area constraint [KK08]), wind shifts, gusts, microbursts, wind shear, tornados, waterspouts,
snow, blizzards, surface icing, and volcanic ash. However, not very many of these weather phenomena occurring at
or near airports have been studied to determine the causality or ATM-impact (as surveyed in [K10]), and in many
*
Analyst, Research and Engineering Department, 45300 Catalina Ct, Suite 101
Sr. Analyst, Research and Engineering Department, 45300 Catalina Ct, Suite 101, AIAA Member
‡
Sr. Analyst, Research and Engineering Department, 45300 Catalina Ct, Suite 101, AIAA Member
§
Sr. Engineer, Research and Engineering Department, 45300 Catalina Ct, Suite 101, AIAA Associate Fellow
1
American Institute of Aeronautics and Astronautics
†
Copyright © 2011 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
cases, there simply are insufficient data available to do the analysis. While related literature [Bu78, FAR08, Le07,
Pa94, Tu95] provides information on pilot guidelines for addressing such aviation weather hazards in the NAS, most
of the scientific analyses based on operational data for ATM impacts (see [K10]) focuses on en route convective
weather hazards since they have such a large impact on the NAS over many days in the year. This paper aims at
studying the surface/terminal area to better understand and compare the different weather hazards and to quantify
how they impact the terminal/surface domain in terms of ATM performance.
This paper is organized as follows. After this brief introduction, we provide background material on the process
of analyzing ATM impacts. Then we present supporting analyses for a number of weather phenomena according to
this process, including a section comparing all weather phenomena studied. Finally, we state our conclusions, and
provide references. An appendix included at the end of the paper discusses results of a complementary analysis of
weather impacts at major airports incurred during Ground Delay Programs (GDPs).
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
II. Background
Next, we provide background material related to this effort. We describe the evaluation process and the data sources
used for the impact analysis.
A. Evaluation Process
The process used to evaluate weather impacts on NAS resources is as follows.
1. Describe the type of weather whose impact is evaluated. In this paper, we focus on a subset of those weather
phenomena identified in the Introduction. In particular, we address how ceiling, visibility, surface winds,
freezing precipitation, and thunderstorms reduce clear weather throughput.
2. Determine the frequency of occurrence of weather states that cause an impact, expressed either as a number or,
if available, the total annualized duration of events in the NAS. This may range from enumerating episodic
cases for events that occur infrequently (e.g., volcanic ash on runways), or a full climatology for weather events
that occur frequently throughout the year (e.g., convection or C&V).
3. Analyze the impacts on NAS performance for each weather phenomenon. The first aspect of this is to determine
the clear weather baseline for the NAS resource, in this case, for each of the 35 busiest airports in the US,
formerly known as Operational Evolution Plan (OEP) airports (Figure 1). Currently, 30 out of OEP-35 airports
are included in the set of “core” airports.
4. Given the clear weather baseline and the frequency of occurrence, quantify the total ATM impact considering a
weighted impact, weighting the ATM impact with the frequency of occurrence. The impact considered in this
paper is the departure and arrival throughput degradation. These throughput reductions can be attributed to
delays, GDPs (discussed in the Appendix), cancellations and diversions. Other types of surface delays, such as
gate delay and taxi delay, are discussed where applicable. Cancellations and diversions are not discussed but
may be analyzed as part of future work.
B. Sources of Data
The sources of traffic and weather
data in this study include the
following.
•
ASPM. The FAA currently
provides Aviation System
Performance
Metrics
(ASPM) data for flight
metrics at 77 major airports,
referred to as the ASPM
airports. This set of 77
airports includes all OEP-35
airports. ASPM includes
airport weather, runway
configuration, and arrival and
departure rates.
Weather
observations recorded for
each of the ASPM airports
come from Meteorological
Aviation Reports (METARs).
Figure 1. OEP-35 airports used in analysis (HNL airport not shown).
2
American Institute of Aeronautics and Astronautics
•
METARs. METARs are aviation routine weather reports that are typically issued once per hour at airports and
permanent weather observation stations throughout the world. This standardized format is approved by the
International Civil Aviation Organization (ICAO) to be readily understood worldwide. The primary users of
METAR information are the aviation community. Hourly observations of temperature, dew point, wind speed
and direction, precipitation, cloud cover and heights, visibility, and barometric pressure are typically reported in
METARs. Standardized codes contained in a METAR message are summarized in Table 1 [FMH95].
• TFMS. The FAA’s Traffic Flow Management System (TFMS) provides detailed traffic demand information,
including scheduled and filed flight plans, along with actual flight tracks.
Table 1. METAR Code Descriptions.
Qualifier
Weather Type
Intensity or Proximity
Descriptor
Precipitation
Obscuration
Other
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
-
Light
a
Moderate
+ Heavy
VC In the Vicinityb
MI
PR
BC
DR
BL
SH
TS
FZ
Shallow
Partial
Patches
Low Drifting
Blowing
Shower(s)
Thunderstorm
Freezing
DZ
RA
SN
SG
IC
PE
GR
GS
Drizzle
BR Mist
PO Well
Developed
Rain
FG Fog
Dust/Sand
Snow
FU Smoke
Whirls
Snow Grains
VA Volcanic Ash
SQ
Squalls
Ice Crystals
DU Widespread
FC Funnel
Dust
Ice Pellets
Cloud(s)
SA Sand
Hail
(Tornado, or
HZ Haze
Small Hail/ Snow
b
Waterspout)
Pellets
PY Spray
SS Sandstorm
UP Unknown
DS Duststorm
Precipitation
Note: The weather groups are constructed by considering columns 1 to 5 in sequence, i.e., intensity, followed by
description, followed by weather phenomena
a
To denote moderate intensity no entry or symbol is used.
b
Tornados and waterspouts are coded +FC
C. Clear Weather Baseline
Throughput is highly variable due to a multitude of factors, including weather, congestion, and demand, so
establishing a cause-and-effect relationship, and quantifying such a relationship, is challenging. Furthermore,
analyzing capacity reductions due to weather impacts, we must compare weather impacted performance with
performance that occurs during peak capacity conditions. Since weather is a large cause of degraded performance of
the NAS, we consider clear weather as the most important peak capacity condition. We defined clear weather as the
following: visibility of 10 miles; ceiling above the minimum visual approach; wind less than 10 knots; and no
precipitation. The average departure and arrival throughput counts, stratified by airport, hour, weekday/weekend
and season (summer, winter, and other (spring and fall combined)), is calculated for the entire year of 2009, for all
hours with clear weather conditions. Thus, with 35 airports, 24 hours a day, and 6 sets of profiles (e.g., summerweekday) for each airport, a lookup table was created that represents the average of clear weather departure and
arrival throughput counts per airport, per summer-weekday-hour, etc. The departure and arrival throughput
observed during a weather impacted day/hour/airport is then compared with the clear weather baseline for that
airport, accounting for weekday/weekend, season and hour; the difference in throughput represents the throughput
degradation.
D. Use of CDF Plots
A CDF plot displays the proportion of X
values less than or equal to x. In particular,
X=x and Y=0.4 will mean that for 40%
cases, X ≤ x.
Consider the example of arrival
throughput CDF plot shown in Figure 2.
The x-axis represents arrival throughput rate
(ratio of actual arrival count to the average
arrival count, multiplied by hundred); the yaxis shows the percentile. The median,
extended from the y-axis with the dashed
horizontal line, intersects with the orange
Figure 2. Example CDF Plot.
3
American Institute of Aeronautics and Astronautics
and black CDF curves at x-axis values of 60% and 100%, respectively. In this example, the black CDF curve is the
clear weather baseline; the distribution is symmetric and similar to a Gaussian and, expectedly, the median rate is
100%. The median of the weather impacted throughput rate distribution, shown in orange, is 60%. This means that
half of all hours with the associated weather condition had arrival throughput that was 60% or less of the typical
arrival throughput at the same airport and the same season, weekend/weekday, and hour. Multiple curves in the
same throughput CDF plot can be compared in this way. For same value of X=x, a higher value of Y means more
cases had less than of equal to x throughput percent. Similarly, for same value of Y=y, a smaller value of X conveys
that same proportion of cases had lesser throughput percent. Thus a curve C1 that lies to the left and above curve C2
represents reduced throughput.
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
III. Impact Analysis
The impact analysis focuses on frequency of occurrence and airport arrival throughput. We distinguish airport
departure and arrival throughput from airport departure rate (ADR) and airport arrival rate (AAR); the latter being
operational terms that represents the number of aircraft an airport can accept, whereas the former represents the
actual throughput (departure or arrival) defined as the number of flights per hour at an airport.
A. Ceilings
Description. A ceiling is defined as “the height above the earth’s surface of the lowest layer of clouds or obscuring
phenomena that is reported as “broken,” “overcast,” or “obscuration,” and not classified as “thin” or “partial”
[AIM10]. A broken layer, the least restrictive of the ceiling definitions, is defined as a layer of the atmosphere with
5/8 to 7/8 sky cover. Full sky coverage is defined as “overcast.”
Frequency of Occurrence. Actual hourly weather observations at OEP-35 airports were analyzed for year 2009
METAR data. Figure 3 shows the count of hourly observations with low ceiling, defined as 3000 feet (ft) or less.
To ensure we are addressing time periods of significant ATM activity, only those hours where NAS-wide demand is
typically the highest was compared with the clear weather baseline. Analysis of the aggregate departure and arrival
counts across all airports showed that the hours of 0700 to 2000 local times are the busiest hours across the whole
NAS - only observations occurring within this time period were used for the analyses of throughput in this paper.
We group ceiling data into three tiers: 2000-3000 ft, 1000-2000 ft, less than 1000 ft; the breakdown of
occurrence, by ceiling tier and airport, is shown in Figure 3. Since low ceilings are sometimes associated with
SEA, SAN, ATL and LAX observe most hours with low ceilings,
although for SAN ceilings below 1000 ft are relatively infrequent.
Figure 3. Frequency of Low Ceiling for OEP-35 airports.
4
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
precipitation, the impact on throughput rate
during low ceiling with and without
precipitation are both analyzed.
Impacts. We quantify the degradation in
throughput observed during hours with low
ceiling by comparing observed throughput
during hours with low ceiling and comparing it
against the clear weather baseline. The hours
that are analyzed are restricted to 0700 and
2000 local time, in order to calculate
degradation in performance when the system –
in this case, the NAS – is busiest. The data is
Percent of Average Clear Weather Arrival Throughput
segregated in terms of weekend/weekday and
Figure
4. Impact of low ceilings on arrival throughput.
summer/winter/other, as the average demand is
different in these periods.
The throughput rate during the presence of low ceiling, as a percentage of the clear-weather baseline for similar
hours at affected airports, is shown in Figure 4. The CDF of the arrival throughput during clear weather is compared
to the CDF of throughput in the presence of low ceiling. Data for low ceilings is split into three ranges: between
2000 ft and 3000 ft, between 1000 ft and 2000 ft and less than 1000 ft. For both departure and arrival throughput,
there is approximately 3-5% reduction in throughput in presence of low ceilings. This is measured by comparing the
integral (or cumulative sum) of the low ceiling CDF curve and the integral of the baseline CDF curve. Since the low
ceiling curve is to the left of the baseline, there is associated throughput degradation. The relatively small
throughput reduction during low ceiling time periods is because a majority of OEP-35 airports have flexibility to
mitigate those weather conditions without impacting scheduled demand. However, airports with scheduled demand
approaching operational limits are much more sensitive, and more disproportionately affected, to weather conditions
of all types. As such, the regional impact of low ceiling can vary greatly, even if the aggregated statistics across all
OEP-35 airports only show a small impact. GDPs are used to mitigate these circumstances at congested airports; a
complementary analysis of GDPs is discussed further in the Appendix
B. Visibility
Description. Visibility is the ability, as determined by atmospheric conditions and expressed in units of distance, to
see and identify prominent unlighted objects by day and prominently lighted objects by night. Visibility is reported
as statute miles (SM), hundreds of feet (ft) or meters (m) [AIM10].
Frequency of Occurrence. Actual hourly weather observations were analyzed to determine hourly counts when low
visibility (below 3 SM) was observed as shown in Figure 5. We separate the hours when only low visibility was
observed from hours when low visibility occurred in presence of other weather factors (i.e. precipitation) that could
affect the airport throughput. Figure 5 shows that low visibility co-exists with some other weather event, such as
rain or snow. Because low visibility can be thought of as a secondary effect, or consequence, of a weather event, it is
difficult to analyze in isolation. Thus, virtually all causes of low visibility are already analyzed in this paper with
the analyses of the multitude of other weather types (rain, snow, fog, drizzle, etc.)
Impacts As previously mentioned, low visibility co-exists as a secondary effect of a variety of weather conditions.
Most of these weather conditions are already analyzed separately in this paper: rain, snow, drizzle, fog,
thunderstorms. The impacts of these conditions are compared in Section IV.
C. Surface Winds
Description. Surface winds are winds reported by METAR observations. Surface winds are conventionally measured
at a height of 10 m above ground in an area where the distance between the anemometer and any obstruction is at
least 10 times the height of the obstruction. In reference to a particular runway, surface winds can act as headwinds,
tailwinds, or crosswinds. High surface winds are typically accompanied by gusts in which the wind speed increases
suddenly.
Frequency of Occurrence. Hourly counts of actual weather observations when strong winds (speed > 20 kts) were
observed are shown in Figure 6. We separate the hours when only strong winds were present and when strong
winds existed in presence of other factors that could affect the airport throughput. Other factors included low C&V,
moderate or heavy rain, snow, fog, thunderstorms, and other conditions in the vicinity of the airport.
5
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Figure 5. Frequency of Low Visibility for OEP-35 airports.
The data shows most hours with strong surface winds at OEP-35 airports do not have another co-existing
weather condition. This is in stark contrast to what was shown for visibility. Thus, winds are a good candidate for an
independent evaluation for ATM impact, as it is easily isolated from other impacting weather conditions.
Figure 6. Frequency of Strong Winds for OEP-35 airports.
Crosswind and tailwind components were analyzed separately. For each airport, we analyzed the wind
direction, wind speed, and the runway configuration in use at that hour to compute the tailwind and crosswind
components. Figure 7 shows the frequency density plot of tailwind (headwind) vs crosswinds. As expected, airports
mostly operate with tailwinds < 8 kts; however, crosswinds go as high as 20 kts.
6
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Impacts. We quantify the degradation of throughput during periods of strong winds. Clear weather baseline
consisted of average arrival throughput counts observed during times when wind speed < 10 kts. The data is
segregated in terms of weekend/weekday and summer/winter/other as described earlier. We consider the crosswind
impact separately from the tailwind impact. The baseline data represents clear weather averages and is same as that
is used in analyses of all other weather types in this paper.
Dark black density
represents data
point with number
of hours > 5000,
with numbers
linearly
decreasing with
the shade of grey.
Tailwinds (kts)
Headwinds (kts)
Figure 7. Headwinds vs Crosswinds frequency density plot for OEP-35 airports for 2009.
Figure 8 shows the arrival
throughput rate during the presence of
strong winds, as a percentage of the
clear-weather baseline for similar
hours at affected airports. It shows that
arrival throughput is reduced by 3-5%
in
presence
of
high
tailwinds/crosswinds. Crosswinds > 15
kts seem to have more impact on
throughput as compared to tailwinds >
5 kts; however, it must be noted that
airports often change their runway
configurations shortly after suboptimal
Percent of Average Clear Weather Arrival Throughput
wind conditions, such as tailwinds,
Figure 8. Arrival throughput in presence of strong winds.
develop; thus there is a much lower
sample size to analyze for tailwinds.
Figure 9 shows the number of hours, after a sudden change in wind speed and/or direction is detected, until a
runway configuration change (RCC) takes place. A majority of the RCCs takes place within four hours of the wind
profile change, including more than 25% that occur within one hour.
D. Freezing Precipitation
Description. Freezing precipitation (i.e. snow, ice pellets, freezing rain, or a combination thereof) can cause icing on
the surfaces of aircraft and runways, posing a safety hazard for aircraft. It is a common occurrence from December
through March in the middle latitudes of the Continental US (CONUS).
Frequency of Occurrence. Hourly METAR observations at OEP-35 airports were analyzed for 2009. Hours where
one of the following weather conditions present were identified:
ƒ Light Snow
ƒ Freezing Drizzle
ƒ Moderate Snow
ƒ Freezing Rain
ƒ Heavy Snow
7
American Institute of Aeronautics and Astronautics
RCC Counts
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Hours Following Wind Event
Figure 9. Number of Hours between Wind Event ** and Runway Configuration Change.
The counts of hourly observations having each of these weather types are shown in Figure 10. Light snow is the
predominant frozen precipitation that was observed.
Impact Analysis. We quantify the degradation in throughput, defined as the number of flight departures and arrivals
per hour at an airport, observed during periods of freezing precipitation at the surface.
Snow observations were categorized based on precipitation intensity. METAR categorizes precipitation
intensity into three levels: light (-), moderate ( ), and heavy (+). For example, -SN, SN and +SN represent light
snow, moderate snow and heavy snow, respectively. Freezing rain and freezing drizzle are denoted in the METARs
as FZRA and FZDZ, respectively.
Figure 10. Number of Hourly Observations for Frozen Precipitation Types.
**
Wind Event is defined as a sudden change in direction and/or speed
8
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
The departure and arrival throughput rates during the presence of frozen precipitation, as a percentage of the
clear-weather baseline for similar hours at affected airports, are shown in Figure 11. Throughput degradation
shown in the CDFs can generally be assessed and quantified as follows: the further left the CDF of throughput rate
for a particular weather type is from the clear weather baseline CDF, the greater the throughput degradation
observed. For example, in Figure 11a, light snow corresponds to some departure throughput degradation, as it is to
the left of the clear weather baseline CDF; however, moderate and heavy snow correspond to much higher
throughput reductions. In fact, there was no departure throughput for 10% of hourly observations with moderate
snow and more than 20% of hourly observations with heavy snow. This type of airport shutdown only occurred with
light snow in 2% of the hourly observations. The ranking for impact of winter weather types on departure and
arrival throughput, in descending order of highest magnitude of impact to lowest magnitude, is also displayed in
Figure 11a and Figure 11b, respectively.
Heavy Snow
Moderate Snow
Freezing Drizzle
Freezing Rain
Light Snow
Baseline
Heavy Snow
Moderate Snow
Freezing Drizzle
Freezing Rain
Light Snow
Baseline
(a) Departures
(b) Arrivals
Figure 11. Throughput Rate in the Presence of Frozen Precipitation.
The reasons for the significant arrival throughput degradation in the presence of frozen precipitation include
GDPs, along with strategic cancellations by the airlines in anticipation of the winter weather event. Cancellations of
scheduled departures are also a significant component of the departure throughput reduction. Another significant
reason is the gate and taxi-out delays that occur because of deicing. The precipitation type and the surface
temperature determine the length of holdover time for deicing. If the precipitation rate, or liquid water equivalent,
of the frozen precipitation is high, aircraft may have trouble meeting the holdover times that are required in order to
takeoff. If holdover times are exceeded, aircraft must deice again, further increasing delays.
Figure 12 shows a box plot for the distribution of gate delays and taxi out delays for each flight that departed
during hours with frozen precipitation. The gate and taxi out delays for each flight were obtained from ASPM and
93,274 flights were analyzed. The horizontal line in the middle of the box represents the median gate or taxi-out
delay; the lower and upper side of the box represents the 25th and 75th percentile, respectively. The length of the box
represents the Interquartile Range (IQR). . The upper “fence” is 1.5*IQR above the 75th percentile; the lower fence
is 1.5*IQR below the 25th percentile. Data outside of these ranges are outliers and are not shown.
Figure 12 shows that there is an increase in gate and taxi out delays during frozen precipitation. In particular,
there variance of the gate and taxi-out times markedly increase during hours of moderate or heavy snow, such that
there are some flights absorbing an extraordinary amount of gate and taxi out delays. This large increase is likely
due to deicing along with impaired movement at the airport due to weather conditions.
E. Thunderstorms
Thunderstorms. A thunderstorm is a local storm produced by a cumulonimbus cloud and accompanied by lightning
and thunder, usually with strong gusts of wind, heavy rain, and hail that occur as a consequence of atmospheric
instability. A strong convective updraft is present in the early stages of a thunderstorm, while a strong downdraft
occurs in a column of precipitation during its dissipation.
Frequency of Occurrence. Actual hourly weather observations were analyzed and the counts of hourly observations
when a light (-TS), moderate (TS), severe (+TS) thunderstorm, or a thunderstorm in the vicinity of an airport
9
American Institute of Aeronautics and Astronautics
(VCTS) was reported are shown in Figure 13. MCO and FLL experienced the longest duration of overall
thunderstorm activity while FLL, MEM, and ATL were the airports most affected by severe thunderstorm activity.
Heavy Snow
Moderate
Snow
Light Snow
Freezing
Drizzle
Freezing
Rain
Clear Weather Baseline
Figure 12. Distribution of Gate and Taxi-Out Delays (for all flights departing during frozen precipitation).
Thunderstorm in the Vicinity of an Airport (VCTS)
Light Thunderstorms (-TS)
Moderate Thunderstorms (TS)
Heavy Thunderstorms (+TS)
Total Hours Observed
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Gate Delay
Taxi Delay
Figure 13. Frequency of thunderstorms for OEP-35 airports
10
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Impacts. We quantify the degradation in
throughput observed during hours of light,
moderate, and severe thunderstorms, and
(+TS) Heavy
thunderstorms in the vicinity of an airport
Thunderstorms
against the clear weather baseline. This clear
weather baseline is computed in the same way
as described in our ceiling analysis. Figure 14
shows that arrival throughput is reduced by
(-TS) Light
about 8-15% in presence of thunderstorms. As
Thunderstorms
(VCTS) Thunderstorms
expected, severe thunderstorms (+TS) cause
in the Vicinity of an
the largest throughput reduction as the
Airport
corresponding CDF curve is most shifted to
the left. Figure 14 also indicates that light
(TS) Moderate
thunderstorms (-TS) appear to have more
Thunderstorms
adverse effect on airport throughput than
moderate thunderstorms (TS). However,
median throughput reductions for light and
Clear Weather
moderate thunderstorms are very similar (both
Baseline
curves cross the horizontal line at y=0.5 at
very similar values of x). Finally,
thunderstorm activity in the vicinity of airport
(VCTS) has the least impact on throughput.
Interestingly, in some case (see upper right
Figure 14. Arrival Throughput in the Presence of
part of Figure 14) thunderstorm activity in the
Thunderstorm Activity.
vicinity of the airport leads to arrival
throughput increase as compared to the clear weather baseline. This may be explained by changes in operational
procedures employed by air traffic controllers when a thunderstorm approaches the airport. Air traffic controllers
typically attempt to streamline arrival procedures to accommodate the largest number of arriving flights and avoid
airborne delays and diversions. This hypothesis is further confirmed by an analysis of the departure throughput (not
shown) which does not show any throughput increases when thunderstorms appear in the vicinity of airports. On the
contrary, in this case the departure throughput reduction is much more profound and very close to departure
throughput reduction caused by moderate thunderstorms. Hence, when a thunderstorm approaches an airport, air
traffic controllers tend to prioritize arriving flights thus increasing arrival throughput and decreasing departure
throughput and with respect to clear-weather baseline.
IV. Comparison
In this section, we consider both the frequency of occurrence and the ATM impact, comparing all weather impacts
to each other using CDFs as the basis of comparison. Figure 15 compares the CDFs for various weather types. The
CDF of the clear weather baseline is also included for comparison purposes. The ranking of weather type, from
highest to lowest impact, is displayed in the plot. Heavy snow has the highest impact on departure throughput
degradation, while low ceiling (with no precipitation) having the least impact. Likewise, a comparison of arrival
throughput with the clear weather baseline is shown in Figure 16. Frozen precipitation categories are in the top 4
list, in terms of impacting throughput, for both departures and arrivals.
Figure 17 compares the magnitude of impact with frequency of occurrence, for each weather type studied in
this paper. While moderate/heavy snow and freezing rain/drizzle has the highest impact on throughput degradation,
it occurs much less frequently over the entire year compared with various other weather types. The magnitude of
throughput degradation relative to the clear weather baseline is calculated by taking the integral (cumulative sum) of
the CDF curve for each weather condition and dividing that by the integral of the CDF of the clear weather baseline.
This provides an estimate of the average throughput degradation across all OEP-35 airports for each weather type.
The frequency of occurrence is calculated by aggregating all hourly observations for each weather type across all
OEP-35 airports during 2009. The magnitude of impact multiplied by the frequency of occurrence provides some
measure of cost of surface weather hazards to the NAS system as a whole; however, there are many other variables
involved in the systemic cost of surface weather hazards on the NAS, such as delay propagation, cancellations, fleet
mix at impacted airports, etc.
11
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Figure 15. Comparison of Departure Throughput Rate with Clear Weather Baseline.
Figure 16. Comparison of Arrival Throughput Rate with Clear Weather Baseline.
12
American Institute of Aeronautics and Astronautics
Magnitude of Impact
Frequency of Occurrence
Heavy Snow
Moderate Snow
Freezing Drizzle
Freezing Rain
Heavy Thunderstorm
Fog
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Low Ceiling with Rain
Arrivals
Departures
Wind (30 kts or greater)
Light Snow
Drizzle
Light/Moderate Thunderstorm
Low Ceiling (No Precip)
Average Percent Reduction in Throughput
(relative to Clear Weather Baseline)
Total OEP-35 Airport Hours
Figure 17. Comparison of Magnitude of Impact with Frequency of Occurrence
V. Conclusions
This study analyzes the ATM impacts of surface/terminal weather constraints at the 35 largest (OEP-35) airports.
The degradation in airport departure and arrival throughput is investigated during time periods where ceiling,
visibility, surface winds, freezing precipitation, and thunderstorms are impacting the airport. Using CDFs, we
compare time periods when weather hazards are present to time periods when there is a clear weather. Comparing
against the clear weather baseline across multiple weather hazard types, we rank the effects of weather hazards.
Heavy/moderate snow, freezing drizzle/rain, and heavy thunderstorms have the greatest magnitude of impact on
both departure and arrival throughput at OEP-35 airports. However, weighing by frequency of occurrence, low
ceiling has the most aggregated impact over the entire year. This is corroborated by an analysis showing that the
most impacting condition for GDPs is low ceilings. Another interesting result from our analyses shows that during
hours with light snow or thunderstorms approaching an airport, there is sometimes an increase in throughput relative
to what typically occurs during clear weather. In the case of light snow, this may happen because the onset of light
snow typically (in case of synoptic scale winter storms) precedes moderate or heavy snow and airlines may increase
their operations before the worst of the weather arrives. Similarly, in the case of a thunderstorm approaching an
airport, air traffic controllers may prioritize arriving flights thus increasing arrival throughput with respect to clearweather baseline to avoid airborne delays and diversions.
The results of our analyses should be useful for developing predictive models of the impact of terminal weather
on airport operations. We are planning to develop such models for major types of terminal weather hazards,
including snow and freezing precipitation as such models are needed to improve traffic flow management planning.
Acknowledgments
This research was funded in part by NASA Ames Research Center, under contract NNA11AA17C, Weather
Translation Models for Strategic Traffic Flow Management, and by FAA contract DTFAWA-10-D-00033, Weather
Integration Technology. The authors appreciate the support of the NASA contract monitor, Mr. William Chan, and
the FAA contract monitor, Mr. Dave Pace, in the course of this work.
13
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Appendix
This Appendix contains an analysis of GDPs. It is common for the Air Traffic Control System Command
Center (ATCSCC) to establish GDPs to address problems caused by reduced capacity at airports. GDPs work by
metering the rate at which traffic can use a NAS resource. The scheduled demand for that resource may exceed its
capacity over some time period, necessitating this Traffic Flow Management (TFM) action. In a GDP, the resource
being rationed is an airport, for flights arriving at that airport.
Using data from the FAA’s TFMS, we analyzed all the GDPs issued during 2009. Aggregate delay statistics
categorized by various impacting conditions are shown in Table 2. The cause of each GDP, represented by the field
“IC_CAUSE”, was counted for each OEP-35 airport. The distribution of GDPs, categorized by OEP-35 airport and
impacting condition, is shown in Figure 18. Figure 19 shows the percentage of hours between 12 p.m. and 10 p.m.
local time where the scheduled demand is at least 90% of the airport’s AAR limit. A strong relationship between
frequency of GDPs and airports that often have scheduled arrival demand approaching their AAR limits can be seen
amongst the airports highlighted in red font †† . This relationship also shows that weather has a more severe impact at
airports with scheduled operations approaching operational limits.
Type
Thunderstorms
Low Ceiling
Low Visibility
Wind
Snow/Ice
Table 2. Aggregate GDP Statistics for 2009
Total EDCT Controlled Flights
Total EDCT Delay Minutes
101,579
10,105,167
281,137
23,268,239
40,089
3,197,686
184,777
14,297,538
57,580
6,835,458
Figure 18. Number of GDPs by Impacting Condition for OEP-35 airports.
††
SFO is a special case, where GDPs are often implemented due to routine instances of low ceiling and fog
14
American Institute of Aeronautics and Astronautics
Downloaded by UNIVERSITY OF ADELAIDE on October 27, 2017 | http://arc.aiaa.org | DOI: 10.2514/6.2011-6385
Figure 19. Comparison of Scheduled Arrival Demand relative to operational AAR.
References
[AIM10] Aeronautical Information Manual-Official Guide to Basic Flight Information and ATC Procedures, U.S. Department of
Transportation, Federal Aviation Administration, Feb. 11, 2010.
[Bu78] Buck, R., Weather Flying, Macmillan Pub., New York, NY, 1978.
[FAA03] Federal Aviation Administration, 2003 Aviation Capacity Enhancement Plan, Federal Aviation Administration Office
of System Capacity, Washington, DC, 2003.
[FAR08] Federal Aviation Regulations / Aeronautical Information Manual, ASA Pub., Newcastle, WA, 2008.
[FMH95] Federal Meteorological Handbook No. 1, Surface Weather Observations and Reports, Dec., 1995.
[K10]
Krozel, J., “Survey of Weather Impact Models used in Air Traffic Management,” AIAA Aviation Technology,
Integration, and Operations Conf., Ft. Worth, TX, Sept., 2010.
[KC03] Krozel, J., Capozzi, B., Andre, A., and Smith, P., “The Future National Airspace System: Design Requirements
Imposed by Weather Constraints,” AIAA Guidance, Navigation, and Control Conf., Austin, TX, Aug., 2003.
[KH03] Krozel, J., Hoffman, B., Penny, S., and Butler, T., “Aggregate Statistics of the National Airspace System”, AIAA
Guidance, Navigation, and Control Conf., Austin, TX, Aug., 2003.
[KJL07] Klein, A., Jehlen, R., and Liang, D. “Weather Index with Queuing Component for National Airspace System
Performance Assessment,” 7th USA/Europe Air Traffic Management R&D Seminar, Barcelona, Spain, July, 2007.
[KK08] Krishna, S., and Krozel, J., “Impact Analysis for In-Flight Icing Hazards,” AIAA Guidance, Navigation, and Control
Conf., Chicago, IL, Aug., 2009.
[KKL09] Klein, A., Kavoussi, S., and Lee, S. R. “Weather Forecast Accuracy: Study of Impact on Airport Capacity and
Estimation of Avoidable Costs,” 8th USA/Europe Air Traffic Management Research and Development Seminar, Napa,
CA, June/July, 2009.
[KM07] Krozel, J., and Murphy, J.T., Jr., “Weather Hazard Requirements for NGATS Aircraft,” Integrated Communications,
Navigation, and Surveillance Conf., Herndon, VA, May, 2007.
[KMP08] Krozel, J., McNichols, W., Prete, J., and Lindholm, T., "Causality Analysis for Aviation Weather Hazards," AIAA
Aviation Technology, Integration, and Operations Conf., Anchorage, AK, Sept., 2008.
[Le07] Lester, P., Aviation Weather, Jeppesen Pub. Co., Englewood, CO, 2007.
[Pa94] Padfield, R., Flying in Adverse Conditions, TAB Books, McGraw Hill Pub., New York, NY, 1994.
[Tu95] Turner, T., Weather Patterns and Phenomena, A Pilot’s Guide, TAB Books, McGraw-Hill Pub., New York, NY, 1995.
15
American Institute of Aeronautics and Astronautics