Embed Size (px)
Citation preview
NASA Reference Publication 1404
Surface-Based Observations of ContrailOccurrence Frequency Over the U.S., April1993 - April 1994____________________________________________________________________________________Patrick MinnisLangley Research Center • Hampton, Virginia
J. Kirk AyersAnalytical Services and Materials, Inc. • Hampton, Virginia
Steven P. Weaver88th Weather Squadron• Wright-Patterson Air Force Base, Ohio
■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■■
June 1997
Summary
Surface observers stationed at 19 U.S. AirForce Bases and Army Air Stations recorded thedaytime occurrence of contrails and cloudfraction on an hourly basis for the period April1993 through April 1994. Each observationuses one of four main categories to reportcontrails as unobserved, non-persistent,persistent, and indeterminate. Additionalclassification includes the co-occurrence ofcirrus with each report. The data cover much ofthe continental U.S. including locations nearmajor commercial air routes. The mean annualfrequency of occurrence in unobstructedviewing conditions is 13 percent for these sites.Contrail occurrence varied substantially withlocation and season. Most contrails occurredduring the winter months and least during thesummer with a pronounced minimum duringJuly. Although nocturnal observations are notavailable, it appears that the contrails have adiurnal variation that peaks during mid morningover most areas. Contrails were most oftenobserved in areas near major commercial aircorridors and least often over areas far removedfrom the heaviest air traffic. A significantcorrelation exists between mean contrailfrequency and aircraft fuel usage above 7 kmsuggesting predictive potential for assessingfuture contrail effects on climate. Additionalsurface observations and a concerted satelliteobservation effort are needed to accuratelyassess the climatic effect of aircraftcondensation trails.
Introduction
Condensation trails or contrails havebecome a common feature in the NorthernHemisphere since World War II. Theseanthropogenic clouds represent the most visiblebyproduct of jet fuel combustion at highaltitudes. The mechanism for contrail formationis complex, depending on a variety ofparameters including the type of jet engine, thesort of fuel, and the ambient temperature andhumidity (Karcher, 1994; Schumann, 1996).The exhaust may produce numerous sulfateaerosols that act as cloud condensation nucleiwhich initiate tiny droplets that subsequentlyfreeze. The resulting cloud usually containslarge concentrations of small ice crystals (e.g.,
Murcray, 1970). They generally form attemperatures less than -30°C at high relativehumidities or below -50°C at moderate to lowmoisture levels (e.g., Appleman, 1953). Ifformed in clear air, contrails can spread,developing into cirrus indistinguishable fromnatural clouds. Their persistence and growthdepend on the available moisture and ambienttemperature. When aircraft fly through existingclouds they can produce contrails or distrails(cloud-free trails) depending on the conditions(e.g., Scorer, 1972). In either case, they producean immediate effect by altering themicrophysical properties of the existing cloud.
Increases in cloud cover or cloud particleconcentrations due to contrails can alter thelocal radiative balance by reflecting more solarradiation and absorbing and emitting longwaveinfrared radiation (e.g., Kuhn, 1970). Theoverall effect of contrails on climate depends ona number of factors including frequency andtiming of occurrence, areal coverage, lifetime,altitude, location, and microphysical properties.The upper troposphere is a relatively clean(aerosol-free) environment so that the additionof high concentrations of cloud condensationnuclei have the potential for making a largerimpact than they would in the lowertroposphere. With commercial air trafficexpected to increase by more than 200 percentby 2015 (Baughcum, 1996), the effects ofaircraft exhaust on the atmosphere have becomea subject of considerable interest leading to theNASA Atmospheric Effects of Aircraft Program(AEAP) which has sponsored the SubsonicAssessment (SASS) Project (Thompson, et al.,1996). One of the goals of the AEAP SASSProject is to evaluate the effect of contrails onclimate. This paper presents the results of astudy of contrail occurrence frequencies overthe U.S based on recent surface observations.
Background
Evaluations of contrail coverage oroccurrence have been made either directly orindirectly from surface and satellite observationssince the 1980's. These efforts have beensporadic and generally confined to a fewparticular areas. Examples of inferred contrailcoverage include the conclusions of Chagnon(1981) and Angell et al. (1984) that decreased
2
sunshine and increased cloudiness since 1960and between 1950 and 1972, respectively, areattributable to contrails. Seaver and Lee (1987)found that the number of cloudless days over thecontinental United States (US) decreasedsignificantly for the period 1950-1982 comparedto 1900-1936 possibly due to the appearance ofcontrails during the latter period. In a follow-upstudy, Angell (1990) found that US cloudinesscontinued to increase through 1988 whilesunshine duration decreased. The relativemagnitude of the change in sunshine was not asgreat as the cloudiness increase. This findingsuggests an increase in thin cirrus due, mostlikely, to contrail-generated cirrus. Significantdecreases in insolation were also observed inGermany during the past 20 to 40 years. Weber(1990) suggested that increased cyclonic activityincreased cloud cover and decreased sunshineover Germany. Liepert et al. (1994) estimatedthat contrail coverage, based on a surface studyof contrails over a single site, was too small toaccount for the diminished sunshine.Discrepancies between the conclusions of thesevarious studies highlight the uncertainties in thecurrent assessment of the climate impact ofcontrails.
Satellite data have been used in a variety ofways to study contrails over larger areas andlonger time periods. Joseph et al. (1975) usedtwo photographs from an Earth resourcessatellite to demonstrate the detection of contrailsfrom space over the Mediterranean. Frompreliminary studies of Defense MeteorologicalSatellite Program (DMSP) imagery, Carletonand Lamb (1986) found that infrared (IR) datawere more valuable for detecting contrails thanvisible data. Lee (1989) showed that brightnesstemperature differences between the splitwindow channels on the NOAA Advanced VeryHigh Resolution Radiometer (AVHRR) couldbe used to detect contrails more easily thansimply examining infrared window channelimagery. DeGrand et al. (1990) used the single-channel IR imagery from the Sun-synchronousDMSP satellites to develop a climatology ofcontrail occurrence over US for the period 1977-1979. Although they provided some estimatesof the relative magnitudes of the mean seasonaland diurnal cycles over US, the actualfrequencies of occurrence were not reported.Engelstad et al. (1992) added image processingtechniques to the brightness temperature
difference imagery to automatically detectcontrails without human intervention. Theirmethod, however, has not yet been applied tosignificant amounts of satellite data. Schumannand Wendling (1993) also developed anautomated technique but they have reportedonly preliminary results from 99 AVHRRimages over central Europe. Bakan et al. (1994)used visual analysis of thousands of quicklookAVHRR IR images taken over the northeastAtlantic and Europe to estimate contrailcloudiness for 1979-1981 and 1989-1992. Theyfound a distinct seasonal cycle with a southwarddisplacement of the contrail maximum duringwinter. Maximum contrail coverage in theiranalysis occurred during summer centered alongthe North Atlantic air routes. The coverageincreased in that area during the 10-year interim.Similar analyses over the air corridors of the UShave not yet been performed. The surfaceobservations reported here represent the firststep to better defining the contrail-based cirruscoverage over the US.
NomenclatureAEAP Atmospheric Effects of AircraftProgramAFB Air Force BaseAVHRR Advanced Very High ResolutionRadiometerc mean annual contrail frequencyDMSP Defense Meteorological SatelliteProgramf mean annual fuel usage above 7km, 106 lbsIR infraredLAFB Langley Air Force BaseLT local timeNASA National Aeronautics and SpaceAdministrationNOAA National Oceanographic andAtmospheric AdministrationR linear correlation coefficientRH relative humiditySASS Subsonic AssessmentTf monthly average temperature atmean flight levelU.S. United StatesUS Continental United StatesUTC Universal Coordinated TimeWPAFB Wright-Patterson Air Force Basezp monthly mean tropopausealtitude
3
Data
The U.S. Air Force established a uniquenetwork of surface observers to develop aclimatology of contrail occurrences over the USfor studying contrail formation and forecastmodels. This dataset was made available toNASA in its raw form: log sheets of hourlymeteorological observations with distinctcontrail codes or special log sheets used only forrecording the contrail codes. The contrailobservations were listed as one of fourclassifications: no contrails, non-persistentcontrails, persistent contrails, and indeterminatecontrails. In addition, each classifier wasqualified as being with or without cirrus.Finally, the contrail observations were notalways taken even though weather data wererecorded. Thus, there is a no-observationcategory yielding a total of nine possiblecontrail classifications. These contrail codes aswell as the sky cover in tenths were transposedto a computer spreadsheet format for additionalanalysis.
The observation network, comprising 19U.S. Air Force Bases and Army Air Facilities,was spread over the US (Figure 1) to determinethe spatial variability of contrail formation. Thenominal period of observation for this specialeffort was April 1, 1993 through March 31,1994. The actual period varied with reportingstation. The earliest month is January 1993,while the latest is May 1994. As many as 15and as few as 7 months of data were taken at agiven location. Most sites recorded data for 13months. Contrail observations were taken everyhour at the same time as the standardmeteorological data, but only during thedaytime. Nocturnal data were not recordedbecause of the ambiguities associated withtaking high-cloud observations in the dark (e.g.,Hahn et al., 1995). Sky cover was recorded atmost locations for much of the period, althoughthe special contrail logs were the only availabledata for a few stations. The indeterminatecategory was selected if the sky was overcastbelow the level of cirrus clouds. Persistentcontrails were defined for this study as thoseextending at least several miles behind theaircraft with no tendency for dissipation. Anonpersistent contrail is defined as one thattends to evaporate and only extends a shortdistance behind the aircraft.
Figure 2 shows an example of the raw datain the form of a coded summary of the hourlycontrail observations for Langley Air ForceBase (LAFB) taken during December 1993. Atnight, no contrail data were taken althoughcirrus occurrences were recorded. Indeterminatecontrail conditions dominated during at least 10days, especially December 4, 5, 15, 16, 28, and29. A period of persistent contrails with cirruson the 2nd was followed the next day withsporadically persistent and nonpersistentcontrails and no cirrus. A few temporary ornonpersistent contrails were seen early duringDecember 6 followed by several hours withoutcontrails or cirrus. After 1200 UTC, cirrusoccurred every hour during the next day with afew temporary contrails and 2 hours ofindeterminate contrails. The other significantperiods of persistent contrails include December9, 10, 14, 20, and 24. The total number of hourswhen contrails were or were not observed were52 and 85, respectively. Persistent contrailsaccount for 65 percent of all of the contrailobservations. Of these, only 34 percent wereunaccompanied by cirrus.
Figure 3 summarizes the 24-hour, dailyobservations over LAFB for the period fromMarch 1993 through April 1994. The colors atthe top of graph correspond to indeterminate(light gray) and contrail-free cases, while theblack area in the middle registers the percentageof missing and nighttime data. Along thebottom of the graph, the colors refer to variouscontrail conditions including indeterminatecontrails with cirrus (gray at bottom). No datawere taken during November 1993 and January,February, and March 1994. Persistent contrailsoccurred most often during a given day in thespring during both 1993 (days 110-150) and1994 (days 465-480) and least during the latesummer of 1993 (days 190-220). This latterperiod also has the most observations of nocontrails with cirrus.
Temperatures and humidities at thestandard meteorological levels were taken from12-hourly National Meteorological Centeranalyses. These data are available on a 2.5°latitude-longitude grid. They were bi-linearlyinterpolated to match the location of eachsurface site. Only data from July 1993 andJanuary and April 1994 are considered here.
4
Results
The data for each of the categories exemplifiedin Figures 2 and 3 and for the cloudobservations were averaged for each month andhour in Coordinated Universal Time (UTC).
Seasonal Variations
Examples of the monthly means are plottedin Figure 4 for LAFB. Contrails occurred mostfrequently during April 1993 with ~32 percentof the total observations (Figs. 4a). A similarnumber of contrail observations was reportedduring the following April. Nonpersistentcontrails were most frequent during June 1993and April 1994. The fewest contrails (~5percent) were seen during July 1993 when cirrusclouds were most abundant. The worst viewingconditions were found during March andDecember 1993 when the indeterminate levelswere greatest. Figure 4b combines thecategories into four classes that do not considerthe occurrence of cirrus. The indeterminateclassifications, more easily discerned in thisfigure, generally correspond to the occurrencefrequencies of 90 and 100 percent cloud cover(Figure 5a). If indeterminate data are removedand the classifications are normalized to thenumber of remaining observations, the relativetemporal pattern of contrail occurrence remainsmuch the same except for the substantialincrease in contrails during December (Figure4c). These normalized percentages may be amore accurate accounting of the contrailsbecause the indeterminate data are almostequivalent to missing data. The months havingthe most clear days are October and December1993 and April 1994 (Figure 5a). Resultscorresponding to Figure 4 are provided inAppendix A for each observing site. Themissing data evident in Figure 5 for November1993 and for January, February, and March,1994 are not typical for most of the sites. Bettersampling was obtained for 14 other sites.
Table 1 lists the mean contrail occurrencesfor each site and the corresponding period ofobservation. Figure 6 shows the means andstandard deviations based on monthly averagesof the combined persistent and nonpersistentcontrails from data like those in Figure 4a whichincludes indeterminate data. The fewest number
of persistent contrails occurred over Eglin AFB(3.6 percent) and Minot AFB (3.8 percent),while the greatest number were seen overWright-Patterson AFB (WPAFB; 15.1 percent)and Edwards AFB (14.9 percent). Ifindeterminate cases are omitted, then contrailswere most frequent over WPAFB (28.5 percent)and LAFB(19.9 percent). Without theindeterminate observations, the sites with thefewest persistent contrails are Eglin (5.1percent) and Luke (5.4 percent) AFB's.Nonpersistent contrails were most oftenobserved over Luke AFB (9.2 percent) andLAFB (~5 percent), while the fewest were seenover Kelly AFB. Both persistent andnonpersistent contrails are most likely to occurwhen cirrus clouds are present. The meanprobability for seeing a contrail when cirrus arepresent is 24.8 percent, compared to only 7.3percent when cirrus are not observed. Contrailsare also more likely to persist when cirrus arepresent. When cirrus and contrails occurredtogether, 75 percent of the contrails wereclassified as persistent compared to 55 percentwhen cirrus were absent. The greatest monthlyvariability (Figure 6) occurs over McClellanAFB, while contrail occurrence over Eglin isconsistently low from month to month.
Figure 7, which summarizes the results inAppendix A, reveals that the maximum contrailoccurrence occurs between January (Figure 7a)and April (Figure 7b) for most sites. A notableexception is Minot AFB where October (Figure7d) is the most favored month for contrails.Minimum contrail occurrence is generally foundduring July (Figure 7c). The seasonal variationsin contrail events averaged for all sites areshown in Figure 8. If the events are referencedto the total number of observations (Figure 8a),then there is a distinct maximum duringFebruary and an apparent secondary peak duringOctober. Although the October value isstatistically different from the September mean,it does not differ significantly from theNovember result. Overall, contrails are scarcestduring July. If indeterminate observations areexcluded from the total (Figure 8b), the seasonalcurve becomes smoother. The maximum occursduring March or between February and Marchwith no secondary peak during October.Regional variability is considerable. In absoluteterms, it is greatest during February. Ifcomputed relative to the mean occurrence
5
values, the greatest geographical variabilityoccurs during July and November.
Diurnal VariabilityFigure 9 depicts the diurnal variations in
contrail frequency over LAFB. The greatestnumber of contrails was observed (Figure 9a)during midmorning at 1400 UTC (0900 LT). Abroad secondary maximum covers the periodfrom 1800-2100 UTC. Normalization to thetotal number of observations (Figure 9b) revealsa substantial peak at 2400 UTC. This maximummay not be representative, however, because itis based on a limited sampling of ~110observations (Figure 9a) that are confined toparticular months. This number of observationsis about half of the total between 1200 and 2200UTC. The differences in hourly sampling areprimarily due to changes in the day length withseason. It is important to consider the samplingwhen examining these diurnal results. If onlyhours having more than 150 samples areconsidered, the peak contrail frequency remainsat 1400 UTC and the lowest frequency ofcontrails is apparent at 1200 and 1700 UTC.When indeterminate data are removed (Figure9c), the relative diurnal cycle is the samealthough the morning maximum is enhancedslightly. The primary minima occur at 1200(0700 LT) and 1700 UTC (local noon). Plots ofthe mean hourly contrail statistics are providedin Appendix B for each observation site.
By excluding the indeterminate data andusing only those hours with more than half themaximum number of hourly samples for eachsite, it is possible to determine the primary andsecondary diurnal maxima in contrailoccurrence. Here, the primary maximum isdefined as that hour having the greatest contrailfrequency. The secondary maximum is the hourwith the next highest frequency that is at least 3hours removed from the primary maximum.The amplitude of these maxima is half of thedifference between a given maximum and theminimum divided by the mean total contrailoccurrence. In Figure 8c, the primary peak at1400 UTC is 39 percent and the minimum is 20percent at 1800 UTC. The amplitude of thismaximum is 40 percent because the mean is 24percent (Table 1). The secondary peak at 1800UTC with a magnitude of 19 percent may not beparticularly significant. Table 2 summarizes thediurnal characteristics of contrail frequency for
each site. Loring AFB has the greatestamplitude with a primary peak in the evening.The smallest amplitude is 23 percent atBarksdale AFB, also with an early eveningmaximum. The average primary and secondaryamplitudes are 58 and 46 percent, respectively.Amplitude does not appear to have alongitudinal dependence. Plots of the primaryand secondary maxima times in Figure 10 alsoshow some interesting features. In Figure 10a,the primary maxima are concentrated between1500 and 1800 UTC and between 2300 and0100 UTC, while many of the secondarymaxima occur between 1800 and 2300 UTC.When converted to local time (Figure 10b), theprimary maxima cluster around 0900 and 1700LT, while the secondary peaks primarily occurduring the afternoon. Figure 11, a plot of themaxima as a function of longitude, shows that,except for Loring AFB, the times of primarymaximum are found near 0830 LT over theeastern US, during the late morning or lateafternoon over the central US, and during theearly morning or late afternoon in the west. Thesecondary maxima generally fall in theafternoon except for some of the westernmostsites..
Cloud CoverThe fractional cloudiness frequencies and
monthly mean areal cloud coverage are given inFigure 5 for LAFB. The monthly frequency of agiven cloud fraction (Figure 5a) was discussedearlier. The mean annual diurnal variation ofeach fractional cloud amount is shown in Figure5b. The times are given in UTC, but theabscissa scale is shifted such that the hourclosest to local noon is at the center of thegraph. In this particular case, cloudobservations were taken for the completediurnal cycle. Clear skies are most frequentnear local midnight (5 - 8 UTC) and seen leastoften around local noon (17 - 21 UTC). Bothfractional and overcast cloudiness follow adiurnal cycle complementary to the clear skies.The monthly mean cloudiness in Figure 5cvaries from a low in June 1993 to a peak duringMarch 1993. The missing months prevent amore complete determination of the annualcycle. Data corresponding to Figure 5 havebeen plotted in Appendix C for each site. Nocloud observations were available at five sites:Griffis, Hill, Kelly, McClellan, and Tinker Air
6
Force Bases. Complete 24-hour sampling isavailable at the remaining sites.
The monthly mean cloudiness for the 14sites with observations is summarized in Figure12 with the mean US surface-observed cloudamounts from Hahn et al. (1986) for the period1971- 1981. The average for both datasets is ~55 percent. In both cases, cloud amount peaksduring the winter with a minimum during latesummer and autumn although the annual rangefor the contrail dataset is ~26 percent comparedto 12 percent from the US average. Figure 13shows the monthly averages of mean cirrusfrequencies. These results show that cirrus wasobserved over the surface sites least often duringthe winter and most frequently during the lateautumn months. Thus, the cirrus frequencies areslightly out of phase with the correspondingcloud amounts. Overall, cirrus was observed in~55 percent of the observations. Figure 13balso plots seasonal mean US cirrus frequenciesbased on surface observations during 1971-1980(Hahn et al., 1984). The seasonal meanscorrespond well with the present results forwinter and spring. Cirrus clouds were observed4 and 10 percent more often in the currentdataset than in 10-year average for summer andautumn, respectively. The general agreementbetween the cloud amounts and cirrusfrequencies suggest that the selected sites arerepresentative of the US as a whole.
The mean monthly variations of persistentcontrails with and without cirrus are shown inFigure 14. In general, both curves follow thetotal contrail seasonal trends seen in Figure 8b.The ratio of seasonal contrail frequencies withcirrus to those without cirrus are 3.1, 4.3, 3.6,and 5.1 for winter, spring, summer, and fall,respectively. These co-occurrence ratios areconsistent with the minimum and maximumfrequencies of cirrus in Figure 13b. Figure 15shows the contrail frequencies as a function ofcoincident cloud amount averaged for the sitesreporting hourly cloud amounts. Few persistentcontrails were seen in otherwise clear skies(Figure 15a), while cloud amounts near 75percent correspond to the most frequentoccurrence of persistent contrails. This result isconsistent with the frequent occurrence ofcontrails with cirrus. Conversely, non-persistentcontrails were observed most often in almostclear skies (Figure 15b) suggesting that the
nonpersistent contrails form primarily in drierconditions. When all contrails are considered,the maxima arise in mostly cloudy and partlycloudy conditions with minima in almost clearor overcast conditions and when cloud fractionis around 50 percent (Figure 15c).
Discussion
The occurrence of contrails is primarilydetermined by two factors: the presence ofaircraft exhaust and the ambient conditions atflight level. An observation of a contrailrequires both proper timing and a sufficient lineof sight from the observer to the contrails.These factors and their relationships to theobservations are discussed here.
Aircraft Fuel UseA preliminary dataset of fuel usage was
developed from the estimates of commercialscheduled and other non-scheduled and militaryair traffic by Baughcum et al. (1993) for May1990. The data were compiled on a 1° x 1°latitude-longitude grid with a vertical resolutionof 1 km. It was assumed that the May data arerepresentative of the annual mean. A lateranalysis by Baughcum (1996) confirms thatassumption. Figure 16 shows the mean fuelusage as a function of altitude for the nine 1°boxes centered over the 19 contrail sites. Themaximum fuel use above the boundary layeroccurs at flight levels between 10 and 12 km(33,000 - 38,000 ft) for all of the sites. The low-altitude fuel is primarily expended on therunway and during takeoffs. To account for allflight levels in which most contrails are likely tooccur, the data were summed for all altitudesabove 7 km. The distribution of the sums inFigure 17 reveals the main flight corridors overthe US with a primary maximum over theMidwest between Chicago and New York.Other routes to Florida, Atlanta, Dallas, andsouthern California from the northeast andMidwest are also evident. The geographicalvariation of contrail occurrence on an annualbasis (Figure 18) roughly coincides with the fueldata in Figure 17. For example, maximumcontrail frequency occurs over WPAFB in theheart of the Chicago-New York corridor. Offutand Edwards AFBs, which have relatively highfrequencies, are under the Chicago-Los Angelesjetways and Eglin, Loring, and Minot AFBs,
7
where contrails are not often observed, are offthe main air thoroughfares. Presumably, thediscrepancy between the Beale and McClellanAFB contrail frequencies occurs because thelatter is closer to the edge of the San Francisco-East Coast airway.
Although the quantity of consumed jet fuelprobably increased between 1990 and 1993, therelative pattern of air traffic likely changed littleduring the interim. Therefore, it should bepossible to correlate the 1990 fuel usage withthe data in Figure 18 to determine therelationship between the fuel use and contrailoccurrence. A surface observer can see high-altitude clouds that are a considerable distancefrom the surface position. Furthermore,contrails can advect rapidly from their formationlocation. For example, the climatological meanzonal wind velocities at 300 mb range from ~40km/hr in July to nearly 100 km/hr in Januarybetween 30°N and 45°N (Sadler, 1977). Todetermine how fuel usage relates to contrailfrequency, the effective viewing area for thesurface observer's hourly reading must bedetermined. This area was estimated bycorrelating the mean annual contrail frequenciesto fuel-use averages computed from arrays of 1°boxes. It was assumed that the optimal areacorresponds to the array size yielding thegreatest correlation coefficient. A 3° boxcentered at each site produced a linearcorrelation coefficient, R = 0.73, the maximumcorrelation between mean fuel use for a squarearray and persistent contrail frequency. Thevalues of R for 1°, 5°, and 7° boxes are 0.64,0.62, and 0.41, respectively. Total contrailfrequency including both persistent and non-persistent contrails shows a stronger relationshipto fuel usage with R = 0.78. Scatterplots andlinear regression fits forced through the originare shown in Figure 19 for the 3°-box fuelaverages and total and persistent contrailfrequencies determined without indeterminatedata. According to this fit, the mean annualtotal and persistent contrail frequencies are
ct = 0.00176 f,(1)
andcp = 0.00127 f,
(2)
respectively, where f is the mean fuel useabove 7 km of the nine 1° boxes centered over agiven location. Fuel consumption is given in1 0 6 lbs yr- 1 . While these correlationdemonstrate the obvious, that the likelihood ofobserving a contrail increases as the number ofplanes at altitude increases, it also quantifies, forthe first time, the relationship between aircraftfuel usage and contrail frequency. Moreover, itshows that over the US, fuel expenditure canaccount for 61 percent (R2) of the variance inmean annual persistent contrail occurrence andthat contrail occurrence will increase as airtraffic volume expands.
Using the mean 3° regional US fuel usageabove 7 km, 4.8 x 106 lbs yr-1, in Eq. (1) yieldsa mean occurrence frequency of 0.085 for thecountry as a whole. This result suggests that, onaverage, an observer situated at a randomlocation and time in the US will have an 8.5percent chance of seeing a contrail if the sky isnot totally obscured. Before the commercial jetage began in earnest during the 1960's, contrailswere a rare sight. In some regions like themidwestern US, especially during winter, thelikelihood of observing a contrail is on the orderof 40 percent, an almost every other dayoccurrence.
Meteorological ConditionsFuel use cannot account for all of the
variability in contrail occurrence. Most of theremaining variance is probably due to thediverse temperature and relative humidity RHconditions at flight level, although engine andfuel type as well as the operating conditions alsoinfluence contrail formation. While a detailedexamination of the meteorological conditionsaffecting contrail occurrence is beyond thescope of this study, monthly averages of certainparameters are examined to demonstrate howatmospheric profiles may affect the contrailfrequencies in this dataset.
Because of the typically low relativehumidities in the stratosphere, a plane isunlikely to produce a significant contrail if itflies above the tropopause. Figure 20 shows thevariation of mean tropopause altitude zp withobserving site for 3 months. During January1994, zp varies from 9.6 km in Maine to over 12km in Texas. Most of the heights are between10.5 and 11.5 km. The tropopause heightgenerally increases during April 1994 to
8
between 10.8 and 14 km. During July 1993, therange is 10.8 to 16 km. If 10.5 km is assumed tobe the average flight level, then most of the airtraffic over the US takes place in thetroposphere, even during much of the winter.The two exceptions are Loring AFB and MinotAFB where zp = 9.6 and 10.1 km, respectively,during January 1994. As seen in Appendix A,the maximum contrail frequencies primarilyoccur during the winter and early spring monthsexcept over these two sites. The maximum forLoring occurs during May and June, while thepeak contrail frequency over Minot is seenduring October. The lowered tropopause duringwinter in northern latitudes is also the likelysource for the southward displacement of thecontrail maximum during winter over the NorthAtlantic air traffic routes reported by Bakan etal. (1994).
The mean temperatures Tf at the averageflight level provide further explanation of theseasonal variability. In Figure 21, Tf increasesfrom January through July at all locations.During July 1993, Tf is greater than 225 K overall sites but Fairchild AFB in Washington. Themean flight level temperatures are 225 K or lessduring January 1994. According to Hanson andHanson (1995), contrail formation at 10.5 km ornear 250 mb requires temperatures lower than226 K (-47°C) for RH less than 100 percent.As the temperature decreases, the relativehumidity required for contrail formation alsodecreases. Thus, the probability for contrailoccurrence increases as the temperature drops.The winter maximum in contrail frequency,therefore, is primarily due to the coldertemperatures at flight level. Contrails wereobserved over all of the sites during July 1993when Tf generally exceeded 225 K. The non-zero contrail occurrence may be attributed tovariations in Tf over the month or to contrailsoccurring at higher levels. The variability in Tfcan probably account for July contrails over thenorthern sites but not over the southernlocations. The Hanson and Hanson (1995)calculations indicate that contrails can form attemperatures as warm as 244 K but only in verymoist conditions at much lower altitudes. Forexample, they found that the critical temperaturefor contrail formation at 500 mb for a lowbypass engine is -40°C at RH = 70 percent withrespect to liquid water. Thus, contrails may beformed when jet aircraft fly through moist layersat lower altitudes. However, this phenomenon
is not likely to occur during the summer. Asseen in Figure 16, there is still considerable airtraffic at 12 km. These flights at the higheraltitudes (colder temperatures) are probably thesource of contrail development over thesouthern sites during July.
Figure 22 shows the mean relativehumidities for each site during the 3 monthsconsidered earlier. These values generally rangebetween 33 and 42 percent. July appears to bethe driest month overall. The increased RHduring January and April is not much greaterthan the July values, however. Thus, thesehumidity values are not likely to explain muchof the variation in contrail occurrence. Despitethis apparent lack of humidity dependence, thefrequent co-occurrence of contrails and cirrus isa clear indication that contrails form more oftenwhen water vapor is more abundant. Theabsence of an association between R H andcontrail occurrence may reflect the oft expressedneed for better measurements of humidity in theupper troposphere (e.g., Schumann, 1995).
Earlier Contrail ObservationsAs noted earlier, contrail observations over
the US have been limited to either a small areaor time period except for the satellite study byDeGrand et al. (1990). That brief reportdiscussed the occurrence frequencies and foundthe greatest density over southern California,central Arizona and New Mexico, and over theMidwest. The current results are consistent withthat finding except for the southwestern USrelative maximum. This discrepancy may bedue to a change in flight patterns, the location ofLuke AFB (the only southwestern site) south ofthe primary flight corridors (Figures 1 and 17),or to differences in the interpretation ofpersistent and non-persistent contrails. It isunlikely that flight patterns have changedconsiderably since 1978 and, although this siteis not immediately underneath the flightcorridors, a considerable amount of fuel isexpended within 150 km of Luke AFB (seeFigures 1 and 17) resulting in the potential formany advected contrails. If non-persistentcontrails are included for Luke AFB, the overallfrequency of occurrence increases from 5.4percent to 14.6 percent, a value more consistentwith the fuel usage around Luke and close tovalues obtained for the nearest contrail site,Edwards AFB. Given that the frequency of non-persistent contrails over Luke AFB is more than
9
twice that of any other base and that satelliteobservers generally only see persistent contrails,it is possible that the observers at Luke AFBused a different criterion for determiningcontrail persistence than those at the other sites.Nevertheless, if the non-persistent contrails areincluded in the average, then the current resultsare qualitatively consistent with the earliersatellite study.
DeGrand et al. (1990) also found a seasonalvariation that differs from the present results.Their maximum frequency occurs duringOctober with a minimum during July. In Figure8, the maximum contrail frequency occursduring February or March regardless ofconsideration of indeterminate data. A minor,statistically insignificant, secondary peak isevident during October (Figure 8a) in the rawdata but it is less than both of the January andApril values. If indeterminate data are notconsidered, October ranks seventh for contrailoccurrence. This difference from the DeGrandet al. (1990) results is difficult to reconcile. It ispossible, but unlikely, that the difference is dueto sampling. Although only an average of 13months of data was used for the present study,the seasonal results are consistent for almost allsites over the US and with the meteorologicaltrends. Whether this seasonal cycle is typical ona climatological scale is a question that can onlybe addressed with further observations.
The discrepancy between the DeGrand etal. (1990) and the current results, however, mayarise from the differing viewing perspectives ofthe satellite and surface observers. The surfaceobserver can easily recognize a thin contrailshortly after its formation. If it undergoessignificant growth, however, it may beinterpreted as a cirrus by a surface observer butcould still be recognized as a contrail in infraredsatellite imagery, especially against a warmbackdrop, because of its linear characteristics.Thus, advecting contrails that have had time togrow may often be identified as cirrus clouds bya surface observer while counted as contrails ininfrared image analysis. It is possible that muchof the excess cirrus cloudiness detected in theautumn surface observations relative toclimatology (Figure 13b) may be contrail cirrusrather than natural cirrus. Such an interpretationwould be compatible with the results of Angell(1990) who found that cloudiness increases overthe US were greatest during autumn and weremost likely due to thin cirrus.
Conversely, contrails may be easilydistinguished against a blue sky from the surfacebut may be difficult to detect in morning andevening satellite imagery when the backgroundis cold. Because contrails are optically thin,they provide minimal contrast in infraredimagery unless the background is significantlywarmer than the cloud. Bispectral brightnesstemperature difference techniques (e.g., Lee,1989) rely more on the contrail's small particlesize and should be more effective for detectingcontrails in low contrast conditions. Becausethe Degrand et al. (1990) data were limited to asingle infrared channel, it is probable that manycontrails were not detected during winter andearly spring when the background issignificantly colder than it is during summer andautumn. This contrast problem would beexacerbated by the lack of DMSP satellite dataduring the afternoon when the contrast betweenclouds and the surface is greatest during allseasons. Furthermore, because the satelliteresolution is between 1 and 8 km, only thelargest persistent contrails can be detected. It ispossible that the differences in the seasonalcycle of contrails are due to a combination ofviewing perspective and contrail growth.Coordinated surface and satellite analyses wouldbe needed to better reconcile the differences.
Diurnal VariabilityBecause the observations were limited to
sunlit hours , the diurnal variations areincomplete and the mean frequencies may be inerror. Commercial aircraft frequently operate atnight over the US, especially before localmidnight and after 0600 LT. Contrails fromearly morning flights will be missed in theobservations during the winter months, inparticular. Inclusion of nocturnal observationscould change the results for the diurnal statisticsand would affect the overall mean frequencies.Because air traffic is generally heavier duringthe daytime, the times of maximum contrailoccurrence found here are probably accurate formost of the sites. The minimum hourlyfrequencies, however, would probably be lowerif 24-hour observations were used. For the samereasons, the mean contrail frequencies would besmaller than the current daytime values.
The diurnal maxima seen in Figure 11 mayreflect, to some extent, the timing of flights overthe US. Contrails observed over a particular site
10
are probably due to flights that originated orterminated at least one half hour from the sitebecause of the time needed to reach or descendfrom cruising altitude. Primary morningmaxima over the east (75-85°W) suggest thatmost flights commence early in the day before0800 - 0900 LT. The secondary afternoonmaxima are probably the result of lateroriginating and the arrival of eastbound flights.A similar breakdown of flights occurs over thewest coast with a mixture of afternoon andprimary maxima. The number of long distanceflights in either direction plus the north-southtraffic would shift both the primary andsecondary peaks to the late morning and lateafternoon over the center of the country. Theexception to this general pattern is Loring AFBwhich is primarily affected by US-European airtraffic. Much of the eastbound traffic originatesduring the late afternoon and early evening formorning arrival while the westbound flightsarrive earlier in the day. While the connectionbetween flight times and contrail occurrence iscomplex and cannot be fully explained here, theobservations are consistent with the generalconstraints imposed by commercial air traffic.A complementary analysis of satellite datacovering the entire day would help complete thedepiction of contrail diurnal variability.
Other ConsiderationsRelying on the interpretation of surface
observers, these data are subject to some errorsbased on the judgment of a particular observer.Distinguishing a contrail from a natural cloudcan be difficult a short time after the contrail'sformation. As a consequence, some bias towardunderestimation of contrail frequency isprobable because only those contrails that canbe confidently identified will be included in theobservations. It is unlikely that any oldercontrails missed in the statistics will be offset bycirrus clouds mistaken as contrails. Thethreshold between the persistent and non-persistent contrails is also subjective to somedegree and will result in uncertainty in theactual ratio of persistent to non-persistentcontrails. It was noted earlier that the highincidence of non-persistent contrails over LukeAFB compared to other sites may be the resultof different criteria used for determiningpersistence.
The viewing conditions can also influencethe detection of contrails. Observations
represent the conditions at one instant.Therefore, the line of sight to contrails may beblocked at a given time by scattered or brokencloud conditions. In overcast situations, thiseffect is recognized in the indeterminatecategory, but it is possible that contrails canremain unobserved in other circumstances.Although Figure 15 indicates that persistentcontrails were most often seen during mostlycloudy conditions, it is possible that the actualfrequency is even greater due to line-of-sightobstruction. This effect would also causeadditional underestimation of the true contrailfrequency. In this analysis, it was assumed thatthe frequency in the indeterminate cases is thesame as in the determinate viewing conditions.This assumption has not been tested yet. Properaccounting for the cases in which the contrailsare potentially obscured will require analyses ofsatellite data coincident with the surfaceobservations.
Concluding Remarks
This paper provides the most completeinventory of contrail frequencies over the US todate. It is just the first step, however, inassessing the impact of aircraft condensationtrails on climate. Because only 1 year ofobservations was available, it is not possible tounequivocally conclude that this dataset is areliable climatology of US contrails. Data fromother years are needed to develop such aclimatological picture. Much additionalinformation is also required to confidentlyestimate the radiative effects of contrails.Statistics regarding the lifetimes, areal coverage,optical properties, and advection of contrails andcontrail cirrus are essential to properlycharacterize changes in the mean radiationfields. Assessment of climate change due toincreasing air traffic appears feasible, however,because of the strong relationship foundbetween fuel usage and contrail occurrence andthe consistency between seasonal meteorologyand contrails. The contrail dataset presentedhere can be exploited to refine the relationshipsbetween contrails and fuel consumption andmeteorology. Correlations between temperatureand humidity from soundings coincident withhourly contrail observations will be critical toempirically quantify the trends found here. A
11
more detailed seasonal analysis of fuel use andcontrail frequency could be useful forsimulating contrail occurrence over a givenlocation in the US. With supplementalnocturnal data, it may be possible to realisticallysimulate the diurnal cycle as well. The hourlyobservations can also serve as validation datafor coincident satellite retrievals of contrailoccurrence. Combination of satellite retrievalsand these surface observations will be requiredto completely depict the entire diurnal cycle.
Contrails have become a prevalent featureof American skies. The relationshipsestablished here indicate that they will becomeeven more common in the future as airlineservice expands. Because contrails add directlyto cloud cover, they will affect the radiationbudget at some magnitude. Even if the impactis determined to be small on a global scale, thelocal effects may still be substantial. Thus, it isimportant to determine the relationshipsbetween contrail frequency and changes incloud cover. This surface analysis of contrailsshould also be repeated in a few years when airtraffic has increased significantly to detect anychanges in contrail occurrence and to test anyprognostication schemes developed from thisdataset. Commercial air traffic is growingworldwide with the potential for an increase incontrails over many areas outside of the US. Tofully assess contrails on a global scale, theirdetection and reporting should be made aroutine part of standard meteorologicalobservations. Because they are a distinct type ofcloud, they could easily be included as part ofthe cloud type codes currently used in the globalmeteorological observing system. An accurateevaluation of the climatic impact of contrailswill require an effort that combines surface andsatellite observing systems.
References
Angell, J. K. 1990: Variation in United StatesCloudiness and Sunshine Duration Between1950 and the Drought Year of 1988. J .Climate, 3, 296-308.
Angell, J. K.; Korshover, K. J.; and Cotton, G.F. 1984: Variation in United StatesCloudiness and Sunshine. J. Climate Appl.Meteorol., 23, 752-761.
Appleman, H., 1953: The Formation of ExhaustCondensation Trails by Jet Aircraft. Bull.Am. Meteorol. Soc., 34, pp. 14-20.
Bakan, S.; Betancour, M.; Gayler, V.; andGrassl, H. 1994: Contrail Frequency OverEurope From NOAA-Satellite Images. Annal.Geophys., 12, 962-968.
Baughcum, S. L. 1996: Subsonic AircraftEmission Inventories. Atmospheric Effects ofAviation: First Report of the SubsonicAssessment Project, NASA RP-1385, pp. 15-29.
Baughcum, S. L.; Metwally, M.; Seals, R. K.;and Wuebbles, D. J., 1993: EmissionsScenarios Development: CompletedScenarios Database. Atmospheric Effects ofStratospheric Aircraft: A Third ProgramNASA RP-1313, pp. 185-208.
Carleton, A. M.; and Lamb, P. J.: 1986: JetContrails and Cirrus Cloud: A FeasibilityStudy Employing High-Resolution SatelliteImagery. Bull. Am. Meteorol. Soc., 67, 301-309.
Changnon, S. A. 1981: Midwestern Sunshineand Temperature Trends Since 1901: PossibleEvidence of Jet Contrail Effects. J. Appl.Meteorol., 20, pp. 496-508.
DeGrand, J. Q.; Carleton, A. M.; and Lamb, P.J. 1990: A Mid-Season Climatology of JetCondensation Trails From High-ResolutionSatellite Data. Proc. of the AMS 7th Conf. onAtmos. Radiation, San Francisco, CA, July23-27, 309-311.
Engelstad, M.; Sengupta, S. K.; Lee, T.; andWelch, R. M. 1992: Automated Detection ofJet Contrails Using the AVHRR SplitWindow. Intl. J. Remote Sensing, 13, 1391-1412.
Hahn, C. J.; Warren, S. G.; London, J.; Chervin;and Jenne, R. L., 1984: Atlas of SimultaneousOccurrence of Different Cloud Types OverLand. NCAR Technical Note TN-241+STR,Boulder, Colorado, 212 pp.
Hahn, C. J.; Warren, S. G.; London, J.; Chervin;and Jenne, R. L., 1986: Global Distribution of
12
Total Cloud Cover and Cloud Type AmountsOver Land. NCAR Technical Note T N-273+STR, Boulder, Colorado, 288 pp.
Hahn, C. J.; Warren, S. G.; and London, J. 1995:The Effect of Moonlight on Observation ofCloud Cover at Night, and Application toCloud Climatology. J. Climate, 8, pp. 1429-1446.
Hanson, H. M.; and Hanson, D. M. 1995: AReexamination of the Formation of ExhaustCondensation Trails by Jet Aircraft. J. Appl.Meteorol., 34, pp. 2400-2405.
Joseph, J. H.; Levin, Z.; Mekler, Y.; Ohring, G.;and Otterman, J. 1975. Study of ContrailsFrom the ERTS 1 Satellite Imagery. J .Geophys. Res., 80, 366-372.
Karcher, B., 1994: Transport of ExhaustProducts in the Near Trail of a Jet EngineUnder Atmospheric Conditions. J. Geophys.Res., 99, pp. 14,509-14,517.
Kuhn, P. M. 1970: Airborne Observations ofContrail Effects on the Thermal RadiationBudget. J. Atmos. Sci., 27, 937-942.
Lee, T. F. 1989: Jet Contrail IdentificationUsing the AVHRR Infrared Split Window. J.Appl. Meteorol., 28, 993-995.
Liepert, B.; Fabian, P.; and Grassl, H. 1994:Solar Radiation in Germany - ObservedTrends and an Assessment of Their CausesPart I: Regional Approach. Beitr. Phys.Atmosph., 67, 15-29.
Murcray, W. B., 1970: On the Possibility ofWeather Modification by Aircraft Contrails.Mon. Wea. Rev., 98, pp. 745-748.
Sadler, J. C., 1977: The Upper TroposphericCirculation Over the Tropics, Part II -Statistics. University of Hawaii TechnicalReport UHMET-77-02 , Department ofMeteorology, University of Hawaii,Honolulu, NSF Grant No. GA-36301, April,24 pp.
Seaver, W. L.; and Lee, J. E. 1987: A StatisticalExamination of Sky Cover Changes in the
Contiguous United States. J. Clim. Appl.Meteorol., 26, pp. 88-95.
Schumann, U. 1996: On Conditions for ContrailFormation From Aircraft Exhausts.Meteorol. Zeitschrift, 5, pp. 4-23.
Schumann, U.; and Wendling, P. 1993:Determination of Contrails From SatelliteData and Observational Results. Air Trafficand the Environment, Proc. of DLR Intl.Colloquium, Bonn, Germany, Nov. 15-16,Springer-Verlag, NY, 138-153.
Scorer, R. S., 1972: Clouds of the World. David& Charles (Publishers) Ltd., 176 pp.
Thompson, A. M., Friedl, R. R.; and H. L.Wesoky, 1996: Introduction and Overview.Atmospheric Effects of Aviation: First Reportof the Subsonic Assessment Project, NASARP-1385, pp. 1-14.
Weber, G.-R. 1990: Spatial and TemporalVariation of Sunshine in the Federal Republicof Germany. Theor. Appl. Climatol., 41, 1-9.
Eglin
Hill
Cairns
Langley
Offut
Whiteman
Loring
Luke
Minot
Mt. Home
Kelly
Wright-Pat.
Tinker
Barksdale
Beale
McClellan
Fairchild
Griffis
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
Edwards
Figure 1. Contrail observation network, comprised of 19 U.S. Air Force Bases and Army AirFacilities.
~
Cirrus present• Contrails, Persistant
¯
Contrails, Not Persistant
˘
No Contrails
?
Indeterminant Contrails
1~ ? ? ? ? ? ? ~? ~˘ ~? ~ ~ ~
2~˘
~ •~˘
~ • ~ •~˘
~ • ~ •~? ? ?
3˘ ˘ ¯
•˘
•? ? ? ? ˘
4? ? ? ? ? ? ? ? ?
5? ? ? ? ? ? ? ? ?
6˘ ¯ ¯ ˘ ˘ ˘ ˘ ˘ ˘ ~˘ ~ ~
7? ~¯ ~˘ ~? ~˘ ~? ˘ ~˘ ~¯ ~¯ ~
8~ ~ ? ? ˘ ˘ ˘ ? ˘ ˘ ˘ ~˘ ~
9 •¯
• • • •? ?
• •1 0
~ ~
~ • ~ • ~ •~? ~? ~? ~? ~? ?
1 1? ? ? ? ˘ ? ? ? ?
1 2˘ ˘
1 3~ ~ ~ ~
~ •~¯ ~˘ ~˘ ~¯ ~¥ ~˘ ~˘ ~˘
~ •1 4
~ ~ ~ ~ ~ ~ ~
~ • ~ •˘ ˘ ˘ ˘ ˘ ? ?
1 5? ? ? ? ? ? ? ? ?
1 6? ? ? ? ? ? ? ? ? ?
1 7~ ~ ~ ~ ~ ? ˘ ¯ ¯ ¯ ~¯ ~˘ ~¯ ~¯
~ • ~ •~ ~
1 8~ ~? ? ? ? ? ~? ~? ? ? ?
1 9~ ~ ~ ~ ˘ ~˘ ˘ ˘ ˘ ~˘ ~˘ ~˘ ~¯ ~˘
2 0~ ~ ~ ~ ~ ˘
~ • ~ • ~ •~? ~? ~? ~? ?
2 1? ? ? ? ˘ ˘ ˘ ˘ ˘
2 2~ ~˘ ~˘ ~˘ ˘ ˘
•? ? ?
2 3? ? ? ? ? ? ? ? ˘
2 4~ ~ ~ ~ ~ ˘
• •?
• •˘ ˘ ~?
~ •~ ~
2 5~ ~ ~ ˘ ? ? ? ? ? ? ? ˘ ˘
2 6~ ? ?
2 7~ ~ ~ ~ ~ ~ ~ ~ ~ ? ?
2 8? ? ? ? ? ? ? ? ?
2 9? ? ? ? ? ? ? ? ?
3 0~ ~ ~ ~ ~ ˘ ˘ ˘ ˘ ˘ ˘ ˘ ˘ ~¯ ~ ~ ~
3 1~ ~ ~ ~ ~ ˘ ˘ ˘ ˘ ˘ ¯ ˘ ¯ ˘ ˘
1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 2 1 2 2 2 3 2 4
UTC
Day
Figure 2. Coded summary of Langley AFB December 1993 contrail and cirrus observations.
Sheet1
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
60 80 100
120
140
160
180
200
220
240
260
280
300
320
340
360
380
400
420
440
460
480
Day Number (3/93 - 4/94)
Contrails, Persistent Contrails, Persistent, w/CirrusContrails, Not Persistent Contrails, Not Persistent, w/CirrusNo Data Indeterminate ContrailsIndeterminate Contrails, w/Cirrus No ContrailsNo Contrails, w/Cirrus
Page 1
Figure 4. Summary of monthly observations for Langley AFB, Virginia from January 93 to May 94:(a) relative frequency of contrails and cirrus, (b) relative frequency and persistence of contrails, and (c)relative frequency of contrails and cirrus with indeterminate data removed.
J F M A M J J A S O N D J F M A M0
0.2
0.4
0.6
0.8
1F
requ
ency
Month
Contrails, Persistent
Contrails, Persistent, w/Ci
Contrails, Not Persistent
Contrails, Not Persistent, w/Ci
No Contrails
No Contrails, w/Ci
Indeterminate Contrails
Indeterminate Contrails, w/Ci
J F M A M J J A S O N D J F M A M0
0.2
0.4
0.6
0.8
1
Fre
quen
cy
Month
Contrails, Persistent
Contrails, Not Persistent
No Contrails
Indeterminate Contrails
J F M A M J J A S O N D J F M A M0
0.2
0.4
0.6
0.8
1
Fre
quen
cy
Month
Contrails, Persistent
Contrails, Persistent, w/Ci
Contrails, Not Persistent
Contrails, Not Persistent, w/Ci
No Contrails
No Contrails, w/Ci
a)
b)
c)
J F M A M J J A S O N D J F M A M0
0.2
0.4
0.6
0.8
1F
requ
ency
Month
0%
10% to 20%
30% to 40%
50% to 60%
70% to 80%
90% to 100%
7 8 9 101112131415161718192021222324 1 2 3 4 5 60
0.2
0.4
0.6
0.8
1
Fre
quen
cy
UTC
Figure 5. Summary of cloud cover for Langley AFB, Virginia from January 93 to May 94: (a) relativefrequency of cloud cover, (b) diurnal cycle of cloud cover relative frequency centered at local noon,and (c) monthly mean cloud cover.
a)
b)
c)
J
JJ
J J J J
J J
J
J F M A M J J A S O N D J F M A M0
102030405060708090
100
Clo
ud C
over
(%
)
Month
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B
B B
Min
ot
Fai
rchi
ld
Lorin
g
Grif
fis
Mt.
Hom
e
Offu
t
Hill
Wrig
ht P
at.
Bea
le
McC
lella
n
Whi
tem
an
Lang
ley
Tin
ker
Edw
ards
Luke
Bar
ksda
le
Cai
rns
Egl
in
Kel
ly
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Fre
quen
cy
BASE
Figure 6. Mean and standard deviations of combined persistent and non-persistent monthly contrailfrequencies, sites ordered North to South according to lattitude.
Loring
Griffis
Eglin
Minot
Hill
Beale
Tinker
Kelly
Whiteman
Edwards
McClellan
Fairchild
Wright Patterson
Barksdale
Offut
Mountain Home
Loring
Griffis
Langley
Minot
Hill
Beale
Cairns
TinkerLuke
Kelly
Whiteman
Edwards
McClellan
Fairchild
Wright Patterson
Barksdale
Offut
Mountain Home
0 5 10 15 20 25 30 35 40%
Loring
Griffis
Langley
Minot
Hill
TinkerLuke
Kelly
Whiteman
Edwards
McClellan
Fairchild
Wright Patterson
Barksdale
Offut
Mountain Home
Loring
Griffis
Langley
Minot
Hill
Beale
Cairns
TinkerLuke
Kelly
Whiteman
Edwards
McClellan
Fairchild
Wright Patterson
Barksdale
Offut
Mountain Home
Figure 7. Comparison of persistent contrail frequency with indeterminate data excluded for several months. a) January. b) April. c) July. d) October.
a)
b)
c)
d)
January
April
July
October
B
BB B
B
B
BB B
BB
B
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
ME
AN
CO
NT
RA
IL F
RE
QU
EN
CY
MONTH
a)
b)
Figure 8. Monthly mean persistent contrail frequency for 19 sites. Bars indicate standard deviations: a) Alldaytime data. b) Indeterminate data excluded.
B
B B
B
B
B
BB B
B B
B
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
ME
AN
CO
NT
RA
IL F
RE
QU
EN
CY
MONTH
Figure 9. Summary of hourly contrail and cirrus observations from Langley AFB, Virginia, centered atlocal noon: (a) number of observations, (b) relative frequency of occurrence, and (c) relative frequencyof occurrence with indeterminate data removed.
7 8 9 101112131415161718192021222324 1 2 3 4 5 60
50
100
150
200
250
300H
ours
of O
bser
vatio
n
UTC
Contrails, Persistent
Contrails, Persistent, w/Ci
Contrails, Not Persistent
Contrails, Not Persistent, w/Ci
No Contrails
No Contrails, w/Ci
Indeterminate Contrails
Indeterminate Contrails, w/Ci
7 8 9 101112131415161718192021222324 1 2 3 4 5 60
0.2
0.4
0.6
0.8
1
Fre
quen
cy
UTC
7 8 9 101112131415161718192021222324 1 2 3 4 5 60
0.2
0.4
0.6
0.8
1
Fre
quen
cy
UTC
a)
b)
c)
6 7 8 9 10 11 12 13 14 15 16 17 180
1
2
3
4
5
6
7
FR
EQ
UE
NC
Y
LOCAL TIME (hr)
PRIMARY
SECONDARY
12 13 14 15 16 17 18 19 20 21 22 23 24 250
1
2
3
4
5
6
7F
RE
QU
EN
CY
UTC (hr)
PRIMARY
SECONDARY
(a)
(b)
Figure 10. Times of maximum contrail occurrence for the 19 study sites.
áá
á
á
áá á
á
á
á
á
á
á
á
á
á
á
á
á
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
H
6
8
10
12
14
16
18
20
125 115 105 95 85 75 65
LOC
AL
TIM
E (
hr)
WEST LONGITUDE (°)
á PRIMARY
H SECONDARY
Figure 11. Longitudinal dependence of times of maximum contrail occurrence.
B
B
B
B
B
BB
BB
B
B
B
B
É ÉÉ
É
É É
É
É ÉÉ
É
É
É
20
30
40
50
60
70
0 2 4 6 8 10 12
CLO
UD
AM
OU
NT
(%
)
MONTH
Fig. 12. Comparison of U.S. climatological average and contrail-data monthly mean cloud cover.
BB
B
BB
B
B
B
B
B
B
B
É É
É É É
É É É
É É É
É
30
40
50
60
70
0 2 4 6 8 10 12
CIR
RU
S F
RE
QU
EN
CY
(%
)
MONTH
B
B
B
B
BB
B
B
B
B
B
B
B
B
B
B
B
Jan-
93F
eb-9
3M
ar-9
3A
pr-9
3M
ay-9
3Ju
n-93
Jul-9
3A
ug-9
3S
ep-9
3O
ct-9
3N
ov-9
3D
ec-9
3Ja
n-94
Feb
-94
Mar
-94
Apr
-94
May
-94
30
40
50
60
70C
IRR
US
FR
EQ
UE
NC
Y (
%)
MONTH
a) For all AFB contrail sites
b) Monthly averages
AFB sites
US seasonal mean(1971-80)
Figure 13. Frequency of cirrus occurrence over the contrail-datasites and the U.S. climatological mean.
B
B
B B
BB
BB B
B
B
B
E
E
E
E
E
E
E
E
E
E
E
E
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec0
0.05
0.1
0.15
0.2
Per
sist
ent C
ontr
ail F
requ
ency
Month
B PC w/o CIrrus
E PC w/Cirrus
Figure 14. Monthly mean persistent contrail frequency with and without cirrus for all 19 surface sites.
J
J JJ
J
J
J
J J
JJ
0 10 20 30 40 50 60 70 80 90 1000
0.05
0.1
0.15
0.2
Mea
n C
ontr
ail F
requ
ency
Cloud Amount (%)
J
J
J
J
J
JJ
J
J
J J
0 10 20 30 40 50 60 70 80 90 1000
0.05
0.1
0.15
0.2
Mea
n C
ontr
ail F
requ
ency
Cloud Amount (%)
J
J
J J
J
J
J
J J
J J
0 10 20 30 40 50 60 70 80 90 1000
0.05
0.1
0.15
0.2
Mea
n C
ontr
ail F
requ
ency
Cloud Amount (%)
Figure 15. Contrail frequency as a function of cloud amount: a) Persistent contrails only. b) Non-persistentcontrails. c) All contrails.
a)
b)
c)
0
2
4
6
8
10
12
14
16
18
20
0 5 10 15 20
ALT
ITU
DE
(km
)
FUEL (106 kg)
Fig. 16. Vertical distribution of annual mean fuel use for 3° latitude-longitude boxescentered over the 19 U.S. contrail observation sites based on May 1990 data fromBaughcum et al. (1993).
Figure 17. Fuel use estimate for 1990 for altitud
es above 7 km.
Loring
Griffis
Langley
Eglin
Minot
Hill
Beale
Cairns
TinkerLuke
Kelly
Whiteman
Edwards
McClellan
Fairchild
Wright Patterson
Barksdale
Offut
Mountain Home
0 5 10 15 20 25 30 35 40%
Figure 18. Comparison of annual mean persistent contrail frequency excluding indeterminate data for 19 surface observation sites.
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
JJ
J
0
0.1
0.2
0.3
0.4
0.5
0 50 100 150 200 250
Per
sist
ent C
ontr
ail F
requ
ency
Fuel Use (106lbs/yr)
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
J
JJJ
0
0.1
0.2
0.3
0.4
0.5
0 50 100 150 200 250
Tot
al C
ontr
ail F
requ
ency
Fuel Use (106lbs/yr)
ct = 0.00176 f
r = 0.78
a)
b)
cp = 0.00127 f
r = 0.73
Figure 19. Contrail frequency as a function of fuel use: a) Total contrail frequency. b) Persistent contrailfrequency.
B
B
B
B
B
B
B
B
B B
BB
BB B
BB B
B
J
J
J
J
J
J
J
J
J J
J
J
J
JJ J
JJ
J
HH
H
H
H
H
H
H
H H
H H
HH
H
HH
H H
Min
ot
Fai
rchi
ld
Lorin
g
Grif
fis
Mt.
Hom
e
Offu
t
Hill
Wrig
ht P
at
Bea
le
McC
lella
n
Whi
tem
an
Lang
ley
Tin
ker
Edw
ards
Luke
Bar
ksda
le
Cai
rns
Egl
in
Kel
ly
8
9
10
11
12
13
14
15
16
17
18
Hei
ght(
km)
AFB
B July 93 J Jan. 94 H Apr. 94
Figure 20. Variation of mean tropopause altitude for July 1993, January 1994, and April 1994. Sitesordered North to South according to latitude.
B
B
B
B
B
B
B
B
B B
B
BB
B
BB
B B
B
J
J
JJ
JJ
J
J
J J
J
JJ
J
J
J
J
J
J
H
H
H
HH
H
H
H
H H
H HH
H
H
H
H
H H
Min
ot
Fai
rchi
ld
Lorin
g
Grif
fis
Mt.
Hom
e
Offu
t
Hill
Wrig
ht P
at
Bea
le
McC
lella
n
Whi
tem
an
Lang
ley
Tin
ker
Edw
ards
Luke
Bar
ksda
le
Cai
rns
Egl
in
Kel
ly
210
215
220
225
230
235
240
245
250
Tem
pera
ture
(K)
AFB
B July 93 J Jan. 94 H Apr. 94
Figure 21. Comparison of temperatures at 10.5 kilometer altitude for July 1993, January 1994, and April 1994.Sites ordered North to South according to latitude.
B
BB
B BB
B
B
B B
B
B
B
B
B
B
B
B
B
J
JJ
JJ
J
J
J
J JJ
J
J
J J
J
J
J
J
H
H
H
H
H
H
H
H
H H
HH
H
H
H H
H
H
H
Min
ot
Fai
rchi
ld
Lorin
g
Grif
fis
Mt.
Hom
e
Offu
t
Hill
Wrig
ht P
at
Bea
le
McC
lella
n
Whi
tem
an
Lang
ley
Tin
ker
Edw
ards
Luke
Bar
ksda
le
Cai
rns
Egl
in
Kel
ly
25
30
35
40
45
50
Rel
ativ
e H
umid
ity (
%)
AFB
B July 93 J Jan 94 H April 94
Figure 22. Comparison of mean relative humidities at 10.5 kilometer altitude for July 1993, January 1994, andApril 1994. Sites ordered North to South according to latitude.
