88 JOURNAL OF APPLIED METEOROLOGY VOLUME 34Aircraft Icing Measurements in East Coast Winter Storms STEWART G. COBER, GEORGE A. ISAAC, AND J. w. STRAPPCloud Physics Research Division, Atmospheric Environment Service, Downsview, Ontario, Canada (Manuscript received 21 January 1994, in final form 21 April 1994)ABSTRACTAnalysis of the aircraft icing environments of East Coast winter storms have been made from 3 1 flights duringthe second Canadian Atlantic Storms Program. Microphysical parameters have been summarized and are com-pared to common icing intensity envelopes and to other icing datasets. Cloud regions with supercooled liquidwater had an average horizontal extent of 4.3 km, with average droplet concentrations of 130 ~m-~, liquid watercontents of 0.13 g m-3, and droplet median volume diameters of 18 pm. In general, the icing intensity observedwas classified as light, although moderate to severe icing was observed in several common synoptic situationsand several cases are discussed. Freezing drizzle was observed on four flights, and represented the most severeicing environment encountered.1. IntroductionThere were two aircraft icing research objectives ofEast Coast winter storms are well known for severeweather. On the Canadian Atlantic coast they can pro-duce gale force winds, heavy precipitation rates, andlarge quantities of freezing rain and freezing drizzle.Consequently they can represent a significant hazardto communications, transportation, and aviation. Tobetter understand such storms, the Canadian AtlanticStorms Program (CASP) was undertaken by the At-mospheric Environment Service (AES). The projectconsisted of two parts; CASP I, from 15 January to 15March 1986, based out of Halifax, Nova Scotia (Stewartet al. 1987); and CASP 11, from 15 January to 15 March1992, based out of St. John's Newfoundland (Stewart1991).The CASP I1 project was designed to study the me-soscale characteristics of East Coast winter storms. Theprimary research aircraft was a National ResearchCouncil (NRC) Convair-580, which performed flightsover the Atlantic Ocean within a radius of 800 km ofthe operating base in St. John's Newfoundland (Fig.1). Consequently, the project was ideally suited formaritime aircraft icing research. This encouragedfunding for icing research from the Canadian NationalSearch and Rescue Secretariat, and led to the involve-ment of Boeing Commercial Airplane Group of Seattleand Airbus Industrie of France. Boeing and Airbus wereinterested in characterizing the maritime Atlantic icingenvironment to help substantiate the aircraft require-ments for cross-oceanic flights of twin-engined aircraft(Patnoe et al. 1993).Corresponding author address: Dr. Stewart Cober, Cloud PhysicsResearch Division, ARMP, Atmospheric Environment Service, 4905Dufferin Street, Downsview, Ontario, Canada M3H 5T4.the AES: to investigate the microphysical and dynam-ical properties of East Coast winter storms and the cor-responding potential for aircraft icing within suchstorms; and to determine techniques for validating andimproving icing forecast models. The former repre-sented an extension of the microphysical measurementsmade during CASP I (Isaac 199 1). The microphysicalproperties of icing conditions measured during CASPI1 are reported here. These will be compared to com-mon icing intensity envelopes and to icing measure-ments from other field projects.2. Icing envelopes and existing datasetsThe Federal Aviation Regulation Part 25 AppendixC (FAR 25-C) icing envelopes detail the probable max-imum icing conditions expected in winter storms. Theenvelopes were defined in the late 1940s and were basedon data obtained from approximately 1038 icing mea-surements across the United States. For certificationfor flight into icing conditions, aircraft must demon-strate the ability to fly in an environment characterizedby the maximum envelopes. Although the envelopeshave been criticized for various limitations (Newton1978; Sand et al. 1984; Jeck 1983), they are widelyused as a measure of the most severe icing conditions.An alternate set of envelopes was discussed by New-ton (1978), who defined the icing intensity in terms ofthe rate of accumulation (g cm-* h-') on a 7.6-cm (3in.) cylinder. These envelopes have a more physicalbasis than those in FAR 25-C and can be used to de-scribe a wider range of icing conditions.Measurements of icing from several field projectshave been summarized by Jeck (1983). Most of theprojects were conducted over the continental United0 1995 American Meteorological SocietyJANUARY 1995 COBER ET AL. 89Longitude (Degrees West)FIG. 1. Map of the Newfoundland area showing the approximate800-km return trip range of the Convair-580 research aircraft. Flightswere based out of St. John's, Newfoundland.States, with Jeck's analysis focused on summarizingthe icing environment below 10 000 ft above groundlevel. This dataset was characterized into a set of icingenvelopes by Masters (1983).A thorough analysis of icing events encountered bythe University of Wyoming Super King Air aircraftduring 109 1 h of field project flights has been given bySand et al. (1984). Again, most of the research wasconducted over the continental United States. Re-cently, the Winter Icing and Storms Project (Rasmus-sen et al. 1992) was conducted with the objective ofstudying the processes leading to the formation anddepletion of supercooled liquid water content (SLWC)and of improving icing forecasts. The project concen-trated on winter storms in Colorado.Relatively few measurements of icing conditionshave been reported for maritime winter storms. Con-sequently, the CASP I1 icing conditions reported herewill make a valuable addition to the existing icing da-tasets.3. Aircraft and instrumentationThe National Research Council Convair-580 was theprimary research aircraft during the CASP II field proj-ect. The aircraft was fully equipped for cloud physicsresearch, with most instruments mounted on pylonsunder the aircraft wings as shown in Fig. 2. Instrumentsapplicable to aircraft icing research are listed inTable 1.Calibration of AES and other King hot wire liquidwater content (LWC) probes in a high-speed wind tun-nel has been described by King et al. (1985). Throughcomparison with icing cylinder measurements, theydetermined that the King probes collectively had anFIG. 2. The National Research Council Convair-580 research air-craft. Measuring instruments can be seen mounted on pylons underthe wings.estimated error of k 15% for cloud droplets. Biter et al.(1987) have shown that the response ofthe King probeis poor for droplets larger than approximately 50 pmin diameter. King et al. (1985) further concluded thatthe probe response to liquid water is stable over longperiods, and that since this response can be calculateddirectly, the need for frequent wind tunnel calibrationsis eliminated. A wind tunnel calibration performed inApril 1993 (one year after the CASP II project) revealeddifferences between the AES short wire King probe andicing cylinders of less than 10% in the LWC range from0.3 to 1 .O g m-3. This provides an example of the long-term accuracy of the AES King probes, and allowsconfidence in the SLWC measurements taken duringCASP II. Removal of the instrument dry power wasperformed in a manner similar to that suggested byKing et al. (1978). Baseline drift resulting from im-perfect dry power removal was estimated by King etal. (1978) as less than 0.03 g m-3, which agrees withTABLE 1. Summary of instruments mounted on the NRC Convair-580 that were applicable to aircraft icing research. PMS designatesinstruments built by Particle Measuring Systems.Instrument NotesPMS King probePMS King probeJohnson-Williams LWC meterPMS FSSP lO0X extended rangeRosemount 87 IFA22 1 B icing detectorPMS 2DC MonoPMS 2DC GreyPMS 2DP MonoRosemount 858 gust probeCambridge dewpoint hygrometerRosemount temperature sensorReverse flow temperature sensorGPS, INS, Loran navigation systemsIcing accumulation cylinder (1 in.)Short-wire versionLong-wire versionSelectable to 5-95 pm25-800 pm25-1600 pm200-6400 pm2 sensorsNot on all flights90 JOURNAL OF APPLIED METEOROLOGY VOLUME 341.0 7ov0 0.63J 0.5 0.4? 0.3P0.20.10.0oJ0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 King Short LWC (g ma)FIG. 3. Comparison of liquid water content (LWC) measurementsof the two King probes. The best fit is given by KL = 0.0067+ 0.96(KS), where KL and KS represent the liquid water contents(g m-3) of the long- and short-wired probes, respectively. The cor-relation coefficient is 0.987 with a standard error of 0.01 5 g m-3. Thedata points shown represent 10-s averages of liquid water content.our observations. To minimize the errors caused bythis drift during research flights, the output was arti-ficially zeroed for FSSP (forward-scattering spectrom-eter probe) concentrations less than 6 droplets per cubiccentimeter. Figure 3 shows a comparison of the SLWCscalculated for the two AES King probes mounted sideby side on the Convair during CASP 11. The two probesagreed to within +-15% for 85% of the SLWC mea-surements, although the scatter is significantly higherfor LWCs lower than 0.1 g m-3. There is a small biasin the best fit of the order of 4% from unknown sources.The increased scatter at lower SLWCs is presumablycaused by uncertainties in the baseline removal and inthe response of the probes to ice crystals. Ice crystalconcentrations measured by the Particle MeasuringSystems (PMS) 2DP probe (nominally larger than 200pm) during CASP I1 commonly exceeded 30 L-', andcould cause both the King probes and the FSSP to giveresponses that were difficult to interpret. It was esti-mated that the King probe response in the presence ofice crystals tended to be less than 0.025 g m-3, althoughcould go as high as 0.1 g m-3. Measurements smallerthan 0.025 g m-3 were difficult to distinguish betweenSLWC and ice crystals and were ignored. The AESshort-wired King probe was used for all analysis in thisarticle. It was considered more accurate because it hada smaller baseline drift, and because the other Kingprobe incorporated some modifications and was flownin an experimental mode.The FSSP was calibrated frequently with glass beads.If calibrations revealed under- or oversizing, a uniformgain change in the response of the probe was assumed,and bin diameters were redefined from simple Miescattering calculations in a manner similar to that usedby the manufacturer to originally set up the probe.Calibration errors of this sort were usually systematicand largely represented a shift caused by an increasedbuildup of residue in the optics. The maximum di-ameter correction from this effect was 30% at 10 pm.Particle concentrations, which were usually low, werecorrected for dead time and coincidence followingBaumgardner et al. (1985). The error in measurementof droplet concentration has been estimated at &20%by Baumgardner (1983). A comparison between FSSPand King LWCs (Fig. 4) shows no indications of majormalfunction of the FSSP probe. The FSSP LWCs weresystematically higher than the King, with a scatter off25%, worsening at low LWCs. Most of the FSSPmeasurements of LWC agree within ?35% of the King,which is consistent with the error analysis of Baum-gardner (1 983). On occasion, the FSSP fogged duringdescent, which caused the FSSP LWC measurementto be significantly lower than that of the King. Con-versely, the FSSP response to ice crystals could causethe LWC measurement to be up to 10 times that ofthe King. FSSP response to crystals has been discussedby Gardiner and Hallett (1985). Figure 5 shows theFSSP response to an ice crystal concentration of 19L-' (as measured by a 2DP), where it is suspected thatno liquid droplets were present. Crystals tend to causeT- 0.7Eo 0.6y 0.5IL 0.4 0.3 0.2 0.1 0.0vJ2. . .. . . . ." - . .*...1 . .:. - .. . *. .. I. . . . . . .. .. , , .I. .. . . . . . . . . .0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 King Short LWC (g nT3)FIG. 4. Comparison of liquid water content (LWC) measurementsof the short-wired King probe and the FSSP. The best fit is given byFS = 0.045 + 1.08(KS) where FS is the liquid water content (g m-3)of the FSSP. The correlation coefficient is 0.85 with a standard errorof 0.06 g m-3.JANUARY 1995 COBERET AL. 91c3E5r".-X1 .o0.90.80.70.6.c00c 0.50 0.42a, 0.3.-.I-LL>.-4-+ 0.2a0.10.00 5 10 15 20 25 30 35 40 45 50 Bin Diameter (microns)FIG. 5. FSSP spectrum of liquid water content and concentrationsfor a 60-s measurement from 155759 UTC 14 March 1992. The yaxis represents the relative fraction ofthe FSSP bin with the maximumliquid water content or concentration (Le., the maximum concen-tration occurred in the first FSSP channel, whereas the maximumLWC occurred in the last FSSP channel). The weighting of the massspectra to higher diameters is indicative of the FSSP response to icecrystals. The average droplet concentration was 14 ~m-~, with a Kingprobe measured SLWC of 0.04 g m-3.relatively more counts in the higher FSSP channels(than for droplets), which leads to a mass spectrumweighted to the larger sizes. Figure 6 illustrates a caseof what is suspected to be mixed cloud, substantiatedby a 2DP concentration of 1.1 L-', and an indicationof icing from the Rosemount icing detector (RID). Thespectrum includes a conventional droplet spectrum anda suspected ice crystal mass mode at the larger sizes.All measurements of supercooled liquid water thatappeared to be contaminated by ice crystals were elim-inated from the dataset, since no accurate measure-ments of small SLWCs could be made under theseconditions. For each 10-s measurement of apparentSLWC, the magnitudes of the King SLWC and FSSPconcentration, and the shape of the FSSP spectra wereused to assess if the King and FSSP measurements werecontaminated by ice crystals. The RID was sometimesused to help distinguish between SLWC and ice crystals.Approximately 15% of the apparent SLWC measure-ments were considered to be dominated by crystal con-tamination, 5% were assessed as showing both dropletand crystal contributions, and 5% were undetermined.For the remaining 75%, crystal contamination wasconsidered to be within the experimental errors of themeasuring instruments and was consequently ignored.Although the RID was tested in a high-speed windtunnel, in a manner similar to Baumgardner and Rodi(1989), it was difficult to use their technique to interpretthe field measurements of icing intensity as a SLWC.The RID showed significantly different responses todifferent droplet distributions at temperatures between0" and - 10C, and was not always a sure indicator oficing. Temperature was measured by two deiced Rose-mount temperature probes and a reverse flow temper-ature probe, which normally agreed within 1 "C. SinceSLWCs were usually less than 0.5 g mP3, errors causedby in-cloud wetting (Lawson and Cooper 1990) areexpected to be less than 0.4OC.During some flights, the ice accumulation on a 1-in.-diameter icing rod was filmed. Ice accumulationcould be measured to +-2 mm, and the rod could beheated to shed accreted ice. The icing rod was veryuseful for estimating average liquid water contentsduring the heaviest icing events, when the reliabilityof some probes may have been reduced by the intensityof the icing.4. Icing datasetDuring CASP 11, 3 1 flights totaling 1 19 h were rel-evant to aircraft icing research. Fourteen of these flightswere specifically designed for icing studies. The 31flights could be divided roughly into two groups; systemflights (25) and low-level stratus flights (6). Systemflights included flights through synoptic-scaled systemssuch as warm fronts, cold fronts, and low pressure re-1 .o0.9C3.- E 0.7Z 0.65 o.82.c00c 0.58 0.4IL'a, 0.3.-w>.-.I-% 0.2ce0.10.00 5 10 15 20 25 30 35 40 45 50 Bin Diameter (microns)FIG. 6. FSSP spectrum for a 60-s average from 1725:08 UTC 20February 1992, where both a droplet and crystal signal are apparent.The droplet spectra is clearly shown by the droplet peak in mass andconcentration in the 5- 10-fim range. The weighting ofthe mass spec-tra in the 30-50-fim range is likely caused by the FSSP response toice crystals. The average concentration was 128 cm-3 with a Kingprobe measured SLWC of 0.14 g m-3.92 JOURNAL OF APPLIED METEOROLOGY VOLUME 34gions. These flights tended to consist of long pathsacross such a system, or shorter tracks back and forthat different altitudes across a region of interest. Thelow-level stratus flights studied stratus or stratocumulusclouds that formed over the ocean, and were designedeither for icing studies or for calibration of microwaveliquid water path algorithms from the Defence Mete-orological Satellite Program Special Sensor Microwave/Imager (similar to the study of Curry and Liu 1992).The low-level stratus clouds typically had bases between300 and 1200 m and tops at 600-2100 m, with nooverlying cloud formations.Data from each flight were analyzed in 1-s intervals(approximately 100 m of flight path). An interval wasconsidered to have SLWC if the static temperature wasequal to or less than 0.0"C and the SLWC was at least0.025 g m-3. The I-s data were averaged into 10- and300-s intervals. There were 3745 10-s intervals in whichthe average SLWC was at least 0.025 g m-3. This rep-resented 9.0% of the in-flight time during the 3 1 flights.In general, there were significant differences betweenthe system clouds and the low-level stratus clouds, withthe former tending to have lower droplet concentra-tions, larger droplet sizes, and larger crystal concentra-tions. The low-level stratus had larger droplet concen-trations presumably because they formed at the top ofthe boundary layer where the aerosol spectra containedmore cloud condensation nuclei. System clouds hadlarger crystal concentrations because they were moredeveloped (deeper) and had colder cloud-top temper-atures. Alternately, the system clouds may have hadlower droplet concentrations because their higher crys-tal concentrations may have caused more scavengingof droplets. System clouds may have had larger dropletmedian volume diameters (MVDs) because the averageflight track in system clouds was 1000-1 500 m abovecloud base, while the average flight track in low stratusclouds was 500 m above cloud base. It is well knownthat the MVD can increase with height above cloudbase (Pruppacher and Klett 1978), with a MVD changeof 5 pm for an altitude change of 1000 m not consideredunreasonable. Note that although there was a differencein the average MVDs for system and low-level stratusclouds, there was no evidence of an increase in theMVD with altitude when the data from several flightswas examined collectively. Figures 7a and 7b show ex-amples of droplet spectra that are characteristic of thesystem and low-level stratus clouds, respectively. Fur-ther comparisons of system and low-level stratus cloudswill be presented in the following sections.5. Analysisa. Icing environment statisticsA histogram of average SLWCs is shown in Fig. 8.The data are 10-s averages, which have correspondingaverage pathlengths of 0.97 km with a standard devia-tion of 0.13 km. The average SLWC was 0.13 g m-3,(a) 1.0 0.9C.-m.E 0.7z5 0.6(I: 0.50 0.42(u 0.3s 0.2 0.1 o.n5 Om8.I-O0_-4-LL>.-.I-a"-.-0 5 10 15 20 25 30 35 40 45 50Bin Diameter (microns)(b) 1.0 0.95.- E 0.7z5 0.6C5 o.8u-00c 0.58 0.4I;I(u 0.3rr 0.2a 0.1 0.0.-4->0.-+.I0 5 10 15 20 25 30 35 40 45 50Bin Diameter (microns)FIG. 7. (a) Example of an FSSP spectrum from a characteristic"system" cloud (60-s average from 165258 UTC 20 February 1992).The average droplet concentration was 48 ~m-~, with a Kmg measuredSLWC ofO. 14 g m-3. The 2DP concentration was 0.5 L-'. (b) Exampleof an FSSP spectrum from a characteristic low-level stratus cloud(60-s average from 2154:47 UTC IO March 1992). The droplet con-centration was 120 cm-', with a King-measured SLWC of 0.26g m-3.which was relatively independent of cloud type. Thestandard deviation in the average value of SLWC was0.1 1 g m-3 for system clouds and 0.07 g m-3 for low-level stratus clouds. Figure 8 also shows the cumulativepercentiles for SLWC measurements. The medianSLWC was 0.1 1 g m-3, with SLWCs exceeding 0.94g m-3 for 0.1 % of the measurements. The SLWC char-JANUARY 1995 COBER100300 I n250 1v)c4-E 200202 1503Y-O2 100E,z500cKa,$z3na,5m.-c-0.0 0.1 0.2 0.3 0.4 0.5Supercooled Liquid Water Content (g ma)FIG. 8. Histogram of 3745 supercooled liquid water content(SLWC) measurements from 31 flights. Each data point representsa 10-s average. The cumulative frequency is shown on the right yaxis. All measurements larger than 0.5 g m-3 are included in the lastbin.acteristics are similar to those reported by Sand et al.(1984) for winter Great Lakes and winter Californiaclouds, where the 0.1% percentile SLWC was 0.6 and1.2 g mp3, respectively. The relatively low maximumSLWCs result from subfreezing cloud-base tempera-tures and limited convective activity, both of whichare characteristics of East Coast winter storms.The drop in frequency at SLWCs below 0.05 g m-3is not an intuitively expected result and is likely causedby the limitations in instrument measurements de-scribed earlier. The frequency must drop to zero at0.025 g m-3 since no values below this were recognizedin this study. As well, it is likely that many of the mea-surements rejected because of ice crystal contaminationfall into this SLWC range. Frequencies of cloud watercontents less than 0.05 g m-3 have been given by Mazin( 1994) at approximately 40% for measurements be-tween -5" and -10C. Consequently, it is importantto stress that the statistics reported here apply only toregions with average SLWCs larger than 0.025 g mP3.Figure 9 shows a histogram of the droplet concen-trations for low-level stratus clouds and system clouds.The average droplet concentration was 96 cm-3 forsystem clouds and 174 cmP3 for low-level stratus cloudswith standard deviations of 9 1 and 137 cmP3, respec-tively. For system clouds the median concentration was64 cmP3, which is generally characteristic of maritimeclouds (Pruppacher and Nett 1978). These observa-tions are similar to those reported by Isaac (199 l), whoreported medians from 72 to 90 ~m-~. The medianconcentration is significantly smaller than the value of1 10 cm-3 reported by Sand et al. (1984). The differencelikely results from the fact that most of Sand's mea-surements were taken in continental clouds.ET AL. 934oo m loo75 c$5C0) n50 .$cm3-02500 100 200 300 400 500 Droplet Concentration (crn3)FIG. 9. Histogram of droplet concentrations, averaged over 10-sintervals. The all-flights histogram includes the system flights andthe low-level stratus flights. All concentrations larger than 500 cm-3are included in the last bin. The cumulative percentile curve representsdata from all flights.A histogram of the droplet median volume diameter(MVD) is given in Fig. 10. The MVD is the diameterfor which half the liquid water content occurs in smallerdroplets. It was calculated from the FSSP data, whichwas set on the 3-47-pm range for 84% of the mea-surements. The average MVD was 18 pm for all clouds,20 pm for the system clouds, and 16 pm for the low-level stratus clouds. These results are similar to those8007000 5 10 15 20 25 30 35 40 45 Median Volume Diameter (microns)FIG. 10. Histogram of median volume diameters, averaged over 10-s intervals.JOURNAL OF APPLIED METEOROLOGY VOLUME 34-'. -20 6 Ta < :lOC. Ta < -20OCooc940.8& 0.62" 0.5E 0.4.-0.308 0.2..PA.Moderate6 g cm" hr'Light 1 g an2 hr'4 0.1v)0.00 5 10 15 20 25 30 35 40Median Volume Diameter (microns)FIG. 1 I. Comparison of SLWC versus MVD for each IO-s average.The icing intensity envelopes of Newton (1978) are superimposedfor comparison. Newton's envelopes are calculated assuming an air-craft velocity of 100 m s-', the droplet collision efficiencies of Lang-muir and Blodgett (1946), and a collection efficiency of I.of Isaac (I 99 I), who reported median MVDs of 15-17 ym, and to measurements in winter Californiaclouds by Sand et al. (1984), where the median MVDwas 20 pm. The median concentration of droplets withdiameters from 40 to 47 pm was 20 droplets per literfor all cloud types observed. As discussed by Prup-pacher and Klett ( 1978), collision-coalescence modelshave shown that similar concentrations could lead tothe development of drizzle and precipitation-sizeddrops through a "warm rain" process. This may explainthe observations of freezing drizzle in several of theCASP I1 flights (section 5d).Ninety percent of the SLWC was observed between0" and -12"C, although only 55% of the flight pathswere at temperatures warmer than -12C. The tem-perature data cannot be considered random becausethe flight levels were often chosen to correspond toparticular temperatures. For example, eleven of the ic-ing flights were designed to fly between -5" and-15C. Icing was measured during 78 10-s intervalsbetween -20" and -35C. This icing was always as-sociated with thin layers of stratus cloud, or with thinlayers of SLWC at the tops of glaciated clouds. Thincloud-top SLWC layers were frequently observed attemperatures colder than -10C (coldest at -35"C),with the cloud below each layer typically fully glaciated.Such layers have been previously observed (Hobbs andRangno 1985; Ranber and Tokay 1991). These layerswere estimated to be less than 100 m thick and pre-sumably are of minimum danger to aircraft.b. Icing envelope comparisonFigure 1 1 shows a plot of the average SLWC versusthe average median volume diameter for each 10-s in-terval in which SLWC was measured. Superimposedon Fig. I 1 are the icing intensity envelopes of Newton( 1978). Newton's envelopes represent potential accu-mulation rates of 1 , 6, and 12 g cm-' h-' (correspond-ing to light, moderate, and severe icing) on a 7.6-cmcylinder mounted on an aircraft flying at 100 m s-'.The potential accumulation rate dM/dt is calculatedasdMdt-_- ELCKwhere E is the bulk collision-collection efficiency, L,,the liquid water content, and V the true airspeed. Theefficiency E depends on the droplet spectrum medianvolume diameter and was calculated from Finstad andLozowski (I 988). For each MVD, assuming a givenvelocity, the corresponding Lc can be found that cor-responds to a given potential accumulation rate. Pairsof MVD and L,, define each curve in Fig. 1 I. In com-paring the 3745 aircraft measurements to Newton'senvelopes, 12 (0.3%) corresponded to severe icing and138 (3.7%) corresponded to moderate icing.The 1-s data were averaged into 300-s intervals,which corresponded to flight legs of approximately 30km (16 n mi). This allowed direct comparison with theFAR 25-C icing envelopes for maximum continuousicing (based for flight over 17.6 n mi). Figure 12 showsa plot of SLWC versus MVD for 182 intervals in whichthe average SLWC exceeded 0.0 15 g mP3 for the 300-sinterval. Superimposed on Fig. 12 are the FAR 25-Cenvelopes. Most of the data points fell well short of the1 .om?E 0.9EI)vE 0.86 0.70& 0.62 0.5.- U$ 0.4; 0.3(ucc.-4-02 0.24 0.1VJ0.0IA -IO< Ta < OC0 5 10 15 20 25 30 35 40 Median Volume Diameter (microns)FIG. 12. Comparison of SLWC versus MVD for each 300-s average.The maximum continuous icing envelopes of FAR 25-C are super-imposed for comparison. There were 182 data points where the300-s average SLWC exceeded 0.0 15 g m-3.JANUARY 1995 COBER ET AL. 95Patch Length (km)0 10 20 30 40tl Inn U1009080706050 .s40 330 02010n0+m51 .orE 0.85 0.70 0.6r" 0.5 0.41g 0.300 0.22E 0.90,vaccU.--3 0.1v)0.0Severe12 g cm* hr'Moderate6 g cm4 hr"Light1 g cm-* hr'0 5 10 15 20 25 30 35 40 Median Volume Diameter (microns)0 100 200 300 400 500- FIG. 14. Comoarison of SLWC Venus MVD for each SLWC oatch.Patch Duration (seconds)FIG. 13. Histogram of SLWC patch durations. There were 801SLWC patches with durations between 5 and 800 s. Since the averagetrue airspeed of the aircraft was 97 m s-' a patch duration of 100 scorresponds to a horizontal distance of 9.7 km. All patches withdurations larger than 500 s are included in the last bin.maximum envelopes, although the - 10C envelopewas exceeded on 1 occasion by icing at -9.5"C (point1 on Fig. 12). The five heaviest icing cases are num-bered. Points 1 and 2 occurred consecutively during aflight through a cold-frontal region, while points 3 and4 occurred consecutively during flight at 760 m througha low-level stratus cloud. A case study of the icing en-counter involving points 1 and 2 is described in section5e. Point 5 occurred at -0.1 "C and was therefore oflittle interest for aircraft icing purposes. The Convair-580 experienced no difficulty in any of the icing con-ditions shown in Fig. 12. These results are quite similarto those reported by Isaac (199 l), although his mea-surements were mainly over Nova Scotia while thesewere predominately over the North Atlantic Ocean.c. Continuous icing patchesIcing encounters were typically less than 5 min induration, with a single 5-min average liable to containseveral smaller distinct patches of SLWC. It was in-formative to investigate the cloud microphysics overcontinuous intervals of SLWC. A "patch" was definedas having a SLWC of at least 0.025 g m-3 for at least5 s (0.5 km of flight) and was terminated when theSLWC dropped below 0.025 g m-3 for 5 s. A histogramof 80 1 SLWC patches measured during the 3 1 flightsis shown in Fig. 13. The average patch duration was44 s (4.3 km), while the median was 18 s (1.7 km). Theduration statistics were weighted heavily by the flat tailshown in Fig. 13, with 50% of the 1-s SLWC data con-tained in patches longer than 80 s ( 10% of the patches).The potential accumulation rate of each SLWC patch can be judgedby its position relative to the icing intensity envelopes of Newton(1978), which are superimposed for comparisonFigure 14 shows the SLWC versus MVD for eachicing patch compared to the intensity envelopes ofNewton (1978). There are relatively fewer instances ofmoderate and severe icing than shown in Fig. 11 be-cause 10-s intervals of moderate or severe icing wereoften averaged into longer patches where the averageintensity was significantly lower. The potential accu-mulation of ice (g cm-2) on an exposed aircraft surfaceis shown in Fig. 15 for each SLWC patch. These were1000cn0)cmaOal13z0 100cY-L5. 101 0.1 1 .oPotential Accumulation (g cm-2)FIG. 15. Total potential accumulation of ice (g cm-') on an exposedaircraft surface for each SLWC patch. The collision-collection effi-ciency was assumed to equal 1.VOLUME 3496 JOURNAL OF APPLIED METEOROLOGYcalculated assuming a collision-collection efficiency of1. The maximum potential accumulation was 1.4g cm-2, which corresponds to 1.6 cm of ice, assumingan accretion density of 0.85 g ~m-~. The average po-tential accumulation was 0.05 g cm-2, which repre-sented approximately 0.06 cm of accretion. The max-imum potential accumulation for an entire flight was6.3 g cm-2 for a low-level stratus flight and 4.2 g cmp2for a system flight, assuming no sublimation of accretedice during flight in subsaturated regions, and no meltingand shedding in regions warmer than 0.0"C. Table 2shows the total potential accumulations for the tenheaviest icing flights during CASP 11, and indicates thatfour of the six heaviest icing cases occurred in low-levelstratus clouds. This may reflect a bias in the dataset,in that low-level stratus clouds were only picked fortheir SLW potential, while system clouds were ofteninvestigated for other reasons.d. Freezing drizzle encountersDuring four flights, the aircraft encountered regionsof supercooled drizzle-sized droplets between 0.1 and1.0 mm in diameter. These cases had similarities tocases described by Politovich (1 989). On two of theseoccasions, a rapid accumulation of icing built up onthe pilots window, which necessitated aborting the flightpath and exiting the icing region. These cases are notincluded in the previous data because the King proberesponse to droplets larger than 50 pm is unknown(Biter et al. 1987), while the FSSP measured dropletsonly to a maximum diameter of 47 pm. Consequentlythe SLWC and median volume diameter could not beaccurately determined. Analysis of these cases reliesmainly on analysis of the 2D probe's imagery.A characteristic of both of the heavy drizzle en-counters was that no warm (>O.O"C) region existedaloft, implying that the supercooled drizzle dropletsformed through a "warm rain" collision-coalescenceprocess. As discussed in section 5a, there were sufficientdroplets larger than 40 pm to make drizzle formationfeasible. Table 3 contains a description of the icingenvironment for each freezing drizzle region measured.A case study of a freezing drizzle icing encounter isdiscussed in section 5e.e. Case studies of heavy icing situationsDuring the CASP I1 project, the majority of icingencounters were of short duration (average 44 s or 4.3km) and of small SLWC (average 0. I3 g m-3). Thesecharacteristics partially result from the clouds beingprimarily glaciated, with SLW regions tending to besmall or isolated within larger regions of glaciatedcloud. While the 3 1 flights could not sample every pos-sible icing condition within East Coast winter storms,the icing data reported here should be biased toward agreater severity than might be expected by randomflight through such storms systems. However, moderateto severe icing encounters over 5-10 min of flight wereobserved in each type of common synoptic feature. Itis informative to briefly describe several case studiesof these more intense icing situations.On 8 February 92 a flight was conducted into thecloud and precipitation region north of a warm front.At the leading edge of the cloud region, freezing drizzlewas measured at 3650 m at -7C for approximately30 min. During the subsequent 50 min (1630-1720UTC) several moderate icing patches were encounteredbetween -4" and -2C. Figure 16 shows the Kingprobe SLWC measurements, FSSP droplet concentra-tions, RID response, and 2DP total concentrations(>200 pm) of ice crystals from 1650 to 1720 UTC.Crystal concentrations peaked above 20 L-', with con-centrations greater than 1 Lp' often observed simul-taneously with SLW. Icing patches were isolated andseparated by regions of higher crystal concentrations.The SLWC exceeded 0.3 g m-3 intermittently, withmost patches having an average between 0. I and 0.25g m-3. Droplet concentrations were relatively small,ranging from 30 to 70 ~m-~, with median volume di-ameters of approximately 25 pm. The potential ac-cumulation from droplets smaller than 50 pm duringthe 50-min flight segment was 1.4 g cm-2. OnboardTABLE 2. Summary of the IO flights during CASP I1 with the largest total potential accumulations (PA) of ice. Totals are based on in-tegration of the King. orobe measurements of SLWC over the entire flight and do not account for sublimation or melting. of accreted ice.CaseI23456189IODate (1992)10 MarchIO March20 February20 February2 February4 March14 March29 February28 February8 FebruaryPA(g cm-76.325.414.203.442.162.212.092.062.041.95CommentsLow-level stratus with freezing drizzle (ZI)Low-level stratusFlight through a convective cold-frontal regionLow-level stratusCloud region north of an intensifying low pressure regionLow-level stratusFlight along/through trough of a redeveloping low region (ZI)Flight across a cold-frontal cloud regionLow-level stratusCloud and precipitation region north of a warm front (with ZI)JANUARY 1995 COBER ET AL. 97TABLE 3. Summary of the icing environment measured during freezing drizzle encounters. In flights on 5 February and 14 March the icing was moderate to severe and the pilots aborted the flight plan to get out of the icing region. FSSPKing SLWC concentration TemperatureDate Time and duration (g m-3) (cm-') ("C) Notes5 Feb 1992 0210 UTC, 5 min 0.3-0.5 60-80 -13" to -8" 600-m-thick St deck over airport topped at 1200 m8 Feb 1992 1600 UTC, 30 min 0.05-0.2 30-200 -7O to -5" St cloud at 3300 m at northern edge of a warm-frontal region10 Mar 1992 1330 UTC, 120 min 0.05-0.2 20-70 -5" to -2" Low-level stratus cloud over ocean, base at 300 m, tops at 1200 m14 Mar 1992 1328 UTC, 5 min 0.05-0.2 20-50 -8" to -1 I" Flight below and into St cloud along small-scale trough (see section 5e)observers reported 2 cm of clear ice on portions of theinstruments under the wings (some of which was causedby the freezing drizzle). This case represents the heaviesticing encounter in a warm frontal region, during theCASP II project.On 20 February 92 the Convair-580 crossed a coldfrontal band of cells oriented southwest to northeastacross St. John's. The cells were weakly convective andwere evident on the 4-km CAPPI (constant-altitude0 180 360 540 720 900 1080 1W 1440 1620 1603 0.49 0.3u, 0.1e* 0.01^ I3 0.25 l50F 1plan position indicator) of the 5-cm C-band Holyroodsurface radar. Between 1545 and 1645 UTC the aircraftcrossed two cells in which several patches of moderateand severe icing were measured. The flight level was3960 m at -9.5"C. The maximum 300-s icing averagescorrespond to points 1 and 2 on Fig. 12, and repre-sented an exceedance of the FAR 25-C maximum icingenvelopes. Figure 17 shows King, FSSP, RID, and 2DPmeasurements starting at 1545 UTC. Peak SLWCswere greater than 0.5 g mp3 and droplet concentrationsranged from 30 to 200 cmp3. The maximum dropletI- 0.49 0.3y 0.26 0.1e" nnI"0 180 360 540 7% 9M) 1080 1260 1440 1620 1800 Elasped Time from 16:50:00 GMT (seconds)FIG. 16. Thirty-minute time histories of King probe SLWCs, FSSPdroplet concentrations, Rosemount icing detector (RID) voltage sig-nals, and 2DP total concentration starting at 1650 UTC 8 February1992. The RID sheds its ice build up when its voltage has reachedapproximately 5000 mV. The average temperature ranged from - 1 Oto -3C during this period.1wo' I. I. U9-E IO'" idB-! Id"'0 180 38) 540 720 900 1080 1260 1440 1620 18M1Elasped Time from 15:45:00 GMT (seconds)FIG. 17. Same as Fig. 16, starting at 1545 UTC 20 February 1992.The FSSP did not work between 1100 and 1680 s. The average tem-perature was approximately -9.5"C during this period.98 JOURNAL OF APPLIED METEOROLOGY VOLUME 340 180 360 540 720 9M 1080 1260 1440 1620 18000.4030.20.10.015010050nI" ~~~~0 180 381 540 720 9M) 1080 128) 1440 1620 18w Elasped Time from 19:50:00 GMT (seconds)FIG. 18. Same as Fig. 16, starting at 1950 UTC 2 February 1992.The temperature was approximately -6C during this period. TheFSSP was fogged from 570 to 1080 s.median volume diameter was 33 pm (averaged over300 s). Note the poor RID response during periodswhen the droplet concentrations were low and medianvolume diameters were high. Crystal concentrationswere relatively low and exceeded 1 L-' only on a fewoccasions. The total potential accumulation was 2.3g cm-2 with onboard observers estimating that 3 cmof ice built up on the measuring probe during the 30-min period. This case represented the heaviest icingevent during CASP 11, not directly associated withfreezing drizzle.On 2 February 92, the Convair-580 was flown intothe northern leading edge of a large-scale circulationaround a low pressure region. Flight levels ranged from3050 to 3650 m with corresponding temperatures be-tween -6" and -8C. Clouds were based at 1200 mand topped above 6700 m although thin gaps betweenlayers were observed. During both the penetrating andexiting flight paths (along roughly the same north-southroute) a 15-min period was encountered in which sev-eral patches of moderate icing were measured. Figure18 shows the King, FSSP, RID, and 2DP measurementsfrom 1950 to 2020 UTC. Momentary maximumSLWCs were 0.4 g m-3 with the maximum 1-min av-erage exceeding 0.3 g mP3. Droplet concentrationsranged from 50 to 200 cm-3 with the droplet medianvolume diameter averaging 21 pm. The maximum300-s average would be rated just below the moderateicing envelope of Newton (1978). The potential ac-cumulation was 1.4 g cm-2 between 1957 and 2012UTC. This case represented the heaviest icing en-countered in a low pressure region during the CASP I1project.On 14 March 92 a flight was conducted into a troughline associated with a small-scale wave and a redevel-oping low pressure region. At 1328 UTC the aircraftentered a region of freezing drizzle (Zl) at 3050 m at-8C. The drizzle caused a rapid accumulation of iceon the pilots window and the pilot increased altitudeto exit the icing region. The Z1 region was exited at3800 m at - 10C. This region was topped by a regionof supercooled cloud extending from 3600 to 4 I50 m.Figure 19 shows the instrument responses during theicing encounter. The King measured SLWCs did notexceed 0.2 g m-3, although only a partial response tothe drizzle size drops is expected. In the cloud regionthe droplet concentrations did not exceed 50 cmP3 withan average median volume diameter of 25 pm. In theZl region, the 2DC Mono and 2DC Grey spectra in-dicated drizzle droplets with sizes between 100 and 500pm. Figure 20 shows images of drizzle droplets as mea-sured by the 2DC Grey, whereas Fig. 21 shows thedroplet spectra. The icing cylinder accumulated 1.1cm of ice between I329 and 1334 UTC. This wouldcorrespond to an average SLWC of 0.3 g m-3, assumingan ice density 0.85 g cm-3 and a collision-collectionefficiency of 1 .O. The corresponding potential accu-mulation was 10.5 g cm-2 h-' implying that underNewton's envelopes the icing intensity was betweenmoderate and severe. Between 1329 and 1331 UTCthe King probes failed to measure SLWCs above 0.05g m-3, indicating that the SLWC was mainly in dropletslarger than 50 pm. Freezing drizzle is considered torepresent a greater danger because the droplets can runback under the wings before freezing and cause a sig-nificant decrease in aircraft performance (Sand et al.1984). It is difficult to compare the freezing drizzlecharacteristics to the FAR 25-C envelopes because thelatter do not extend beyond 50 pm.0 180 360 540 720 900 1080 1260 1440 1620 18w04030.201---5010'103102in1'-0 180 360 540 720 900 1080 1260 1440 1620 1800 Elasped Time from 13:20:00 GMT (seconds)FIG. 19. Same as Fig. 16, starting at 1320 UTC 14 March 1992.Between 660 and 1000 s the aircraft ascended from 3050 m (-8OC)to 3950 m (- 1 1 "C). The aircraft flew through Z1 between approxi-mately 540 and 800 s.JANUARY 1995 COBERI I I I I. ...I I IFIG. 20.2DC Grey images of freezing drizzle. These were measuredbetween 1330 and 1331 UTC 14 March 1992. The largest dropletsshown have a diameter of approximately 420 pm. The temperaturewas between -8" and -llC with no warm layer aloft, implyingthat the droplets may have formed through a collision-coalescenceprocess.6. Conclusions and future workMeasurements of aircraft icing encounters, madeduring 31 flights into East Coast winter storms overthe North Atlantic Ocean allow the following conclu-sions:1) SLWC regions were characterized by limitedhorizontal extent (average 4.3 km), low droplet con-centrations (average 1 30 ~m-~), relatively large dropletmedian volume diameters (average 18 pm), and lowliquid water contents (average 0.13 g m-3). On a 1-kmscale, the relative frequencies of light, moderate, andsevere icing observations were 0.960,0.037, and 0.003,respectively (using the definition of Newton 1978). Thesmall horizontal icing scale of 4.3 km resulted becausethe clouds measured were primarily glaciated, with onlysmall pockets of SLWC. Droplet concentrations andMVDs were quite characteristic of maritime clouds,while the low liquid water contents resulted from lim-ited convective activity and characteristic cold cloud-base temperatures.2) Freezing drizzle was observed from the aircrafton several occasions, and may be a frequent phenom-enon in East Coast winter storms representing a sig-nificant aviation hazard. The median cloud dropletconcentration for system clouds was 64 ~m-~, whichwas representative of maritime cloud concentrations.The median droplet concentration for droplets largerthan 40 pm in diameter was 20 droplets per liter. Thelow droplet concentrations and relatively large con-ET AL. 99centrations of large droplets are suspected to have con-tributed to the formation of freezing drizzle through a"warm rain" collision-coalescence process.3) Moderate to severe icing was observed over 30-120-km regions during flights across warm fronts, coldfronts, low pressure regions, and low-level stratusclouds. The case studies presented in this paper describethe nature and severity of icing events encountered incommon synoptic phenomena in East Coast storms.4) This dataset should serve as a solid reference formicrophysical icing conditions in a maritime winterenvironment. Consequently, it should make a signifi-cant addition to the datasets used by the U.S. FederalAviation Administration and Transport Canada in de-fining icing environments for aircraft safety and cer-tification. Emphasis must be placed on the high qualityof the dataset, primarily because of the extensive andcareful quality control applied to the instrument cali-bration and data analysis. Current projects for whichthis dataset is being used as a basis include: a compar-ison between the aviation forecast of the CanadianMeteorological Centre and the aircraft measurementsfrom 34 flights in CASP 11; an assessment of the al-gorithms for determining SLW over the ocean fromthe Defence Meteorological Satellite Program SpecialSensor Microwave/Imager; possible certification of aConvair-5800 for flight into icing conditions; redefiningof safety requirements in cross-oceanic commercialflights (Patnoe et al. 1993), and the development ofimproved microphysics parameterizations for the im-provement of aircraft icing forecasts.1000m?Evc0.-CI24-CC8 1008.wa,a-En100.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Droplet Diameter (mm)FIG. 21. Droplet spectra for droplets between 125 and 700 pm asmeasured with the 2DC Mono probe. The spectra represent averagesover a 300-s penod from 1329 to 1334 UTC 14 March 1992. Thefirst four channels were ignored and channels with less than 10 mea-surements were not considered.100 JOURNAL OF APPLIED METEOROLOGY VOLUME 34Acknowledgments. This work was funded by the Ca-nadian National Search and Rescue Secretariat. Fund-ing for the field project was also provided by the In-stitute for Aerospace Research (IAR) of the NationalResearch Council of Canada, Boeing Commercial Air-plane Group, and Airbus Industrie. The authors wouldlike to acknowledge Ron Stewart, Mike Patnoe, andIan MacPherson for their invaluable assistance duringthe flight project. Thanks also to the technicians ofAES and IAR for their work on keeping the aircraftand instruments operational, and to the forecasters ofAES for their thorough forecasts. REFERENCESBaumgardner, D., 1983: An analysis and comparison of five water droplet measuring instruments. J. Climate Appl. Meteor., 22,-, and A. Rodi, 1989: Laboratory and wind tunnel evaluations of the Rosemount icing detector. J. Atmos. Oceanic Technol.,-, W. Strapp, and J. E. Dye, 1985: Evaluation of the forward scattering spectrometer probe. Part 11: Corrections for coinci- dence and dead time losses. J. Atmos. Oceanic Technol., 2,626- 632.Biter, C. J., J. E. Dye, D. Huffman, and W. D. King, 1987: The drop- size response of the CSIRO liquid water probe. J. Atmos. Oceanic Technol., 4, 359-361.Curry, J. A,, and G. Liu, 1992: Assessment of aircraft icing potential using satellite data. J. Appl. Meteor., 31, 605-62 1.Finstad, K. J., and E. P. Lozowski, 1988: A computational investi- gation of water droplet trajectories. J. Atmos. Oceanic Technol.,Gardiner, B. A,, and J. Hallett, 1985: Degradation of in-cloud forward scattering spectrometer probe measurements in the presence of ice particles. J. Atmos. Oceanic Technol., 2, 171-180.Hobbs, P. V., and A. L. Rangno, 1985: Ice particle concentrations in clouds. J. Atmos. Sci., 42, 2523-2549.Isaac, G. A,, 199 1: Microphysical characteristics of Canadian Atlantic storms. Atmos. Res., 26, 339-360.Jeck, R. K., 1983: A new data base of supercooled cloud variables for altitudes up to 10,000 feet AGL and the implications for low altitude aircraft icing. U.S. Dept. of Transportation Report891-910.6,971-979.5, 160- 170.DOT/FAA/CT-83/2 I, 137~~.King, W. D., D. A. Parkin, and R. J. Handsworth, 1978: A hot wire liquid water device having fully calculable response character- istics. J. Appl. Meteor., 17, 1809- 18 I 3.-, J. E. Dye, J. W. Strapp, D. Baumgardner, and D. Huffman, 1985: Icing wind tunnel tests on the CSIRO liquid water probe. J. Atmos. Oceanic Technol., 2, 340-352.Langmuir, I., and K. B. Blodgett, 1946: Mathematical investigation of water droplet trajectories. Vol. 10, Collected Works of Irving Langmuir, Pergamon Press, 348-393.Lawson, R. P., and W. A. Cooper, 1990: Performance of some air- borne thermometers in clouds. J. Atmos. Oceanic Technol., I,Masters, C. O., 1983: A new characterization of supercooled clouds below 10,000 feet AGL. U.S. Dept. of Transportation ReportMazin, I. P., 1994: Cloud water content in continental clouds of middle latitudes. Atmos. Res., in press.Newton, D. W., 1978: An integrated approach to the problem of aircraft icing. J. Aircraft, 15, 374-380.Patnoe, M. W., W. G. Tank, G. A. Isaac, S. G. Cober, and J. W. Strapp, 1993: Airplane icing research at the Boeing Company: Participation in the Canadian Atlantic Storms Program. Fifth Int. Conf on Aviation Weather Systems. Vienna, VA, Amer. Meteor. SOC., 432-434.Politovich, M. K., 1989: Aircraft icing caused by large supercooled droplets. J. Appl. Meteor., 28, 856-868.Pruppacher, H. R., and J. D. Klett, 1978: Microphysics ofCIouds and Precipitation. Reidel, 7 14 pp.Ranber, R. M., and A. Tokay, I99 I: An explanation for the existence of supercooled water at the top of cold clouds. J. Atmos. Sci.,Rasmussen, R., M. Politovich, J. Marwitz, W. Sand, J. McGinley, J. Smart, R. Pielke, S. Rutledge, D. Wesley, G. Stossmeister, B. Bernstein, K. Elmore, N. Powell, E. Westwater, B. B. Stankov, and D. Burrows, 1992: Winter Icing and Storms Project (WISP). Bull. Amer. Meteor. Soc., 13, 95 1-974.Sand, W. R., W. A. Cooper, M. K. Politovich, and D. L. Veal, 1984: Icing conditions encountered by a research aircraft. J. Climate Appl. Meteor., 23, 1427-1440.Stewart, R. E., 199 I: Canadian Atlantic Storms Program: Progress and plans of the meteorological component. Bull. Amer. Meteor. Soc., 12, 364-371.-, R. W. Shaw, and G. A. Isaac, 1987: Canadian Atlantic Storms Program: The meteorological field project. Bull. Amer. Meteor. Soc., 68, 338-345.480-494.DOT/FAA/CT-83/22, 36 pp.48, 1005-1023.