VOLUME 15 WEATHER AND FORECASTING APRIL 2000

Utilization of Automatic Weather Station Data for Forecasting High Wind Speeds at Pegasus Runway,

R. E. HOLMES AND C. R. STEARNS Space Science and Engineering Center, University of WisconsinÐMadison, Madison, Wisconsin

G. A. WEIDNER AND L. M. KELLER Department of Atmospheric and Oceanic Sciences, University of WisconsinÐMadison, Madison, Wisconsin

(Manuscript received 14 December 1998, in ®nal form 20 September 1999)

ABSTRACT Reduced visibility due to blowing snow can severely hinder aircraft operations in the . Wind speeds in excess of approximately 7±13 m sϪ1 can result in blowing snow. The ability to forecast high wind speed events can improve the safety and ef®ciency of aircraft activities. The placement of automatic weather stations to the south (upstream) of the Pegasus Runway, and other air®elds near McMurdo Station, Antarctica, can provide the forecaster the information needed to make short-term (3±6 h) forecasts of high wind speeds, de®ned in this study to be greater than 15 m sϪ1. Automatic weather station (AWS) data were investigated for the period of 1 January 1991 through 31 December 1996, and 109 events were found that had high wind speeds at the Pegasus North AWS site. Data from other selected AWS sites were examined for precursors to these high wind speed events. A temperature increase was generally observed at most sites before such an event commenced. Increases in the temperature difference between the Pegasus North AWS and the Minna Bluff AWS and increasing pressure differences between other AWS sites were also common features present before the wind speed began to increase at the Pegasus North site. Many times, changes in one or more of these parameters occurred hours before the wind began to increase at the Pegasus North site. Monitoring of these parameters can lead to an improved 3±6-h forecast of these high wind speed events at Pegasus Runway, Antarctica.

1. Introduction whiteout, the vast snow ®eld of the Ross Ice Shelf, while the wheeled aircraft can only land on runways con- The National Science Foundation's Of®ce of Polar structed on ice. The reliance on an for Programs operates the United States Antarctic Pro- wheeled aircraft increases the importance of a good gram's (USAP) year-round stations in the Antarctic at Anvers Island (Palmer Station), Ross Island (McMurdo weather forecast for these ¯ights because after passing Station), and at the South Pole (Amundsen±Scott Sta- the point of safe return, the only landing sites that can tion). While Palmer Station is generally serviced via accommodate wheeled aircraft are the two ice runways USAP research vessels from South America, McMurdo near McMurdo Station. Station is partially supplied by aircraft from Christ- Figure 1 shows the three landing sites, church, New Zealand, and Amundsen±Scott Station is ski-way, Pegasus blue-ice runway, and the sea-ice run- supplied entirely by aircraft ¯ights from McMurdo Sta- way, in the vicinity of McMurdo Station. The sea-ice tion. Most of the personnel at McMurdo and Amund- runway is located on the annual sea ice west of the sen±Scott Stations travel by air to and from New Zea- southernmost point of Hut Point Peninsula although its land. location annually varies. The sea-ice runway is used by The aircraft used for ¯ights between Christchurch and both ski-equipped and wheeled aircraft. Its season of McMurdo Station are the ski-equipped LC-130, and the use begins in August and continues until early Decem- wheeled C-130, C-141, and C-5. In the area of McMurdo ber, when the sea ice becomes unsafe for aircraft op- Station, the ski-equipped LC-130 can land on ice, com- erations. At this time, all aircraft operations shift to pacted snow, or, in extreme circumstances such as a Williams Field and all ¯ights, whether between Mc- Murdo Station and New Zealand, Amundsen±Scott Sta- tion, or remote ®eld camps, are made exclusively by the ski-equipped LC-130. Corresponding author address: Robert Holmes, Space Science and Engineering Center, 1225 West Dayton St., Madison, WI 53706. The Pegasus blue-ice runway is used for ¯ights by E-mail: [email protected] wheeled aircraft late in the austral summer ®eld season,

᭧ 2000 American Meteorological Society 137

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 138 WEATHER AND FORECASTING VOLUME 15

TABLE 1. Locations of AWS units used in support of this study. Elevation Site Lat Long (m) Pegasus North 77.95ЊS 166.51ЊE 10 Minna Bluff 78.56ЊS 166.69ЊE 920 Elaine 83.15ЊS 174.46ЊE 60 Willie Field 77.85ЊS 167.08ЊE 40 Ferrell 78.02ЊS 170.80ЊE 45 Lettau 82.52ЊS 174.43ЊW 55 Gill 80.03ЊS 178.63ЊW 55 Linda 78.50ЊS 168.35ЊE 50 Schwerdtfeger 79.94ЊS 169.83ЊE 60 Marilyn 79.98ЊS 165.03ЊE 75

AWS site than that at Ferrell AWS site. The wind di- rection associated with the maximum monthly wind speed at Pegasus was from the direction of Minna Bluff, which lies 55 km to the south of the Pegasus AWS site. Data for 1989 for Ferrell and Pegasus AWS sites are given by Stearns and Weidner (1990). Plans were made to install AWS units at the north and south ends of the Pegasus Runway, on top of Minna Bluff, at Williams Field, and on the Ross Ice Shelf east of Minna Bluff based on the information initially pro- vided by the Pegasus AWS unit. The purpose of the AWS site on top of Minna Bluff, at an elevation of 920 m, was to detect the strong winds ¯owing over Minna Bluff and continuing toward the Pegasus Runway. The AWS site east of Minna Bluff (Linda AWS site) was installed to investigate the air¯ow near the east end of Minna Bluff. The AWS units on each end of the Pegasus Runway were installed to observe any differences in the FIG. 1. Location map of the Pegasus runway site (adapted from wind speed, wind direction, atmospheric pressure, and Blaisdell et al. 1995). air temperature along the runway. The AWS unit on the west end of the Williams Field ski-way was installed to provide a nearby AWS test site and to obtain meteo- approximately late January into March. During most of rological data for the entire year. Previously, meteoro- the austral summer, the runway is covered with snow logical data were collected only when Williams Field to maintain the integrity of the ice surface, and then the was in operation. snow is removed prior to the commencement of air op- The Pegasus North AWS site was installed in January erations. The use of the Pegasus runway also allows the 1990 at the north end of the now established Pegasus ski-equipped LC-130 to operate on wheels and not on runway. The AWS sites at Minna Bluff, Linda, and Peg- skis, which allows for as much as a 50% increase in asus South were established in January 1991. The Peg- the payload. asus South AWS unit also measures a vertical temper- The ®rst automatic weather station (AWS) unit near ature pro®le in the ice. Stearns and Weidner (1991, all three runway facilities was installed near the wrecked 1992) reported that while the wind directions at both aircraft named Pegasus in January 1989. It provided the Pegasus North and South AWS sites were most com- atmospheric pressure, air temperature, and wind speed monly from the east-northeast, winds with speeds great- and direction data until the unit was removed in No- er than 8 m sϪ1 had a strong tendency to blow from the vember 1989. The data record over this time period south. showed that the Pegasus AWS site monthly mean air Several AWS units have also been deployed near Mc- temperature was approximately 1ЊC warmer than the Murdo and to the south (upstream) of the Ross Island monthly mean temperature at Ferrell AWS site, ap- area to aid in operational forecasting. Fleming (1983) proximately 88 km to the east and at approximately the utilized data from many of these units to forecast fog same latitude (Table 1). The monthly average wind at the Williams Field runway. He found that during the speed was 1±2 m sϪ1 less at Pegasus AWS site than that period from 1 November 1982 to 13 January 1983 there at Ferrell AWS site, but the maximum monthly wind was a strong correlation between a wind from the south- speed had a tendency to be 8±10 m sϪ1 greater at Pegasus east quadrant (90Њ±170Њ) at Ferrell AWS site and dense

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 139

The data for this study were obtained from the AWS sites, whose geographic location and elevation are found in Table 1; Fig. 4 shows the locations for these AWS sites. Data from 1991 through 1996 were chosen for this study. Unfortunately, not all of the units were op- erating at all times. Missing data can be the result of any number of events that occur while the AWS unit is unattended.

3. Statement of the problem Because of its remote location, air travel in the Ant- arctic can be dangerous. A downed aircraft could lead to loss of life not only from the crash itself, but from exposure or because of the lack of nearby medical fa- cilities. Therefore, safety is a high priority for those involved in air transportation. FIG. 2. Layout of the AWS unit used in the Antarctic. The installed Most of the personnel working for USAP are trans- AWS unit has a 3-m tower with a horizontal boom supporting the antenna, aerovane, air temperature thermometer, upper themopile, and ported to the continent by air, using one of three runway the relative humidity sensor. The electronics enclosure is mounted at facilities near McMurdo Station. The sea-ice runway is the midpoint of the tower. The gel cell batteries are placed at the located just to the west of McMurdo on the annual sea tower base. The 10-W solar panel faces north. ice and can accommodate conventional (wheeled) air- craft. The annual ice begins to deteriorate in early De- fog at Williams Field. Sinkula (1993) and Holmes cember, at which time air operations shift to Williams (1994) have done previous studies on the use of AWS Field (Fig. 1). Williams Field, a compacted snow fa- data for forecasting high wind speeds in the vicinity of cility, is located approximately 15 km to the east-south- Ross Island. Other examples of the use of AWS data as east of McMurdo Station and is used primarily by ski- a forecasting aid are shown by Holmes et al. (1993) and equipped aircraft. The use of skis limits the amount of Stearns and Sinkula (1994). This study focuses on the cargo and passengers that can be carried on each air- development of empirical rules based on AWS data to plane because friction with the snow surface limits the forecast high wind speed events at the Pegasus Runway. ability to reach takeoff speed. More fuel must be used because the skis create more drag in the air. These fac- tors led to a search for a runway that could accommodate 2. Data acquisition wheeled aircraft throughout the austral summer. During The development and deployment of automatic the summer of 1989±90, the site known as Pegasus was weather stations have made possible the collection of chosen for the development of such a runway. near-surface meteorological data in remote areas of Ant- The U.S. Antarctic Program opened the Pegasus run- arctica. The automatic weather station is a surface me- way on 6 February 1993. The runway is located about teorological data collection unit capable of operating 12 km south of McMurdo and 16 km west of Williams throughout the year without any intervention. Each Field. It lies just to the east of a loosely de®ned region AWS unit measures air temperature, horizontal wind between the area of net snow and ice accumulation and speed, and wind direction at a nominal height of 3 m. ablation (Fig. 1). Therefore, the ice surface of the Peg- Atmospheric pressure is measured at the electronics en- asus runway has a light snow cover that is removed closure (Fig. 2). Some AWS units also measure the rel- before runway operation. ative humidity at 3 m and the air temperature difference The weather is ever changing near the Pegasus run- between 1.0 and 3.9 m. The height at which each pa- way. Visibility is probably the most important forecast rameter is measured varies with time because of the element at the Pegasus runway, and poor visibility can accumulation or ablation of snow. be caused by fog and/or blowing snow. Wind speeds in The AWS units have been deployed in Antarctica excess of 13 m sϪ1 can cause blowing snow in the austral since 1980 by the U.S. Antarctic Program (Stearns et summer, and during the winter, speeds in excess of 7 m al. 1993). As of 25 May 1998, 54 AWS units are op- sϪ1 can result in reduced visibilities from blowing snow erating in Antarctica, enhancing meteorological research (Bromwich 1988). The difference can be attributed to (Fig. 3). The units have operated successfully at tem- surface melting and greater adhesion in the summer. peratures as low as Ϫ80ЊC and at wind speeds as high Abrupt changes in wind speed and direction can also as 50 m sϪ1. The units are serviced as needed each have an impact on safety by affecting the lift of the austral summer. Stearns et al. (1994) and Holmes and aircraft, thereby hindering its ability to land or take off. Stearns (1995) describe the activities during two pre- Therefore, the ability to accurately forecast the wind vious ®eld seasons. would enhance the operational safety at Pegasus.

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 140 WEATHER AND FORECASTING VOLUME 15

FIG. 3. Antarctic automatic weather station locations as of 25 May 1998. Locations are identi®ed by site name.

The variations in the wind at Pegasus can be attributed 1 January 1991 through 31 December 1996 were to many factors and therefore can be dif®cult to forecast. scanned for high wind speed events. These events were The passage of a synoptic-scale disturbance can affect de®ned as those where the observed wind speed became the wind, as may the local topography. Air¯ow between greater than 15 m sϪ1 and was maintained for at least White and Black Islands can be accelerated by a jet 1 h. One hundred nine such events were found. Data effect (Savage and Stearns 1985). In fact, maximum preceding these events at other AWS sites were inves- wind speeds at Pegasus North tend to be greater than tigated for the purpose of developing empirical rules for those at nearby sites (Keller et al. 1996, 1997). The forecasting such events. onset of a barrier wind parallel to the Transantarctic Mountains or katabatic ¯ow can also change the wind 4. Observations and results conditions at Pegasus North. Changes in static stability can promote or prevent vertical mixing, and mesoscale A temperature increase with southerly winds seems cyclogenesis may also play an important role. Typically, to be common at many of the AWS sites before and changes in wind speed and direction can be attributed during high wind speed events at Pegasus North. This to one or more of these factors and because most of may be due to turbulent mixing of an elevated, neutrally these factors are on a relatively small spatial scale, glob- buoyant katabatic airstream with the underlying layer al numerical models are of limited use. Examination of of air, which initially has a strong surface inversion surface pressure analyses from the European Centre for (Bromwich et al. 1992). The temperature increase was Medium-Range Weather Forecasts (ECMWF) revealed usually most notable at the Pegasus North site itself. no speci®c larger-scale synoptic features that can be Signi®cant temperature increases were also observed at used as a signature of high wind speed events. both Marilyn and Schwerdtfeger. Observations of wind speed at Pegasus North from On many occasions Pegasus North recorded the high-

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 141

FIG. 4. Location map for the AWS sites used in this study. est wind speeds among the AWS sites used in this study the wind shifted southerly before the wind speed ex- during these events. This fact should not be surprising ceeded 15 m sϪ1. because, as previously stated, the local topography of Minna Bluff and White and Black Islands creates a jet a. Estimated stability and its relationship to Pegasus effect in that area. There were two AWS sites, Linda winds and Minna Bluff, that generally had winds of similar speeds as Pegasus North. However, many times the wind Automatic weather stations have increased the speed at Minna Bluff was greater than that at Pegasus amount of Antarctic surface meteorological data, lead- North. Other sites usually experienced wind speeds sig- ing to a better understanding of surface conditions in ni®cantly less than that observed at Pegasus North. the Antarctic. But the lack of upper-air observations still In all of the cases studied, the high wind speeds at presents a problem for forecasters. McMurdo Station, Pegasus North site coincided with winds from the south. Antarctica, releases radiosondes every 12 h during the The wind direction was usually from the northeast be- summer months, and nearly every day during the winter, fore the wind speed began to increase. The direction of but those are the only vertical measurements taken in

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 142 WEATHER AND FORECASTING VOLUME 15 the area. Because of its elevation (920 m), Minna Bluff resulting in a cooling of the surface layer from radiation can provide some insight into changes of meteorological loss and an increase in static stability. This increase in parameters with height. static stability further decouples the surface ¯ow from To investigate this possibility, the temperature dif- the ¯ow aloft. Eventually, the kinetic energy of the sur- ference was calculated from data recorded by Pegasus face ¯ow is insuf®cient to overcome the effects of the North AWS and that from Minna Bluff: (Pegasus North increasing negative buoyancy (Sinclair 1988), and the temperature) Ϫ (Minna Bluff temperature). Using this air is restrained to ¯ow around the topography. How- de®nition, a negative result would indicate inversion ever, a slight increase in the larger-scale ¯ow may be conditions (temperature increasing with height) and stat- suf®cient to sweep away the cold layer, resulting in ically stable air, and a temperature difference greater stronger, and usually southerly, winds at the surface. than 8.9ЊC may indicate statically unstable air. This in- The mixing process associated with these stronger winds crease in the temperature difference between these two may contribute to the further warming of the air near AWS sites can often be attributed to an increase in the the surface, further enhancing the instability of the lower temperature at Pegasus North. However, the temperature atmosphere. at Minna Bluff also plays an important role. Many times With some evidence of a relationship between the the temperature at Minna Bluff increased coincidentally stability and the surface wind decoupling, it would seem with that at Pegasus North, resulting in a temperature reasonable that changes in stability may give advanced difference that would indicate stable conditions, even warning to an increase in wind speed. To investigate though there was a large temperature increase at Pegasus this possibility, time series graphs were made of the North. There were also cases in which there was a rel- Pegasus North±Minna Bluff temperature difference atively small increase in the temperature at Pegasus (TD) and Pegasus North wind speed during the hours North but a decrease in temperature at Minna Bluff, preceding and during the high wind speed events. Be- resulting in a temperature difference indicating statically cause of missing data at either Pegasus North or Minna unstable conditions. Bluff, this hypothesis could not be tested for every high This stability estimate was chosen because the cal- wind speed event, but the data were available for 87 of culation can be made quickly by the forecaster using the events studied. data that is readily available. It must be stated that, An increase in the TD was a common feature pre- because the results are calculated from AWS units more ceding high wind speed events at Pegasus North. In all than 60 km apart, this temperature difference is only an but eight cases that had the data necessary for computing estimate of the conditions present over this general area. the TD, an increase to levels in excess of 8.9ЊC was Nevertheless, the results have given some useful in- observed during the 24 h prior to the increase in wind sights to the meteorology of the area. The effects of speed. The amount of time between when the TD be- diurnal temperature changes, especially at Pegasus came greater than 8.9ЊC and the increase in wind speed North, must also be taken into account. The diurnal to 15 m sϪ1 ranged from Ϫ6.7 to 33.8 h, with the average temperature changes in the summer make the use of the being nearly 6 h. The range within one standard devi- Pegasus North±Minna Bluff temperature difference as ation of the mean was Ϫ2.0 to 13.6 h. The frequency an indicator of static stability less reliable. distribution and various statistical data are shown in Fig. The temperature difference method was similar to the 5a. approach taken by Sinclair (1988) in which he compared An example of this relationship can be seen during weather data observed at Scott Base, Antarctica, to that the high wind speed event of 15 April 1992 (Fig. 6). recorded at Castle Rock, which has an elevation of 300 The wind at Pegasus North increased to 10 m sϪ1 at m and overlooks Scott Base. He observed that during 0700 UTC, and then underwent a gradual decrease. At inversion conditions, the wind tended to follow local 1800 UTC, the wind speed increased abruptly from ap- topography, resulting in a northeast wind at Scott Base. proximately 7 m sϪ1 to greater than 16 m sϪ1 in less However, during neutral or unstable conditions, the sur- than 1 h. Wind speeds remained generally at or above face ¯ow closely followed that at 300 m. These ®ndings 15msϪ1 until 0230 UTC 16 April. are consistent with simple physical arguments based on The TD was less than 0ЊCkmϪ1 throughout the day buoyancy considerations, which indicates that this meth- of 14 April. During this period of temperature inversion, od provides a realistic estimate of local static stability. the wind speed at Pegasus North remained less than 4 As surface winds decrease, turbulent mixing is reduced, msϪ1. Beginning at 0000 UTC 15 April, the TD began

→

FIG. 5. Frequency distributions for the amount of time that the trend of the AWS parameter was complete before the wind speed at Pegasus North increased to 15 m sϪ1 and associated statistics. The vertical dashed line denotes the 0-min line. Frequency distributions are in 60-min class widths and are for (a) Pegasus North±Minna Bluff temperature difference, (b) Marilyn±Schwerdtfeger pressure difference, (c) Elaine± Marilyn pressure difference, (d) Marilyn temperature, (e) Marilyn wind direction, (f) Schwerdtfeger temperature, and (g) Schwerdtfeger wind direction.

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 143

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 144 WEATHER AND FORECASTING VOLUME 15

FIG. 6. Pegasus North wind speed and Pegasus North±Minna Bluff FIG. 7. Pegasus North wind speed and Pegasus North±Minna Bluff temperature difference vs time, 14±16 Apr 1992. Pegasus North wind temperature difference vs time, 26±28 Feb 1993. Pegasus North wind speed is denoted by solid circles. Open squares represent the Pegasus speed is denoted by solid circles. Open squares represent the Pegasus North±Minna Bluff temperature difference. North±Minna Bluff temperature difference. to increase and continued to do so for the next 19 h. Mountains. The Schwerdtfeger site is 95 km to the east The TD increased to above 8.9ЊC by 1700 UTC, ap- of the Marilyn site. proximately 1 h before the abrupt increase in the Peg- The pressure differences were determined between asus North wind speed. The TD remained above 8.9ЊC Marilyn and Schwerdtfeger for the periods prior to the until 0600 UTC 16 April, shortly after the wind speed onset of high wind speed events at the Pegasus North had decreased below 10 m sϪ1. site. The pressure at Schwerdtfeger was subtracted from The TD behaved in a similar manner prior to the high the pressure at Marilyn, and the result was then plotted wind speeds observed at Pegasus North on 27 February as a time series. Station pressure was used instead of 1993 (Fig. 7). The TD increased from Ϫ2.7Њ to 8.8ЊC sea level pressure because it is the trend of this differ- and then quickly decreased back to Ϫ2.3ЊC between ence that is important, and station pressure is readily 0000 and 1600 UTC 26 February, without an apprecia- available to the forecaster. ble increase in wind speed. A second increase in TD Of the 109 high wind speed events studied, the Mar- commenced shortly after 1600 UTC, when the TD in- ilyn±Schwerdtfeger pressure differences were available creased to greater than 10.9ЊC by 0800 UTC 27 Feb- for 76 of the cases. The Marilyn±Schwerdtfeger pressure ruary. The TD increased to 8.9ЊC by 0720 UTC, ap- difference increased prior to the high wind speeds ob- proximately 1 h before the wind speed at Pegasus North served at Pegasus in 61 of those cases. The time between increased to speeds greater than 15 m sϪ1. The TD re- when the Marilyn±Schwerdtfeger pressure difference mained above 8.9ЊC until 1600 UTC 28 February, nearly began to increase and the wind speed at Pegasus North 10 h after wind speeds had decreased below 10 m sϪ1. increased to greater than 15 m sϪ1 varied from Ϫ0.2 h to as much as 57 h and averaged just over 19 h. The increase of the pressure difference ended between 39.8 b. The barrier wind and horizontal pressure h before and 17.7 h after the winds increased at Pegasus, gradients with the average time being about 5.5 h before the wind As mentioned previously, a barrier wind (Schwerdt- event. The range within one standard deviation of the feger 1984) may contribute to the predominately south- mean was 15.0 h before and 4.0 h after the wind event erly ¯ow over the western portions of the Ross Ice Shelf. (Fig. 5b). A synoptic-scale feature, such as a low pressure dis- An example of this phenomenon can be found on 17 turbance over the Ross Sea, can force cold, stable air September 1991. During most of the day of 16 Septem- against the Transantarctic Mountains. As this air be- ber, the Marilyn±Schwerdtfeger pressure difference comes dammed up by the Transantarctic Mountains, a ¯uctuated between Ϫ0.5 and 1.0 hPa, but at 1500 UTC west±east pressure gradient is formed and a southerly it began to increase (Fig. 8). By 1700 UTC it had in- ¯ow along the western edge of the Ross Ice Shelf results. creased to 3.0 hPa. The wind speed at Pegasus North By examining AWS data from sites located near the began to increase at 1900 UTC, quickly reaching speeds Transantarctic Mountains, evidence of this phenomenon of near 16 m sϪ1. The pressure difference remained and its relationship to high wind speed events at Pegasus steady at about 3.0 hPa until 1900 UTC, after which it North was investigated. underwent a second increase to 4.0 hPa by 2100 UTC. Marilyn and Schwerdtfeger lie roughly along 80ЊS, The pressure difference remained near 4.0 hPa until which also places these units along a line nearly per- 0130 UTC 18 September when it then began to decrease, pendicular to the Transantarctic Mountains (refer to Fig. diminishing to near 0.0 hPa by 0600 UTC. The wind 4). Marilyn is closest to the mountain chain, located speed at Pegasus North remained relatively strong approximately 100 km to the east of the Transantarctic throughout the day of 18 September.

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 145

FIG. 8. Pegasus North wind speed and Marilyn±Schwerdtfeger pres- FIG. 9. Pegasus North wind speed and Elaine±Marilyn pressure sure difference vs time, 16±17 Sep 1991. Pegasus North wind speed difference vs time, 12±14 Jun 1996. Pegasus North wind speed is is denoted by solid circles. Open squares represent the Marilyn± denoted by solid circles. Open squares represent the Elaine±Marilyn Schwerdtfeger pressure difference. pressure difference.

A northward-directed horizontal pressure gradient sure difference began to increase, rising from Ϫ3.1 to force would also indicate an environment suitable for 3.2 hPa by 0250 UTC 13 June. The wind speed at Peg- observing southerly winds at Pegasus North. Initially, asus North began to increase at 0900 UTC 13 June, such a pressure regime would result in easterly winds approximately 27 h after the increase of pressure dif- on the Ross Ice Shelf. But as the air encountered the ference began. The wind speed had increased to 15 m Ϫ1 Transantarctic Mountains, a barrier wind scenario would s by 1020 UTC 13 June, 7.5 h after the pressure dif- initiate a southerly wind in the region near and adjacent ference had reached its maximum. The wind speed con- to the Transantarctic Mountains. Elaine and Marilyn lie tinued to increase, even though the pressure difference Ϫ1 in nearly a south-southeast to north-northwest line and decreased, and reached a maximum of 22.1 m s at are roughly parallel to the Transantarctic Mountains, 1700 UTC, approximately 14 h after the maximum pres- with Elaine positioned 381 km to the south-southeast sure difference was reached. of Marilyn (refer to Fig. 4). The formation of a north- ward-directed horizontal pressure gradient force can be c. AWS data from selected sites observed using these two AWS sites. Again, pressure 1) MARILYN SITE differences between these two sites were calculated by subtracting the air pressure at Marilyn from that at As mentioned earlier, a temperature increase was a Elaine. A positive result would indicate that the station common phenomenon observed at many of the AWS pressure at Elaine was higher than that at Marilyn, hence sites before an increase in the wind speed at Pegasus a northward-directed horizontal pressure gradient force. North. In all but six cases the Marilyn site temperature The results of this simple calculation were then plotted increased during the 18 h preceding the high wind as a time series and compared to the wind speed ob- speeds at Pegasus North. The amount of temperature served at Pegasus North. increase ranged from 3.5Њ to 16.0ЊC, and the average Of the 109 high wind speed events studied, the increase was 9.7ЊC. Sometimes this was a gradual in- Elaine±Marilyn pressure differences were available for crease, of the order of 0.5Њ±1.0ЊChϪ1, and other times 66 of the cases. The Elaine±Marilyn pressure difference the increase was more rapid. The largest rate of tem- increased prior to the high wind speed in 57 of those perature increase observed was 5.5ЊChϪ1. The temper- cases. The amount of time between when the pressure ature usually remained relatively high throughout the differences began to increase and when the wind speed early portion of the event and then began to decrease. increased at Pegasus North also varied widely from case A decrease in wind speed at Pegasus North usually be- to case. The time between when the Elaine±Marilyn gan shortly thereafter. Also, there seemed to be no cor- pressure difference began to increase and the wind speed relation between the magnitude of the rate of temper- at Pegasus North increased to greater than 15 m sϪ1 ature increase at Marilyn site and the magnitude of the varied from 6 h to as much as 57 h and averaged just wind speed at Pegasus North. Six of the cases studied over 20 h. The increase of the pressure difference ended were accompanied by a temperature decrease that oc- between 46.5 h before and 9 h after the winds increased curred soon after the initial increase. There seemed to at Pegasus, with the average time being about 5.7 h be no correlation between those with an overall increase before the wind event. The range within one standard in temperature or those with a temperature increase soon deviation from the mean was 13.7 h before and 2.4 h followed by a similar decrease and the magnitude of the after the event (Fig. 5c). wind speed at Pegasus North. The range within one An example of this relationship can be seen in Fig. standard deviation of the mean between the time when 9. At 0550 UTC 12 June 1996, the Elaine±Marilyn pres- the temperature increase was complete and the wind

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 146 WEATHER AND FORECASTING VOLUME 15 speed increased to 15 m sϪ1 at Pegasus North was Ϫ2.1 occurred over a short period of time, averaging 4.2 h, to 16.5 h. The frequency distribution and various sta- but six of the cases had veering winds that occurred tistical data are shown in Fig. 5d. over a period of more than 10 h. The magnitude of the The winds at Marilyn site had a tendency to veer, veering ranged from 30Њ to 180Њ, but was generally be- de®ned in the Southern Hemisphere as a counterclock- tween 60Њ and 120Њ, and averaged 84Њ. The average wise rotation with respect to time, prior to an increase period of time between the completion of the veering in wind speed at Pegasus North. The wind direction and when the wind speed at Pegasus North increased generally veered from a westerly wind to a southerly to greater than 15 m sϪ1 was nearly 7.5 h, and the range wind. This veering of the wind direction may be due to within one standard deviation of the mean was Ϫ1.1 to a change in forcing mechanisms, from possible katabatic 16.0 h (Fig. 5g). ¯ow down the glaciers of the nearby Transantarctic Mountains, to one of a mesoscale or synoptic-scale dis- turbance. Also at times, the veering was followed by 5. Case study, 19 March 1991 the temperature increase, most likely due to warm air advection, but at other times the temperature increase The general synoptic pattern for this case had a low began hours before the winds began to veer. During most moving slowly southeastward into the Ross Sea and the of the events, this veering occurred rather quickly, av- northeast corner of the Ross Ice Shelf. Sea level pressure eraging 3.5 h, but ranged from 0.5 to 27 h. The amount on the southern end of the ice shelf began to rise during of veering ranged from 40Њ to 145Њ, but was generally 18 December 1991 while pressure on the northern end between 60Њ and 110Њ, and averaged 84Њ. The average of the ice shelf fell slightly due to the approach of the amount of time between when the veering was complete low. By 19 March the sea level pressure had risen dra- and the wind speed at Pegasus had increased to 15 m matically over the ice shelf especially along the Trans- sϪ1 was about 7.1 h, and the range within one standard antarctic Mountains. This pressure rise and nearly sta- deviation of the mean was Ϫ2.7 to 16.8 h (Fig. 5e). tionary low over the Ross Sea had caused a tight pres- sure gradient across the ice shelf and especially in the Pegasus runway area (Fig. 10). 2) SCHWERDTFEGER SITE The weather conditions at Pegasus North were fairly Data for 75 of the 109 cases were available from tranquil on 18 March 1991. The wind speed was less Schwerdtfeger site. A temperature increase was ob- than 8.0 m sϪ1 for most of the day, but the wind speed served at Schwerdtfeger site prior to the increase in wind began to slowly increase at 2000 UTC and continued speeds at Pegasus North in 68 of these cases. Usually, to increase the next day (Fig. 11a). The wind speed the temperature increase occurred in less than 18 h and reached 15 m sϪ1 by 0630 UTC 19 March and then was complete before the wind speed at Pegasus North increased much more rapidly. The speed increased to had increased to 15 m sϪ1. In those cases in which the 23.4 m sϪ1 in 1.5 h and reached a maximum of 25.7 m temperature increase occurred over a longer time period, sϪ1 at 0830 UTC. Wind speeds remained greater than the increase tended to be complete after the wind speed 15.0 m sϪ1 until 1230 UTC and then decreased rapidly. at Pegasus North had increased to 15 m sϪ1 or greater. There was also a marked temperature increase ob- Again, at times this temperature increase may be due served at Pegasus North prior to the increase in wind to warm air advection associated with a shift to southerly speed (Fig. 11a). The temperature began to rise from winds, but frequently, the temperature increase occurs Ϫ30.4ЊC at 1110 UTC on 18 March and reached long before any change in the wind direction. The mag- Ϫ20.0ЊC at 2130 UTC that same day. The temperature nitude of the temperature increases ranged from 5Њ to remained near Ϫ20ЊC throughout the event and then 39ЊC and began from 52.7 h before and 1.5 h after the decreased rapidly. wind speed at Pegasus North had increased to 15 m sϪ1. A shift in wind direction occurred during the event. The average temperature increase was 13.3ЊC, and the The wind direction was between 60Њ and 120Њ during average duration of the increase was 12.5 h. The average much of the day on 18 March (Fig. 11b). The wind period of time between the completion of the increase began to shift from 50Њ at 1930 UTC to 150Њ at 2230 and when the wind speed at Pegasus North increased UTC. As the wind speed continued to increase, the wind to greater than 15 m sϪ1 was 6.1 h, and the range within direction continued to shift and was 195Њ during the one standard deviation of the mean was from Ϫ4.0 to period of the highest wind speeds. As the wind speed 16.1 h (Fig. 5f). began to decrease, the direction then returned to about The winds at Schwerdtfeger site also had a tendency 60Њ. to veer prior to an increase in wind speed at Pegasus Temperatures at Marilyn were decreasing throughout North. Veering winds were observed in 54 of the 62 the day on 18 March 1991 (Fig. 11c). At 0010 UTC 19 cases in which data was available. The wind direction March the temperature increased from Ϫ39.1Њ to generally veered from a westerly wind to a southerly Ϫ29.4ЊC in 4 h and then began to decrease. The max- wind, again possibly due to a change in the forcing imum temperature observed at Marilyn occurred 2.5 h mechanisms. During most of the events, this veering before the wind speed at Pegasus North increased to 15

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 147

FIG. 10. ECMWF surface analyses for (a) 0000 UTC 18 Mar 1991, (b) 1200 UTC 18 Mar 1991, (c) 0000 UTC 19 Mar 1991, and (d) 1200 UTC 19 Mar 1991. Light solid lines are isobars (mb) plotted at an interval of 4 mb. The black dot denotes the location of the Pegasus runway. msϪ1, and nearly 4 h before the maximum wind speed Ϫ29.0ЊC for 4 h and then began to decrease. The tem- was observed at Pegasus North. perature increase occurred as the wind direction began The change in wind direction exhibited the veering to veer from north-northwest (330Њ) to south (180Њ)in observed during many of these events. At 1600 UTC about 6 h, beginning at 1300 UTC 18 March (Fig. 11f). 18 March, the wind began to veer from approximately The wind direction remained southerly throughout the 280Њ to 180Њ in 5 h (Fig. 11d). The wind direction re- event and then backed to north-northwesterly afterward. mained between 180Њ and 220Њ throughout most of the The Marilyn±Schwerdtfeger pressure difference be- event, and then shifted abruptly to 325Њ shortly after the gan to increase from Ϫ3.5 hPa at 1920 UTC 18 March wind speed at Pegasus North decreased. to 1.6 hPa at 0650 UTC 19 March (Fig. 11g). The pres- The temperature at Schwerdtfeger increased from sure difference then decreased to less than Ϫ4.0 hPa by Ϫ41.9Њ to Ϫ27.0ЊC in 14 h, beginning at 1310 UTC 18 1600 UTC 19 March. The maximum wind speed ob- March (Fig. 11e). The temperature remained near served at Pegasus North occurred nearly 2 h after the

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 148 WEATHER AND FORECASTING VOLUME 15

FIG. 11. Time series of AWS data for 18 and 19 Mar 1991. Pegasus North wind speed is denoted by solid circles on all graphs. Open squares represent (a) Pegasus North temperature, (b) Pegasus North wind direction, (c) Marilyn temperature, (d) Marilyn wind direction, (e) Schwerdtfeger temperature, (f) Schwerdtfeger wind direction, (g) Marilyn±Schwerdtfeger pressure difference, and (h) Pegasus North±Minna Bluff temperature difference.

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 149

Marilyn±Schwerdtfeger pressure difference reached its maximum. Unfortunately, the data necessary for deter- mining the Elaine±Marilyn pressure difference were un- available for this case. The temperature difference, as calculated from Peg- asus North and Minna Bluff temperatures, also increased during this event. Between 0530 and 2240 UTC 18 March, the TD increased from Ϫ3.9Њ to 11.2ЊC (Fig. 11h). The TD remained between 10.0Њ and 10.9ЊC for the next 6 h before exhibiting some ¯uctuation and then increasing further to 16.2ЊC at 0900 UTC 19 March. Soon after, the TD decreased to less than 0.0ЊC by 1600 UTC 19 March. In this particular case, the Pegasus North±Minna Bluff TD began to increase nearly a day before the wind FIG. 12. Success rates of AWS parameters for both a signi®cant speed at Pegasus North reached the threshold of 15 m wind speed increase and a wind speed increase in excess of 15 m sϪ1 and increased to statically unstable levels at ap- sϪ1 at Pegasus North AWS site. proximately the same time as the Pegasus North wind speed began to increase. The wind direction at both Schwerdtfeger and Marilyn had veered from 330Њ to as a percentage of the wind speed increases that occurred 180Њ in approximately 12 and 9.5 h, respectively, before after the de®ned parameter was met. the wind speed at Pegasus North reached threshold lev- The results obtained were categorized in both a year- els. The temperature at Schwerdtfeger began to increase round tabulation as well as a seasonal grouping, and approximately 7 h before the wind speed at Pegasus both varied among the individual parameters. The TD North, while the temperature at Marilyn increased co- underwent the de®ned increase 299 times during the incidentally with the wind speed at Pegasus North. 6-yr time period, and an increase in the Pegasus North wind speed to levels above 15 m sϪ1 occurred after 81 of the TD increases, resulting in a success rate of 27.1%. 6. Success rates Also, the wind speed at Pegasus North underwent a We have shown that on a large majority of occasions, signi®cant increase in 244 of those cases, a success rate certain changes occur in various meteorological param- of 81.6%. Year-round success rates for each parameter eters obtained from AWS data before the wind speed at are shown in Fig. 12. Other parameters had success rates Pegasus North AWS site increases to 15 m sϪ1. For this near or slightly below the rates calculated for the TD, information to be useful as a forecasting aid, the reverse but the temperature increase at both Schwerdtfeger and must also be evident. We must answer the question, Marilyn AWS sites had success rates much lower for ``How often do these changes occur and no increase in both the Pegasus North wind speed increase to 15 m the Pegasus North wind speed follows?'' To answer this sϪ1 and the signi®cant wind speed cases. Many of the question, the data were scanned for changes in the pa- temperature increases at both Marilyn and Schwerdt- rameters discussed previously. Speci®cally, changes feger are most likely attributed to the diurnal cycle and were considered that occurred in 18 h or less and were therefore are not indicative of a meteorological event as follows: a TD increase from near 0.0ЊC to at least that could cause a substantial increase in the surface 8.9ЊC; an increase of the Marilyn±Schwerdtfeger pres- wind speed at Pegasus North. The largest seasonal var- sure difference of at least 2.5 hPa, and an increase of iation in the success rate occurred with the TD (Fig. the Elaine±Marilyn pressure difference of at least 6.0 13). The success rate for the TD in the summer months hPa; a shift in the wind direction at both Marilyn and (December, January, and February) of 61.5% was sig- Schwerdtfeger sites from a westerly direction to a south- ni®cantly less than the rate of 96.3% in the winter erly direction; and an increase in the temperature at both months (June, July, and August). Conversely, the var- Marilyn and Schwerdtfeger sites of at least 6ЊC. Results iations in the success rate were much smaller for both were tabulated for two scenarios. It was ®rst determined the two wind direction parameters and the pressure dif- how often a change in a parameter was observed and ference parameters. Also, the success rates for both the an increase in the Pegasus North wind speed to 15 m Marilyn and Schwerdtfeger temperature parameters are sϪ1 followed within a 24-h period. It was also deter- relatively smaller than the other parameters in each sea- mined how often a signi®cant increase in the wind speed son. at Pegasus North occurred within the 24-h period. A signi®cant increase in the wind speed at Pegasus North 7. Conclusions was de®ned as at least a doubling of the wind speed that ultimately resulted in a wind speed of at least 6 m The primary purpose for the deployment of the AWS sϪ1. For each parameter, a success rate was determined units near Ross Island and on the Ross Ice Shelf was

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC 150 WEATHER AND FORECASTING VOLUME 15

occurred only when the TD was unstable. It was shown that an increase in the TD precedes a high wind speed event at Pegasus North by several hours. This was ev- ident in nearly every case examined and is probably the most reliable predictor obtained in this study. Several empirical guidelines for forecasting high wind speeds at Pegasus were developed from this study. No single guideline is adequate to forecast a high wind speed event at Pegasus North. Both the magnitudes of the individual parameters as well as the duration of the changes of those parameters can vary greatly from one event to another. Many times these changes occur abruptly, while at other times, the change may occur over a period of 24 h or more, with the majority of these changes occurring in less than 18 h. By monitoring the FIG. 13. Seasonal success rates of AWS parameters for a signi®- trends of the parameters outlined in this study and using cant wind speed increase at Pegasus North AWS site. them in conjunction with the other data available at McMurdo Station, better forecasts of these events can be made. to support operational forecasting. The purpose of this The following will generally precede the onset of high study was to derive empirical rules based on automatic wind speeds at Pegasus by up to6hifthey occur over weather station data for forecasting high wind speed a period of not more than 18 h: events at Pegasus Runway, Antarctica. The results of 1) an increase in the TD (estimated from the temper- the work presented here give several guidelines to im- atures at Minna Bluff and Pegasus North) from in- prove both the accuracy and lead time of high wind version conditions (less than 0.0 C) to statically un- speed and, hence, visibility forecasts. While these guide- Њ stable conditions (greater than 8.9 C); lines lack robustness, they do demonstrate the merit of Њ 2) pressure differencesÐ(a) an increase in the Marilyn± the use of AWS data for this type of forecast. The 109 Schwerdtfeger pressure difference, usually on the cases examined and the case study discussed have order of 2.5 hPa or greater, and (b) an increase in shown that the data from AWS sites can be used for the Elaine±Marilyn pressure difference on the order speci®c forecasting problems. of 6.0 hPa or greater; and A substantial increase in temperature was observed 3) a veering of the wind direction with time, from west at most of the AWS sites before a high wind speed event or southwesterly to southerly at the Marilyn and/or at Pegasus North. The temperature increases were usu- Schwerdtfeger sites. ally greatest at Schwerdtfeger, Marilyn, and Pegasus North, and those at Schwerdtfeger and Marilyn sites Although the use of the Marilyn and Schwerdtfeger were generally complete at least 3 h before the wind temperature increases may have some merit, the rela- speed increased at the Pegasus North site. Additionally, tively poor success rate excludes these two parameters a shift in the wind direction at both Marilyn and from inclusion as forecasting aids. These guidelines, Schwerdtfeger sites can lead to a forecast for high wind used in conjunction with radiosonde data, synoptic speeds at Pegasus North. The wind directions at these charts, satellite data, and the forecaster's knowledge of two AWS sites had a tendency to veer from a westerly the local climatology, should improve both the lead time direction to a southerly direction. This veering usually and accuracy of high wind speed forecasts at Pegasus occurred 3±6 h before wind speeds at Pegasus North North. increased. The station pressure differences computed between Acknowledgments. This research was funded by the the stations near the Transantarctic Mountains also pro- National Science Foundation's Of®ce of Polar Programs vided data that could lead to a forecast of high wind under Grant 9303569. speeds at Pegasus North. The Marilyn±Schwerdtfeger and Elaine±Marilyn pressure differences proved useful predictors of high wind speeds at Pegasus. REFERENCES The temperature difference, as calculated between Blaisdell, G. L., V. D. Klokov, and D. Diemand, 1995: Compacted Minna Bluff and Pegasus North sites, also proved a snow runway technology on the Ross Ice Shelf near McMurdo, useful indicator. The time series of TD versus wind Antarctica. Contributions to Antarctic Research IV, D. H. Elliot speed and wind direction revealed that when the TD and G. L. Blaisdell, Eds., Antarctic Research Series, Vol. 67, Amer. Geophys. Union, 153±174. was unstable (greater than 8.9ЊC), the winds at Pegasus Bromwich, D. H., 1988: Snowfall in high southern latitudes. Rev. North tended to blow from the south, and that wind Geophys., 26, 149±168. speeds at Pegasus North greater than 10 m sϪ1 generally , J. F. Carrasco, and C. R. Stearns, 1992: Satellite observations

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC APRIL 2000 HOLMES ET AL. 151

of katabatic wind propagation for great distances across the Ross Schwerdtfeger, W., 1984: Weather and Climate of Antarctica. Elsev- Ice Shelf. Mon. Wea. Rev., 120, 1940±1949. ier, 261 pp. Fleming, D. A., 1983: Antarctic automatic weather stations as fore- Sinclair, M. R., 1988: Local topographic in¯uence on low-level wind casting aids. Antarct. J. U.S., 18 (5), 302±303. at Scott Base, Antarctica. N. Z. J. Geol. Geophys., 31, 237±245. Holmes, R. E., 1994: An investigation into the use of automatic Sinkula, B. B., 1993: Application of AWS data for Antarctic oper- weather station data for the forecasting of high wind speed events ational forecasting and climate research. M.S. thesis, Department at Pegasus Runway, Antarctica. M.S. thesis, Department of At- of Atmospheric and Oceanic Sciences, University of Wiscon- mospheric and Oceanic Sciences, University of WisconsinÐ sinÐMadison, 117 pp. [Available from Department of Atmo- Madison, 111 pp. [Available from Department of Atmospheric spheric and Oceanic Sciences, University of Wisconsin, 1225 and Oceanic Sciences, University of WisconsinÐMadison, 1225 West Dayton St., Madison, WI 53706.] W. Dayton St., Madison, WI 53706.] Stearns, C. R., and G. A. Weidner, 1990: Wind speed events and wind , and C. R. Stearns, 1995: Antarctic automatic weather stations: direction at Pegasus site during 1989. Antarct. J. U.S., 25 (5), Austral summer 1994±1995. Antarct. J. U.S., 30 (5), 327±329. 258±262. , G. A. Weidner, and C. R. Stearns, 1993: Antarctic weather , and , 1991: Wind speed, wind direction, and air temper- forecasting: 1992±1993. Antarct. J. U.S., 28 (5), 337. ature at Pegasus North during 1990. Antarct. J. U.S., 26 (5), Keller, L. M., G. A. Weidner, C. R. Stearns, M. T. Whittaker, and R. 251±253. E. Holmes, 1996: Antarctic automatic weather station data for , and , 1992: Wind speed, wind direction, and air temper- the calendar year 1994. Department of Atmospheric and Oceanic ature at Pegasus North during 1991. Antarct. J. U.S., 27 (5), Sciences, University of WisconsinÐMadison, 465 pp. [Avail- 285±287. able from L. M. Keller, Department of Atmospheric and Oceanic , and B. B. Sinkula, 1994: Antarctic weather forecasting: 1993. Sciences, University of WisconsinÐMadison, 1225 West Dayton Antarct. J. U.S., 29 (5), 284. St., Madison, WI 53706.] , L. M. Keller, G. A. Weidner, and M. Sievers, 1993: Monthly , , , , and , 1997: Antarctic automatic weather mean climatic data for Antarctic automatic weather stations. Ant- station data for the calendar year 1995. Department of Atmo- arctic Meteorology and Climatology: Studies Based on Auto- spheric and Oceanic Sciences, University of WisconsinÐMad- matic Weather Stations, D. H. Bromwich and C. R. Stearns, Eds., ison, 539 pp. [Available from L. M. Keller, Department of At- Antarctic Research Series, Vol. 61, Amer. Geophys. Union, 1± mospheric and Oceanic Sciences, University of WisconsinÐ 22. Madison, 1225 West Dayton St., Madison, WI 53706.] , G. A. Weidner, and R. E. Holmes, 1994: Antarctic automatic Savage, M. L., and C. R. Stearns, 1985: Climate in the vicinity of weather stations: Austral summer 1993±1994. Antarct. J. U.S., Ross Island, Antarctica. Antarct. J. U.S., 20 (1), 1±9. 29 (5), 281±284.

Unauthenticated | Downloaded 09/26/21 02:59 AM UTC