Numerical Prediction of an Antarctic Severe Wind Event with the Weather Researchand Forecasting (WRF) Model
JORDAN G. POWERS
Mesoscale and Microscale Meteorology Division, Earth and Sun Systems Laboratory, National Center for Atmospheric Research,Boulder, Colorado
(Manuscript received 10 August 2006, in final form 30 October 2006)
ABSTRACT
This study initiates the application of the maturing Weather Research and Forecasting (WRF) model tothe polar regions in the context of the real-time Antarctic Mesoscale Prediction System (AMPS). Thebehavior of the Advanced Research WRF (ARW) in a high-latitude setting and its ability to capture asignificant Antarctic weather event are investigated. Also, in a suite of sensitivity tests, the impacts of theassimilation of Moderate Resolution Imaging Spectroradiometer (MODIS) atmospheric motion vectors onARW Antarctic forecasts are explored. The simulation results are analyzed and the statistical significanceof error differences is assessed. It is found that with the proper consideration of MODIS data the ARW canaccurately simulate a major Antarctic event, the May 2004 McMurdo windstorm. The ARW simulationsilluminate an episode of high-momentum flow responding to the complex orography of the vital Ross Islandregion. While the model captures the synoptic setting and basic trajectory of the cyclone driving the event,there are differences on the mesoscale in the evolution of the low pressure system that significantly affectthe forecast results. In general, both the ARW and AMPS’s fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5) tend to underforecast the windmagnitudes, reflecting their stalling and filling of the system near Ross Island. It is seen, however, that bothtargeted data assimilation and grid resolution enhancement can yield improvement in the forecast of the keyparameter of wind speed. It is found that the assimilation of MODIS observations can significantly improvethe forecast for a high-impact Antarctic weather event. However, the application to the retrievals of a filteraccounting for instrument channel, observation height, and surface type is necessary. The results indicatebenefits to initial conditions and high-resolution, polar, mesoscale forecasts from the careful assimilation ofnontraditional satellite observations over Antarctica and the Southern Ocean.
1. Introduction
The Weather Research and Forecasting (WRF)model (Skamarock et al. 2005) has been developed as anext-generation mesoscale modeling system for bothoperational prediction and atmospheric research. Thenonhydrostatic WRF (information available online athttp://www.wrf-model.org/index.php) has arguably thelargest user base of any current mesoscale model and isreplacing the fifth-generation Pennsylvania State Uni-versity–National Center for Atmospheric Research(PSU–NCAR) Mesoscale Model (MM5; Grell et al.
1995) in that user community.1 It is being used by offi-cial forecasting centers [e.g., the National Centers forEnvironmental Prediction (NCEP)], while seeing re-search applications ranging from large eddy simula-tions, to severe weather at convection-permitting reso-lutions, to tropical cyclogenesis, to regional climatemodeling (see, e.g., Moeng et al. 2007; Davis et al. 2006;Done et al. 2006).
The developmental applications of WRF have prima-rily been in the midlatitudes, and to date the high lati-tudes have been largely ignored. In light of this, andgiven the growing use of WRF worldwide and thematuration of the model’s capabilities, this study ini-tiates the application of WRF to Antarctica. AlthoughCorresponding author address: Dr. Jordan G. Powers, Meso-
scale and Microscale Meteorology Division, Earth and Sun Sys-tems Laboratory, National Center for Atmospheric Research,P.O. Box 3000, Boulder, CO 80307.E-mail: [email protected]
1 As of this writing, WRF has over 5000 registered users in over90 countries.
3134 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
addressing the interest of the WRF user community inthe model’s behavior in the high latitudes, the morespecific concern is investigating its ability to capture asignificant polar weather event. The target is the May2004 windstorm at McMurdo Station, Antarctica (Pow-ers et al. 2005; Steinhoff and Bromwich 2005). Whilethis was an extraordinary episode that inflicted winds ofup to 71 m s�1 (139 kt) on the main American researchbase in Antarctica (McMurdo Station, Fig. 1), the ac-
curate forecasting of surface flow in Antarctica is im-portant for a number of reasons. The lifeline of flightoperations depends on reliable wind forecasts (see, e.g.,Holmes et al. 2000), as flows beyond specified thresh-olds can exceed crosswind tolerances for takeoffs andlandings and blowing snow can obscure runways. Bothsituations are of critical importance in the case of air-craft that absolutely must land at McMurdo after pass-ing their points of safe return. Second, activities in the
FIG. 1. AMPS domains and Antarctic locations. (a) 90- (60)and 30- (20) km grids. (b) 30- (20) km grid with 10- (6.7) and 3.3-(2.2) km grids inset. (c) 10- (6.7) km grid with 3.3- (2.2) km gridinset. Dots mark observation/AWS sites discussed in the text.Grid spacings in parentheses refer to the 60-/20-/6.7-/2.2-km(MOD1_60) setup.
SEPTEMBER 2007 P O W E R S 3135
harsh polar field may face life-threatening conditionsdependent on the strength and duration of surfacewinds. Third, operations at the focal facility of Mc-Murdo may simply be shut down under excessive ve-locities (see footnote 4). Beyond these concerns is theneed to be able to simulate and analyze, for both fore-casting and research purposes, the ubiquitous katabaticand other outflows endemic to Antarctica. A prime ex-ample is the Ross Ice Shelf Air Stream (RAS; Brom-wich and Parish 2002).2
The May 2004 case serves as a vehicle for the first-time Antarctic application of the Advanced ResearchWRF (ARW; Skamarock et al. 2005). Sharing the WRFsoftware framework, this features the ARW dynamicssolver (originally referred to as the “Eulerian masscore”) and its mass-based vertical coordinate, a grid-nesting capability, numerous physical process schemes,and the WRF-Var (variational) data assimilation sys-tem (Barker et al. 2004). In addition to beginning tounderstand and advance the capability of the ARW forpolar prediction, this work is concerned with the im-provement of a real-time Antarctic numerical weatherprediction (NWP) facility known as the AntarcticMesoscale Prediction System (AMPS; Powers et al.2003). Funded by the National Science Foundation,AMPS is an experimental mesoscale modeling systemthat provides forecast guidance in support of the flight,scientific, and logistical activities of the U.S. AntarcticProgram (USAP) and international Antarctic efforts.While originally established to improve the numericalguidance available to the USAP forecasters3 at Mc-Murdo Station, over the years AMPS has expanded toserve a broad range of international groups and activi-ties (including emergency rescues) across Antarctica(see, e.g., Monaghan et al. 2003). Historically AMPShas relied on the MM5 (Powers et al. 2003; Bromwichet al. 2003), but it has begun also to employ the ARW.(AMPS forecasts may be accessed online at http://www.mmm.ucar.edu/rt/amps/wrf_pages.)
A challenge to Antarctic NWP is the lack of conven-tional (e.g., surface observational and radiosonde) dataover the Southern Ocean and the continent. Antarcticahas approximately 10 radiosonde sites, and, except forthe Amundsen–Scott Station at the South Pole, theseare situated around the continent’s edge. Satellite mea-
surements can populate the data void, however, and apromising source is the Moderate Resolution ImagingSpectroradiometer (MODIS) instrument aboard theNational Aeronautics and Space Administration(NASA) Terra and Aqua polar-orbiting platforms.Over the polar regions MODIS imagery can yield at-mospheric motion vectors (AMVs; Velden et al. 2005;Key et al. 2003) offering vector winds at varyingheights. Available in near–real time, MODIS polarwinds have shown promise in improving performancemetrics in global NWP models (Key et al. 2003; Bor-mann and Thépaut 2004). Given the potential for thesemeasurements to enhance polar forecasting, this studyalso investigates the impact of the assimilation ofMODIS winds on the skill of a mesoscale model, (viz.the ARW) in the Antarctic.
This investigation thus provides the growing WRFuser community with an initial test and analysis of thisnew modeling capability in the polar regions. The May2004 McMurdo windstorm is the vehicle for the ARW’sapplication to a challenging forecast for this criticalarea, for the ARW’s comparison with the MM5 inAMPS, and for sensitivity experiments to exploremodel performance with different MODIS data assimi-lations and grid resolutions. The foci are how well theARW can simulate a high-impact polar weather eventand whether mesoscale ARW forecasts in AMPS maybe improved by MODIS data assimilation. Section 2 ofthis paper describes the observed event, while section 3covers the ARW configuration and experiments. Sec-tion 4 examines the ARW event simulations and ana-lyzes the MODIS sensitivity experiments. Section 5pursues a statistical evaluation of the results, and sec-tion 6 presents a summary and conclusions.
2. The May 2004 McMurdo windstorm
On 15 May 2004 extreme winds battered the Mc-Murdo Station, Antarctica region (see Fig. 1 for mapsof Antarctica and the McMurdo area) and locked downbase activities as “condition 1” status4 was declared.The event winds, out of the south, were sustained atover 44 m s�1 (86 kt) and exceeded 52 m s�1 (102 kt) ingusts in town. One building on a hill above the basecenter claimed gusts to 71 m s�1 (139 kt). The windspounded structures and equipment, blew in doors, toreup roofs, and peeled siding off dormitories. The veloci-
2 The RAS is the broad, northward-moving airstream adjacentto the Transantarctic Mountains. It extends across the Ross IceShelf from its southern edge and through the western Ross Sea. Itis a primary transport channel between Antarctica and the South-ern Hemisphere (Bromwich and Parish 2002).
3 The forecasters are employed/contracted by the Space andNaval Warfare Systems Center (SPAWAR), Charleston, SC.
4 Condition 1 is defined by visibility less than 100 ft (31 m), orwinds exceeding 55 kt (28 m s�1), or wind chill temperaturescolder than �100°F/�73°C. A declaration of condition 1 meansthat personnel must remain indoors where they are at the time.
3136 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
ties hit abruptly, and the peak hours were 1800 UTC 15May–0000 UTC 16 May 2004. Figure 2 presents timeseries of wind speeds at sites in the area. A number ofthe observational records from the vicinity are trun-cated, reflecting the winds’ damaging of the meteoro-logical instruments [e.g., Crater Hill and Helo Pad (he-licopter pad, in town)].
The event was motivated by the passage of a deepsynoptic low pressure system to the east of Ross Island.Figure 3 shows the track of this system, based on theanalysis of satellite imagery.5 Over 14–15 May 2004,after having moved in from the Amundsen Sea, the
5 The orientation of Fig. 3 and subsequent figures—with theSouth Pole at the top—corresponds to the available satellite im-agery perspective (Fig. 4). Figure 3 (and Fig. 17) provides direc-tional information (N, S, E, W indicated relative to Ross Island)to assist in the references to compass directions.
FIG. 2. Observed wind speeds (m s�1) at sites in the McMurdo area for 0000 UTC 15 May–0000 UTC17 May 2004. Cosray, Arrival Heights, Helo Pad, and Crater Hill are all in the immediate McMurdovicinity; Black Island AWS is about 34 km to the south.
FIG. 3. Track of the observed low. Times (UTC) of centralposition of low (marked “L”) indicated. Italic L reflects positionsduring the observed wind event at McMurdo.
SEPTEMBER 2007 P O W E R S 3137
center traveled from Marie Byrd Land, across the SipleCoast, and onto the Ross Ice Shelf (Fig. 1b). The pres-ence of the low in the Siple Coast region is consistentwith that area exhibiting a maximum in cyclone densityfor the winter months (Simmonds et al. 2003). Aftercrossing 180° at around 1500 UTC 15 May 2004, the lowturned northward. Seasoned operational forecasters atMcMurdo know that that area is subject to its strongestwinds when low pressure systems track from the southover the ice shelf and remain east of Ross Island (R.Hennig, Space and Naval Warfare Systems Center,2005, personal communication; see also Holmes et al.2000).
Figure 4 presents IR imagery of the system. At 1455UTC 15 May 2004 (Fig. 4a) the low sits in midshelf. The2125 UTC image reveals the center during the event,south and east of McMurdo, positioned east of MinnaBluff and White Island (Fig. 4b). The system propa-gates northward east of Ross Island, and after 0000UTC 16 May 2004 it weakens and moves across Frank-lin Island to the Terra Nova Bay area.
While the migrating cyclone is essential for the windsin the McMurdo area, the forcing pressure gradient,enhanced on the mesoscale, is a reflection of the syn-
optic conditions established by the low and a stationaryhigh pressure system over East Antarctica. Thisbroader picture appears in Figs. 5 and 6, which presentsurface and 500-hPa analyses for 15–16 May 2004.These analyses are from the European Centre for Me-dium-Range Weather Forecasts (ECMWF) globalmodel. Note in Fig. 5a the strong (�1032 hPa) surfacehigh over East Antarctica at 1200 UTC 15 May 2004.At this time the surface low sits downstream of the500-hPa trough over the western the Ross Ice Shelf;that trough is rotating around the upper-level cutoffover the Ross Sea (Figs. 5a and 6). Sea level pressuresover the Ross Sea sector are low relative to the EastAntarctic high. Because of the potential for cumulativeerror in deriving below-ground SLP analyses over theelevated Antarctic continent, Fig. 5b presents the 600-hPa height analysis to confirm the synoptic gradient.
The synoptic-scale pressure gradient across the iceshelf, the Transantarctic Mountains (TAM), and EastAntarctica is thus established by the high and the mi-grating low, and the wind event at McMurdo attendsthe imposition of an enhanced mesoscale gradient onthe Ross Island region. The axis of the 500-hPa trough(dashed line in Fig. 6) remains upstream of the surface
FIG. 4. IR satellite imagery (MODIS, 5 km) for Ross Sea sector, 15 May 2004. The L indicates the center of the surface low: (a)1455 and (b) 2125 UTC. MB and WI indicate Minna Bluff and White Island.
3138 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
low until approximately 0000 UTC 16 May 2004. Thisoffers upper-level support through divergence and posi-tive vorticity advection aloft in accordance with quasi-geostrophic theory (see, e.g., Holton 1992; Bluestein1992; Carlson 1998).
3. Model configurations and experiments
The model experiments here employ the ARW, ver-sion 2 (Skamarock et al. 2005), while the real-timeAMPS forecast for the event used the MM5. Figures1a,b show the grids for the experiments. The primaryfour-domain configuration has horizontal grid spacingsof 90 km (Southern Hemisphere/Southern Ocean), 30km (Antarctic continent), 10 km (western Ross Sea),
and 3.3 km (Ross Island; “90/30” setup). All nesting istwo-way interactive. One higher-resolution ARW ex-periment is run with the same grid areas as in Fig. 1, butwith 60-, 20-, 6.7-, and 2.2-km spacings (“60/20” setup).In the vertical the model is run with 31 half-levels witha model top at 50 hPa. The initialization time is 0000UTC 15 May 2004. The 90/30 grid configuration em-ployed here has been chosen to be consistent with thatof the real-time AMPS running at the time of the event.
The experiments enlist the WRF single-moment,five-species microphysics scheme (WSM5) (Hong et al.2004), the Mellor–Yamada–Janjic (aka Eta) planetaryboundary layer (PBL) scheme (Mellor and Yamada1982; Janjic 2002), and the Kain–Fritsch cumulus pa-
FIG. 5. (a) SLP and (b) 600-hPa heights at 1200 UTC 15 May2004 from ECMWF analyses. The letters L and H in (a) mark thesurface low and the East Antarctica high and in (b) mark theupper-level Ross Sea and East Antarctica centers. Contour inter-val is 6 hPa in (a) and 60 gpm in (b).
FIG. 6. 500-hPa heights and winds from ECMWF analyses.Heights (solid lines); contour interval is 60 gpm. Wind vectors(arrows); magnitudes are 25 m s�1 (vector length interval)�1.Dashed line marks trough axis. (a) 1200 UTC 15 May and (b) 0000UTC 16 May 2004.
SEPTEMBER 2007 P O W E R S 3139
rameterization (Kain and Fritsch 1993; see also Skama-rock et al. 2005). The 3.3-km (2.2 km) domain is runfully explicit. This scheme configuration is chosen be-cause the counterpart package was that which was runsuccessfully in AMPS (see Bromwich et al. 2005). TheNoah land surface model (Chen and Dudhia 2001) isalso active. Capabilities to ingest sea ice analyses and torepresent sea ice concentration fractionally in grid cellsdo not exist at present in the ARW. Thus, for theseexperiments full sea ice coverage is assumed at oceanicpoints where the skin temperature is less than 271.4 K.
Initial and boundary conditions are derived from theNCEP Global Forecast System (GFS) output. In thedata assimilation experiments, the GFS first-guess fieldis reanalyzed with observations using the WRF-Vardata assimilation system (Barker et al. 2004; Skamarocket al. 2005). WRF-Var offers a three-dimensional varia-tional data assimilation (3DVAR) capability within theWRF software framework. It is run on the outer (90/30and 60/20 km) grids, as in the real-time AMPS. Thebackground error covariances were generated via theNMC method. For the assimilation, the observationsreferred to as “standard” consists of the conventionaldata acquired from the Global Telecommunication Sys-tem (GTS) circuit and used in a regular AMPS run:reports from manned surface stations (e.g., SYNOP,METAR), surface automatic weather stations (AWSs),upper-air stations, ships, buoys, pilot and Aircraft Me-teorological Data Relay (AMDAR) reports, and (geo-stationary) satellite cloud-track winds.
The real-time AMPS MM5 grid configuration fea-tured the 90-, 30-, 10-, and 3.3-km domains shown inFig. 1, as well as 10-km grids over the South Pole andthe Antarctic Peninsula (not shown). The MM5 was runwith 31 half-� levels from the surface to 50 hPa, and itsinitial and boundary conditions were derived from theGFS. WRF-Var assimilated the standard GTS observa-tions. Sea ice analyses from the National Snow and IceData Center (NSIDC) initialized the sea ice coverage.
Note that AMPS employs the “Polar MM5” (Brom-wich et al. 2001; Cassano et al. 2001). This is a modi-fied version of the model that was developed by thePolar Meteorology Group of the Byrd Polar ResearchCenter (at The Ohio State University) and containsadjustments to improve performance in the polar re-gions and to better capture features unique to extensiveice sheets. The Polar MM5 modifications include ac-counting for a separate sea ice category with specifiedthermal properties, use of forecast cloud species in theradiation scheme, representing fractional sea ice cover-age in grid cells, using the latent heat of sublimation forcalculations of latent heat fluxes over ice surfaces, and
assuming ice saturation when calculating surface satu-ration mixing ratios over ice.
This study also investigates the assimilation ofMODIS wind data on ARW forecasts of a significantAntarctic weather episode. This is motivated by thepotential of MODIS wind measurements for high-latitude forecast improvement, given their focus on therelatively data-sparse polar regions, and the absence ofan examination of MODIS data’s impacts on a high-resolution, mesoscale model simulation of a criticalevent (cf. global models or period studies; e.g., Key etal. 2003; Velden et al. 2005). The MODIS wind data areproduced from sequential imagery tracking of height-assigned cloud and water vapor features detected bythe MODIS instrument’s infrared (IR) and water vapor(WV) channels (Velden et al. 2005; Key et al. 2003).For this study, the Cooperative Institute for Meteoro-logical Satellite Studies (CIMSS; at the University ofWisconsin) MODIS data are used. Examining the as-similation of a 30-day MODIS dataset on the ECMWFglobal model and the NASA Data Assimilation Officemodel, Key et al. (2003) found significant improve-ments in skill scores for geopotential height forecasts.In a subsequent study, Bormann and Thépaut (2004)found that the incorporation of MODIS data usingfour-dimensional variational data assimilation (4DVAR)had a positive effect on medium-range forecasts by theECMWF global model. Neither of these studies, how-ever, dissected case impacts in a high-resolution meso-scale model forecast.
We performed ARW experiments involving the as-similation of the conventional (or standard) AMPS ob-servations and the CIMSS MODIS data. The followingtests are conducted:
• CTRL—No data assimilation;• STD—Standard AMPS data only;• ALL—Standard AMPS data plus all MODIS data;• MOD1—Standard AMPS data plus filtered MODIS
data; and• MOD1_60—As in MOD1, but for a 60-, 20-, 6.7-, and
2.2-km domain setup.
CTRL is a run involving no data assimilation. In STDthe standard AMPS data only are assimilated. ALLentails the incorporation of all standard data plus all ofthe MODIS observations in an assimilation windowaround 0000 UTC 15 May 2004. There is no exclusionof any MODIS data apart from the possible rejectionby WRF-Var’s quality control (QC) criteria.
In MOD1 a subset of MODIS observations, one re-maining after the application of a filter, is used with thestandard data. Reflecting lower confidence in the esti-
3140 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
mates in certain regimes, the filtering follows the sug-gestion of Key et al. (2003), later applied by Bormannand Thépaut (2004), in which retrieval height, surfacetype, and source MODIS channel (IR or WV) are con-sidered in accepting a measurement. Specifically, theprobability of poorer-quality retrievals below certainlevels is the reason for the restrictions (Key et al. 2003).For MOD1, the filtering criteria are as follows: overland, both IR and WV data above 400 hPa are retained;over ocean, IR data above 700 hPa and WV data above550 hPa are retained; other measurements are rejected.A number of operational forecasting centers (e.g.,ECMWF, the Met Office, Japan MeteorologicalAgency, Deutscher Wetterdienst) have adopted modi-fications of these based on their own models and expe-rience (see, e.g., Forsythe and Berger 2004). For theexperiments here, the filtering reduces the number ofmeasurements considered by approximately 33%.
The impact of the filtering may be seen by comparingthe MODIS data used in ALL and MOD1 and thedifference in their initializations. Figures 7a,b revealhow the filtering reduces the number of the MODISwinds, at all heights, considered for ingest in MOD1(e.g., over East Antarctica). Aloft (Fig. 7c), the differ-ences are manifested in ALL’s greater 500-hPa heightsover Queen Maud Land and East Antarctica and lowerheights over West Antarctica. The positive height dif-ference over Queen Maud Land is collocated with 500-hPa low centers in both analyses (not shown) and re-flects a slightly weaker upper-level low in ALL (ALLheights 22 gpm higher than a 4994-gpm low center inMOD1). The maximum wind differences at this levelreach about 10 m s�1.
A final experiment, MOD1_60, mimics MOD1, buthas horizontal grid spacings enhanced 33%: its grids are60, 20, 6.7, and 2.2 km. This experiment investigates theimpact of such enhanced resolution on the forecast of amajor weather event in the Ross Island region. Theeffectiveness of such resolution increase has been aquestion of interest to the USAP forecasters and inter-national users of AMPS.
←
FIG. 7. MODIS observations for ALL and MOD1 and 500-hPaheight/wind analysis differences for ALL and MOD1. (a) MODISdata available for ALL. Circles mark wind data points at allheights. (b) Same as in (a), but for MOD1. (c) Difference in500-hPa analyses (ALL � MOD1) for 0000 UTC 15 May 2004.Positive (negative) geopotential height differences are indicatedby solid (negative) contours; interval is 5 gpm. Wind differencevectors (arrows); magnitudes are 8.5 m s�1 (vector length inter-val)�1. Maximum vector magnitude is �10 m s�1.
SEPTEMBER 2007 P O W E R S 3141
4. Model results
Satellite and surface observations from the McMurdoarea for the 15 May 2004 event indicate the passage ofa deep cyclone through the region, with local AWS datashowing minimum pressures (�960 hPa) occurringfrom 1200 to 1600 UTC 15 May 2004. After 1500 UTC,with the center passing east of Ross Island (Fig. 3), thetrajectory results in the mesoscale enhancement ofthe pressure gradient and the barrier influence of theTAM, producing intense southerly flow through theMcMurdo region.
The ARW experiments and the AMPS MM5 forecastsimulate the transit of a strong low pressure systemfrom Marie Byrd Land across the ice shelf (Fig. 8).While the model trajectories on the synoptic scale aresimilar to that observed, meso-� scale (Orlanski 1975)differences from the observed evolution in the RossIsland region have a crucial bearing on the wind event.Consider first the AMPS MM5 forecast (Fig. 8f). TheMM5 low is driven too quickly westward, to south ofMinna Bluff by 1500 UTC 15 May 2004, when in con-trast the observed system was still in midshelf. The cen-ter stalls east of Minna Bluff at 1800 UTC, and fillingfollows. This results in a cyclone that is rapidly weak-ening as it approaches Ross Island. While the AMPSsystem persists north of Ross Island, the low is but aremnant.
In the WRF results in CTRL (no data assimilation;Fig. 8a) one sees a system driven too far westward, intothe coast south of Minna Bluff. In STD (Fig. 8b), thecounterpart to the AMPS MM5 forecast in terms ofdata assimilated, the low remains east of Minna Bluff.This shift compared to CTRL will be reflected in animproved wind event depiction. ALL (Fig. 8c) producesthe poorest results, sending the low into Minna Bluffwhere it fills in situ. The filtered MODIS experimentsMOD1 and MOD1_60 (Figs. 8d,e) yield the low trackand evolution closest to observation. The system turnsnorthward in the best agreement with the observed mo-tion, and the simulations correctly center the low off ofRoss Island’s east end at 2300–0000 UTC 15–16 May2004.
As seen in the low positions after 1200 UTC (hour12), all of the ARW runs move the system too swiftlyacross the shelf. This westward bias in the experimentsis the largest in CTRL. ALL’s track performance suf-fers compared to both MOD1 and STD, suggesting thatperhaps MODIS observations should be filtered priorto assimilation (MOD1), and that if not, it may be pref-erable to exclude them (STD).
Sea level pressure (SLP) analyses from MOD1 give afuller picture of the path and evolution of the system.
The run initializes with a 941-hPa center in MarieByrd Land (Fig. 9a) By hour 12 (1200 UTC) the cy-clone (955 hPa) has traveled to midshelf, sitting overthe 180° meridian (Fig. 9b). The packing of contourspoleward and southwest of the center is associated withthe strongest surface winds, which are 20–30 m s�1
south of the center on the shelf boundary. The localizedmaximum along the TAM is consistent with an en-hanced, barrier flow. Adams (2005) examined this casewith the University of Wisconsin Nonhydrostatic Mod-eling System and concluded the enhanced flow and anattendant surface cold wind surge along the TAM tohave characteristics of a trapped wave. An analysis ofthe wind event at McMurdo, however, was not the fo-cus of that broader investigation, which targeted theRAS.
By hour 18 (1800 UTC) the low, at 967 hPa, hasreached the Ross Island region and sits east of MinnaBluff (Fig. 9c). An enhanced pressure gradient lies tothe south and west of the system (i.e., toward the coast/TAM). The strongest surface winds are in this zone,with speeds of 20–30 m s�1 over the shelf and over 40m s�1 in the mountains. Marilyn AWS (79.95°S,156.13°E; Fig. 1c) is in this area and recorded winds ofapproximately 24 m s�1 from 180° at this time, compar-ing well to MOD1’s simulation of 26 m s�1 from 169°.
The model’s slowing and filling of the cyclone in theRoss Island area are evident in MOD1 at hour 21 (Fig.9d). The center has moved but little to the north fromits hour-18 position shelfward (east southeast) of MinnaBluff, and the central pressure has risen 10 hPa to 977hPa. The strong SLP gradient to the south and south-west is still present, however. Note that the gradientalong the south face of Ross Island is a fingerprint ofstagnating flow and higher surface pressure on thewindward side of this obstacle, the southern embay-ment of Windless Bight. As documented in O’Connorand Bromwich (1988), static stability in the PBL airencountering the steep topography results in a stagna-tion zone with relatively high pressure. At hour 23 (Fig.9e) the low is losing its identity, having further filled to984 hPa. While all of the experiments display this rapidcyclolysis, satellite imagery shows that the observedsystem maintained a circulation through 1105 UTC 16May 2004.
The ARW’s ability to recreate the synoptic low isthus apparent. The further question for forecasting,however, is how well the event winds are simulated.Figure 2 presents time series of wind speeds in the Mc-Murdo region (locations in Fig. 1c). The sites aroundMcMurdo proper are Helo Pad, Arrival Heights, CraterHill, and Cosray, while Pegasus North and Black Islandlie to the south. Arrival Heights and Pegasus North are
3142 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
FIG. 8. Tracks of ARW and AMPS MM5 lows. Times (UTC) of the central position of the low (marked L) are indicated. Italic Lreflects the time during the observed wind event at McMurdo. (a) CTRL, (b) STD, (c) ALL, (d) MOD1, (e) MOD1_60, and (f) AMPSMM5.
SEPTEMBER 2007 P O W E R S 3143
FIG. 9. SLP from MOD1. (a), (b) Window of 30-kmdomain; (c), (d), (e) 10-km domain. Contour interval in(a), (b) is 2 hPa to 1000 hPa, 4 hPa above; contourinterval in (c), (d), (e) is 1 hPa to 1000 hPa, 2 hPaabove. South Pole is denoted by SP. (a) Hour 0 (0000UTC 15 May 2004), (b) hour 12 (1200 UTC), (c) hour18 (1800 UTC), (d) hour 21 (2100 UTC), and (e) hour23 (2300 UTC).
3144 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
chosen here for highlighted analysis.6 As for verifica-tion of wind directions (not presented), the modelevent winds are, as observed, consistently southerly.
Figure 10 presents the results for Arrival Heights.Note first that all of the simulations delay and under-predict the winds representing the event at this loca-tion. CTRL (Fig. 10a), for example, misses the onsetand magnitude of the southerly blast. CTRL’s strongestwind is 21.6 m s�1, compared to a measured 48 m s�1.In STD (Fig. 10b) the assimilation of standard AMPSdata improves the flow intensity, which reaches28 m s�1. ALL (Fig. 10c) simulates conditions less ac-curately, both in terms of the maximum velocities(26 m s�1) and their duration and in the poorer corre-spondence of the early wind trace (0000–1800 UTC 15May 2004).
Among the 90/30 runs MOD1 (Fig. 10d) best ap-proximates the intensity, timing, and duration of thehigh winds. The strong flow arrives at 2000 UTC andhits 33.6 m s�1. The higher-resolution version ofMOD1, MOD1_60 (Fig. 10e), yields a somewhat im-proved profile, with a truer pre-event trace (1200–1800UTC), a stronger, slightly longer event, and a maximumof 35.5 m s�1. In CTRL and STD the event is about 5–6h delayed; ALL reduces this to 4 h; MOD1 andMOD1_60 exhibit the least delay, about 2 h.
Figures 11a–e present the traces for Pegasus North.CTRL (Fig. 11a) produces an event at Pegasus, butunderforecasts and delays it. The maximum simulatedwind speed is 24.6 m s�1, compared to 39.6 m s�1 ob-served. In contrast to CTRL, STD produces a signifi-cantly stronger event (Fig. 11b). After a first pulse[O (15 m s�1)] at 1900 UTC, the winds decrease, thenthe strong southerlies hit at up to 31.5 m s�1. Again,there is a timing lag. ALL (Fig. 11c) forecasts a weaker,more delayed episode.
MOD1 (Fig. 11d) produces a strong wind episode oftiming and magnitude that compares reasonably wellwith the observations. The event begins just after 1900UTC and peak at 36.6 m s�1 (cf. 39.6 m s�1 observed).MOD1 also simulates the increase in flow seen in thepre-event period of �1000–1800 UTC. MOD1_60 (Fig.11e) is similar to MOD1, but its peak velocity is slightlyhigher (37.2 m s�1), and the profile through the eventis, overall, closer to that observed. In short, from theresults of these representative McMurdo area sites (andothers, not shown), MOD1 best reproduces the event ofthe 3.3-km grid experiments. In addition, in MOD1_60
the resolution increase to 2.2-km spacing appears to bea further improvement.
While the focus has been on the ARW experiments,consider now the performance of the operationalAMPS MM5. For Arrival Heights and Pegasus North(Fig. 12), one first sees that the MM5 significantly un-derpredicts the wind maxima. At Arrival Heights (Fig.12a) there is a wind speed increase at observed onset,but this is not sustained. At Pegasus North (Fig. 12b)there are sustained relatively strong winds after 1600UTC, although they peak at only 21.6 m s�1 (2100UTC). Comparing the counterpart experiment STD inFigs. 10b and 11b with the MM5, the ARW produces amore distinct and less underestimated, albeit more de-layed, event at these sites.
One may also examine the evolution of the windsassociated with the synoptic system (i) away from theimmediate Ross Island area and its complex topogra-phy and (ii) indicative of conditions on the ice shelf. Forthis, consider Marilyn AWS. The winds peaked at Mari-lyn from 1600 to 2200 UTC, reaching 26.7 m s�1 (Fig.13). CTRL (Fig. 13a) reproduces the onset of the eventwinds and the observed magnitude (26.2 m s�1), al-though it does not sustain the high winds for as long asobserved (e.g., 2000–0000 UTC). STD is better at thissite (Fig. 13b). The timing of the increase is captured, aswell as the peak winds, with even a bit of an overfore-cast (28.8 m s�1). An arguable improvement over this,in terms of wind strength from before the event (0000–1800 UTC) through the strongest wind phase, is MOD1(Fig. 13d). MOD1 avoids an overforecast, and the winddecrease after the peak hours agrees with the observedtrace. MOD1_60 (Fig. 13e) is similar to MOD1. ALLdisplays the poorest results (Fig. 13c). It suffers from alag and a significant underforecast. Last, although inthe AMPS MM5 (Fig. 12c) the amplitude of themaxima is not captured as well as in STD and MOD1,the mean profile does approximate the episode to thelevel of those ARW runs.
To illuminate the wind event and flow pattern on themesoscale, as opposed to the point views from the timeseries, Fig. 14 presents the ARW surface (lowest modellevel) winds from the Ross Island grids at 2300 UTC(hour 23) 15 May 2004. CTRL exhibits strong souther-lies striking Minna Bluff and the TAM (Fig. 14a), butthis momentum is not reaching McMurdo. In contrast,STD (Fig. 14b) generates a well-defined flow aroundRoss Island, with a broader and stronger (�40 m s�1)wind field. One revelation from the high-resolutionoutput is the strong shadowing effect that Minna Bluffand the associated TAM topography can exert on HutPoint Peninsula (the extension of Ross Island on whichMcMurdo sits, marked in Fig. 14a) and the McMurdo
6 These two sites, and Marilyn site (discussed below), have an-emometers at 3 m. Accuracies are O(0.25 m s�1) for PegasusNorth and Marilyn and 5% for Arrival Heights.
SEPTEMBER 2007 P O W E R S 3145
FIG. 10. Observed (solid) vs ARW (dashed) windspeed (m s�1) at Arrival Heights. Abscissa shows thetime in hours from 0000 UTC 15 May 2004. Missingobserved values are not plotted. (a) Observed vsCTRL, (b) observed vs STD, (c) observed vs ALL, (d)observed vs MOD1, and (e) observed vs MOD1_60.
3146 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
FIG. 11. Same as in Fig. 10, but at Pegasus North.
SEPTEMBER 2007 P O W E R S 3147
area. Due to this, in STD McMurdo is not yet experi-encing the southerlies. And, ALL (Fig. 14c) shows onlya narrow stream of higher-velocity flow (�12.5 m s�1)getting into McMurdo.
MOD1 (Fig. 14d) simulates a 984-hPa low center eastof Cape Crozier and an associated pressure gradientthat has engaged the McMurdo area. MOD1’s windspeeds are over 32 m s�1 at Arrival Heights (32 m s�1
observed) and over 34 m s�1 at Pegasus North(33 m s�1 observed) (marked in Fig. 1c). The southerlymomentum field has engulfed Ross Island. This is in-
ducing notable secondary phenomena—flow splitting inthe surface layer and a von Kármán vortex (e.g.,Heinemann 1986) to the north (marked “V” in Fig.14d). Model-generated von Kármán vortices resultingfrom strong southerly flow around Ross Island havebeen previously described by Powers et al. (2003).
MOD1_60 (Fig. 14e) displays conditions at Mc-Murdo similar to MOD1, but with even greater cover-age and intensity of the high-momentum flow envelop-ing Ross Island, Minna Bluff, and the TAM. Charac-teristics of these two most successful experiments,
FIG. 12. Observed (solid) vs AMPS MM5 (dashed) wind speed(m s�1) at (a) Arrival Heights, (b) Pegasus North, and (c) Mari-lyn. Abscissa shows time in hours from 0000 UTC 15 May 2004.Missing observed values are not plotted.
3148 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
FIG. 13. Same as in Fig. 10, but at Marilyn.
SEPTEMBER 2007 P O W E R S 3149
FIG. 14. Surface winds and SLP from ARWexperiments for hour 23 (2300 UTC 15 May2004). Wind speed shaded (scale at right), me-dium gray shading is �15 m s�1. Arrows indi-cate wind directions and magnitudes are 22m s�1 (vector length interval)�1. SLP contourinterval is 1 to 988 hPa, 2 hPa for SLP � 988hPa; maximum SLP contour is 1008 hPa. Theitalic L indicates the surface low and V indi-cates the von Karman vortex. (a) CTRL withMinna Bluff (MB) and the TAM indicated, theasterisk shows the location of McMurdo,which lies on Hut Point Peninsula; (b) STD;(c) ALL; (d) MOD1, the open circle shows thelocation of Cape Crozier; and (e) MOD1_60.
3150 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
MOD1 and MOD1_60, are the low’s maintaining itsintegrity and the associated pressure gradient beingrelatively strong as it skirts Ross Island. Note, as inMOD1, the well-defined, leeside von Kármán vortex(Fig. 14c). Although the satellite imagery does not re-veal such vortices at this time, midlevel and highercloud are shielding the lower levels north of Ross Island(see, e.g., Fig. 4b).
The AMPS MM5 hour-23 forecast is seen in Fig. 15.While over the Ross Island area in general there isstrong southerly flow, McMurdo itself is not experienc-ing a wind event, as Hut Point Peninsula and most ofthe south side of the island are sheltered. The MM5result is most similar to that of STD (Fig. 14b). From aforecast perspective, the MM5 wind field depiction isnot as good as that of MOD1, which does capturestrong winds impacting McMurdo at this time (Fig.14d).
To briefly investigate the variations in the low tracks,and thus the simulated wind events in the experiments,the upper-level synoptic differences in the runs are nowconsidered. As 500 hPa is taken as a representativelevel for the steering flow (SPAWAR forecasters 2005,personal communication), Figs. 16a,b show theECMWF analysis and the CTRL 12-h forecast at thislevel for 1200 UTC 15 May 2004. First apparent is thatthe trough over the Ross Ice Shelf is a sharper, higher-amplitude feature in the analysis than in CTRL. Cor-respondingly, downstream of the trough axis the ana-lyzed flow is more southerly and less easterly over theeastern ice shelf than the forecast. The model, further-more, displays less ridging over the eastern edge of theshelf and over the TAM.
Figure 17a shows the 500-hPa vector wind differencebetween CTRL and the analysis at 1200 UTC. In theanalysis the midshelf flow is more southerly, while theeastern shelf flow is more westerly (from the TAMside). The actual track would thus tend to be more tothe north and away from the western shelf edge, whileconversely, in the model, the trajectory would be lessnortherly and more westward (i.e., toward the MinnaBluff coast and the TAM). This is indeed seen com-pared to observations (cf. Figs. 3 and 8a).
Figures 17b,c isolate the westerly and southerly windcomponent differences between the analysis andCTRL. The broad arrows indicate the momentum inthe analysis that is not represented in the model. TheARW lacks a relative westerly component, one thatwould retard the propagation of the system to the RossIsland side of the shelf (Fig. 17b), as well as the strongersoutherly flow (Fig. 17c). In short, the steering was rela-tively northward and eastward for the observed systemcompared to the ARW systems, and the CTRL track
most distinctly exemplifies the behavior of the latter.The improvement in low trajectory in MOD1 andMOD1_60 indicates that the assimilation of filteredMODIS AMVs aloft can better the forecast on themesoscale from a modified upper-level synoptic initial-ization.
5. Statistical evaluations
The ARW’s performance in simulating the local flowamplitude is now analyzed statistically. To this endwind speed errors have been quantified through verifi-cation at six locations across the Ross Island region:Arrival Heights, Pegasus North, Black Island, MinnaBluff, Marilyn, and Schwerdtfeger (located in Fig. 1c).Wind speed bias [or mean error (ME)], mean absoluteerror (MAE), and root-mean-square error (RMSE)have been calculated for two periods: 0000 UTC 15May–UTC 17 May 2004 and 1200 UTC 15 May–0600 16May 2004. The former is the whole period of simulation(hours 0–48), while the latter represents the subperiodcentered on the event, from 6 h prior through 6 h af-terward (hours 12–30). The purpose in considering twointervals is to see whether the picture of model perfor-mance varies for the event proper versus the entireforecast.
FIG. 15. Surface winds and SLP from the AMPS MM5 forecastfor hour 23 (2300 UTC 15 May 2004). Shading, vectors, and con-tours are the same as in Fig. 14.
SEPTEMBER 2007 P O W E R S 3151
FIG. 16. 500-hPa heights and winds for 1200 UTC 15 May 2004. Heights (solid lines);contour interval is 50 gpm. Wind vectors (arrows); magnitudes are 18 m s�1 (vector lengthinterval)�1. (a) ECMWF analysis and (b) CTRL 12-h forecast.
3152 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
Tables 1 and 2 present the results for the full simu-lation and event periods. For both periods the lowestbiases, MAEs, and RMSEs are seen for experimentsSTD, MOD1, and MOD1_60. The AMPS MM5 com-pares well with these ARW runs. Note that the windspeed biases are consistently negative: the ARW andthe MM5 underpredict the wind speeds, both for theevent and the whole forecast period.
Table 3 presents the averages of the errors. For theARW experiments, MOD1 and MOD1_60 have the low-
est error means. Following these is STD, while the ALLresults are relatively poor. The AMPS MM5 scores areamong the best and are comparable to those for MOD1.
To objectively determine whether the errors and er-ror differences are significant, statistical significancetesting has been performed. The first analysis is that ofwhether each average error can be concluded to bestatistically distinct from zero. A two-tailed Student’s ttest (see, e.g., Walpole and Myers 1985) reflects the nullhypothesis (H0) that the error populations of the ex-
FIG. 17. 500-hPa wind and wind component differences foranalysis � CTRL, 1200 UTC 15 May 2004. (a) Wind vectordifferences, magnitudes are 9 m s�1 (vector length interval)�1.(b) The u-component differences; contour interval is 2 m s�1;solid contours reflect reduced easterly/stronger westerly compo-nent, dashed contours reflect stronger easterly/reduced westerlycomponent, and large arrows indicate net momentum relative toCTRL. (c) The �-component differences; contour interval is 2m s�1; solid contours reflect stronger southerly component,dashed contours reflect weaker southerly/stronger northerlycomponent, and large arrows indicate net momentum relative toCTRL.
SEPTEMBER 2007 P O W E R S 3153
periments have means of zero and the alternative hy-pothesis that the means do not equal zero. The resultsfind that for both the 48- and the 18-h periods, all theaverage MAEs for the experiments are significantly dif-ferent from zero at the 95% confidence level.
To assess the significance of experiment differences,testing on the differences of the mean errors has beenperformed. The null hypothesis is that the difference inerror means is zero, while the alternate hypothesis isthat the error mean of one experiment is less than thatof the other. The t value is computed as
t � d � d0�v�sd, 1
where d is the difference of the means, d0 is the nullhypotheses difference in means (here 0), � is the de-grees of freedom, and sd is the standard deviation of thesample of mean differences (see, e.g., Walpole and My-ers 1985; Panofsky and Brier 1968; Wilks 1995). Thecomputed t value is compared to the critical value oft /2, where the primary confidence level consideredhere is 95% ( � 0.05).
Table 4 presents the results. Here the error means ofthe experiments in the first column are compared withthose in the second. The experiment with the concludedlower mean error at the 95% level is shown under theerror type (bias, MAE) column. If the null hypothesiscannot be rejected the table entry is “I.” If the alternatehypothesis may be accepted for experiment 1 or 2 at the90% level (but not the 95% level), then the entry givenis E190 or E290 (for experiment 1 or experiment 2 hav-ing the lower mean error).
For the first pair of STD and CTRL, the mean biasesare significantly different at the 95% level over both thefull 48-h forecast and the 18-h event subperiod, with themean error for STD less than that of CTRL. As seen inresults such as in Figs. 11, 12, and 14, STD did verifybetter than CTRL, and such a difference is significant.In terms of wind speed MAEs, however, the mean er-rors in STD and CTRL are statistically indistinguish-able. For MOD1 and MOD1_60 compared to CTRL,these filtered MODIS experiments are both concludedto have significantly lower mean biases and MAEs forboth periods considered. In contrast, in comparing theapproach of assimilating all of the available MODIS
TABLE 1. Model wind speed errors (m s�1) for hours 0–48 (0000UTC 15 May–0000 UTC 17 May 2004) for McMurdo region: Ar-rival Heights, Minna Bluff, Pegasus North, Black Island, Marilyn,and Schwerdtfeger.
Wind speed error: Hours 0–48
0000 UTC 15 May–0000 UTC 17 May 2004 (values in m s�1)
3154 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
data without filtering (ALL) with either no data(CTRL) or conventional observation (STD) assimila-tion, the results argue against the former approach(ALL). STD exhibits significantly lower biases andMAEs than ALL for both periods, while CTRL exhib-its such lower errors for the wind event subperiod.
Comparing MOD1 and STD it is seen that the addi-tion of filtered MODIS data yields an improvementover the entire simulation and results in significantlyimproved biases and MAEs for the 48-h period. Com-pared to ALL, the MOD1 filtering yields better modelperformance for both periods. In terms of relative per-formance among the 90/30 ARW experiments, overallMOD1 shows the lowest mean errors for both periods.It is followed by STD, then CTRL. ALL is statisticallythe poorest. It is thus again seen that MODIS data canimprove the ARW simulations for this extreme polarevent, but under the conditions of filtering.
With respect to the increase in resolution from a 90/30/10/3.3- to a 60/20/6.7/2.2-km configuration, the finergrids do improve scores somewhat. MOD1_60 exhibitsa significantly lower wind speed bias than MOD1 forthe episode at the 95% level, while at a lower 90% con-fidence level it is better in terms of MAE for both periods.
With respect to the performance of the AMPS MM5,the MM5 has lower mean biases and MAEs, in general,than ALL (Table 4, bottom). Compared to CTRL, theMM5 is superior in terms of bias, but no conclusion canbe made for MAE. Compared to those of its ARWcounterpart run, STD, the MM5’s biases for the 48-h
period are significantly better (95% level), with theconfidence in this conclusion reduced to 90% consid-ering just the 18-h event subperiod. The MM5 MAE,however, is not significantly better than the STD MAE.Compared to the best ARW 90/30 experiment, MOD1,it cannot be concluded that the MM5’s errors are sig-nificantly different. The deduction is mostly the samewith respect to MOD1_60, although at the 90% confi-dence level MOD1_60 is superior in terms of MAE.
6. Summary and conclusions
In the setting of the Antarctic Mesoscale PredictionSystem (AMPS), the Weather Research and Forecast-ing (WRF) model has been applied for the first time toAntarctica. In this study the abilities of the ARW toforecast a major Antarctic weather event and of MODISretrievals to improve such forecasts are explored insimulations of the windstorm that struck McMurdo Sta-tion on 15 May 2004. The suite of tests primarily uses anested 90/30/10/3.3-km domain setup, with an addi-tional run using higher-resolution grids of 60/20/6.7down to 2.2 km over the critical Ross Island area. Inaddition, a comparison of the AMPS MM5 forecastwith the ARW simulations provides a first look at therelative performance in a polar case study of these twomodels, currently both in real-time use.
Considering first its ability to capture the synopticsetting and evolution of the event, the ARW simulates
TABLE 4. Statistical comparisons of experiments. Under “Bias”and “MAE,” the listed experiment’s mean error for the periodindicated (forecast hours 12–30 or 0–48) is concluded to be lowerthan that of the compared experiment at the 95% confidencelevel. EXPT90 indicates that mean error is lower at 90% confi-dence level. The I indicates that the test is inconclusive at eitherthe 95% or 90% confidence level. M1_60 indicates MOD1_60.
Comparison of experiment mean errors
Expts compared Hours 12–30 Hours 0–48
Expt 1 Expt 2 Bias MAE Bias MAE
STD CTRL STD I STD IALL CTRL CTRL CTRL I IMOD1 CTRL MOD1 MOD1 MOD1 MOD1M1_60 CTRL M1_60 M1_60 M1_60 M1_60ALL STD STD STD STD STDMOD1 STD I I MOD1 MOD1M1_60 STD M1_6090 M1_6090 M1_60 M1_60ALL MOD1 MOD1 MOD1 MOD1 MOD1M1_60 MOD1 M1_60 M1_6090 I M1_6090
TABLE 3. Average McMurdo region wind speed errors in modelruns. Errors averaged over values from locations in Tables 1 and2. Full simulation (hours 0–48) and event period (hours 12–30)results shown.
the evolution of the motivating low pressure system,and the accuracy of the track forecasts on the largerscale is confirmed. The best experiments with respect tothese forecast elements are MOD1 and MOD1_60,which both assimilate filtered MODIS data. A meso-scale examination, however, reveals that the ARW (aswell as the AMPS MM5) tends to move the low acrossthe Ross Ice Shelf too strongly and rapidly to the west,not precisely capturing the timing of the backing to thenorth. Analyses show the discrepancy in the synoptictrajectory to be associated with long-wave pattern er-rors and weaker westerly and southerly upper-levelflow components in the model. The model also fills thesystem relatively abruptly near Ross Island. Both ofthese developments significantly affect the wind simu-lations in the McMurdo region.
The high-resolution (3.3 and 2.2 km) ARW grids re-veal the regional character of the 15 May 2004 wind-storm and show how strong southerly momentum asso-ciated with a low pressure system moving to the east ofRoss Island impacts McMurdo. The surface wind de-pictions from the most successful simulations (MOD1,MOD1_60) present a pattern of high-momentum flowencountering and responding to the local topography(i.e., Minna Bluff, the Transantarctic Mountains, andRoss Island). As the surge arrives in the McMurdoarea, the wind velocities at the base increase substan-tially. The realistic model results further show that, far-ther afield, circulations such as von Kármán vorticesmay be spawned from the interaction of such flow andRoss Island. While it is possible that the latter are ubiq-uitous in strong, stable, southerly flow regimes, theirprevalence cannot as yet be confirmed.
The ARW can successfully forecast the strong south-erly flow defining the event [e.g., provide guidancealerting forecasters to condition-1 (�55 kt) and condi-tion-2 level (48–55 kt) winds]. The 90/30 run best re-producing the wind timing and intensity, MOD1, re-flects the assimilation of the filtered MODIS AMVs.From the 60/20 MOD1_60 experiment, however, it isfound that increasing the ARW’s grid resolution by33% does yield statistically significant improvement inthe wind event simulation. Despite the model’s ability,in general the surface wind speed amplitudes for theevent tend to be underforecast. As noted, this reflectstrack errors on the mesoscale and a stalling and fillingof the cyclone responsible for the event near Ross Is-land.7 The assimilation of filtered MODIS data, how-ever, mitigates the errors in low trajectory and evolu-
tion and in the resultant McMurdo wind forecast. Thisshows the potential for benefits to the initial conditionsand forecasts of Antarctic high-resolution NWP sys-tems from the assimilation of nontraditional satelliteobservations over the continent and the SouthernOcean.
The performance of the ARW and the AMPS MM5has also been examined statistically, through signifi-cance testing of event wind speed errors. It is found thatthe assimilation of conventional observations and selectMODIS AMV data can improve the mesoscale forecastin statistically significant terms. The application of afilter to the MODIS retrievals, however, is necessaryfor such benefit, as the assimilation of unfiltered mea-surements is found to actually degrade model perfor-mance. The filtering criteria used here follow from pre-viously published suggestions and account for instru-ment channel, surface type, and observation height.
In summary, this study has provided an initial inves-tigation of the behavior of the emerging Advanced Re-search WRF model in a polar region. It has also illu-minated the impact of the high-potential MODIS AMVdata on polar mesoscale NWP and on the reproductionof a major Antarctic weather event. Most importantly,it is found that the ARW can realistically simulate suchan event and that assimilation of MODIS polar windscan significantly improve the forecast on the mesoscale.That the application of a data filter is necessary wouldconfirm—in the context of a specific, high-impact fore-cast and for a mesoscale model—previous work involv-ing global model applications (Key et al. 2003; Bor-mann and Thépaut 2004). Overall, it is seen that WRFshows promise for both research and operational appli-cations over Antarctica.
Acknowledgments. This study and AMPS have beensupported by the National Science Foundation, Officeof Polar Programs. The author thanks Michael G. Dudaof NCAR for his assistance in data processing andgraphics support. The author thanks Udo Voight andRalf Brauner of Deutscher Wetterdienst for their pro-vision of ECMWF analyses. The author also thanksSPAWAR for computing hardware and the SPAWARforecasters for helpful discussions.
REFERENCES
Adams, A. S., 2005: The relationship between topography and theRoss Ice Shelf air stream. Ph.D. thesis, University of Wiscon-sin—Madison, 125 pp.
Barker, D. M., W. Huang, Y. R. Guo, and Q. N. Xiao, 2004: Athree-dimensional (3DVAR) data assimilation system for usewith MM5: Implementation and initial results. Mon. Wea.Rev., 132, 897–914.
7 While the PBL and land surface representations may also befactors, sensitivity investigations of the associated schemes (keptconstant in the experiment suite here) are left for future work.
3156 M O N T H L Y W E A T H E R R E V I E W VOLUME 135
Bluestein, H. B., 1992: Synoptic-Dynamic Meteorology in Midlati-tudes. Vol. 1. Oxford University Press, 431 pp.
Bormann, N., and J.-P. Thépaut, 2004: Impact of MODIS polarwinds in ECMWF’s 4DVAR data assimilation system. Mon.Wea. Rev., 132, 929–940.
Bromwich, D. H., and T. R. Parish, 2002: Ross Island Meteorol-ogy Experiment (RIME) detailed science plan. BPRC Mis-cellaneous Series M-424, Byrd Polar Research Center, TheOhio State University, 39 pp. [Available from Byrd PolarResearch Center, The Ohio State University, 1090 CarmackRd., Columbus, OH, 43210.]
——, J. J. Cassano, T. Klein, G. Heinemann, K. M. Hines, K.Steffen, and J. E. Box, 2001: Mesoscale modeling of katabaticwinds over Greenland with the Polar MM5. Mon. Wea. Rev.,129, 2290–2309.
——, A. J. Monaghan, J. G. Powers, J. J. Cassano, H.-L. Wei,Y.-H. Kuo, and A. Pellegrini, 2003: Antarctic Mesoscale Pre-diction System (AMPS): A case study from the 2000/2001field season. Mon. Wea. Rev., 131, 412–434.
——, ——, K. W. Manning, and J. G. Powers, 2005: An evaluationof the Antarctic Mesoscale Prediction System (AMPS). Mon.Wea. Rev., 133, 579–603.
Carlson, T. N., 1998: Mid-Latitude Weather Systems. Amer. Me-teor. Soc., 507 pp.
Cassano, J. J., J. E. Box, D. H. Bromwich, L. Li, and K. Steffen,2001: Evaluation of Polar MM5 simulations of Greenland’satmospheric circulation. J. Geophys. Res., 106, 33 867–33 890.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land-surface/hydrology model with the Penn State/NCAR MM5modeling system. Mon. Wea. Rev., 129, 569–585.
Davis, C. A., and Coauthors, 2006: Advanced Research WRF de-velopments for hurricane prediction. Extended Abstracts,Seventh WRF Users’ Workshop, Boulder, CO, National Cen-ter for Atmospheric Research.
Done, J. M., L. R. Leung, and B. Kuo, 2006: Understanding errorin the long-term simulation of warm season rainfall using theWRF model. Extended Abstracts, Seventh WRF Users’ Work-shop, Boulder, CO, National Center for Atmospheric Re-search.
Forsythe, M., and H. Berger, 2004: MODIS winds: Latest resultsat the Met Office. Proc. Seventh Int. Winds Workshop, Hel-sinki, Finland, EUMETSAT.
Grell, G. A., J. Dudhia, and D. R. Stauffer, 1995: A description ofthe fifth-generation Penn State/NCAR Mesoscale Model(MM5). NCAR Tech. Note TN-398�STR, 122 pp. [Availablefrom UCAR Communications, P.O. Box 3000, Boulder, CO80307.]
Heinemann, G., 1986: Lee-vortices in the Antarctic. Beitr. Phys.Atmos., 59, 599–602.
Holmes, R. E., C. R. Stearns, G. A. Weidner, and L. M. Keller,2000: Utilization of automatic weather station data for fore-casting high wind speeds at Pegasus Runway, Antarctica.Wea. Forecasting, 15, 137–151.
Holton, J. R., 1992: An Introduction to Dynamic Meteorology.Academic Press, 511 pp.
Hong, S.-Y., J. Dudhia, and S.-H. Chen, 2004: A revised approachto ice microphysical processes for the bulk parameterizationof clouds and precipitation. Mon. Wea. Rev., 132, 103–120.
Janjic, Z. I., 2002: Nonsingular implementation of the Mellor–
Yamada level 2.5 scheme in the NCEP Meso Model. NCEPOffice Note 437, 61 pp.
Kain, J. S., and J. M. Fritsch, 1993: Convective parameterizationfor mesoscale models: The Kain–Fritsch scheme. The Repre-sentation of Cumulus Convection in Numerical Models, K. A.Emanuel and D. J. Raymond, Eds., Amer. Meteor. Soc., 165–170.
Key, J. R., D. Santek, C. S. Velden, N. Bormann, J.-N. Thépaut,L. P. Riishojgaard, Y. Zhu, and W. P. Menzel, 2003: Cloud-drift and water vapor winds in the polar regions fromMODIS. IEEE Trans. Geosci. Remote Sens., 41, 482–492.
Mellor, G. L., and T. Yamada, 1982: Development of a turbulenceclosure model for geophysical fluid problems. Rev. Geophys.Space Phys., 20, 851–875.
Moeng, C.-H., J. Dudhia, J. B. Klemp, and P. P. Sullivan, 2007:Examining two-way grid nesting for large eddy simulation ofthe PBL using the WRF model. Mon. Wea. Rev., 135, 2295–2311.
Monaghan, A. J., D. H. Bromwich, H.-L. Wei, A. M. Cayette,J. G. Powers, Y.-H. Kuo, and M. Lazzara, 2003: Performanceof weather forecast models in the rescue of Dr. Ronald She-menski from the South Pole in April 2002. Wea. Forecasting,18, 142–160.
O’Connor, W. P., and D. H. Bromwich, 1988: Surface airflowaround Windless Bight, Ross Island, Antarctica. Quart. J.Roy. Meteor. Soc., 114, 917–938.
Orlanski, I., 1975: A rational subdivision of scales for atmosphericprocesses. Bull. Amer. Meteor. Soc., 56, 527–530.
Panofsky, H. A., and G. W. Brier, 1968: Some Applications ofStatistics to Meteorology. The Pennsylvania State UniversityPress, 224 pp.
Powers, J. G., A. J. Monaghan, A. M. Cayette, D. H. Bromwich,Y.-H. Kuo, and K. W. Manning, 2003: Real-time mesoscalemodeling over Antarctica: The Antarctic Mesoscale Predic-tion System (AMPS). Bull. Amer. Meteor. Soc., 84, 1533–1546.
——, K. W. Manning, and M. M. Lamberston, 2005: Applicationof the Weather Research and Forecasting (WRF) model inAntarctica. Preprints, Eighth Conf. on Polar Meteorologyand Oceanography, San Diego, CA, Amer. Meteor. Soc.,CD-ROM, 3.5.
Simmonds, I., K. Keay, and E.-P. Lim, 2003: Synoptic activity inthe seas around Antarctica. Mon. Wea. Rev., 131, 272–288.
Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M.Barker, W. Wang, and J. G. Powers, 2005: A description ofthe Advanced Research WRF, version 2. NCAR Tech. NoteNCAR/TN-468�STR, 88 pp. [Available from UCAR Com-munications, P.O. Box 3000, Boulder, CO 80307.]
Steinhoff, D. F., and D. H. Bromwich, 2005: Extreme winds andrapid degeneration of the May 2004 McMurdo Antarcticastorm: Analysis using the AMPS forecast model, AWS data,and MODIS data. AMPS Users’ Workshop 2005, Columbus,OH, Byrd Polar Research Center, 43–51.
Velden, C., and Coauthors, 2005: Recent innovations in derivingtropospheric winds from meteorological satellites. Bull.Amer. Meteor. Soc., 86, 205–223.
Walpole, R. E., and R. H. Myers, 1985: Probability and Statisticsfor Engineers and Scientists. Macmillan, 639 pp.
Wilks, D. S., 1995: Statistical Methods in the Atmospheric Sciences.Academic Press, 467 pp.
