SEARCH

SEARCH BY CITATION

Keywords:

  • AMPS;
  • Antarctica;
  • RAS;
  • Ross Ice Shelf;
  • SOM;
  • wind

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

[1] The low-level wind field across the Ross Ice Shelf region, Antarctica is studied through the method of self-organizing maps (SOMs). Emphasis is placed on identifying and characterizing the Ross Ice Shelf air stream (RAS). The RAS is a dominant northward stream of air resulting in significant transport of atmospheric mass from the interior of Antarctica. Two SOMs are created for the lowest model level (∼13 m) and the 5th lowest model level (∼150 m), from archived real-time numerical weather prediction output of the Antarctic Mesoscale Prediction System (AMPS). A verification of the SOM for the lowest model level is completed by comparing select nodes to automatic weather station wind rose plots. The winds depicted in the lowest model level SOM are dominated by the position of the cyclone in the Ross Sea and the changes in wind speeds on the Antarctic plateau. The 5th lowest model level provides a better depiction of the RAS. The RAS is found to be present primarily when associated with a cyclone in the Ross Sea. There is a strong seasonality to the RAS with the RAS occurring frequently during the winter months and rarely during the summer months. The structure and position of the RAS is strongly dependent on the position of the Ross Sea cyclone.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

[2] The purpose of this study is to gain a greater understanding of the low-level wind field of the Ross Ice Shelf and surrounding regions, and in particular the Ross Ice Shelf air stream (RAS), through the use of self-organizing maps (SOMs). Five years of numerical weather prediction output from the Antarctic Mesoscale Prediction System (AMPS) is evaluated to complete the study. The winds at the lowest model level and the 5th lowest model level, at a height of approximately 150 m, are investigated. This study is in contrast to previous research (described below) that has focused on the surface and/or lowest model level winds and either annual, seasonal, or case study evaluations for the winds in the region. The focus of the study is on better characterizing the RAS that has received increased attention over the past 10 years.

[3] The RAS is a dominant stream of air flowing northward from the interior of Antarctica across the western to central Ross Ice Shelf and the Ross Sea. The RAS has been associated with significant northward atmospheric mass transport from Antarctica to the midlatitudes [Parish and Bromwich, 1998]. The RAS has been directly studied by Parish et al. [2006] using 1-yr of AMPS output, and by Steinhoff et al. [2009] performing a case study of a RAS event using satellite imagery, automatic weather station data, and AMPS output. The RAS is fed by three primary sources of air from the continental interior: the Siple Coast region of West Antarctica; through the Byrd Glacier valley, and nearby glacier valleys, of the Transantarctic Mountains; and air flowing from the Reeves Glacier valley. The amount of contribution of atmospheric mass transported from the interior through these three regions varies over time. The most pronounced source of the airflow feeding the RAS is the winds along the Siple Coast. Winds in the Siple Coast region are the result of a complex interplay of katabatic winds from West Antarctica and East Antarctica, blocking effects of the Transantarctic Mountains, and enhancements from synoptic-scale cyclones to the south [Bromwich and Liu, 1996; Liu and Bromwich, 1997].

[4] The RAS is not a defined wind type with a specific atmospheric forcing. Instead it is a compilation of one or more of the following: katabatic winds, a barrier wind, mesocyclones, and a synoptic-scale cyclone. A katabatic wind is a negatively buoyant airflow down a slope under the influence of gravity [Parish and Cassano, 2003]. Parish and Bromwich [1987] identified confluence zones, regions of enhanced katabatic winds due to the underlying topography, above Reeves Glacier, Byrd Glacier, and the Siple Coast. Bromwich [1989] presents results on the Reeves Glacier katabatic winds based on a study using thermal infrared satellite imagery and AWS observations. The study analyzed the propagation of the katabatic wind from the base of Reeves Glacier across the Ross Sea under the influence of synoptic forcing. Parish and Bromwich [1989] studied the katabatic winds of Reeves Glacier using an instrumented aircraft and confirmed the presence of a confluence zone upstream of the glacier and the fact that the katabatic flow was negatively buoyant compared to the surrounding ambient environment. Breckenridge et al. [1993] studied the katabatic winds along the Transantarctic Mountains, including Byrd Glacier, using thermal infrared satellite imagery and AWS observations. Case studies for an intense katabatic event and an event with minimal katabatic activity were presented to provide a more in-depth understanding of katabatic winds in the region. Bromwich et al. [1994] presents a numerical study using a hydrostatic primitive equation model for the katabatic winds at Siple Coast. The results identify the katabatic winds along the Siple Coast as being influenced by synoptic forcing. Schwerdtfeger [1984] provides a description of the barrier wind as a statically stable low-level airflow that lacks sufficient kinetic energy to pass over a topographic barrier. The result is a blocking of the flow at the base of the barrier and the development of a pressure gradient directed away from the barrier and a barrier parallel geostrophic wind.

[5] The ambient pressure gradient forcing, associated with synoptic-scale cyclones and mesocyclones in the region, is the primary forcing mechanism for the RAS [Parish et al., 2006]. The katabatic flow loses its forcing when it reaches the Ross Ice Shelf and needs additional large-scale forcing to transport the atmospheric mass over large distance across the Ross Ice Shelf and the Ross Sea. The motion and circulation of the cyclones and mesocyclones toward the Transantarctic Mountains creates a blocking situation that drives the barrier wind dynamics [O'Connor et al., 1994]. Bromwich et al. [1992], Carrasco and Bromwich [1993], and Parish and Bromwich [1998] study the propagation of katabatic winds across the Ross Ice Shelf. The synoptic-scale analyses presented in these studies indicated that the ambient pressure gradient force associated with the synoptic-scale cyclones enhances the propagation of katabatic winds across the Ross Ice Shelf. Mesoscale cyclones, commonly referred to in the region as mesocyclones, are less than 1000 km in diameter and play an active role in the atmospheric circulation of the Ross Ice Shelf and Ross Sea regions. Carrasco et al. [2003] assesses the presence, distribution, and characteristics of mesocyclones over the Ross Sea/Ross Ice Shelf region through a one-year study of satellite imagery. Areas of maximum distribution of mesocyclones correspond to katabatic wind drainage regions on the Ross Ice Shelf and over the Ross Sea.

[6] The RAS is predominantly a feature of the western to central Ross Ice Shelf. This is true especially when viewed from a climatological perspective [Parish et al., 2006]. However, the RAS does not have to be a western to central Ross Ice Shelf feature as it is foremost defined as a northward directed corridor of atmospheric mass transport from the continental interior. When the RAS is adjacent to the Transantarctic Mountains there are generally three wind speed maxima seen along the Transantarctic Mountains [Parish et al., 2006; Seefeldt and Cassano, 2008]. The wind speed maxima are found over the Siple Coast and along the Prince Olav Mountains, from Byrd Glacier extending to the eastern edge of Ross Island, and from Reeves Glacier extending north along the Transantarctic Mountains. The RAS is a low-level feature generally in the lowest 1000 m of the atmosphere [Parish et al., 2006]. The exact height of maximum wind speed likely varies depending on the latitude, nearby topography, and intensity of the RAS.

[7] An understanding of the local geography (Figure 1) is necessary as the low-level wind field is strongly influenced by the regional topography. The Ross Ice Shelf is a relatively flat region surrounded by elevated terrain to the west, east, and south, and the Ross Sea to the north. The Transantarctic Mountains are to the south and west of the Ross Ice Shelf rising from sea level to elevations ranging from 2000 m to peaks over 4000 m. The massive East Antarctic Plateau lies to the west of the Transantarctic Mountains. The East Antarctic ice sheet descends steeply to the Ross Ice Shelf and the Ross Sea through large glacial valleys in the Transantarctic Mountains such as the Reeves, Skelton, Mulock, Byrd, and Beardmore glacier valleys. Siple Coast borders the Ross Ice Shelf to the east, and it has a more gradual ascent to the West Antarctic Plateau. Ross Island is located in the northwest corner of the Ross Ice Shelf and is a part of the complex topography of that region.

image

Figure 1. Geographic location map for the Ross Ice Shelf, Ross Sea, and surrounding regions. AWS locations are indicated by three-letter IDs. Topography contours are in 250 m intervals. The SOM domain is indicated by the large bold rectangle. The boxes over the East Antarctic Plateau and West Antarctica indicate the location of model points in creating the node-averaged vertical profiles.

Download figure to PowerPoint

[8] The following section provides background information on the numerical model, AWS data sources, and a description of the method of SOMs and how SOMs were used to further characterize the low-level wind field of the Ross Ice Shelf region. Section 3 provides a description of the SOM results, a comparison of the SOM to the associated synoptic patterns, and a verification against AWS observations. The fourth section provides a description of the SOM for the horizontal wind at the 5th lowest model level. A more complete analysis is covered including analyzing seasonal fluctuations and a concentration on the RAS. A summary of the results is provided in the final section.

2. Data Sources and Methodology

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

2.1. Model Data: Antarctic Mesoscale Prediction System (AMPS)

[9] Data from the years 2001 to 2005 from AMPS is used to characterize the low-level wind field. During this time period AMPS was based on a version of the Fifth-generation Pennsylvania State University/NCAR Mesoscale Model, that had been modified for use in the polar regions, referred to as the Polar MM5. Bromwich et al. [2001] and Cassano et al. [2001] provide a detailed description of the Polar MM5 and an evaluation of simulations over the Greenland ice sheet. Guo et al. [2003] indicates that the Polar MM5 is reasonably accurate in simulating the atmospheric state over the Antarctic continent on the synoptic scale. Bromwich et al. [2005] evaluated two years of AMPS forecasts in comparison to surface manned and AWS observations and upper-level observations from radiosondes. The results indicate that AMPS produces a good representation of the atmosphere at synoptic time scales. The modeled winds had higher correlation coefficients for wind speed (∼0.6–0.7) on the plateau while a slightly lower value (∼0.5–0.6) for coastal stations. Powers et al. [2003] and Bromwich et al. [2005] provide a detailed description of the configuration and operation of the AMPS simulations used in this study.

[10] This study uses the output from the AMPS forecasts as extracted from the NCAR – Mass Storage System. The 30 km domain is used to characterize the atmosphere as it is the highest resolution domain that covers the entire Ross Ice Shelf, Ross Sea, and surrounding regions. The beginning of the AMPS archive is January 2001 and the end of the 90-30-10-3.3 km AMPS resolutions is December 2005. The analysis presented here uses the AMPS forecasts valid 12, 15, 18, and 21 h after the model initialization time. Each forecast of AMPS model fields is referred to as a time slice and it represents a snapshot of the conditions in three-hourly intervals. The initial 12 h of each AMPS forecast are not analyzed in order to provide the model atmospheric state 12 h to adjust from the coarse resolution initial fields (provided by Global Forecasting System (GFS) analysis) to the higher resolution AMPS grids and topography. This is similar to the methodology in Guo et al. [2003] and Bromwich et al. [2005]. For example, the three-hour forecasts from 12–21 h from the 0000 UTC 25 March 2005 model run are used to represent the 1200–2100 UTC 25 March 2005 conditions. Gaps in the AMPS archive exist due to missing forecasts, which typically reflect them not being originally run due to hardware and/or software problems. The 24–33 h forecasts from the previous model run, or if needed the 36–45 h forecasts from two previous model runs, are used to fill in the missing forecasts. A gap of more than three model runs results in missing time slices in the model archive time series. The total time series is comprised of 14,273 (98.5%) time slices out of the possible 14,488 three-hour time slices from 12 UTC 05 January 2001 to 09 UTC 21 December 2005 (the entire duration of the 30 km archive).

2.2. Observation Data: Antarctic Automatic Weather Stations

[11] Data from the University of Wisconsin Antarctic automatic weather station (AWS) program are used to provide observations to supplement the model analysis and for further characterization of the wind field. The AWS sites provide year-round measurements at locations throughout Antarctica [Stearns et al., 1993]. Each AWS measures wind speed, wind direction, temperature, and atmospheric pressure. The wind speed, wind direction, and temperature are measured at the top of the tower, at a nominal height of 3.9 m. Station atmospheric pressure is measured at the electronics enclosure, at a nominal height of 1.5 m. The heights of the sensors change over time due to snow accumulation at the site. Oftentimes, the AWS unit or specific sensors may fail in the harsh Antarctic environment, and this limits the data collected from a specific site. Measurements from the sensors are made every 10 min. A semi-automated quality control process is applied to the AWS 10-min data to produce three-hourly observations using the closest valid observation within 40 min of the hour. Figure 1 shows the location of selected AWS sites for the greater Ross Ice Shelf region. These AWS surface observations are the sole source of in situ measurements available to characterize the wind field across the Ross Ice Shelf. There are no regular atmospheric observations above the surface of the Ross Ice Shelf. The upper air observations from McMurdo Station have little value to this study due to the extreme local topographic influences on the low-level winds in the northwest Ross Ice Shelf region [Seefeldt et al., 2003].

2.3. Self-Organizing Maps: Background

[12] The method of self-organizing maps (SOMs) is used to identify patterns in the modeled low-level wind field. SOMs are one of several techniques that can be used to stratify large volumes of data (such as high temporal resolution fields of the atmosphere) into a small number of recurring patterns on a physically meaningful basis. The SOM methodology uses an unsupervised and objective classification procedure to group events into common patterns or clusters, known as nodes. The patterns are then displayed as a two-dimensional array of nodes that is known as a map. Used in this manner the SOM technique is similar to other cluster analysis methods in that it seeks to define common patterns in the input data. Kohonen [2001] provides a theoretical discussion on the SOM technique. Sheridan and Lee [2011] provide a review of the application of SOMs in synoptic climatology.

[13] The method of SOMs was chosen for this study because of the distinct strengths in using SOMs [Hewitson and Crane, 2002; Sheridan and Lee, 2011]. One distinct advantage of SOMs is a better visualization of the node patterns across the resultant SOM. Another strength is that patterns identified by the SOM training cover the entire continuum of events depicted in the training data (i.e., the 3-hourly AMPS model output for this study). This provides confidence that the resulting SOM is capturing the entire spectrum of the patterns in the low-level wind field. The SOM training also places similar nodes next to each other and contrasting nodes further apart, resulting in an easy way to analyze similar and dissimilar patterns on the map. This resulting distribution of nodes thus provides both node-by-node analysis of the depicted events as well as area-by-area analysis if desired. Very different synoptic states map to the corners and edge of the SOM map. Last, more nodes are clustered in regions with a higher density of time slices across the data space. Each resulting node represents an approximation of the mean of the selected training variable for the time slices comprising that node.

2.4. Application of the SOM Technique to the Low-Level Wind Field Over the Ross Ice Shelf

[14] The domain (Figure 1, bold rectangle) for the SOM analysis of the low-level wind field is defined to identify the primary low-level airflows descending into the Ross Ice Shelf region. The western, eastern, and southern limits were selected based on the edges being at the approximate location of ridgelines for the East Antarctic Plateau and the West Antarctic Plateau. The ridgelines were chosen as they roughly represent the drainage region for the katabatic winds across the plateaus [see Parish and Bromwich, 2007, Figure 3]. The northern limit is just beyond the northern edge of the continent to limit the influence of the winds associated with the frequent passage of cyclones in the southern ocean. The resulting domain from the AMPS 30 km grid is 86 points x 69 points.

[15] The software package and associated algorithms used for the SOM analysis in this study is freely available (http://www.cis.hut.fi/research/som-research). An in-depth description of the software is provided by Kohonen et al. [1996]. This software was applied to a total of 14,273 time slices of the zonal and meridional components of wind from the AMPS 30 km grid. No normalization was applied to the zonal and meridional components of the wind, contrary to some previous SOM studies based on sea level pressure [e.g., Nigro et al., 2011]. A series of 4 × 3, 5 × 4, 6 × 4, 6 × 5, 7 × 5, 8 × 6, and 9 × 7 maps were created to determine the best size map to represent a variety of wind field patterns and also not being too selective to result in nodes representing a limited number of events. Numerous SOMs were created for each map size using a variety of settings in the SOM analysis software. A final map for each of the seven different SOM map sizes was selected based on having the lowest, or one of the lowest, qerror values (a measure of the cumulative squared difference between the training data and the resulting SOM) in combination with a subjective selection based on having a preferred flat Sammon map (a two-dimensional representation of the relationship of the higher-dimensional SOM space [Hewitson and Crane, 2002]), and a SOM with a variety of representative nodes. The different map sizes were analyzed among themselves and evaluated in terms of value gained by going to more nodes in relation to the increasing complexity of a larger number of nodes. The 5 × 4 SOM was selected based on this evaluation.

[16] Once the SOM is selected, data can be mapped to the SOM by identifying the node which has the smallest cumulative squared difference in zonal and meridional wind components, over all grid points in the analysis of the domain, to the data sample of interest. This is repeated for all time slices and results in all time slices being associated with a single node on the SOM. From this the frequency of occurrence of each node in the data set can be calculated. Further, averages of other variables of interest (such as sea level pressure) can be calculated for each node by averaging that variable for all time slices that map to a particular node. This averaging provides a way to evaluate other atmospheric properties associated with a specific pattern.

3. Analysis: Wind at the Lowest Model Level

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

3.1. The 2001–2005 Average

[17] Figure 2 is a plot of the average horizontal wind at the lowest model level (approximately 13 m AGL over the Ross Ice Shelf) from the AMPS 2001–2005 30 km archive. The contours represent the average vector wind speed and the arrows represent the average wind direction. The plot indicates similar features as shown in previous research studies including Schwerdtfeger [1984, Figure 3.17], Parish and Bromwich [1987, Figure 3], and Parish and Bromwich [2007, Figure 3]. Confluence zones, present above the primary glacier valleys, including Byrd and Reeves Glaciers, are distinctly noticeable in the 5-year average of the near-surface wind field. The Siple Coast confluence zone has air flowing into it from the East and West Antarctic Plateaus. Strong southerly winds are located parallel to the Transantarctic Mountains, along the western Ross Ice Shelf, indicating the influence of barrier winds. The air flows around the complex topography of the northwest Ross Ice Shelf including a region of light winds downwind of Ross Island. Such features are in agreement with the high resolution study of the Ross Island region by Seefeldt et al. [2003]. The 2001–2005 average also shows how the Ross Ice Shelf serves as a climatological outflow region for the low-level winds across this sector of the Antarctic continent.

image

Figure 2. Average horizontal wind for the lowest model level (approximately 11–13 m AGL) for the greater Ross Ice Shelf region from the AMPS 30 km archive for 2001–2005 Contour shading indicates the average vector wind speed and the arrows indicate the wind direction. Contour lines are in intervals of 2.5 m s−1.

Download figure to PowerPoint

3.2. SOM of the Wind at the Lowest Model Level

[18] The method of SOMs is used to analyze the horizontal wind for the lowest model level from the AMPS 2001–2005 30 km simulations. The resultant 5 × 4 SOM (Figure 3) shows 20 different nodes that highlight common patterns in the near-surface wind field that makes up the more general 2001–2005 average (Figure 2). Each node is labeled in brackets above the top-left corner. The frequency of occurrence for each node, also referred to as node frequency, is included in parentheses next to the label for each node. The node frequency is determined by the number of time slices mapped to a given node divided by the total number of time slices from the 2001–2005 AMPS data set. A simplified analysis follows for the SOM of the horizontal wind for the lowest model level. A more extensive SOM analysis for the horizontal wind at the fifth-lowest model level will be presented in the next section.

image

Figure 3. Self-organizing map of the horizontal wind for the lowest model level (approximately 11–13 m AGL) from the AMPS 30 km domain for 2001–2005. Contour shading indicates the average vector wind speed and the arrows indicate the wind direction. Contour lines are in intervals of 2.5 m s−1.

Download figure to PowerPoint

[19] There are several dominant patterns that can be identified by grouping together nodes with similar features. The node-averaged sea level pressure plot (Figure 4) provides an understanding of the corresponding synoptic structure for the general patterns as well as individual nodes. Light winds are found covering the center nodes of the top two rows (nodes [2,1] to [4,1] and [2,2] to [4,2]). Below average wind speeds are found over the entire domain for these six nodes. Nodes [2,1] and [3,1] show the weakest node-averaged winds. The plot of node-average sea level pressure (Figure 4) indicates little to no synoptic structure for nodes [2,1] to [4,1]. A weak cyclone progressing across the Ross Sea is indicated for nodes [2,2] to [4,2]. The nodes covering the left column and bottom row show the progression of the Ross Sea cyclone from Cape Adare, across the Ross Sea, and to Cape Colbeck. In node [1,1] there are high wind speeds near Cape Adare as the cyclone is modified with the edge of the continent. Strong barrier parallel winds are found over the western Ross Sea along the Transantarctic Mountains in nodes [1,2] to [1,4] and [2,4]. Such a barrier parallel flow is expected as the clockwise circulation around the cyclone forces the air up against the steep barrier of the Transantarctic Mountains. As the cyclone advances toward the eastern Ross Sea the winds near Cape Colbeck increase as the airflows over the increasing elevation of West Antarctica. The winds increase in intensity over the West Antarctica Plateau and the Siple Coast confluence zone as the cyclone reaches Cape Colbeck. Nodes [2,3] to [4,3] are very similar to that of nodes [2,4] to [4,4] except that the cyclone is not as intense in [2,3] to [4,3] as it is in [2,4] to [4,4]. The light wind nodes of [2,2] to [4,2] show a similar but even weaker series of patterns.

image

Figure 4. Sea level pressure averaged for each node of the SOM of horizontal wind for the lowest model level from the AMPS 30 km 2001–2005 archive. Isobars are in intervals of 2 hPa. The map projection is larger for the node-averaged sea level pressure than the SOM of wind.

Download figure to PowerPoint

[20] The SOM of the horizontal wind for the lowest model level shows little variation in wind direction over the East Antarctic plateau. A difference between the node-average wind direction and the 2001–2005 average (not shown) indicates that the node-averaged wind direction varies by less than 30° across all nodes over the continent. Over most of the East Antarctic Plateau the difference in wind direction is generally less than 7.5°. Such wind constancy is an established feature of the Antarctic katabatic winds that are highly dependent on the underlying topography. This is in agreement with the results of Parish and Cassano [2003] that indicated that the low-level wind field over the Antarctic ice sheet shows little variability as both synoptic forcing and katabatic forcing varies. The differences in the node-averaged wind direction over West Antarctica is larger than that of the East Antarctic Plateau but still generally within 30° except near Cape Colbeck where significant influences are experienced from the passing Ross Sea cyclones.

[21] An analysis of the difference of the vector wind speed from the 2001–2005 mean (not shown) highlights the unique patterns across the range of SOM nodes. The largest differences in vector wind speed are found across the East Antarctic Plateau, West Antarctica, and the western Ross Sea. The opposing nodes [1,1] and [5,1] in Figure 3 highlight the dramatic difference in vector wind speed on the East Antarctic Plateau under similar synoptic forcing (Figure 4). There is also enhanced katabatic activity over the East Antarctic Plateau, and through the glacier valleys, when a cyclone is positioned near Cape Colbeck. A similar contrast in opposing nodes is seen with node [1,4] being defined by high vector wind speeds over the western Ross Sea while, node [5,4] has high wind speeds across West Antarctica. A careful analysis of the SOM shows that the location of the high vector wind speeds is strongly dependent on the position of the Ross Sea cyclone.

[22] The SOM of the wind for the lowest model level provide a wealth of information over that of the 5-year average. Dominant patterns from within the 5-year average can be identified and analyzed. However, the largest differences in the horizontal wind are dependent on the characteristics in association to a synoptic environment with a weak pressure gradient across the Ross Ice Shelf, or with the progression of cyclones across the Ross Sea. The variation in vector wind speed and wind direction away from the primary cyclone is relatively small in comparison to the variation near the location of the cyclone. This then indicates that the resultant SOM training is more dependent on the position of the cyclone, because it is the dominant and most varying feature across the SOM domain, than the unique characteristics of the near-surface horizontal wind across the region. The activity of the cyclones and the katabatic winds plays a large role in shaping the near-surface wind field of the Ross Ice Shelf. However, those dominant features at the lowest model level also minimize the identification of the RAS in the SOM training, since the SOM training captures the areas of highest variability. An analysis at a level above the surface is necessary in order to highlight and characterize the features of the RAS across this SOM domain and will be presented in section 4.

3.3. Comparison of the SOM to AWS Observations

[23] Comparing the SOM patterns to AWS observations of the Ross Ice Shelf region provides a qualitative verification of the SOM for the horizontal wind at the lowest model level. The AMPS 30 km resolution used in this study will have limitations in a direct comparison to AWS observations due to the close proximity and significance of the nearby topography in shaping the near surface wind field. In particular, many of the glacial valleys in the Transantarctic Mountains are not adequately resolved with the 30-km grid spacing used in this version of AMPS. Figure 5 is a collection of wind rose plots for the Ross Ice Shelf region based on AWS observations for the year 2005. The data for 2005 was selected as it represents the best and most continuous collection of AWS data for the region. Previous years do not include some of the AWS sites and hardware difficulties resulted in other sites not providing continuous data. All plotted wind roses meet the requirement of having over 50% valid observations out of the available observations corresponding to each node. Most plotted wind roses are based on greater than 95% of the available observations. The length of each petal represents the frequency of occurrence in each of the 16 wind direction sectors. The color of the petal indicates wind speed. Purple is less than 2.0 m s−1, green is between 2.0 and 5.9 m s−1, blue is between 6.0 and 9.9 m s−1, and red is greater than 10.0 m s−1.

image

Figure 5. Wind rose plots of AWS data averaged by node corresponding to the SOM for the lowest model level. (a) Node [1,1], (b) node [3,1], (c) node [1,4], and (d) node [5,4]. The length of each petal indicates the frequency. Each circle around the center indicates a frequency increment of 5%. The color of the petal indicates wind speed. Magenta is less than 2.0 m s−1, green is between 2.0 and 5.9 m s−1, blue is between 6.0 and 9.9 m s−1, and red is greater than 10.0 m s−1. See Figure 1 for AWS names. The SOM node plots are the same as in Figure 3.

Download figure to PowerPoint

[24] Wind rose plots for four selected nodes are included in Figure 5. The wind rose plots were limited to four nodes in order to ensure readability of the wind roses for each AWS site. The selected nodes ([1,1], [3,1], [1,4], and [5,4]) were chosen in order to capture well-defined and contrasting nodes for the verification. The SOM node plots in Figure 5 are the same as that in Figure 3, but are on a map projection corresponding to that of the AWS wind rose plots, and are included for easier side-by-side verification. Node [1,1] (Figure 5a) represents the conditions with a cyclone located near Cape Adare. There are light to moderate winds in the Siple Coast confluence zone with winds approximately 2–10 m s−1 (green and blue wind rose petals). The winds across the Ross Ice Shelf are light and variable. The AWS wind rose plots compare favorably to the SOM node with light winds across the entire Ross Ice Shelf as well as the Siple Coast region. Node [3,1] (Figure 5b) shows slightly stronger winds in the Siple Coast confluence zone and light katabatic winds across the Ross Ice Shelf (indicated by westerly winds with green and blue petals). Such features correspond nicely to the SOM node [3,1] that shows more westerly wind directions at the base of Byrd Glacier and extending northward. Node [1,4] (Figure 5c) shows a strong barrier wind pattern. The Siple Coast confluence zone has strong (mostly red petals) easterly flow extending onto the Ross Ice Shelf. The increased wind speeds for this region are indicated in both the wind rose plot and the SOM node. Moderate to strong (blues and reds) southerly winds at all locations dominate the AWS sites on the Ross Ice Shelf. This barrier parallel flow, shown in the AWS wind rose plots, corresponds favorably to that of the SOM node. The SOM node does indicate some katabatic drainage from Byrd Glacier and this is indicated with some westerly winds at the Marilyn AWS site. Node [5,4] (Figure 5d) continues to indicate the strong winds over West Antarctica at the AWS sites near the Siple Coast (Erin and Briana AWS). Moderate to strong westerly winds characterize the AWS sites on the Ross Ice Shelf. These winds are indicating strong katabatic flow down the glacier valleys. Such conditions are in agreement with the SOM node showing a stream of higher wind speeds flowing down Byrd Glacier and onto the Ross Ice Shelf. The SOM node indicates winds that are more southwesterly than westerly. This is likely a difference due to the relatively coarse AMPS 30 km resolution.

[25] The wind rose plots for the four selected nodes indicate a strong visual correlation between the SOM node plots and the AWS observations for the year 2005. Such verification provides confidence that the descriptions of the near-surface wind based on AMPS are reasonable. The four selected nodes also correspond closely to the wind rose plots for the four dominant wind regimes as presented in the observational study of near-surface winds of the Ross Ice Shelf region by Seefeldt et al. [2007]. In that study, dominant wind regimes were identified based on criteria at select AWS sites. The dominant regimes from the study are light wind (node [1,1]), weak katabatic (node [3,1]), barrier wind (node [1,4]), and strong katabatic (node [5,4]).

4. Analysis: Wind at the 5th Lowest Model Level

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

4.1. The 2001–2005 Average

[26] One of the advantages in working with data from numerical simulations is the ability to look at multiple vertical levels. The 5th lowest model level in AMPS was selected based on the results from Seefeldt and Cassano [2008, Figure 3]. In their study the wind at every model level was averaged across a selected region over the western Ross Ice Shelf and the Siple Coast region of West Antarctica. This region roughly corresponds to the location of the RAS. The 5th lowest model level in AMPS was found to be the level with the highest wind speeds. The level of maximum wind speed of the RAS is dependent on latitude, proximity to nearby topography, and intensity of the RAS [Parish et al., 2006]. Some components of the RAS, such as katabatic flow through the glacier valleys in the Transantarctic Mountains have also been found to be at heights lower than the maximum wind speed of the RAS. Additionally, Steinhoff et al. [2009] concluded that AMPS provides more realistic simulations of conditions aloft than near the surface due to problems with the model planetary boundary layer parameterization.

[27] Figure 6 is a plot of the average horizontal wind for the 5th lowest model level, at an approximate height of 150 m above ground level, from the 2001–2005 AMPS simulations. The presence of the RAS is much more clearly indicated across the western Ross Ice Shelf at the 5th lowest model level than at the lowest model level (Figure 2). The RAS is seen as a continuous corridor of atmospheric mass from the continental interior, crossing the western Ross Ice Shelf, and extending northward across the western Ross Sea. A heavy line is drawn on Figure 6 indicating an approximate location of the RAS. The RAS is being fed by air flowing down Siple Coast and along the Transantarctic Mountains, air flowing down the Byrd Glacier valley, and airflow from the Reeves Glacier valley. A barrier wind component is indicated by the presence of the RAS parallel to and adjacent to the Transantarctic Mountains. The cyclonic wind vectors and zero vector mean wind over the eastern Ross Sea indicates a climatologically averaged presence of a cyclone in the region. There are three wind speed maxima in the RAS positioned adjacent to the Transantarctic Mountains, the same as indicated by Parish et al. [2006] and Seefeldt and Cassano [2008]. Nigro et al. [2012] presents a case study on the localized wind maximum near the Prince Olav Mountains, and identified this feature as a barrier wind corner jet.

image

Figure 6. Average horizontal wind field for the 5th lowest model level (approximately 150–160 m AGL) for the greater Ross Ice Shelf region from the AMPS 30 km archive for 2001–2005. Contour shading indicates the average vector wind speed and the arrows indicate the wind direction. Contour lines are in intervals of 2.5 m s−1.

Download figure to PowerPoint

4.2. SOM of the Wind at the 5th Lowest Model Level

[28] The method of SOMs can be used to further analyze and characterize the RAS at the 5th lowest model level. Figure 7 is a SOM of the horizontal wind for the 5th lowest model level from the AMPS 2001–2005 30 km archive. The same process in creating the SOM for the lowest model level was used in creating this SOM. The most pronounced feature used to identify the RAS in the SOM is the presence of a corridor of atmospheric mass transported from the continental interior northward. The characteristics of the RAS are based on a composite of atmospheric forcing features in the region including synoptic-scale cyclones, mesocyclones on the Ross Ice Shelf and over the Ross Sea, katabatic winds from West Antarctica and the East Antarctic Plateau, and barrier winds. The synoptic influences associated with the different SOM wind patterns can be analyzed by evaluating the node-averaged sea level pressure (Figure 8) for each node of the SOM. Unfortunately, the small scale and transitory movement of the regional mesocyclones makes them almost impossible to identify in this synoptic-climatology study, and the influence of these systems on the RAS likely has to be analyzed at the case-study level. Parish et al. [2006] identifies the presence of blocking, and hence barrier wind dynamics, to be when the largest pressure gradient is located adjacent and parallel to the Transantarctic Mountains. A barrier wind can be identified in this manner in the node-averaged sea level pressure (Figure 8) and verified with the SOM of the horizontal wind at the 5th lowest model level (Figure 7).

image

Figure 7. Self-organizing map of the horizontal wind for the 5th lowest model level (approximately 150–160 m AGL) from the AMPS 30 km domain for 2001–2005. Contour shading indicates the average vector wind speed and the arrows indicate the wind direction. Contour lines are in intervals of 2.5 m s−1.

Download figure to PowerPoint

image

Figure 8. Sea level pressure averaged for each node of the SOM of horizontal wind for the 5th lowest model level from the AMPS 30 km 2001–2005 archive. Isobars are in intervals of 2 hPa.

Download figure to PowerPoint

[29] Five distinct patters in the horizontal wind at the 5th lowest model level can be identified through an examination of the SOM (Figure 7). The 2001–2005 mean wind speed (Figure 6) is subtracted from each node-averaged value of wind speed (Figure 7) to create the wind speed differential plot in Figure 9. Figure 9 helps highlight the features that make the horizontal wind pattern of each node unique. The 2001–2005 mean (Figure 6) is subtracted from each node-averaged value of wind speed to create the wind speed differential plot. Nodes [2,1], [3,1], [2,2], and [3,2] are dominated by light wind conditions across the entire SOM domain. The node-averaged wind speed differences (Figure 9) show lighter wind speeds on the range of 2–6 m s−1 with basically no regions with higher than average wind speeds. The node-averaged sea level pressure (Figure 8) lacks any defined synoptic pattern for these nodes. These four nodes will be referred to as light wind (LW) and have a cyan background color on the SOM. Nodes along the left column ([1,1] to [1,4]) and part of the bottom row ([1,4] to [3,4]) have a horizontal wind pattern that is dominated by the movement of a cyclone from Cape Adare to Cape Colbeck across the Ross Sea. The node-averaged sea level pressure (Figure 8) shows the presence and progression of the large cyclone across the Ross Sea. Such a cyclone track is a common feature of the Ross Sea [King and Turner, 1997]. These six nodes will be collectively referred to as the Ross Sea (RS) cyclone nodes and are indicated with a yellow background color on the SOM. The four nodes making up the upper-right corner ([4,1], [5,1], [4,2], and [5,2]) are characterized by katabatic winds on the East Antarctic Plateau. The differential node-averaged wind speeds (Figure 9) show lighter wind speeds across the Ross Ice Shelf and most of West Antarctica and higher wind speeds across the East Antarctic Plateau. The amount of katabatic activity increases diagonally from node [4,1] to node [5,2]. The node averaged sea level pressure shows minimal synoptic activity in nodes [4,1] and [5,1] and the slight presence of a cyclone in the Ross Sea in nodes [4,2] and [5,2]. These four nodes will be referred to as katabatic (K) and highlighted with a red background. Nodes [4,3], [5,3], [4,4], and [5,4], in the lower-right corner of the SOM, indicate the influence of a significant cyclone near Cape Colbeck and over West Antarctica in addition to strong katabatic winds. The differential node-averaged wind speeds show high wind speed values for these four nodes across the East Antarctica Plateau, West Antarctica, and the Ross Ice Shelf. The intensity of the features increases diagonally from node [4,3] to [5,4]. These four nodes will be referred to as hybrid (H), as they have significant katabatic and synoptic influences, and they have a background color of magenta on the SOM. The remaining two nodes ([2,3] and [3,3]) do not have any pronounced defining characteristics. The node-averaged sea level pressure shows the presence of a weaker cyclone in the Ross Sea than what is seen in the Ross Sea cyclone nodes. Below average wind speeds on the East Antarctic Plateau shows that there is minimal katabatic activity for these nodes. The presence of such moderately defined nodes makes sense given that oftentimes, especially during seasonal transitions, weather patterns are less extreme. These two nodes are referred to as the moderate (M) nodes and are indicated with a blue background color.

image

Figure 9. Difference between node-average wind speed and the 2001–2005 average wind speed (node average minus 2001–2005 average) for the SOM of horizontal wind for the 5th lowest model level. Contour lines are in intervals of 2 m s−1 and contour shading is in intervals of 1 m s−1. The non-shaded region covers values in the −1 m s−1 to +1 m s−1 range.

Download figure to PowerPoint

[30] The defining characteristic of the RAS is a pronounced corridor of atmospheric mass transport northward across the Ross Ice Shelf and Ross Sea from the continental interior. The identification of a RAS in the node patterns is ultimately a subjective classification due to the distinction of when there is a continual stream of air flowing across the Ross Ice Shelf and Ross Sea. An additional factor to evaluate is if the stream of air is being maintained by the three sources feeding air from the continental interior. A review of the SOM for the 5th lowest model level (Figure 7) shows that the light wind nodes do not have RAS signatures due to the minimal amount of air flowing onto the Ross Ice Shelf from the continental interior. The katabatic nodes (K) also do not indicate the presence of a RAS. These nodes show a katabatic flow onto the Ross Ice Shelf and Ross Sea but the airflow lacks continuous strong flow across the entire north-south extent of the Ross Ice Shelf. For these nodes an ambient pressure gradient to drive the katabatic air northward in a continual stream of air is lacking. Node [1,1] of the Ross Sea (RS) cyclone nodes has a similar pattern on the Ross Ice Shelf and the southern Ross Sea as the light wind nodes, with a lighter than average winds across much of this area (Figure 9), and therefore does not indicate the presence of a RAS. Node [1,2] of the Ross Sea (RS) cyclone nodes does show a continual stream of air over the western Ross Sea fed by air from the continental interior at the Byrd and Reeves Glacier valleys. The node can be classified as a partial RAS because it is basically only present over the western Ross Sea. All other nodes ([1,3] to [5,4]), including the remaining Ross Sea cyclone nodes, the moderate (M) nodes, and the hybrid (H) nodes, show a continual northward corridor of atmospheric mass transport across the western Ross Ice Shelf and the western Ross Sea. In some nodes (e.g., [3,3] to [5,3]) the RAS is weak in intensity. In other nodes (e.g., [1,3], [1,4] to [5,4]) the RAS is strong in intensity. Similarly, in some nodes the RAS is fed through all three source regions and in other nodes one of the source regions is weak (e.g., Reeves Glacier in nodes [3,3] to [5,3]). The nodes classified as having a RAS present are indicated by ‘RAS’ being placed above the node in Figure 7.

[31] The associated katabatic influences on the wind regimes can be evaluated by analyzing node-averaged vertical profiles of potential temperature upstream of Byrd and Reeves Glaciers and over the Siple Coast confluence zone of West Antarctica (Figure 10). Katabatic winds are dependent on the strength of the near-surface inversion and the terrain slope. A stronger near-surface inversion, over a given terrain slope, provides more katabatic forcing. Parish and Cassano [2003] estimate the inversion strength by using a linear interpolation of the ambient temperature profile (the profile at and above the 7th lowest model level for this study) and extending the interpolation toward the surface. The difference between the interpolated temperature profile and the actual temperature indicates the inversion strength. Higher inversion strengths indicate a larger amount of katabatic forcing on the plateaus. The regions used to calculate the average potential temperature profiles in Figure 10 are indicated in Figure 1 and are located upstream of Byrd Glacier, upstream of Reeves Glacier, and over West Antarctica. Node-averaged values for six nodes are provided in Figure 10. The selected nodes were chosen to represent the five different dominant SOM patterns. A second node was selected for the Ross Sea cyclone nodes given the variation in conditions across that dominant pattern. A dashed line has been added to the profiles of potential temperature representing a linear interpolation of the potential temperature profile from the 7th to the 12th lowest model levels. The difference between the dashed line (the linear interpolation) and the surface temperature is referred to as the inversion strength. A value indicating the inversion strength (K) is included on each profile next to the dashed line. The light-wind node shows an inversion strength around 6 K over the selected East Antarctic Plateau regions. This value is at least half that of all other selected nodes which is in agreement with the minimal katabatic winds indicated on the SOM. The vertical profile representing the moderate nodes has the next lowest values for the East Antarctic Plateau regions. The selected nodes for the katabatic, Ross Sea cyclone, and hybrid domains indicate an inversion strength of 14–16 K over the East Antarctic Plateaus regions. The high values of inversion strength correspond favorably to the above average winds, mostly katabatic, on the East Antarctic Plateau for these selected nodes. The region over West Antarctica is more complicated. For example in the hybrid node the considerable mixing associated with the nearby cyclone results in a relatively weak inversion strength.

image

Figure 10. Node-averaged vertical profile of potential temperature (K; left side) and wind speed (m s−1; right side) for selected regions and selected nodes representing the different dominant patterns. Wind barbs are plotted at the height of each model level. The number at the base of each temperature profile is the strength of the inversion (K). See text for a description of calculating this value. The selected regions are indicated in Figure 1.

Download figure to PowerPoint

[32] The ambient forcing associated with the synoptic-scale cyclones is the dominant forcing mechanism for determining the presence of a RAS (Figure 7 and 8). Without the ambient pressure gradient there is not sufficient forcing to transport the atmospheric mass northward over large distances across the Ross Ice Shelf and Ross Sea. In the katabatic nodes there is katabatic winds across the East Antarctic Plateau and descending over the Ross Ice Shelf and Ross Sea. Yet the katabatic winds lose the katabatic forcing when it reaches the Ross Ice Shelf and there is not a cyclone present to drive this flow across the Ross Ice Shelf (Figures 8 and 9). In all of the RAS nodes there is a cyclone present in the Ross Sea (Figure 7 and 8). The one Ross Sea cyclone node ([1,1]) that is not classified as a RAS has the cyclone positioned north of Cape Adare. The intensity of the RAS is dependent on the position and associated pressure gradient of the Ross Sea cyclone. The nodes with the largest pressure gradient (see Figure 7) are [1,3], [1,4], [2,4], and [5,4]. These nodes also show the strongest wind speeds across the western Ross Ice Shelf and/or Ross Sea within the RAS.

[33] There is a significant katabatic component to the RAS, particularly near southern Ross Ice Shelf and the Siple Coast. However, this katabatic component must be aided by the presence of the Ross Sea cyclone to drive the air of katabatic origin across the Ross Ice Shelf. Node [5,4], part of the hybrid dominant pattern, shows the additional intensity of the RAS due to a strong katabatic flow from the Byrd Glacier with high node differential wind speed values downstream of Byrd Glacier. The contributions of the katabatic winds for node [5,4] is also indicated by node [5,4] having one of the largest values of inversion strength, as indicated in Figure 10. The amount of katabatic activity on the East Antarctic Plateau is also influenced by the position of the Ross Sea cyclone. The differential of node-averaged wind speed (Figure 9) can be used as a quick estimate of the amount of katabatic activity. As discussed previously, the node-averaged vertical profile of potential temperature and inversion strength can be used to approximate the amount of katabatic forcing. When the cyclone is positioned over the western Ross Sea (e.g., nodes [1,1], [1,2], and [1,3]) the amount of katabatic activity appears to be about that of the 5-year average or slightly less. As the cyclone moves to the eastern Ross Sea the amount of katabatic activity is increased. The katabatic winds are enhanced the most when the cyclone is over Cape Colbeck / West Antarctica (e.g., nodes [4,4] and [5,4]). The inversion strength upstream of Byrd Glacier is 16.1 K for node [1,4] and 15.9 K for node [5,4]. Meanwhile the wind speed values on the East Antarctic Plateau are approximately 4–6 m s−1 higher for node [5,4] than node [1,4]. The inversion strength upstream of Reeves Glacier is larger (16.3 K) for node [5,4] than node [1,4] (15.0 K). The wind speed for node [5,4] is also 6–8 m s−1 higher than node [1,4] indicating that the larger inversion strength and the synoptic-scale cyclone are both enhancing the wind flow in the Reeves Glacier region.

[34] The west-to-east extent of the RAS is dependent on the position of the Ross Sea cyclone. The RAS extends toward the center of the Ross Ice Shelf in nodes [1,3], [2,4], and especially in [1,4]. The node-averaged sea level pressure patterns show the position of the center of the cyclone to be over the central to eastern Ross Sea for these nodes. When the cyclone is located near Cape Adare and near Cape Colbeck / West Antarctica the RAS remains in close proximity of the Transantarctic Mountains.

[35] Seasonal node frequencies are able to describe the SOM patterns and the variations in the RAS over the course of the year. The seasons for this analysis are defined as December–January (DJ), February–March–April (FMA), May–June–July–August (MJJA), and September–October–November (SON) for the summer, fall, winter, and spring seasons. A two-month summer and a four-month winter were chosen as this distribution corresponds more closely to the annual progression of temperature and radiation in the Antarctic. This phenomenon has been referred to as the coreless winter [Schwerdtfeger, 1984]. There is a strong seasonal dependency in the occurrence of the different nodes for the SOM of the 5th lowest model-level. The light wind nodes dominate the summer months. Nearly 60% of the summer (DJ) time slices map to the light wind nodes ([2,1]: 14.7% + [3,1]: 25.5% + [2,2]: 11.4% + [3,2]: 8.1% = 59.7%). An additional 22% of the summer (DJ) time slices map to the katabatic node with the weakest winds ([4,1]), and the Ross Sea cyclone near Cape Adare ([1,1] and [1,2]). All but one of these nodes ([1,2]) has been classified as not being a RAS node. The remaining summer time slices (∼18%) are distributed to the other 13 nodes with less than 1% frequency of occurrence for most of the nodes with high winds speeds along the bottom row and right column. All but three of these nodes, the remaining katabatic nodes, have been classified as nodes with the RAS present. This evaluation of the summer months indicates that the RAS is rarely present during the summer months. 71% of the winter (MJJA) time slices map to the nodes along the right column ([5,1]–[5,4]), bottom row ([1,4]–[5,4]), and node [1,3]. All of these nodes are classified as having a RAS present. The six nodes making up the light wind nodes and the moderate nodes occur with a total of less than 10% frequency during the winter months. Thus the RAS is a nearly permanent feature for the region during the winter months. The RAS nodes are present approximately half of the time during the spring and fall months. The fall patterns are reasonably distributed across all nodes with frequency of occurrence between 2.5% and 6.9%, except for node [1,1] at 7.8%. The frequency of occurrence is relatively evenly distributed across the different dominant patterns. The frequency of occurrences are 19.9% for the four light wind nodes, 8.7% for the two moderate nodes, 24.8% for the four katabatic nodes, 27.1% for the six Ross Sea cyclone nodes, and 19.5% for the four hybrid nodes. The distribution of frequency of occurrence for the spring months is somewhere between that of the fall and the winter. The Ross Sea cyclone patterns occur 37.8% of the time, close to that of the winter. Meanwhile the light wind nodes and moderate nodes occur a combined 25.4% of the time which is considerably more than the 9.1% for the winter months. This makes it consistent with the spring season transitions from the strong baroclinic regions over the southern ocean during the winter months to the more synoptically weak summer months.

5. Conclusion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

[36] The method of SOMs has been used to identify common patterns in the low-level wind field for the Ross Ice Shelf region using the 2001–2005 AMPS 30 km simulations. Particular attention was placed in identifying and characterizing the presence and location of the RAS. The SOM for the lowest model level identifies 20 patterns in the horizontal wind. From the 20 patterns there are a few dominant patterns. Included in the dominant patterns is a collection of nodes with light winds over most of the SOM domain. The most pronounced dominant pattern is found in the nodes associated with the progression of a cyclone across the Ross Sea. There is little variation in the wind direction over the continent, particularly covering the East Antarctic Plateau, across all 20 nodes. There are large differences in wind speed across the different nodes. The location of high wind speeds for a given node is strongly dependent on the position of the Ross Sea cyclone. A verification of the SOM results was completed by comparing the horizontal winds for four nodes to AWS wind rose plots using observations corresponding to the nodes. The wind roses based on AWS observations showed a favorable visual correlation to the horizontal wind patterns from the SOM. Overall, the SOM for the lowest model level is helpful in characterizing the low-level wind field, but it includes several limitations. Much of the node-to-node contrast is related to the varying position of the Ross Sea cyclone and the variation in wind speeds on the Antarctic plateau. This results in the RAS being poorly identified in the SOM patterns for the lowest model level across this SOM domain.

[37] The presence of the RAS is more clearly indicated in the horizontal wind from the AMPS 30 km simulations at the 5th lowest model level. The AMPS 2001–2005 average for the 5th lowest model level shows a pronounced northward corridor across the Ross Ice Shelf and the Ross Sea. The RAS has three source regions of air from the continental interior: Siple Coast and the southern end of the Ross Ice Shelf, Byrd Glacier valley, and Reeves Glacier valley. The SOM for the horizontal wind of the 5th lowest model level provides greater detail and understanding of the composition, structure, and seasonality of the RAS. Five distinct patterns are indicated in the 5 × 4 SOM including light wind, katabatic, Ross Sea cyclone, hybrid, and moderate. The RAS is found to occur in all of the moderate, hybrid, and all but one of the Ross Sea cyclone nodes. The presence of a cyclone in the Ross Sea is a requirement for the presence of a RAS. The additional ambient pressure gradient is necessary in order to transport the atmospheric mass northward across the Ross Ice Shelf and Ross Sea. The intensity of the RAS is dependent on the position of the cyclone, associated pressure gradient of the cyclone, and the amount of katabatic activity from the interior of Antarctica. The west-to-east position and width of the RAS is primarily determined by the position of the cyclone in the Ross Sea. There is a strong seasonality to the presence and composition of the RAS. Approximately 80% of the summer (DJ) time slices correspond to nodes where the RAS is not present. Winter is nearly the opposite with approximately 68% of the time slices corresponding to a node with the RAS present. Fall and spring are in between these values with 48% and 61% time slices matching to the presence of the RAS.

[38] Overall, using a five-year period of AMPS simulations and the method of SOMs has provided an improved understanding of the RAS. The use of the 5th lowest model level provided an improved perspective on the RAS. Future work can focus on selecting different cases from the five dominant patterns in the SOM for a closer analysis on the forcing mechanisms associated with the RAS. The use of SOMs was not able to capture the small and transient patterns associated with the contributions from mesocyclones. Future studies can concentrate on further developing an understanding of the role of mesocyclones. An observational study using in situ measurements of the low-level wind field would provide the greatest benefit in better understanding the RAS.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References

[39] This work was supported in part by NSF grants OPP-0229645 and ANT-0636811. Seefeldt was supported by a University of Colorado–Cooperative Institute for Research in the Environmental Sciences Graduate Research Fellowship during the research. The AMPS data was retrieved courtesy of the National Center for Atmospheric Research–Computational and Information Systems Laboratory. The automatic weather station data was retrieved from the University of Wisconsin–Madison Antarctic Meteorology Research Center. Jonathan Thom completed post-processing of the AWS data.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data Sources and Methodology
  5. 3. Analysis: Wind at the Lowest Model Level
  6. 4. Analysis: Wind at the 5th Lowest Model Level
  7. 5. Conclusion
  8. Acknowledgments
  9. References