1. Introduction
The Joint Precision Airdrop System (JPADS) was developed to provide military aircraft the ability to conduct cargo airdrop operations with greatly improved accuracy, allowing more operational flexibility—resupplying smaller drop zones from higher altitudes. The evolving JPADS capability relies on advanced GPS guidance systems working in conjunction with parachute decelerators and an advanced computer-based mission planning tool, which is able to rapidly assimilate weather data. The JPADS system makes use of a high-resolution 4D wind forecast provided by atmospheric models. This wind information is used during preflight mission planning, but alone it is not reliable enough for consistent, accurate high-altitude airdrop. To correct for errors in the forecast model, aircrews need the ability to update the winds in flight prior to an airdrop. As currently employed, this wind update is obtained by releasing a GPS dropsonde in the vicinity of the planned drop zone and receiving the wind data by radio. The JPADS Mission Planner (JPADS-MP) software package incorporates this new wind profile and recalculates the computed air release point. With the new release coordinates for the updated winds, the crew is prepared return to the objective area, configure the aircraft for airdrop, and navigate to the updated release point for the drop.
As originally conceived, JPADS airdrops rely on steerable, GPS-guided cargo systems that are able to make glide-path corrections to land very close to their intended point of impact on the ground. Because of the high cost of these guidance systems, the difficulty of returning the reusable components from the field, and the recent requirements for frequent high-altitude combat airdrops, guided cargo systems have not been available in sufficient quantities to meet demand. To overcome the guidance system equipment shortage, the wind profiles measured by GPS dropsondes have been successfully used to improve the accuracy of nonguided Container Delivery System (CDS) airdrops from medium altitudes (5000–10 000 feet above ground level). The process of dropping nonguided ballistic parachute cargo systems from a JPADS-MP computed release point is referred to as Improved Container Delivery System (ICDS; Meier 2010).
In the combat environment, the requirement to fly over the same objective area twice in order to release a dropsonde to measure winds increases the risk to the aircraft and crew. Current procedures for both guided JPADS and ICDS airdrops require airdrop of a GPS dropsonde within 25 nautical miles (n mi; 1 n mi = 1.852 km) of the drop zone, a minimum of 20 min prior to the airdrop of cargo. To prevent the increased exposure of the aircraft in the threat environment, alternative wind-measurement techniques are being pursued (Meier 2010).
The approach investigated in this research is to use satellite sounding data to calculate a wind profile that could be used by the JPADS-MP to calculate an airdrop release point. This technique was originally pursued to take advantage of the passive nature of the satellite sounders and the potential for an aircrew to receive the satellite data broadcast in flight, providing the ability to update winds and eliminating the need for the first aircraft pass over the drop zone. The early phases of this research took place in 2009, and at that time in-flight data-link capability on military aircraft was not common, adding to the appeal of the satellite-based solution. In the years since, airborne data-link capability has expanded, and alternative sources for wind updates have become available.
2. Sources of IR/microwave sounder data
The AIRS instrument is a passive sensing system on board NASA’s Aqua spacecraft. Aqua was launched in 2002 as part of the Earth Observing System (EOS) project. The scientific instruments on Aqua collect data related to global water cycles to improve weather prediction and understanding of climate change. The AIRS instrument uses infrared hyperspectral sensing to passively measure temperature and humidity. Based on density profiles of atmospheric constituent gases responsible for molecular absorption, there is a weighting function for each of the 2378 AIRS channels (with wavelengths between 3.7 and 15.4 μm). By measuring the infrared radiance at each of these AIRS channels and calculating temperature using the Planck equation, AIRS provides the data necessary for the calculation of vertical temperature profiles from the surface (or top of thick cloud cover) to more than 70 km in height (JPL 2014).
When cloud cover prevents accurate IR temperature retrieval from the lower atmosphere and surface, measurements made by the Advanced Microwave Sounding Unit (AMSU), also on board the Aqua satellite, are used to fill in missing data. This allows NASA to provide an integrated AIRS/AMSU dataset with full temperature profiles for locations with a range of cloud-cover conditions, only missing data in regions with heavy precipitation. The vertical resolution of the 15-channel AMSU-A measurements does not match the AIRS data, but the AMSU temperatures are interpolated to the AIRS resolution. The substitution of AMSU temperatures for missing AIRS temperatures is made during the level 2 processing of the data, and it is an integrated AIRS/AMSU data file that is downloaded for use in this research. Figure 1 shows a NASA depiction of AIRS scan geometry and AIRS and AMSU horizontal resolution.
NASA illustration of scan geometry and coverage pattern for AIRS and AMSU instruments on board the Aqua satellite. Image courtesy of AIRS Science Team, NASA/JPL–California Institute of Technology (GES DISC 2016).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
NASA illustration of scan geometry and coverage pattern for AIRS and AMSU instruments on board the Aqua satellite. Image courtesy of AIRS Science Team, NASA/JPL–California Institute of Technology (GES DISC 2016).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
NASA illustration of scan geometry and coverage pattern for AIRS and AMSU instruments on board the Aqua satellite. Image courtesy of AIRS Science Team, NASA/JPL–California Institute of Technology (GES DISC 2016).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
The AIRS data are made publicly available by NASA in 6-min data granules. In this research, NASA’s Mirador Earth science data search tool is used to select the appropriate combined AIRS and AMSU level 2 data granules for download (AIRS Science Team/Joao Texeira 2013). Within each data granule, there are 1350 sounding locations (45 points along the direction of the flight path by 30 points wide). The sounding coordinates for one data granule are shown in Fig. 2, depicting a segment of a descending pass on 15 October 2014, as the AIRS instrument tracked over the eastern edge of Alaska.
Depiction of sounding boresight coordinates for one 6-min data granule. In this example, granule number 125 on 15 Oct 2014 represents this segment of a descending AIRS pass over Alaska between 1229 and 1234 UTC.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Depiction of sounding boresight coordinates for one 6-min data granule. In this example, granule number 125 on 15 Oct 2014 represents this segment of a descending AIRS pass over Alaska between 1229 and 1234 UTC.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Depiction of sounding boresight coordinates for one 6-min data granule. In this example, granule number 125 on 15 Oct 2014 represents this segment of a descending AIRS pass over Alaska between 1229 and 1234 UTC.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
The level 2 AIRS data contain atmospheric temperature values for each of 100 reported standard pressure levels for each sounding location. These pressure levels are the same for every AIRS field of view, making it easy to compare one sounding location with another (Won 2008). The first steps in calculating the wind profiles in this research are to specify a date and time and select a set of coordinates to represent a notional drop zone. If AIRS data are available for this location at a time within 1 h of the specified time, the one or two granules covering the surrounding geographic region are downloaded. Each data file is filtered to select vertical temperature profiles for all of the sounding coordinates within 400 km of the notional drop zone (or evaluation point) coordinates. Of the 100 AIRS pressure levels, only the 25 levels below the 500-hPa level are of interest for the cargo airdrop application, but wind analysis is completed for the entire vertical temperature profile. For each sounding field of view, the dataset includes surface pressure, surface geopotential height, and heights and temperature measurements for up to 100 pressure levels between 0.0161 and 1100 hPa (all 100 temperatures may not be available because of terrain elevation and surface pressure). During level 2 processing, each atmospheric temperature calculated from the satellite-measured radiance is assigned an error value. Figure 3 shows how these AIRS-reported errors are distributed in a sample layer of data within a granule collected on 15 October 2015.
Soundings with AIRS-reported temperature-measurement error better than specified threshold. Shown is the 905-hPa level for AIRS data between 1229 and 1235 UTC 25 Oct 2015.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Soundings with AIRS-reported temperature-measurement error better than specified threshold. Shown is the 905-hPa level for AIRS data between 1229 and 1235 UTC 25 Oct 2015.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Soundings with AIRS-reported temperature-measurement error better than specified threshold. Shown is the 905-hPa level for AIRS data between 1229 and 1235 UTC 25 Oct 2015.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
3. AIRS-retrieved temperature error analysis
Because the technique for deriving winds through the thermal wind relations relies on sometimes small horizontal temperature gradients, and because the full wind profile is constructed by adding each layer’s wind gradients to the previous layer, even small errors in measured temperature can compound and lead to large errors in wind speed and direction. For this reason, additional analysis of the error in the AIRS-retrieved temperatures is appropriate. Based on the following analysis, the error in the AIRS temperatures shows dependence on height, amount of cloud cover, and geographic location (IR sounding background).
A useful source of measured temperature profiles for use in the analysis of AIRS-retrieved temperature profiles is the network of rawinsonde observations (raobs). These weather balloon soundings provide wind data for hundreds of locations each day at 0000 and 1200 UTC. Federal Meteorological Handbook No. 3 specifies that the minimum standard for the accuracy of rawinsonde wind measurements is 3 kt, or 1.542 m s−1 (OFCM 1997). For this research, raob data are obtained using the search tool on the website for the Department of Atmospheric Science at the University of Wyoming (Department of Atmospheric Science 2015). The height dependence in the AIRS-reported error can be seen in Fig. 4, which plots an AIRS temperature profile against the raob temperature profile for the 1200 UTC release at the Anchorage, Alaska, station. The AIRS profile is for the field of view closest to the rawinsonde release coordinates as the satellite passed over Anchorage at 1408 UTC.
Comparison of rawinsonde observation and AIRS temperatures—1408 UTC 15 Oct 2014 at Anchorage.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Comparison of rawinsonde observation and AIRS temperatures—1408 UTC 15 Oct 2014 at Anchorage.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Comparison of rawinsonde observation and AIRS temperatures—1408 UTC 15 Oct 2014 at Anchorage.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
To analyze the error in the AIRS temperatures, four locations in Alaska and Hawaii were selected because of their scheduled 0000 and 1200 UTC rawinsonde times closely matching the AIRS overpass times. For each of these locations, AIRS data from 2003 to 2014 were processed, and when the overpass time for the sounding location nearest to the rawinsonde site was within 1 h of the release, the temperature profiles were compared. These comparisons are shown in Figs. 5 and 6. The difference between the AIRS-retrieved temperatures and the rawinsonde measurements are greatest in the lower atmosphere (below 700 hPa) and in the upper atmosphere (above 20 hPa). In the region between 20 and 700 hPa the root-mean-square (RMS) error is between 1.0 and 2.5 K. for each location.
RMS error between AIRS and rawinsonde-measured temperatures for four locations selected for frequent AIRS overpass near the times of 0000 and 1200 UTC rawinsonde releases. Twelve years of AIRS data (2003–14) were processed, and comparison was made in each instance when the AIRS overpass was within 1 h of the rawinsonde release. This figure includes data for 7519 comparisons for Lihue, HI; 7358 for Hilo, HI; 6792 for Anchorage; and 6414 for Fairbanks, AK.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
RMS error between AIRS and rawinsonde-measured temperatures for four locations selected for frequent AIRS overpass near the times of 0000 and 1200 UTC rawinsonde releases. Twelve years of AIRS data (2003–14) were processed, and comparison was made in each instance when the AIRS overpass was within 1 h of the rawinsonde release. This figure includes data for 7519 comparisons for Lihue, HI; 7358 for Hilo, HI; 6792 for Anchorage; and 6414 for Fairbanks, AK.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
RMS error between AIRS and rawinsonde-measured temperatures for four locations selected for frequent AIRS overpass near the times of 0000 and 1200 UTC rawinsonde releases. Twelve years of AIRS data (2003–14) were processed, and comparison was made in each instance when the AIRS overpass was within 1 h of the rawinsonde release. This figure includes data for 7519 comparisons for Lihue, HI; 7358 for Hilo, HI; 6792 for Anchorage; and 6414 for Fairbanks, AK.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Temperature bias between AIRS and rawinsonde-measured temperatures for four locations selected for frequent AIRS overpass near the times of 0000 and 1200 UTC rawinsonde releases. These comparisons include the same 2003–14 AIRS and rawinsonde data as Fig. 5.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Temperature bias between AIRS and rawinsonde-measured temperatures for four locations selected for frequent AIRS overpass near the times of 0000 and 1200 UTC rawinsonde releases. These comparisons include the same 2003–14 AIRS and rawinsonde data as Fig. 5.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Temperature bias between AIRS and rawinsonde-measured temperatures for four locations selected for frequent AIRS overpass near the times of 0000 and 1200 UTC rawinsonde releases. These comparisons include the same 2003–14 AIRS and rawinsonde data as Fig. 5.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
In the lower atmosphere the satellite-retrieved temperatures were most accurate for locations with low or no cloud cover, as the resolution of the AIRS sounder outperforms the AMSU instrument. A different effect was observed in the upper atmosphere for the two locations in Alaska. For heights above 100 hPa the temperatures for cloud covered locations better matched the rawinsonde measurements. This effect is possibly caused by low cloud cover providing a more uniform background for the IR sounder, preventing any errors caused by nonuniform surface temperatures present when the location is cloud free all the way to the surface. This effect is not present in the Hawaii locations as much of the AIRS field of view is water when the location is cloud free. Further investigation including additional locations and surface properties is required to confirm this effect.
4. AIRS-derived wind profiles
The process used to calculate wind profiles relies on gradients in the heights of pressure levels and temperatures. To determine these gradients and minimize the effects of noise and error in the AIRS data, smooth surfaces are fit to both the AIRS heights and temperatures for each of the 100 pressure levels. The variability in both heights and temperatures from one field of view to the next leads to large changes in local gradients over short distances. Using the local gradients from only the nearest AIRS soundings would lead to unrealistic differences in winds for each grid point, so a method is needed to smooth the isobaric surfaces and temperatures to capture the large-scale trends and not the point-to-point fluctuations that could be caused by instrument noise or error.
a. Smoothing the height and temperature data
In this research, individual temperature measurements with AIRS-reported error exceeding 3 K were removed from the dataset. Locally weighted scatterplot smoothing, with robust weighting, to minimize the effect of outliers, was used to create smooth surfaces from the AIRS temperature retrievals and pressure level heights within 400 km of the target coordinates. This smoothing algorithm is described in the documentation for the MathWorks Curve Fitting Toolbox (MathWorks 2015). From these interpolated surfaces, the slope of each isobaric surface and the horizontal rate of change in temperature at each pressure level can be determined.
Figures 7 and 8 illustrate the result of the smoothing process. Figure 7 shows two cross sections through a sounding data granule. The north–south line of blue diamonds represents a column of sounding locations parallel to the satellite’s flight path, and the east–west red squares are a row of locations perpendicular to the satellite track. The intersection of those two lines is the sounding location closest to the notional drop zone, in this case near Dayton, Ohio.
Sample of sounding locations surrounding drop zone—along satellite flight path (blue diamonds) and perpendicular to satellite flight path (red squares). This granule of AIRS data was collected on 24 Oct 2014, as the Aqua satellite overflew Dayton at 1839 UTC. This horizontal slice through the data shows retrieved temperature at the 802-hPa pressure level (heights between 1.7 and 2.1 km).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Sample of sounding locations surrounding drop zone—along satellite flight path (blue diamonds) and perpendicular to satellite flight path (red squares). This granule of AIRS data was collected on 24 Oct 2014, as the Aqua satellite overflew Dayton at 1839 UTC. This horizontal slice through the data shows retrieved temperature at the 802-hPa pressure level (heights between 1.7 and 2.1 km).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Sample of sounding locations surrounding drop zone—along satellite flight path (blue diamonds) and perpendicular to satellite flight path (red squares). This granule of AIRS data was collected on 24 Oct 2014, as the Aqua satellite overflew Dayton at 1839 UTC. This horizontal slice through the data shows retrieved temperature at the 802-hPa pressure level (heights between 1.7 and 2.1 km).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Satellite-measured temperatures along the satellite flight path (blue diamonds) and perpendicular to the satellite flight path (red squares) for the 802-hPa pressure level from the AIRS data granule shown in Fig. 7. The vertical dotted lines indicate the location of the notional drop zone. The solid symbols represent the temperature measurements within 400 km of the notional drop zone location. The solid black lines show cross sections of the smoothed temperature surface fit to the AIRS data.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Satellite-measured temperatures along the satellite flight path (blue diamonds) and perpendicular to the satellite flight path (red squares) for the 802-hPa pressure level from the AIRS data granule shown in Fig. 7. The vertical dotted lines indicate the location of the notional drop zone. The solid symbols represent the temperature measurements within 400 km of the notional drop zone location. The solid black lines show cross sections of the smoothed temperature surface fit to the AIRS data.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Satellite-measured temperatures along the satellite flight path (blue diamonds) and perpendicular to the satellite flight path (red squares) for the 802-hPa pressure level from the AIRS data granule shown in Fig. 7. The vertical dotted lines indicate the location of the notional drop zone. The solid symbols represent the temperature measurements within 400 km of the notional drop zone location. The solid black lines show cross sections of the smoothed temperature surface fit to the AIRS data.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Figure 8 shows the AIRS temperatures at the 802-hPa level along the two paths shown in Fig. 7, and highlights the need for the smoothing process. Horizontal temperature gradients from one point to the next would have large changes in magnitude, and reversals of direction over short horizontal distances. Error and small-scale variation in the temperatures can lead to many different values for temperature gradients, depending on how many data points are included and how a curve is fit to the data.
When the locally weighted scatterplot smoothing technique is applied to the data within 400 km of the notional drop zone, a three-dimensional surface is created for each isobaric surface and temperature distribution. Cross sections of one of these surfaces are shown in Fig. 8. The slope of this smooth surface at the notional drop zone coordinates is used as the local horizontal temperature gradient, and by a similar surface fitting process, the slope of the isobaric surface is determined. This method of smoothing accounts for more data points than just those immediately surrounding the drop zone coordinates and minimizes the effect of error in the satellite-measured temperatures.
The vertical temperature gradients are also required for the windfinding technique. These gradients are found by dividing the difference in temperature at the coordinates of interest between each pair of pressure surfaces by the vertical separation of the surfaces.
b. Construction of the wind profile using the thermal wind equations
c. AIRS-based windfinding results and analysis
The pressure and temperature gradients at each pressure level above the coordinates of interest were used to construct full vertical wind profiles. Initial steps in this research used rawinsonde-measured winds for comparison and analysis of the accuracy of the AIRS-derived winds, but there are only two bands of longitude values for which the twice-daily AIRS passes (early afternoon and early morning local time) are close to the time of the raobs. These longitude bands are near 160°W and 20°E. When the rawinsonde release and AIRS overpass approximately coincide (within 1 h for this research), direct comparison of the wind profiles can be made, as shown in Fig. 9. Figure 9 also includes the corresponding NWP wind profile, in this case from the GFS 0.5° model. This model is used because it is widely available and could be accessed from an aircraft while in flight. GFS wind profiles do not match the vertical resolution of the raob but typically provide an accurate representation of the trends in wind speed and direction. The GFS winds are also available on a global grid, at 3-h time steps. Measuring wind profile error by comparing with GFS winds rather than raob winds may provide a less accurate assessment of the error (because of differences between the GFS winds and raob-measured winds), but the method allows investigation of locations and times of day for which no raob data are available.
Comparison of AIRS-derived, raob-measured, and GFS-modeled wind profiles for Lihue at 0000 UTC 5 Aug 2014.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Comparison of AIRS-derived, raob-measured, and GFS-modeled wind profiles for Lihue at 0000 UTC 5 Aug 2014.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Comparison of AIRS-derived, raob-measured, and GFS-modeled wind profiles for Lihue at 0000 UTC 5 Aug 2014.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
To characterize the performance of the AIRS-based windfinding process, a grid of 210 locations was selected in the Northern Hemisphere, covering the eastern Pacific Ocean and North America (at 5° latitude and longitude intervals). For each of these sets of coordinates, 100 AIRS soundings were selected for download, one day and one night sounding each week throughout 2014. Corresponding GFS model data for the nearest GFS grid point and closest 3-h interval were retrieved for comparison. The GFS forecast data were obtained from the latest 12-h model run that preceded the AIRS overpass (either the 0000 or 1200 UTC model run for the day of interest). Of the possible 21 000 matches, 18 747 had both AIRS and GFS profiles available. Each of these locations was treated as a notional drop zone and a vertical wind profile was calculated for the time of the AIRS overpass. The geographic coverage of these sample soundings is shown in Fig. 10.
AIRS sounding coordinates selected for GFS wind profile comparison.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
AIRS sounding coordinates selected for GFS wind profile comparison.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
AIRS sounding coordinates selected for GFS wind profile comparison.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
When calculating a wind profile, if the coordinates of interest fall near the edge of the AIRS swath, the temperature and pressure gradients used in the wind derivation technique may be less accurate because of missing neighboring AIRS data points. For this reason, it is expected that the AIRS winds should not match the GFS winds as well when the closest AIRS field of view is near an edge. In this research, error is defined as the difference between the GFS-modeled wind and the AIRS-derived wind profiles. Figure 11 shows the average of all wind speed profile RMS errors plotted by corresponding AIRS field-of-view number. Numbers 1 and 30 represent the left and right edges of the AIRS swath. In this figure, only comparisons between AIRS and GFS wind speed values are presented, but similar results are observed for u and υ wind components individually. The least accurate results occur at the edges, as seen in field of view 30 in Fig. 11, but this effect does not appear to impact accuracy for coordinates more than one field of view from the edge of the AIRS swath.
Wind speed profile error by AIRS field-of-view number. The AIRS scan geometry is a left-to-right, whisk-broom type, with each scan line composed of 30 fields of view, numbered from 1 through 30.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Wind speed profile error by AIRS field-of-view number. The AIRS scan geometry is a left-to-right, whisk-broom type, with each scan line composed of 30 fields of view, numbered from 1 through 30.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Wind speed profile error by AIRS field-of-view number. The AIRS scan geometry is a left-to-right, whisk-broom type, with each scan line composed of 30 fields of view, numbered from 1 through 30.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Time of day and time of year are additional factors that are investigated for their effects on the accuracy of the AIRS-derived wind profiles. Figures 12a and 12b show the average RMS error values for all of the retrieved soundings, separated by month and whether it was the local afternoon sounding (day) or just after local midnight sounding (night). The overland soundings do not indicate any clear dependence on time of day or time of year. The overwater comparisons in Fig. 12b appear to show some seasonal variation in accuracy, evident in both the day and night sounding data.
(a) Average AIRS-derived wind speed RMS error for overland soundings within the geographic bounding box depicted in the inset image, plotted by month. (b) Average AIRS-derived wind speed RMS error for overwater soundings within the geographic bounding box depicted in the inset image, plotted by month.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
(a) Average AIRS-derived wind speed RMS error for overland soundings within the geographic bounding box depicted in the inset image, plotted by month. (b) Average AIRS-derived wind speed RMS error for overwater soundings within the geographic bounding box depicted in the inset image, plotted by month.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
(a) Average AIRS-derived wind speed RMS error for overland soundings within the geographic bounding box depicted in the inset image, plotted by month. (b) Average AIRS-derived wind speed RMS error for overwater soundings within the geographic bounding box depicted in the inset image, plotted by month.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
The most significant factor in how well the AIRS-derived wind profile agrees with the GFS wind appears to be cloud cover. The atmospheric temperatures below clouds that are not able to be retrieved by the IR AIRS instrument are retrieved at reduced vertical resolution by the AMSU radiometer. These AMSU temperatures are used to complete the temperature profiles, and the reduced vertical resolution negatively affects the accuracy of the satellite-derived winds.
The AIRS-reported cloud fraction for the soundings sampled ranged from completely clear (8.1%) to completely overcast (5.9%). Each AIRS field of view is made up of 9 separate IR soundings, and the total effective cloud fraction over all cloud layers for these 9 spots is determined by multiplying the cloud fraction for each field of view by the cloud emissivity. The assumption is made that the clouds are spectrally flat, and for a cloud that is partially transmissive, the equivalent opaque fraction is reported. This allows an effective cloud fraction value for each AIRS field of view to have any value between 0 and 1.0 (Manning 2015).
The portion of the AIRS field of view that is cloud obscured affects wind profile accuracy, with the worst average accuracy occurring when cloud fraction exceeds 0.8. Figure 13 shows the average of all wind speed profile RMS errors within each bin plotted against corresponding cloud-cover percentage, in 5% increments. Figure 14 depicts the spread of these RMS errors throughout the range of cloud-cover values, showing that the cloud-cover effect on wind profile accuracy is small relative to the variability from one sample profile to the next.
Average wind speed profile RMS error by AIRS-reported cloud-cover percentage for the 18 747 sampled AIRS soundings and corresponding GFS data points.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Average wind speed profile RMS error by AIRS-reported cloud-cover percentage for the 18 747 sampled AIRS soundings and corresponding GFS data points.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Average wind speed profile RMS error by AIRS-reported cloud-cover percentage for the 18 747 sampled AIRS soundings and corresponding GFS data points.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Wind speed profile RMS error by AIRS-reported cloud-cover percentage for the 18 747 AIRS soundings sampled (logarithmic vertical scale).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Wind speed profile RMS error by AIRS-reported cloud-cover percentage for the 18 747 AIRS soundings sampled (logarithmic vertical scale).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Wind speed profile RMS error by AIRS-reported cloud-cover percentage for the 18 747 AIRS soundings sampled (logarithmic vertical scale).
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Another important factor affecting the accuracy of the AIRS-derived wind profile is the surface type below the sounding. The AIRS weighting functions for many of the lowest-altitude channels intersect the surface, and the emissive properties of the surface determines the amount of radiance contributed by the surface to that channel.
AIRS-derived wind profiles over land are more affected by cloud cover than those over water because the AMSU data incorporated because of the presence of the clouds are more strongly affected by the varying surface emissivities found over land. Figure 15 shows that even when corrected for cloud cover, the surface emissivity variation leads to reduced accuracy in regions where the sounder background is land. In this figure, only soundings with less than 0.2 cloud fraction are included. The average wind speed RMS error for the ocean-only regions is 8.60 m s−1. When land fills part of the box, the average RMS error is 11.2 m s−1, 30% larger.
Average of AIRS-derived wind speed profile RMS error (m s−1), compared with GFS wind profile for the nearest grid point and time. In this figure, the RMS error for the wind profiles for each of the cloud-free or nearly cloud-free (AIRS cloud fraction across the field of view is less than 0.2) AIRS soundings within each 5° × 5° box.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Average of AIRS-derived wind speed profile RMS error (m s−1), compared with GFS wind profile for the nearest grid point and time. In this figure, the RMS error for the wind profiles for each of the cloud-free or nearly cloud-free (AIRS cloud fraction across the field of view is less than 0.2) AIRS soundings within each 5° × 5° box.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
Average of AIRS-derived wind speed profile RMS error (m s−1), compared with GFS wind profile for the nearest grid point and time. In this figure, the RMS error for the wind profiles for each of the cloud-free or nearly cloud-free (AIRS cloud fraction across the field of view is less than 0.2) AIRS soundings within each 5° × 5° box.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
5. Summary and conclusions
This research investigates the potential application of satellite-derived wind profiles to JPADS airdrop operations, with the intent of increasing safety while maintaining the ability to resupply small drop zones. The analysis performed in this research used data from the AIRS instrument, but the techniques developed could be applied to temperature data collected by other space-based, infrared sounders, such as the Cross-track Infrared Sounder (CrIS) or the Infrared Atmospheric Sounding Interferometer (IASI). This solution would have several operational advantages over the current practice of a first pass over the objective area to release a GPS dropsonde. Deriving winds from satellite sounding data is completely passive, gives no indication of the intended airdrop location, and does not rely on GPS availability. But to eliminate the need for the dropsonde for high-altitude airdrops, the accuracy of the satellite-derived winds needs to be proven sufficient. When compared with rawinsonde-measured winds, the wind profiles obtained often capture trends in wind speed and direction, but the satellite-derived wind profiles are typically not as accurate as winds from modern NWP models.
Because this research uses AMSU atmospheric temperatures to complete temperature profiles below cloud cover, winds are still calculated in regions where the AIRS instrument is unable to retrieve temperatures. The reduced vertical resolution of the AMSU sounding affects the accuracy of the winds. The average RMS error between the satellite-derived wind profiles and GFS modeled winds was found to be 45% greater in regions with complete cloud cover than locations that are AIRS reported cloud free.
When this project was initiated in 2009, satellite data-link capability was not common on board aircraft conducting operational airdrop missions. This meant that the most current NWP model data available to a crew might be 8–12 h old by the time the airdrop was conducted. This fact added to the appeal of a method based on weather satellite data received directly from the satellite and processed on the aircraft in flight. Improvements in NWP model resolution and the fact that data links allow easy access to updated NWP forecasts during flight lead the authors to the conclusion that the wind profiles available from NWP models are more operationally useful than the satellite-derived wind profiles investigated in this research.
This research focused on the 0.5° GFS model output, but there are additional NWP models that could be used as the source of the required wind update. The GFS model is now running operationally on a 0.25° grid, and future research could show whether the increased horizontal resolution leads to an improvement in the accuracy of the wind profiles. Another option is the European Centre for Medium-Range Weather Forecasts. This model has the advantages of being run globally on a 0.125° grid, with 91 vertical levels. A limitation for both research and operational applications, however, is access to the output data. The current forecast data are not made freely available (as are the GFS data), limiting the number of users who can access it.
While not found to be as accurate as short-term NWP for the original airdrop application, this technique can provide wind information in regions of the atmosphere not currently covered by the GFS model. The AIRS instrument retrieves temperatures up to 75–80 km (standard pressure level 0.0161 hPa), so derived wind profiles extend throughout the stratosphere and lower mesosphere, above the highest data available from many current NWP models. An example of this is seen in Fig. 16, which shows the full AIRS-derived u and υ wind profiles in comparison with wind components measured by SuperLoki/PWN-12A rockets deploying Rocket Balloon Instrument (ROBIN) Inflatable Falling Spheres. While not matching the vertical resolution of the rocketsonde data in the stratosphere, the AIRS windfinding technique captures the wind component trends, providing some insight into the strength and direction of the wind at heights well above those modeled by GFS and other easily accessed NWP models (Cole 1978).
A comparison of the u and υ wind component profiles derived from AIRS temperature data and measured by rocketsondes launched from the Pacific Missile Range Facility (PMRF) on 8 Jun 2015. The rocketsonde data were obtained through personal communications with M. Bedrick, Air Force Life Cycle Management Center (AFLCMC)/Staff Meteorology Branch (XZIG), Wright-Patterson Air Force Base and Dr. C. O’Farrell, NASA JPL.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
A comparison of the u and υ wind component profiles derived from AIRS temperature data and measured by rocketsondes launched from the Pacific Missile Range Facility (PMRF) on 8 Jun 2015. The rocketsonde data were obtained through personal communications with M. Bedrick, Air Force Life Cycle Management Center (AFLCMC)/Staff Meteorology Branch (XZIG), Wright-Patterson Air Force Base and Dr. C. O’Farrell, NASA JPL.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
A comparison of the u and υ wind component profiles derived from AIRS temperature data and measured by rocketsondes launched from the Pacific Missile Range Facility (PMRF) on 8 Jun 2015. The rocketsonde data were obtained through personal communications with M. Bedrick, Air Force Life Cycle Management Center (AFLCMC)/Staff Meteorology Branch (XZIG), Wright-Patterson Air Force Base and Dr. C. O’Farrell, NASA JPL.
Citation: Journal of Applied Meteorology and Climatology 55, 10; 10.1175/JAMC-D-15-0296.1
The global coverage and wide swaths of 3D temperature data provided by the AIRS instrument, and the close agreement between the satellite-retrieved temperatures and rawinsonde measurements, point to another possible application for these data. Optical turbulence affects laser propagation through the atmosphere and distorts the images obtained by sensors. The strength of this small-scale turbulence, as modeled by the refractive index structure function Cn2, is a quantity of interest for long-range sensor applications and ballistic missile defense. The Cn2 can be related to the temperature structure function, and while it remains to be shown, the AIRS instrument has the potential to provide the volume temperature measurements required to calculate path-weighted Cn2 values for long paths through the atmosphere.
Acknowledgments
This research is based on the first author’s 2010 Air Force Institute of Technology master’s thesis (available at www.dtic.mil/cgi-bin/GetTRDoc?AD=ADA516487). Additionally, we thank the collaborative support of MZA Associates and the Missile Defense Agency through two Small Business Innovation Research Projects (SBIRs). Author Steven Fiorino was supported by Missile Defense Agency SBIR Topic MDA12-017. The views expressed in this paper are those of the authors and do not necessarily reflect the official policy or position of the Air Force, the Department of Defense, or the U.S. government.
REFERENCES
AIRS Science Team/Joao Texeira, 2013: Aqua AIRS level 2 support retrieval (AIRS+AMSU), version 006. NASA GES DISC, accessed 1 January 2014 to 15 November 2015, doi:10.5067/AQUA/AIRS/DATA207.
Bluestein, H. B., 1992: Principles of Kinematics and Dynamics. Vol. 1, Synoptic-Dynamic Meteorology in Midlatitudes, Oxford University Press, 448 pp.
Cole, A. E., 1978: Review of data and models of the middle atmosphere. Proc. Open Meetings of the Working Groups on Physical Sciences of the 21st Plenary Meeting of COSPAR, Innsbruck, Austria, COSPAR, 153–164.
Department of Atmospheric Science, 2015: Worldwide radiosonde soundings of the atmosphere. College of Engineering, University of Wyoming, accessed 1 January 2014 to 15 November 2015. [Available online at http://weather.uwyo.edu/upperair/sounding.html.]
GES DISC, 2016: Atmospheric Infrared Sounder (AIRS) instrument guide. NASA, accessed 16 February 2016. [Available online at http://disc.sci.gsfc.nasa.gov/AIRS/documentation/airs_instrument_guide.shtml#2.]
JPL, 2014: AIRS, Atmospheric Infrared Sounder. NASA, accessed 10 December 2015. [Available online at http://airs.jpl.nasa.gov/mission_and_instrument/instrument/airs-instrument-detail.]
Manning, E. M., 2015: AIRS version 6.1.1 released processing files description. JPL–California Institute of Technology, 280 pp. [Available online at http://disc.sci.gsfc.nasa.gov/AIRS/documentation/v6_docs/v6releasedocs-1/V6_Released_Processing_File_Description.pdf.]
MathWorks, 2015: Curve Fitting Toolbox user’s guide, release R2015a. The MathWorks Inc., 754 pp. [Available online at http://www.mathworks.com/help/pdf_doc/curvefit/curvefit.pdf.]
Meier, D. C., 2010: Application of satellite-derived wind profiles to Joint Precision Airdrop Systems (JPADS) operations. M.S. thesis, Dept. of Engineering Physics, Air Force Institute of Technology, Wright-Patterson Air Force Base, 96 pp. [Available online from www.dtic.mil/cgi-bin/GetTRDoc?AD5ADA516487.]
OFCM, 1997: Rawinsonde and pibal observations. Federal Meteorological Handbook, No. 3, FCMH3-1997. Accessed 14 November 2015. [Available online at http://www.ofcm.gov/fmh3/fmh3.htm.]
Wallace, J. M., and P. V. Hobbs, 2006: Atmospheric Science: An Introductory Survey. Academic Press, 504 pp.
Won, Y. I., 2008: README document for AIRS level-2 version 5 support products. NASA GES DISC, 57 pp. [Available online at http://disc.sci.gsfc.nasa.gov/AIRS/documentation/readmes/README.AIR-2SUP.pdf.]
