• Bedka, K. M., , and J. R. Mecikalski, 2005: Application of satellite-derived atmospheric motion vectors for estimating mesoscale flows. J. Appl. Meteor., 44, 17611772.

    • Search Google Scholar
    • Export Citation
  • Bedka, K. M., , C. S. Velden, , R. A. Petersen, , W. F. Feltz, , and J. R. Mecikalski, 2009: Comparisons of satellite-derived atmospheric motion vectors, rawinsondes, and NOAA Wind Profiler observations. J. Appl. Meteor. Climatol., 48, 15421561.

    • Search Google Scholar
    • Export Citation
  • Berger, H., , R. H. Langland, , C. S. Velden, , C. A. Reynolds, , and P. M. Pauley, 2011: Impact of enhanced satellite-derived atmospheric motion vector observations on numerical tropical cyclone track forecasts in the western North Pacific during TPARC/TCS-08. J. Appl. Meteor. Climatol., 50, 23092318.

    • Search Google Scholar
    • Export Citation
  • Dunion, J. P., , and C. S. Velden, 2002: Application of surface-adjusted GOES low-level cloud-drift winds in the environment of Atlantic tropical cyclones. Part I: Methodology and validation. Mon. Wea. Rev., 130, 13331346.

    • Search Google Scholar
    • Export Citation
  • Franklin, J. L., , J. Kaplan, , C. S. Velden, , and C. M. Hayden, 1990: Some comparisons of VAS and dropwindsonde data over the subtropical Atlantic. Mon. Wea. Rev., 118, 18691887.

    • Search Google Scholar
    • Export Citation
  • Goerss, J. S., , C. S. Velden, , and J. D. Hawkins, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part II: NOGAPS forecasts. Mon. Wea. Rev., 126, 12191227.

    • Search Google Scholar
    • Export Citation
  • Hock, T. F., , and J. L. Franklin, 1999: The NCAR GPS dropwindsonde. Bull. Amer. Meteor. Soc., 80, 407420.

  • Holmlund, K., , C. S. Velden, , and M. Rohn, 2001: Enhanced automated quality control applied to high-density satellite-derived winds. Mon. Wea. Rev., 129, 517529.

    • Search Google Scholar
    • Export Citation
  • Kelly, G., , T. McNally, , J.-N. Thepaut, , and M. Szyndel, 2004: OSEs of all main data types in the ECMWF operation system. Proc. Third WMO Workshop on the Impact of Various Observing Systems on Numerical Weather Prediction, Alpbach, Austria, WMO World Weather Watch (Tech. Doc. WMO/TD 1228), 63–94. [Available online at http://www.wmo.int/pages/prog/www/GOS/Alpbach2004/Proceedings.pdf.]

    • Search Google Scholar
    • Export Citation
  • Key, J. R., , D. A. Santek, , C. S. Velden, , N. Bormann, , J.-N. Thepaut, , L. P. Riishojgaard, , Y. Zhu, , and W. P. Menzel, 2003: Cloud-drift and water vapor winds in the Polar regions from MODIS. IEEE Trans. Geosci. Remote Sens., 41, 482492.

    • Search Google Scholar
    • Export Citation
  • Kiladis, G. N., , C. D. Thorncroft, , and N. M. J. Hall, 2006: Three-dimensional structure and dynamics of African easterly waves. Part I: Observations. J. Atmos. Sci., 63, 22122230.

    • Search Google Scholar
    • Export Citation
  • Langland, R. H., , C. Velden, , P. M. Pauley, , and H. Berger, 2009: Impact of satellite-derived rapid-scan wind observations on numerical model forecasts of Hurricane Katrina. Mon. Wea. Rev., 137, 16151622.

    • Search Google Scholar
    • Export Citation
  • Leslie, L. M., , J. F. LeMarshall, , R. P. Morison, , C. Spinoso, , R. J. Purser, , N. Pescod, , and R. Seecamp, 1998: Improved hurricane track forecasting from the continuous assimilation of high quality satellite wind data. Mon. Wea. Rev., 126, 12481258.

    • Search Google Scholar
    • Export Citation
  • Menzel, W. P., 2001: Cloud tracking with satellite imagery: From the pioneering work of Ted Fujita to the present. Bull. Amer. Meteor. Soc., 82, 3347.

    • Search Google Scholar
    • Export Citation
  • Merrill, R. T., , and C. S. Velden, 1996: A three-dimensional analysis of the outflow layer of Supertyphoon Flo (1990). Mon. Wea. Rev., 124, 4763.

    • Search Google Scholar
    • Export Citation
  • Montgomery, M. T., and Coauthors, 2012: The Pre-Depression Investigation of Cloud-Systems in the Tropics (PREDICT) experiment: Scientific basis, new analysis tools, and some first results. Bull. Amer. Meteor. Soc., 93, 153172.

    • Search Google Scholar
    • Export Citation
  • Poteat, K. O., 1973: A comparison of satellite-derived, low-level and cirrus-level winds with conventional wind observations. J. Appl. Meteor., 12, 14161419.

    • Search Google Scholar
    • Export Citation
  • Pu, Z., , X. Li, , C. S. Velden, , S. D. Aberson, , and W. T. Liu, 2008: The impact of aircraft dropsonde and satellite wind data on numerical simulations of two landfalling tropical storms during the Tropical Cloud Systems and Processes experiment. Wea. Forecasting, 23, 6279.

    • Search Google Scholar
    • Export Citation
  • Rao, P. A., , C. S. Velden, , and S. A. Braun, 2002: The vertical error characteristics of GOES-derived winds: Description and experiments with numerical weather prediction. J. Appl. Meteor., 41, 253271.

    • Search Google Scholar
    • Export Citation
  • Rodgers, E., , and R. C. Gentry, 1983: Monitoring tropical-cyclone intensity using environmental wind fields derived from short-interval satellite images. Mon. Wea. Rev., 111, 979996.

    • Search Google Scholar
    • Export Citation
  • Sears, J., 2011: Investigating the role of the upper-levels in tropical cyclogenesis. M.S. thesis, Dept. of Oceanographic and Atmospheric Science, University of Wisconsin—Madison, 96 pp.

  • Soden, B. J., , C. S. Velden, , and R. E. Tuleya, 2001: The impact of satellite winds on experimental GFDL hurricane model forecasts. Mon. Wea. Rev., 129, 835852.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , and K. M. Bedka, 2009: Identifying the uncertainty in determining satellite-derived atmospheric motion vector height attribution. J. Appl. Meteor. Climatol., 48, 450463.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , C. M. Hayden, , W. P. Menzel, , J. L. Franklin, , and J. S. Lynch, 1992: The impact of satellite-derived winds on numerical hurricane track forecasting. Wea. Forecasting, 7, 107118.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , C. M. Hayden, , S. J. Nieman, , W. P. Menzel, , S. Wanzong, , and J. S. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc., 78, 173195.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , T. Olander, , and S. Wanzong, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part I: Dataset methodology, description, and case analysis. Mon. Wea. Rev., 126, 12021218.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., and Coauthors, 2005: Recent innovations in deriving tropospheric winds from meteorological satellites. Bull. Amer. Meteor. Soc., 86, 205223.

    • Search Google Scholar
    • Export Citation
  • Viezee, W., , R. M. Endlich, , and S. M. Serebreny, 1967: Satellite-viewed jet stream clouds in relation to the observed wind field. J. Appl. Meteor., 6, 929935.

    • Search Google Scholar
    • Export Citation
  • Xiao, Q., , X. Zou, , M. Pondeca, , M. A. Shapiro, , and C. Velden, 2002: Impact of GMS-5 and GOES-9 satellite-derived winds on the prediction of a NORPEX extratropical cyclone. Mon. Wea. Rev., 130, 507528.

    • Search Google Scholar
    • Export Citation
  • Zapotocny, T. H., , J. A. Jung, , J. F. Le Marshall, , and R. E. Treadon, 2007: A two-season impact study of satellite and in situ data in the NCEP Global Data Assimilation System. Wea. Forecasting, 22, 887909.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Composite 2D histogram of 925-hPa dropsonde locations for all 26 flights binned into 1° latitude and longitude grid boxes relative to analyzed mission (disturbance) center. Warm colors denote a higher concentration of observations per grid box.

  • View in gallery

    Example of upper-tropospheric AMV coverage derived every hour during PREDICT from GOES–Meteosat imagery. The AMVs (kt; 1 kt ≃ 0.5 m s−1) are color coded by their height assignments. The inset shows a zoom in over Hurricane Earl.

  • View in gallery

    Vertical distribution of AMVs with vector heights binned every 50 hPa for all AMVs in this tropical validation sample within 1° of coincident dropsondes.

  • View in gallery

    Differences between AMVs and nearly collocated/coincident dropsondes for (left) wind speed (m s−1) and (right) direction (°). Results from two match distance criteria (radii) are shown.

  • View in gallery

    GOES IR image with AMV vectors assigned between 350 and 450 (yellow), and dropsonde vectors from 350 to 450 (cyan) including one vector for 300 hPa (magenta) and one for 500 hPa (red). Green circles show match radii. Example from TD Fiona at 1215 UTC 9 Sep 2010.

  • View in gallery

    Vertical profile of mean AMV QI values for all PREDICT sample AMVs (N = 6733), sorted into 50-hPa bins.

  • View in gallery

    Mean RMSE and bias for AMVs (in comparison with collocated dropsondes) sorted by aircraft missions during PREDICT.

  • View in gallery

    Distribution of match counts for interpolated gridded AMV wind values vs coincident dropsonde winds at standard pressure levels.

  • View in gallery

    Vertical profile of (left) root-mean-square error (m s−1) and (right) bias (m s−1) of interpolated gridded AMV wind analyses vs collocated dropsonde winds at standard levels.

  • View in gallery

    As in Fig. 7, but for AMV gridded wind analyses (in comparison with collocated dropsonde wind profiles) broken down by missions during PREDICT.

  • View in gallery

    Distribution of AMV and associated interpolated gridded analysis match counts by height (pressure) for the homogeneous intercomparison dataset.

  • View in gallery

    Comparison of vertical profiles at standard pressure levels for (a) wind speed root-mean-square error (m s−1), (b) speed bias (m s−1), and (c) wind direction (°) between collocated AMVs, interpolated gridded AMV wind analyses, and interpolated GFS wind analyses vs collocated dropsonde winds. The numbers in (c) indicate match counts at each standard pressure level.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 36 36 5
PDF Downloads 28 28 2

Validation of Satellite-Derived Atmospheric Motion Vectors and Analyses around Tropical Disturbances

View More View Less
  • 1 Cooperative Institute for Meteorological Satellite Studies, and Space Science and Engineering Center, University of Wisconsin—Madison, Madison, Wisconsin
© Get Permissions
Full access

Abstract

Fields of atmospheric motion vectors (AMVs) are routinely derived by tracking features in sequential geostationary satellite infrared, water vapor, and visible-channel imagery. While AMVs produced operationally by global data centers are routinely evaluated against rawinsondes, there is a relative dearth of validation opportunities over the tropical oceans—in particular, in the vicinity of tropical disturbances when anomalous flow fields and strongly sheared environments commonly exist. A field experiment in 2010 called Pre-Depression Investigation of Cloud-Systems in the Tropics (PREDICT) was conducted in the tropical west Atlantic Ocean and provides an opportunity to evaluate the quality of tropical AMVs and analyses derived from them. The importance of such a verification is threefold: 1) AMVs often provide the only input data for numerical weather prediction (NWP) over cloudy areas of the tropical oceans, 2) NWP data assimilation methods are increasingly reliant on accurate flow-dependent observation-error characteristics, and 3) global tropical analysis and forecast centers often rely on analyses and diagnostic products derived from the AMV fields. In this paper, the authors utilize dropsonde information from high-flying PREDICT aircraft to identify AMV characteristics and to better understand their errors in tropical-disturbance situations. It is found that, in general, the AMV observation errors are close to those identified in global validation studies. However, some distinct characteristics are uncovered in certain regimes associated with tropical disturbances. High-resolution analyses derived from the AMV fields are also examined and are found to be more reflective of anomalous flow fields than the respective Global Forecast System global model analyses.

Corresponding author address: John Sears, Space Science and Engineering Center, 1225 West Dayton St., Madison, WI 53706. E-mail: john.sears@ssec.wisc.edu

Abstract

Fields of atmospheric motion vectors (AMVs) are routinely derived by tracking features in sequential geostationary satellite infrared, water vapor, and visible-channel imagery. While AMVs produced operationally by global data centers are routinely evaluated against rawinsondes, there is a relative dearth of validation opportunities over the tropical oceans—in particular, in the vicinity of tropical disturbances when anomalous flow fields and strongly sheared environments commonly exist. A field experiment in 2010 called Pre-Depression Investigation of Cloud-Systems in the Tropics (PREDICT) was conducted in the tropical west Atlantic Ocean and provides an opportunity to evaluate the quality of tropical AMVs and analyses derived from them. The importance of such a verification is threefold: 1) AMVs often provide the only input data for numerical weather prediction (NWP) over cloudy areas of the tropical oceans, 2) NWP data assimilation methods are increasingly reliant on accurate flow-dependent observation-error characteristics, and 3) global tropical analysis and forecast centers often rely on analyses and diagnostic products derived from the AMV fields. In this paper, the authors utilize dropsonde information from high-flying PREDICT aircraft to identify AMV characteristics and to better understand their errors in tropical-disturbance situations. It is found that, in general, the AMV observation errors are close to those identified in global validation studies. However, some distinct characteristics are uncovered in certain regimes associated with tropical disturbances. High-resolution analyses derived from the AMV fields are also examined and are found to be more reflective of anomalous flow fields than the respective Global Forecast System global model analyses.

Corresponding author address: John Sears, Space Science and Engineering Center, 1225 West Dayton St., Madison, WI 53706. E-mail: john.sears@ssec.wisc.edu

1. Introduction

The evolving ability to use satellite imagery to extract quantitative wind information has been of tremendous value to meteorological analyses. Techniques have advanced in sophistication and accuracy from the manually derived methods in the mid-1960s (Viezee et al. 1967) to modern automated processing and assimilation techniques (Menzel 2001; Velden et al. 2005). Cloud and water vapor feature tracking via sequential infrared (IR), water vapor (WV), and visible (VIS) imagery produces atmospheric motion vectors (AMVs), which have been shown to provide reasonable estimates of the ambient tropospheric wind (Franklin et al. 1990; Poteat 1973). The applications of AMVs are wide ranging, with analytic value to forecasters, and as information content for numerical models.

One significant advantage of satellite-derived AMVs is the potential for good coverage in data-sparse regions, particularly over the oceans. When polar AMVs are included (Key et al. 2003), they provide more complete global coverage that can improve model forecasts (Kelly et al. 2004; Velden et al. 2005). AMVs can add accuracy to model analyses such as the development of extratropical cyclones (Xiao et al. 2002), and have also been used to help depict mesoscale features (Bedka and Mecikalski 2005).

AMVs have particular value in the data-sparse tropics for observing both upper-tropospheric features such as tropical cyclone (TC) outflow mass flux and relative angular momentum (Merrill and Velden 1996; Rodgers and Gentry 1983) and lower-tropospheric features such as transient African easterly waves (Kiladis et al. 2006). Improvements in numerical model TC track and intensity forecasts from the assimilation of AMVs show the positive influence they can have both on individual case studies and in bulk seasonal statistics (Soden et al. 2001; Velden et al. 1992, 1997; Leslie et al. 1998; Goerss et al. 1998; Zapotocny et al. 2007; Pu et al. 2008; Langland et al. 2009; Berger et al. 2011; Rao et al. 2002).

Tropical analyses derived from the AMVs have been produced by the Cooperative Institute for Meteorological Satellite Studies (CIMSS) for over a decade, and made available to users through the CIMSS Tropical Cyclones website: http://tropic.ssec.wisc.edu/. Diagnostic fields such as vertical wind shear produced from these analyses are also displayed. Since these products are often relied upon by both the forecasting and research communities, it is important to assess their accuracy along with the input AMVs themselves.

Recent validation studies by Velden and Bedka (2009) and Bedka et al. (2009) illuminate some of the characteristics and shortcomings of AMVs—in particular, the difficulty in assigning precise heights to the vectors for certain feature types being tracked. Reliance is often placed on effective automated quality control of the raw AMV information to filter out the poor vectors (Velden et al. 1998; Holmlund et al. 2001).

Validating AMVs in tropical disturbances is a more difficult task given that in situ full tropospheric wind profiles are usually scarce. Dunion and Velden (2002) looked at the quality of AMVs near Atlantic Ocean tropical cyclones derived from the Geostationary Operational Environmental Satellite (GOES)-East, but were limited to mainly the assessment of vectors in the lower half of the troposphere against low-flying reconnaissance aircraft dropsonde observations. The addition of the Gulfstream-IV jet to the National Oceanic and Atmospheric Administration (NOAA) reconnaissance fleet in the late 1990s now provides for higher-altitude surveillance. However, this aircraft is normally tasked to sample the synoptic environment (outside the convective circulation area) of target TCs, and normally does not venture into the center of circulation. In 2010, an Atlantic basin tropical field program sponsored by the National Science Foundation (NSF), called Pre-Depression Investigation of Cloud-Systems in the Tropics (PREDICT; Montgomery et al. 2012), provided a rare opportunity to validate upper-level AMVs in and around developing and weak TC circulations. Using in situ data collected from the NSF–National Center for Atmospheric Research (NCAR) Gulfstream-V during PREDICT missions, and a special effort to process high-resolution AMV datasets by CIMSS, the authors investigate the quality of the AMVs as well as the resultant gridded analyses derived from them.

2. Data and methodology

The PREDICT Gulfstream-V (GV) flights were primarily focused on Atlantic tropical disturbances resulting from African easterly waves (AEWs) in order to ascertain information about the genesis phase of TCs resulting from these types of tropical disturbances. Twenty-six GV missions were conducted; a summary is presented in Table 1. These 26 flights involved 8 different tropical systems from predepression to tropical storm (TS) intensity observed in either the western Atlantic or Caribbean Sea regions (some of the storms went on to exceed TS intensity after the period of observation). Other aircraft with dropsonde capabilities associated with companion field campaigns [e.g., Genesis and Rapid Intensification Processes (GRIP) and Intensify Forecast Experiment (IFEX)] were flying during phases of PREDICT. However, these aircraft either did not fly high enough [e.g., NOAA P3s and National Aeronautics and Space Administration (NASA) DC-8] or did not penetrate into the central circulation area (NOAA G-IV), thus limiting their utility to a particular focus on high-altitude, disturbance-centered AMVs.

Table 1.

NSF–NCAR GV mission–flight summary during PREDICT, August–September 2010. Under mission designation, the PGI number stands for field projects PREDICT, GRIP, and IFEX and is numbered on the basis of their chronological joint mission identification.

Table 1.

From the 26 GV flights, a total of 558 dropsondes were deployed and used in this study for the AMV validation. A 26-flight composite 2D histogram of the 925-hPa dropsonde locations binned in 1° × 1° boxes (Fig. 1) shows the coverage of observations relative to the mission target (disturbance) centers. The highest concentration of dropsonde reports occurs near the mission target center (usually a closed surface circulation in a Lagrangian framework), where flight operations were normally focused. It can be seen that the majority of the dropsonde observations occur within 5° of the target system center, yielding excellent coverage of the disturbance core in most cases.

Fig. 1.
Fig. 1.

Composite 2D histogram of 925-hPa dropsonde locations for all 26 flights binned into 1° latitude and longitude grid boxes relative to analyzed mission (disturbance) center. Warm colors denote a higher concentration of observations per grid box.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

It is also important to note that the majority of the dropsondes began their descent around 150 hPa, which in most cases was near or slightly above the tropopause level. This is critical for the validation since there is generally a strong concentration of AMVs at the cirrus level (usually found at or just below the tropopause for weak tropical disturbances).

The GV dropsonde observations are treated as “ground truth” in this study, neglecting the small instrument noise and representativeness errors. The dataset underwent a rigorous postexperiment quality-control (QC) process by the NCAR Earth Observing Laboratory (EOL), including automated and manual procedures. In addition to this, further QC of the wind reports was applied by the authors: the first value observed by the dropsonde after release from the aircraft was discarded because of suspect observations in a number of cases. The dropsonde data vertical resolution is dependent on drop rate, based on 0.5-s sampling. The geolocations during descent are derived by global positioning system.

The AMVs to be validated were derived from GOES as a special product for the PREDICT field experiment by CIMSS. The processing algorithm and derivation approach are very similar to the operational algorithm and approach used by NOAA/National Environmental Satellite, Data, and Information Service (NESDIS). This includes the use of rapid-scan imagery for cloud tracking when it was available from GOES and the use of the Global Forecast System (GFS) global model for input to AMV height assignment and postprocessing routines. Therefore, the AMVs produced by CIMSS are representative of the operational GOES product with a couple of exceptions: 1) the datasets were produced hourly (operational datasets are 3 hourly), and 2) in an attempt to enhance the upper-level vector coverage in order to obtain a better depiction of the disturbance-associated mesoscale flow fields, the visible channel was used during daylight hours to obtain vectors throughout the full depth of the troposphere. In operational processing, only low-level VIS vectors (>600 hPa) are attempted. The ramifications of these deviations are discussed in the next section. Together, they allow maximum AMV coverage of tropical systems; for example, see Fig. 2. For further details on the AMV processing, see Velden et al. (1998).

Fig. 2.
Fig. 2.

Example of upper-tropospheric AMV coverage derived every hour during PREDICT from GOES–Meteosat imagery. The AMVs (kt; 1 kt ≃ 0.5 m s−1) are color coded by their height assignments. The inset shows a zoom in over Hurricane Earl.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

The validation methodology relies on collocated “matches” of AMVs and dropsonde winds in three-dimensional space and time. The matching criteria demand the following: 1) observation time tags within ±30 min, 2) AMV within a 1° or 0.5° radius of the latitude and longitude of the dropsonde (both radii are evaluated) based on the latitude and longitude of the dropsonde GPS at each pressure level, and 3) AMV-assigned height is matched to closest dropsonde wind value within ±5 hPa. Multiple AMVs can be compared with a given dropsonde observation upon meeting the above criteria.

As part of the AMV postprocessing algorithm, gridded wind analyses (latitude–longitude at 1.0° spatial resolution) are produced on traditional isobaric surfaces (150, 200, 250, 300, 400, 500, 600, 700, 775, 850, and 925 hPa) using a three-dimensional recursive filter objective analysis that weights the AMVs heavily in the assimilation process (Velden et al. 1998, 2005). The operational GFS model is used to provide background fields for the analyses. To validate the accuracy of these analyses, the gridded values are interpolated to the matching dropsonde location, with the same vertical and temporal match requirements as with the AMVs described above.

3. Results

a. Validation of the AMVs

As mentioned above, AMVs are a crucial observational element of tropical ocean analyses, including those produced by CIMSS for the tropical cyclone community. Before evaluating the gridded analysis products, the authors assess the quality of the tropical AMV observations as a first step. It is informative to look at the vertical distribution of the AMVs in this tropical validation sample. Figure 3 shows a histogram of observed AMVs within 1° of coincident dropsondes, and indicates a clear preference for upper-level observations, which is expected given the abundance of cirrus clouds typical in the tropical North Atlantic summer. AMV counts above 200 hPa are somewhat limited by the dropsonde observations, which generally start near or just below 150 hPa.

Fig. 3.
Fig. 3.

Vertical distribution of AMVs with vector heights binned every 50 hPa for all AMVs in this tropical validation sample within 1° of coincident dropsondes.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

Sparse AMV coverage below 300 hPa can be attributed to the nature of cloud patterns in the tropical systems being observed. Deep and high clouds from convection and resulting cirrus canopies normally prevail in developing tropical disturbances, and often prohibit the retrieval of AMVs below. A weaker secondary maximum in AMV concentration below 700 hPa exists primarily from the sampling of shallow convection tracers that can be observed outside (surrounding) the disturbance’s convective core. This lower-tropospheric maximum is also likely limited in amplitude owing to dropsonde coverage being focused on the disturbance core.

Figure 4 shows the distribution of differences between AMVs and near collocated–coincident dropsonde wind values for both speed and direction. The term “error” is not used here since the authors do not want to imply that these differences are all due to AMV observational error. Some of the discrepancies may be due to dropsonde wind measurement error (0.5–2.0 m s−1; Hock and Franklin 1999), and some due to imperfect AMV–dropsonde matching criteria (Velden and Bedka 2009). This is illustrated in Fig. 4 by presenting two spatial match distance (radii) criteria: <1.0° and <0.5°. It is readily evident that a number of the large difference outliers are eliminated by using the 0.5° matching criterion. However, this also drastically reduces the match sample from 6733 to 1757.

Fig. 4.
Fig. 4.

Differences between AMVs and nearly collocated/coincident dropsondes for (left) wind speed (m s−1) and (right) direction (°). Results from two match distance criteria (radii) are shown.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

Tables 2 and 3 show the comparison statistics for both match radii. The lower-level (<700 hPa) RMSE and bias values are comparable to those found in Dunion and Velden (2002). Interestingly, the low directional deviations for this group indicate the coherent nature of the shallow cumulus clouds as tracking elements. However, the larger negative speed bias (AMVs too slow) might suggest the AMV height assignment methods are not optimal in TC boundary layer environments.

Table 2.

Statistical comparison of AMVs and near collocated–coincident dropsonde winds for the match radius criterion of <1.0°. Bias is determined as AMV minus dropsonde.

Table 2.
Table 3.

As in Table 2, but for the match criterion of <0.5°.

Table 3.

For the most part, the RMSE statistics for the upper-level AMVs are better than bulk global values versus collocated rawinsondes posted by operational AMV data producers. This is likely due to generally lower upper-tropospheric wind speeds and more coherent cloud tracers in the tropics, both of which promote smaller tracking errors relative to conditions more common in midlatitude jet streams. The increased spread in the difference values at upper levels (Fig. 4) is likely due to the divergent nature of the developing outflow patterns (Sears 2011) and the associated strong vertical shear gradients that can amplify small AMV height assignment errors.

Figure 4 also shows that the greatest uncertainty in terms of outliers exists in the midlevels (between 700 and 300 hPa). This region is particularly vulnerable to erroneous height assignments owing to difficulty in differentiating between optically thin cirrus and other midlevel clouds with similar brightness temperature values. If thin cirrus tracers fail semitransparency checks in the AMV processing stream, they can be assigned to levels that are far too low in the troposphere. Despite quality-control procedures in the postprocessing of the AMVs, these vectors can occasionally get through to the final distributed datasets.

A good example of this effect is shown in Fig. 5, where selected AMVs are plotted with collocated dropsonde wind profiles during a mission in Tropical Depression (TD) Fiona. A number of AMVs on the west side cirrus outflow are height assigned to the 350–450-hPa layer, but more closely resemble the 300-hPa dropsonde winds. This highlights the difficulty in tracking very thin cirrus tracers. Further analysis reveals that the majority of these misassigned vectors were derived through tracking features in visible imagery. Since the GOES visible imagery has 1-km spatial resolution, it can often pick up traceable thin cirrus features that the 4-km IR misses. This could be important for picking up mesoscale flow characteristics in the storm canopy and accurately calculating resultant diagnostic fields. Such was the reasoning for the full processing of the VIS AMVs during PREDICT by CIMSS [the normal operational procedure at NESDIS for large-scale coverage only uses the VIS channels for low-level (<600 hPa) cloud tracking, and relies on the IR and WV for upper-level AMVs]. But as pointed out here, problems can arise from the fact that the VIS tracers must still use the IR for their height assignment. The 4-km IR scene often cannot resolve the thin cirrus tracer depicted in the VIS, and therefore IR-based semitransparency cloud corrections for the height assignment fail, allowing the vector to be assigned too low in the troposphere. In utilizing the higher spatial resolution of visible imagery to track features, it necessitates the use of less certain height assignments and explains some of the large errors present in the middle troposphere for both speed and direction (Fig. 4). This naturally leads to the conflict of AMV quantities versus accuracy, although VIS AMVs with height assignments in the upper and lower troposphere have error profiles comparable to their IR and WV counterparts. Therefore, it is believed that the much-improved AMV coverage in the storm cloud canopy using VIS tracking with correctly assigned heights far outweighs the few middle-troposphere outliers that may also result.

Fig. 5.
Fig. 5.

GOES IR image with AMV vectors assigned between 350 and 450 (yellow), and dropsonde vectors from 350 to 450 (cyan) including one vector for 300 hPa (magenta) and one for 500 hPa (red). Green circles show match radii. Example from TD Fiona at 1215 UTC 9 Sep 2010.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

To further address this issue, the AMV postprocessing routine outputs an estimated measure of the quality of each vector. A quality-indicator (QI) value is assigned, which indicates the level of confidence in the precision of each vector’s coherence and representativeness of the ambient flow. The QI values range from 0 to 1, where 1 represents the greatest confidence. AMV QI values must meet a minimum 0.5 in order to be retained in the final dataset and available for the CIMSS gridded analyses (no values below 0.5 are assessed in this study). A vertical profile of mean QIs from the PREDICT AMV dataset (Fig. 6) supports the previous evidence, with lower QI (confidence) values for the middle troposphere AMVs between 350 and 700 hPa. These QIs can be used to effectively thin out bad AMVs before assimilation into objective analyses. In fact, this is already an operational practice at some global NWP centers.

Fig. 6.
Fig. 6.

Vertical profile of mean AMV QI values for all PREDICT sample AMVs (N = 6733), sorted into 50-hPa bins.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

The evaluation of the AMV datasets can also be broken down by experiment mission (designated by PGI# for field projects PREDICT, GRIP, and IFEX, and numbered based on their chronological mission identification). Figure 7 shows the mean AMV bias and RMSE for 25 of the 26 flight missions. (The very first PREDICT mission was a test flight and is not evaluated here. Also, the first two flights of PGI44 occurred on the same day and are counted as one mission day for this figure.) In general, AMV speed biases remain within ±1 m s−1 for the majority of missions. The fluctuations in RMSE are greater and likely more sensitive to varying tropospheric flow conditions. RMSE generally remains near or below 4 m s−1, although a couple of curious spikes exist for mission PGI36L (Fiona). In the first case, the RMSE spike is associated with an abnormally high amount of thin cirrus and associated erroneous VIS AMVs, as discussed above. The spike for the third PGI36L mission can be mostly attributed to a single dropsonde that indicated weak wind speeds between 600 and 800 hPa being compared to a large number of nearly uniform, higher speed AMVs. It remains unclear if the single dropsonde wind values in this case are corrupt, or the collocated fleet of AMVs is erroneous, although animations of cloud movement suggest the former.

Fig. 7.
Fig. 7.

Mean RMSE and bias for AMVs (in comparison with collocated dropsondes) sorted by aircraft missions during PREDICT.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

b. Validation of the AMV gridded fields

As stated in section 2, CIMSS postprocesses the AMVs into gridded analyses that are used extensively by the global tropical cyclone community in both research and operations. The next step in this study is to validate these analyses versus the same dropsonde observations. Unlike the AMV matches, which are single level, the gridded analyses can be matched to dropsonde profiles. It is recognized that these matches may often reflect the background GFS being used in the analyses where AMVs are nonexistent. However, there may still be some AMV influence since the objective analysis methodology being employed is three-dimensional. For match precision, the nearest gridpoint values are interpolated to the latitude and longitudes of the coincident GV dropsondes.

Figure 8 shows the match count distribution for the gridded AMV analyses versus the dropsondes for the PREDICT period dataset. It is shown that below 150 hPa, the match counts at standard levels are all at or above 500. There were originally 558 dropsondes, although QC processes brings this nearer to 500 for most levels. The 150-hPa match count is lower since the GV was not always flying quite that high.

Fig. 8.
Fig. 8.

Distribution of match counts for interpolated gridded AMV wind values vs coincident dropsonde winds at standard pressure levels.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

Figure 9 shows the vertical profile of wind speed bias and RMSE for the above match set. The bias stays at less than 1 m s−1 for most of the troposphere, but becomes negative (AMV analyses slower than dropsonde values) at the very high levels. RMSE values stay primarily between 3 and 4 m s−1 although a maximum around 5 m s−1 occurs at 150 hPa along with the maximum negative bias. It should be noted again that the number of matches at 150 hPa are less than half of those used to validate at the other levels, and thus the statistics are more vulnerable to outliers. The secondary bump of slightly higher differences near 600 hPa is consistent with the region of the lowest AMV concentrations, leading to speculation that this could be due to an increased reliance on the GFS background values used in the assimilation process.

Fig. 9.
Fig. 9.

Vertical profile of (left) root-mean-square error (m s−1) and (right) bias (m s−1) of interpolated gridded AMV wind analyses vs collocated dropsonde winds at standard levels.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

In an analysis identical to that for the AMVs in Fig. 7, the mean gridded wind statistics are broken down by mission (Fig. 10). It is cautioned that a direct comparison of Figs. 7 and 10 would not be appropriate since the match samples are not homogeneous. In other words, Fig. 7 is dominated by upper-level matches (Fig. 2), while the results in Fig. 10 are representative of vertical profiles (Fig. 8). A collocated, homogeneous match sample will be shown and discussed in the next subsection. Figure 10 shows that the gridded AMV analyses maintain a bias of <1 m s−1 for the vast majority of missions, although there is a negative tendency. This suggests some combination of analysis smoothing and an inherent GFS (background field) weak bias in wind speeds. The RMSE trends around 4 m s−1 for the majority of missions, but higher values are noted in a couple of early missions, including PGI36L (Fiona) noted earlier in the AMV results.

Fig. 10.
Fig. 10.

As in Fig. 7, but for AMV gridded wind analyses (in comparison with collocated dropsonde wind profiles) broken down by missions during PREDICT.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

c. Intercomparison

In this section the authors construct a homogeneous dataset of AMVs and associated gridded values in order to facilitate a direct comparison between the AMV observations and their resultant analyses (as described in previous section). Only AMVs within 30 min and 1° of a dropsonde, and with assigned heights ±5 hPa of a standard level constant-pressure analysis, are considered. The gridded analysis values are interpolated to the latitude–longitude location of the AMV. A histogram of the intercomparison dataset (Fig. 11) shows generally the same height-dependent distribution structure as the AMV observations (Fig. 2), as might be expected. This leads to a reduced sampling in the 500–700-hPa layer, where intercomparisons will be suggestive at best.

Fig. 11.
Fig. 11.

Distribution of AMV and associated interpolated gridded analysis match counts by height (pressure) for the homogeneous intercomparison dataset.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

Finally, the authors wish to evaluate the AMVs and gridded analyses relative to the global model (GFS) from which background analyses were provided during the processing. In this way, the authors can assess the “value added” aspect of the AMV information in tropical analyses. For each of the match points in Fig. 11, gridpoint wind values were extracted from the GFS operational 1.0° analyses, spatially interpolated in the same way as the CIMSS gridded analyses, and time interpolated between 6-h analyses. Therefore, for this homogeneous intercomparison dataset, the authors have collocated (in space and time) wind values from dropsondes, AMVs, the CIMSS analyses, and the GFS. It should be noted that the operational GFS analyses did not assimilate the PREDICT dropsondes; therefore, this represents an independent validation.

From the intercomparison dataset, vertical profiles of wind speed RMSE (Fig. 12a) and speed bias (Fig. 12b) and directional RMSE (Fig. 12c) are computed using the collocated dropsonde winds as validation. It is shown that little difference in wind variance (RMSE) exists below 500 hPa, where fewer AMVs factor in the analysis assimilation. Above 500 hPa, the AMV and AMV-aided analyses generally show a slight improvement over the GFS. The improvements are more dramatic in the speed bias comparison. A strong slow bias is noted in the GFS, while the AMVs hover closer to neutral. The CIMSS analyses generally fall in between; however, they tend closer to the AMVs, which reflects the analysis strategy of giving strong weight to the AMV observations. The large spike at 150 hPa is apparently due to a mixture of factors. The environment above TCs is normally highly divergent, meaning comparison errors will inherently be larger (some of which may be collocation error). This is also reflected and supported by the wind direction RMSE profiles in Fig. 12c. There is also the possibility of larger error in the validating dropsonde wind estimates at this height, since the aircraft were dropping from just above this level in most cases. While the first dropsonde report after release was edited by quality control for this reason, it is possible that subsequent winds below that initial reading were also not yet stable. Finally, the spike in the CIMSS analyses at 150 could be in part due to the analysis boundary conditions, given this level is near the top level (100 hPa) in the three-dimensional analysis domain.

Fig. 12.
Fig. 12.

Comparison of vertical profiles at standard pressure levels for (a) wind speed root-mean-square error (m s−1), (b) speed bias (m s−1), and (c) wind direction (°) between collocated AMVs, interpolated gridded AMV wind analyses, and interpolated GFS wind analyses vs collocated dropsonde winds. The numbers in (c) indicate match counts at each standard pressure level.

Citation: Journal of Applied Meteorology and Climatology 51, 10; 10.1175/JAMC-D-12-024.1

4. Summary and conclusions

Atmospheric motion vectors (AMVs) are routinely derived by tracking features in sequential geostationary satellite imagery. While AMVs produced operationally by global data centers are routinely evaluated against rawinsondes, there is a relative dearth of validation opportunities over the tropical oceans; in particular, the vicinity of tropical disturbances when anomalous flow fields and strongly sheared environments commonly exist. A field experiment in 2010 called Pre-Depression Investigation of Cloud-Systems in the Tropics (PREDICT) was conducted in the tropical west Atlantic, and provided an opportunity to evaluate the quality of tropical AMVs and products derived from them. The importance of such verification is that tropical analysis and forecast centers often rely heavily on analyses and diagnostic products derived from the AMV fields.

In this study the authors utilize dropsonde wind information from PREDICT aircraft to identify AMV characteristics and better understand their errors in tropical disturbance situations. It is found that in general, the AMV quality metrics are close or slightly better than those identified in global validation studies. Some distinct characteristics are uncovered in certain regimes associated with tropical disturbances. Occasional AMV height assignment errors can occur in regions with thin tropical cirrus cloud tracers, and these errors will be amplified in high vertical wind shear situations. In many cases these vectors can be identified by their lower-quality indicator values and flagged before user applications or assimilation by objective analysis. This study illuminates areas in which AMV data producers may direct their attention to refine height assignment issues in particular.

High-resolution wind analyses and diagnostic fields derived from the AMV datasets by CIMSS are also examined against the dropsonde wind observations and found to be more reflective of anomalous flow fields than the respective GFS global model analyses, particularly in the upper levels. The CIMSS tropical analyses are employed extensively by the research and forecasting communities; therefore, these findings have relevance to support their use during TC events.

Acknowledgments

The authors thank D. Stettner and S. Wanzong of CIMSS for their help in producing the AMV datasets during PREDICT. EOL provided the quality-controlled GV dropsondes. This research was supported by NSF Grant Award Number 0850709.

REFERENCES

  • Bedka, K. M., , and J. R. Mecikalski, 2005: Application of satellite-derived atmospheric motion vectors for estimating mesoscale flows. J. Appl. Meteor., 44, 17611772.

    • Search Google Scholar
    • Export Citation
  • Bedka, K. M., , C. S. Velden, , R. A. Petersen, , W. F. Feltz, , and J. R. Mecikalski, 2009: Comparisons of satellite-derived atmospheric motion vectors, rawinsondes, and NOAA Wind Profiler observations. J. Appl. Meteor. Climatol., 48, 15421561.

    • Search Google Scholar
    • Export Citation
  • Berger, H., , R. H. Langland, , C. S. Velden, , C. A. Reynolds, , and P. M. Pauley, 2011: Impact of enhanced satellite-derived atmospheric motion vector observations on numerical tropical cyclone track forecasts in the western North Pacific during TPARC/TCS-08. J. Appl. Meteor. Climatol., 50, 23092318.

    • Search Google Scholar
    • Export Citation
  • Dunion, J. P., , and C. S. Velden, 2002: Application of surface-adjusted GOES low-level cloud-drift winds in the environment of Atlantic tropical cyclones. Part I: Methodology and validation. Mon. Wea. Rev., 130, 13331346.

    • Search Google Scholar
    • Export Citation
  • Franklin, J. L., , J. Kaplan, , C. S. Velden, , and C. M. Hayden, 1990: Some comparisons of VAS and dropwindsonde data over the subtropical Atlantic. Mon. Wea. Rev., 118, 18691887.

    • Search Google Scholar
    • Export Citation
  • Goerss, J. S., , C. S. Velden, , and J. D. Hawkins, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part II: NOGAPS forecasts. Mon. Wea. Rev., 126, 12191227.

    • Search Google Scholar
    • Export Citation
  • Hock, T. F., , and J. L. Franklin, 1999: The NCAR GPS dropwindsonde. Bull. Amer. Meteor. Soc., 80, 407420.

  • Holmlund, K., , C. S. Velden, , and M. Rohn, 2001: Enhanced automated quality control applied to high-density satellite-derived winds. Mon. Wea. Rev., 129, 517529.

    • Search Google Scholar
    • Export Citation
  • Kelly, G., , T. McNally, , J.-N. Thepaut, , and M. Szyndel, 2004: OSEs of all main data types in the ECMWF operation system. Proc. Third WMO Workshop on the Impact of Various Observing Systems on Numerical Weather Prediction, Alpbach, Austria, WMO World Weather Watch (Tech. Doc. WMO/TD 1228), 63–94. [Available online at http://www.wmo.int/pages/prog/www/GOS/Alpbach2004/Proceedings.pdf.]

    • Search Google Scholar
    • Export Citation
  • Key, J. R., , D. A. Santek, , C. S. Velden, , N. Bormann, , J.-N. Thepaut, , L. P. Riishojgaard, , Y. Zhu, , and W. P. Menzel, 2003: Cloud-drift and water vapor winds in the Polar regions from MODIS. IEEE Trans. Geosci. Remote Sens., 41, 482492.

    • Search Google Scholar
    • Export Citation
  • Kiladis, G. N., , C. D. Thorncroft, , and N. M. J. Hall, 2006: Three-dimensional structure and dynamics of African easterly waves. Part I: Observations. J. Atmos. Sci., 63, 22122230.

    • Search Google Scholar
    • Export Citation
  • Langland, R. H., , C. Velden, , P. M. Pauley, , and H. Berger, 2009: Impact of satellite-derived rapid-scan wind observations on numerical model forecasts of Hurricane Katrina. Mon. Wea. Rev., 137, 16151622.

    • Search Google Scholar
    • Export Citation
  • Leslie, L. M., , J. F. LeMarshall, , R. P. Morison, , C. Spinoso, , R. J. Purser, , N. Pescod, , and R. Seecamp, 1998: Improved hurricane track forecasting from the continuous assimilation of high quality satellite wind data. Mon. Wea. Rev., 126, 12481258.

    • Search Google Scholar
    • Export Citation
  • Menzel, W. P., 2001: Cloud tracking with satellite imagery: From the pioneering work of Ted Fujita to the present. Bull. Amer. Meteor. Soc., 82, 3347.

    • Search Google Scholar
    • Export Citation
  • Merrill, R. T., , and C. S. Velden, 1996: A three-dimensional analysis of the outflow layer of Supertyphoon Flo (1990). Mon. Wea. Rev., 124, 4763.

    • Search Google Scholar
    • Export Citation
  • Montgomery, M. T., and Coauthors, 2012: The Pre-Depression Investigation of Cloud-Systems in the Tropics (PREDICT) experiment: Scientific basis, new analysis tools, and some first results. Bull. Amer. Meteor. Soc., 93, 153172.

    • Search Google Scholar
    • Export Citation
  • Poteat, K. O., 1973: A comparison of satellite-derived, low-level and cirrus-level winds with conventional wind observations. J. Appl. Meteor., 12, 14161419.

    • Search Google Scholar
    • Export Citation
  • Pu, Z., , X. Li, , C. S. Velden, , S. D. Aberson, , and W. T. Liu, 2008: The impact of aircraft dropsonde and satellite wind data on numerical simulations of two landfalling tropical storms during the Tropical Cloud Systems and Processes experiment. Wea. Forecasting, 23, 6279.

    • Search Google Scholar
    • Export Citation
  • Rao, P. A., , C. S. Velden, , and S. A. Braun, 2002: The vertical error characteristics of GOES-derived winds: Description and experiments with numerical weather prediction. J. Appl. Meteor., 41, 253271.

    • Search Google Scholar
    • Export Citation
  • Rodgers, E., , and R. C. Gentry, 1983: Monitoring tropical-cyclone intensity using environmental wind fields derived from short-interval satellite images. Mon. Wea. Rev., 111, 979996.

    • Search Google Scholar
    • Export Citation
  • Sears, J., 2011: Investigating the role of the upper-levels in tropical cyclogenesis. M.S. thesis, Dept. of Oceanographic and Atmospheric Science, University of Wisconsin—Madison, 96 pp.

  • Soden, B. J., , C. S. Velden, , and R. E. Tuleya, 2001: The impact of satellite winds on experimental GFDL hurricane model forecasts. Mon. Wea. Rev., 129, 835852.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , and K. M. Bedka, 2009: Identifying the uncertainty in determining satellite-derived atmospheric motion vector height attribution. J. Appl. Meteor. Climatol., 48, 450463.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , C. M. Hayden, , W. P. Menzel, , J. L. Franklin, , and J. S. Lynch, 1992: The impact of satellite-derived winds on numerical hurricane track forecasting. Wea. Forecasting, 7, 107118.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , C. M. Hayden, , S. J. Nieman, , W. P. Menzel, , S. Wanzong, , and J. S. Goerss, 1997: Upper-tropospheric winds derived from geostationary satellite water vapor observations. Bull. Amer. Meteor. Soc., 78, 173195.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., , T. Olander, , and S. Wanzong, 1998: The impact of multispectral GOES-8 wind information on Atlantic tropical cyclone track forecasts in 1995. Part I: Dataset methodology, description, and case analysis. Mon. Wea. Rev., 126, 12021218.

    • Search Google Scholar
    • Export Citation
  • Velden, C. S., and Coauthors, 2005: Recent innovations in deriving tropospheric winds from meteorological satellites. Bull. Amer. Meteor. Soc., 86, 205223.

    • Search Google Scholar
    • Export Citation
  • Viezee, W., , R. M. Endlich, , and S. M. Serebreny, 1967: Satellite-viewed jet stream clouds in relation to the observed wind field. J. Appl. Meteor., 6, 929935.

    • Search Google Scholar
    • Export Citation
  • Xiao, Q., , X. Zou, , M. Pondeca, , M. A. Shapiro, , and C. Velden, 2002: Impact of GMS-5 and GOES-9 satellite-derived winds on the prediction of a NORPEX extratropical cyclone. Mon. Wea. Rev., 130, 507528.

    • Search Google Scholar
    • Export Citation
  • Zapotocny, T. H., , J. A. Jung, , J. F. Le Marshall, , and R. E. Treadon, 2007: A two-season impact study of satellite and in situ data in the NCEP Global Data Assimilation System. Wea. Forecasting, 22, 887909.

    • Search Google Scholar
    • Export Citation
Save