University of Washington GPM-Ku Data Set
***** You must first register your email address before accessing any data.
Click here to register. *****
The data are stored by regions of the globe, represented by the
boxes in the map above. By clicking inside a box on the map, the user will go to
the page containing the analysis products for that selected region. One can accomplish
the same thing by going here and clicking on the region of
The NASA GPM 2AKu product is provided in bins along
the slant range of the antenna beam. In the University of Washington database, these
data in radar coordinates are geolocated and interpolated into a three-dimensional
Cartesian grid and stored for analysis. From this dataset it is possible to identify
structures of reflectivity echoes that satisfy certain criteria. Six types of echo objects
are identified: detectable rain areas (DRA), isolated shallow echoes (ISE),
deep convective cores (DCC), wide convective cores (WCC), deep-wide
convective cores (DWC), and broad stratiform rain areas (BSR). Specific,
defining characteristics of each of the objects are defined in section 4, below.
For all the classifications except DRA, the
calculations are based on two different sets of thresholds -
strong (str) and moderate (mod). Each object identified by these
criteria is provided in netcdf format in both gridded (grd) (dimensioned by latitude and
longitude) and tabular (tab) (dimensioned by instance or case) form. These gridded and
tablular netcdf files contain information of the monthly climatology and individual
properties of the identified echo objects. So, for example, the broad stratiform data
from January of 2015 is contained in 4 files - strong thresholds and gridded, strong
thresholds and tabular, moderate thresholds and gridded, and moderate thresholds and tabular.
3. Processed Data
In order to identify the echo objects (DRA, ISE, DCC, WCC, DWC, BSR), the raw
GPM data are subjected to the following, specialized processing.
- Interpolation of the GPM data to a Cartesian Grid The GPM corrected reflectivity values are interpolated to
a regular Cartesian grid. We produce the interpolated values from the GPM archived data
identified as 2AKu (specifially, the sub-product, zFactorCorrected). The GPM archive
provides these data in range bins along
the slant range of the radar beam, i.e. in "radar coordinates." For
each of the above-described regions, the 2AKu data in each
range bin are first geolocated and then interpolated into a regular
three-dimensional Cartesian grid. The interpolated grid size is 0.05°
(approximately 5 km) in the horizontal at 176 vertical levels (separated
by 0.125 km) for the 2AKu reflectivity factor. Details of the methods applied for
correcting the geolocation and the interpolation of the datasets can be found in
Houze et al. (2007) Additionally, nearest neighbor interpolation is used to re-grid
the 2AKu Near Surface Rain and the 2AKu Rain Type products corresponding to
the selected areas into a two-dimensional 0.05° grid. The interpolated
netcdf files contain all of this information with appropriate meta-data
containing accurate descriptions of the stored variables.
- Raintype Designation The
files contain three versions of the interpolated 2AKu Rain Type variable.
- rain_type_raw contains the original v05 values (ranging from 10000000 to 39999999)
and includes values of -1111 for no rain and -99 for missing data.
- rain_type is a simplified version of the categories with values of
(1) for stratiform, (2) for convective, and (3) for other.
- rain_type_uw is based on a correction that we have developed to reclassify some of
the mis-typed convective pixels as stratiform. The need for this variable should diminish with newer
versions of the GPM data which will incorporate our logic into the CSF algorithm.
4. Identification of Echo Objects
The UW database contains information on six types of radar echo
objects derived from the interpolated Cartesian data described above.
- Detectable Rain Area (DRA) An area in the
interpolated data consisting of two or more contiguous reflectivity pixels
containing detectable radar echoes.
- Isolated Shallow Echo (ISE) This feature
is an area of contiguous pixels containing detectable reflectivity and
identified in the 2AKu Rain Type product as the Shallow Isolated type of
echo. It consists of echoes whose tops are at least 1 km below
the 0degC level and may be thought of as regions governed primarily by warm
- Deep Convective Core (DCC) This feature is a
three-dimensional echo object that is convective (as designated by the GPM
rain_type_uw field) and exceeds both an intensity threshold and a height threshold.
Moderate (strong) DCCs are those that exceed a threshold intensity
of 30 (40) dBZ and an echo-top height threshold of 8 (10) km.
This category, whether defined by the moderate or strong threshold, is
associated with young, vigorous convection.
- Wide Convective Core (WCC) This feature is a
three-dimensional echo object that is convective (as designated by the
GPM rain_type_uw field) and exceeds both an intensity threshold and a horizontal
area threshold. Moderate (strong) WCCs are those that at some altitude
exceed a threshold intensity of 30 (40) dBZ and a horizontal echo-area
threshold of 800 km2 (1,000 km2). These echo
volumes represent extremely intense convection that has organized upscale
into mesoscale areas of active, widespread convection.
- Deep-Wide Convective Core (DWC) This feature is
a three-dimensional convective echo object that satisfies both the DCC and
WCC criteria for either moderate or strong thresholds.
- Broad Stratiform Region (BSR) This feature is
a contiguous region of stratiform echo (as designated by the GPM rain_type_uw
field) and exceeds a horizontal area threshold. Moderate (strong) BSRs
are those that exceed a horizontal echo-area threshold of 40,000 km2
(50,000 km2). These echoes represent parts of mature and very
well-developed mesoscale convective systems.
5. Echo Objects Masks
Each of the core echo objects -- BSR, DCC, DWC and WCC -- is usually
embedded in a larger storm. The statistics for each event in the tabular
output describie both the core and its associated storm.
Each core and storm are also identified by 2-dimensional masks in the interpolated
data files. Since some storms contain more than one echo object of a particular
type, each core is identified by a core_id. The core_id is included in both the
tabular data and the appropriate mask variable in the interpolated data. For
example, if a storm contains three different DCC objects, they are numbered 1,
2 and 3 in the tabular output and the correspoinding DCC mask is filled with 1's,
2's and 3's at the appropriate grid points. All remaining points in the mask
are filled with the missing value.
6. File Naming Convention
The file naming convention is consistent throughout the
website. The files are stored in directories first by region and then
by data type.
For the interpolated data, the convention is:
- [version] = currently 'GPM2Ku5_uw3' which means this is based on
GPM version 5 data and the UW's version 3 products
- [start_yyyymmdd] = start year, month and day
- [start_hhmmss] = start hour, minute and second
- [end_yyyymmdd] = end year, month and day
- [end_hhmmss] = end hour, minute and second
- [orbit] = six-digit orbit number
- [region] = three or four letter abbreviation for the
region of interest
For the classified radar echo objects, the convention is:
- [version] = currently 'GPM2Ku5_uw3' which means this is based on
GPM 2Ku version 5 data and the UW's version 3 products
- [classification] = one of 'BSR', 'DCC', 'DRA', 'DWC', 'SHI' or
- [threshold] = one of 'str', 'mod' or 'xxx'
- [type] = one of 'tab' or 'grd'
- [yyyymm] = year and month
- [region] = three letter abbreviation for the
region of interest
So a typical filename might look like this -
'GPM2Ku5_uw3_BSR_str_tab_201412_SAM.nc'. This file would contain
broad stratiform (BSR) data based on strong (str) thresholds in
'tabular' format (tab) for December of 2014 for the South America
DRA is a special case since we do not use different
thresholds. In that instance, threshold is set to 'xxx'.
Since our analyses is based on a series of regional studies
which will eventually span most of the globe, the data is meant to overlap
at the edges of all the global regions. It is suggested that if a user
needs to combine information from two or more regions for a singular
analysis, a careful 'stitching' of the data must be accomplished to avoid
identifying overlapped echo elements multiple times.
7. Note of Appreciation
Erich Stocker of NASA/Goddard Spaceflight Center provided invaluable help
with the processing of the data and products provided via this website.
8. Citation to be used for Data and Products
Obtained from this Site
In all publications and presentations based on the information
provided via this website, please acknowledge the University of Washington
with a statement like this: "Data for this study comes from the University of
Washington GPM-Ku Data Set located at http://gpm.atmos.washington.edu" and
supported by the NASA Earth Sciences PMM Program.
For questions, contact us here.