ODIS
Mexico,
nm.edu
of data fusion
liscussed. The
ern of weather
1e Continental
4 km. Several
nd Fraction of
NPP]) provide
local weather
on response fo
susceptible to
lata (hourly to
ind ecosystem
than 7 million
ily 30 percent
States. These
monitoring is
So Kxnelen
rtt!
| by summ
ulf of Mexico.
n is typicdl
track eastward
on totals during
ation occurs in
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Julv with an average monthly precipitation of 75mm. Peak
temperatures occur between June and August and as such
contribute to the rapid evaporation of available soil moisture in
the root zone.
These factors create grassland conditions that are highly
responsive to local rainfall by virtue or little or no available soil
moisture during the principal growth period.
Sensitivity to local rainfall is further exacerbated by drought
conditions that began 1n late 1999 and continue to the present.
This extended period of subnormal precipitation has strongly
influenced both near and subsurface moisture availability and
has had a significant impact on grassland ecosystems, grassland
productivity and economic use.
summer
2. DATA AND METHODS
2.1 Data
The data used in this study were acquired from publicly
available sources and processed using standard software and
algorithms. MODIS data were acquired from the Land
Processes Distributed Active Archive Center (LPDAAC)
through the Earth Observing System Data Gateway (EOS).
Precipitation data were acquired the National Centers for
Environmental Prediction of the US Weather Service.
2.1.1 MODIS Data. The vegetation data were derived from
the MODIS Terra products, specifically the MODISA2 Leaf
Area Index/FPAR 8-day L4 Global Ikm SIN Grid. MODISA2
provides Leaf Area Index (LAI) and Fraction of
Photosynthetically Active Radiation (FPAR) absorbed by
vegetation with a spatial resolution of 1 km and is updated once
each 8-day period throughout the calendar year (Knyazikhin, et
al, 1999; Myneni, et al, 2003; MCST, 2003). The MODIS
FPAR product (MOD15A2) was selected because it provides
information on the structural property of the plant canopy. A
plant productivity product was selected as more likely to
indicate vegetation response to precipitation in the sparse
grassland environment of Northeastern New Mexico.
Fourteen files of 8-day composite FPAR data were downloaded
from the LPDAAC for the period of study of June 01, 2002
through September 15, 2002. The native hdf data files were
converted into ERDAS Imagine version 8.6 img files and
reprojected to UTM zone 13, using the NAD 27 datum. Eight-
day differences were calculated between the available data sets
to capture the change in FPAR over time. These difference
images were used in ArcGIS to compare changes in FPAR
values to areas of local precipitation derived from the
NEXRAD data (Figure 2).
2.1.2 NEXRAD Data. Weather radar data are systematically
collected over virtually all of the Continental United States and
provide spatially continuous and calibrated rainfall estimates at
a nominal spatial resolution of 4 km.
2.1.3 Processing of NEXRAD Stage IV data. The
conversion of NEXRAD Stage IV data to a format suitable for
use in an image processing and GIS requires several steps. The
native format for NEXRAD data is NWP HRAP GRIB
(National Centers for Environmental Prediction) which must
converted from the radial GRIB format be integrated into raster
based systems. The acquisition and conversion process involved
5 steps:
I. Data Acquisition
7 ‘ :
2. Data Conversion
907
3. Georeferencing
4. Conversion to Raster Format
5. Data Export
2.1.3.1 Data Acquisition. NEXRAD Stage IV data are
available for 1-, 6-, and 24-hour total precipitation summaries
for a continental US grid. These data are mosaiced from local, 4
km polar-stereographic grids for each of the 12 regional River
Forecast Centers (RFCs) in the continental US. These data are
derived from the Stage III precipitation products generated by
the RFCs. Each grid point in the Stage IV product provides
accumulated precipitation measured in mm for the specified
span of time. The nominal ground spacing is approximately 4
km nationally, but varies with actual NEXRAD coverage and
distance from the radar installation. For New Mexico the
nominal ground spacing is approximately 4.8 km. The data for
this analysis were obtained via FTP from the online 6-month
archive of data files ftp://ftpprd.ncep.noaa.gov/pub/gep/precip/
mpe.arch/stage4/.
2.1.3.2 Data Conversion. The downloaded files are provided
as compressed HRAP GRIB files. The compressed GRIB files
were uncompressed and converted from binary to ascii text
format using wgrib (http://wesley.wwb.noaa.gov/wgrib.html).
The ASCII text format generated by wgrib encodes the number
of columns and rows in the grid in the first line of the file and
stores the values for cach row in sequence as a single column of
values. This output was then used to generate the desired output
format through an automated process.
2.1.3.3 Georeferencing. After conversion, the ASCII
precipitation values were merged with their corresponding
geographic coordinates to generate a point dataset from which a
raster representation could be generated. To facilitate automated
processing, a reference ASCII file containing UTM (Zone 13,
NAD27) coordinate pairs for each GRIB grid point was
generated for each point. The order of the coordinate pairs in
the reference file exactly matched the order of the precipitation
values in the ASCII output of wgrib.
A custom Perl script was used to merge the UTM coordinates
pairs with the precipitation values, to subset the data to the
defined area of interest, and generate a GRASS GIS
(http://grass.itc.it/) sites file. A GRASS sites file consists of an
ASCII encoded point GIS coverage.
2.1.3.4 Conversion to Raster Format. The GRASS
s.surf.idw command was used with a | as the specified number
of points for the inverse-distance-weighting algorithm, resulting
in a nearest neighbor conversion of the points in the sites file
into a 1 km raster.
2.1.3.5 Data Export. To maximize portability, the GRASS
raster file was exported as an ASCII ArcGRID using the
r.out.arc GRASS command. The resultant ASCII file could then
be imported into ArcGIS for analysis of the precipitation and
any resultant vegetation response. This entire conversion and
reformatting process was automated using a Perl script that
performed all of the conversion, import, and export processes
on the HRAP NEXRAD GRIB files.
2.2 Methods
After conversion to gridded raster data the NEXRAD
precipitation data was imported into ArcGIS ver. 8.3 for
analysis. For analysis purposes, total precipitation was grouped
into eight classes (Table 1). A comparison of NEXRAD