274
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
for MPE NEXRAD results (http://water.weather.gov/about.php). Further information
about MPE NEXRAD is provided by Seo and Breidenbach (2002). The MPE products are precipitation approximations over a grid of about 4 × 4 km 2 , usually referred
to as a Hydrologic Precipitation Analysis Project (HRAP) grid (Reed and Maidment
1999). MPE analyses generated by the 12 river forecast centers (RFCs; http://water
.weather .gov/precip/rfc.php) are used to create a mosaic for a national Stage IV product at the National Centers for Environmental Prediction over CONUS at hourly, 6-h,
and daily temporal scales (http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/stage4/).
Daily NEXRAD data from 2006 and later across CONUS can also be obtained in
shapefile and network common data form (NetCDF) formats (http://water.weather
.gov). These daily NEXRAD data are derived from hourly NEXRAD precipitation
data (in compressed binary format) provided by the RFCs in the United States.
12.2.2 validation and caliBRation of nexRad
PReciPitation iMageS uSing Rain gauge data
One central question for the application of NEXRAD precipitation data in earth
system modeling is: How good are these estimates? (Krajewski and Smith 2002).
NEXRAD performance is influenced by many factors such as range degradation,
beam blockage in complex terrain, and quality of rain gauge data incorporated in
NEXRAD (Smith et al. 1996; Steiner et al. 1999; Stellman et al. 2001). The errors
associated with rain gauge observations may also lead to uncertainty of evaluation
of NEXRAD performance (Ciach and Krajewski 1999; Ciach et al. 2007; Villarini
et al. 2009). It is important to realize the limitations of using rain gauge observations
to validate NEXRAD data. In this study, due to the difficulty of obtaining true precipitation values, rain gauge observations were assumed to be the “ground truth” and
used to validate and calibrate NEXRAD data.
Many studies evaluated the accuracy of the NEXRAD data using rain gauge data.
For example, Steiner et al. (1999) evaluated hourly NEXRAD products in mountainous
regions and found that underestimation and nondetection of precipitation are significant
concerns. Young et al. (2000) evaluated NEXRAD Stage III products in Oklahoma
and found that the bias of NEXRAD reached about 20%. Jayakrishnan et al. (2004)
compared rain gauge observations with WSR-88D Stage III precipitation data over the
Texas–Gulf basin and found large differences (about 42% of the rain gauge measurements) between the two precipitation data sources. Dyer and Garza (2004) reported significant underestimation of Stage III products at a basin-average scale over Florida. Xie
et al. (2006) evaluated NEXRAD Stage III precipitation data over central New Mexico,
a semiarid area, and found that NEXRAD pronouncedly overestimated seasonal precipitation accumulation by 11%–88% during monsoon season or underestimated by
18%–89% during nonmonsoon season compared with rain gauge observations.
Recent studies showed the improvement of NEXRAD performance through the
transition from Stage III to MPE. Yilmaz et al. (2005) reported the superior performance of MPE to Stage III for basin-average precipitation estimation, especially
in winter. Wang et al. (2008) validated NEXRAD MPE and Stage III precipitation
products using rain gauge observations in the upper Guadalupe River Basin in Texas.
MPE has a higher capability for precipitation detection, higher linear correlation,
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
for MPE NEXRAD results (http://water.weather.gov/about.php). Further information
about MPE NEXRAD is provided by Seo and Breidenbach (2002). The MPE products are precipitation approximations over a grid of about 4 × 4 km 2 , usually referred
to as a Hydrologic Precipitation Analysis Project (HRAP) grid (Reed and Maidment
1999). MPE analyses generated by the 12 river forecast centers (RFCs; http://water
.weather .gov/precip/rfc.php) are used to create a mosaic for a national Stage IV product at the National Centers for Environmental Prediction over CONUS at hourly, 6-h,
and daily temporal scales (http://www.emc.ncep.noaa.gov/mmb/ylin/pcpanl/stage4/).
Daily NEXRAD data from 2006 and later across CONUS can also be obtained in
shapefile and network common data form (NetCDF) formats (http://water.weather
.gov). These daily NEXRAD data are derived from hourly NEXRAD precipitation
data (in compressed binary format) provided by the RFCs in the United States.
12.2.2 validation and caliBRation of nexRad
PReciPitation iMageS uSing Rain gauge data
One central question for the application of NEXRAD precipitation data in earth
system modeling is: How good are these estimates? (Krajewski and Smith 2002).
NEXRAD performance is influenced by many factors such as range degradation,
beam blockage in complex terrain, and quality of rain gauge data incorporated in
NEXRAD (Smith et al. 1996; Steiner et al. 1999; Stellman et al. 2001). The errors
associated with rain gauge observations may also lead to uncertainty of evaluation
of NEXRAD performance (Ciach and Krajewski 1999; Ciach et al. 2007; Villarini
et al. 2009). It is important to realize the limitations of using rain gauge observations
to validate NEXRAD data. In this study, due to the difficulty of obtaining true precipitation values, rain gauge observations were assumed to be the “ground truth” and
used to validate and calibrate NEXRAD data.
Many studies evaluated the accuracy of the NEXRAD data using rain gauge data.
For example, Steiner et al. (1999) evaluated hourly NEXRAD products in mountainous
regions and found that underestimation and nondetection of precipitation are significant
concerns. Young et al. (2000) evaluated NEXRAD Stage III products in Oklahoma
and found that the bias of NEXRAD reached about 20%. Jayakrishnan et al. (2004)
compared rain gauge observations with WSR-88D Stage III precipitation data over the
Texas–Gulf basin and found large differences (about 42% of the rain gauge measurements) between the two precipitation data sources. Dyer and Garza (2004) reported significant underestimation of Stage III products at a basin-average scale over Florida. Xie
et al. (2006) evaluated NEXRAD Stage III precipitation data over central New Mexico,
a semiarid area, and found that NEXRAD pronouncedly overestimated seasonal precipitation accumulation by 11%–88% during monsoon season or underestimated by
18%–89% during nonmonsoon season compared with rain gauge observations.
Recent studies showed the improvement of NEXRAD performance through the
transition from Stage III to MPE. Yilmaz et al. (2005) reported the superior performance of MPE to Stage III for basin-average precipitation estimation, especially
in winter. Wang et al. (2008) validated NEXRAD MPE and Stage III precipitation
products using rain gauge observations in the upper Guadalupe River Basin in Texas.
MPE has a higher capability for precipitation detection, higher linear correlation,