298
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
of number of days that different methods performed best for the four evaluation coefficients (Table 12.9), BA, NEXRAD, SKlm, and RK perform best in terms of EB,
R 2 , EE, and RVar, respectively. Implementing multiple methods to estimate spatial
precipitation maps is a practical way of providing a more accurate spatial precipitation map.
12.6 CONCLUSIONS
NEXRAD has emerged as a valuable precipitation product. In this chapter, we
reviewed the current literature on accuracy evaluation of NEXRAD and calibrated
NEXRAD data using rain gauge observations. The review indicates that the development of user-friendly GIS-based NEXRAD processing and calibration software is
critical for the application of NEXRAD precipitation products in hydrology, ecology,
agriculture, and meteorology. We have also introduced recently developed GIS software (NEXRAD-VC) that can calibrate NEXRAD data with rain gauge observations
using geostatistical approaches and automatically process NEXRAD data for hydrologic and ecological models. NEXRAD-VC can (1) automatically read NEXRAD
data in NetCDF or XMRG format, transforming the projection of NEXRAD data to
match rain gauge observations, (2) apply different geostatistical approaches to calibrate NEXRAD data using rain gauge data, (3) evaluate the performance of different calibration methods using the leave-one-out cross-validation scheme, (4) output
spatial precipitation maps in ArcGIS grid format, and (5) calculate spatial average
precipitation for each spatial modeling unit used by hydrologic and ecological models. NEXRAD-VC is a public-domain software, which is expected to facilitate the
application of NEXRAD.
Two case studies on evaluating the accuracy of NEXRAD and calibrating
NEXRAD data with rain gauge observations were presented. The first case study
examined the performance of NEXRAD in a mountainous region versus the southern plains in the United States and during a cold season versus a warm season. In
addition, we explored the effect of subgrid variability and temporal scale. Results
from these comparisons indicate that (1) NEXRAD performs better in the plains
region than in a mountainous region with complex terrain, (2) NEXRAD performs
better in a warm season than in a cold season, (3) NEXRAD should be evaluated
using a dense rain gauge network to reduce the influence of subgrid heterogeneity
of precipitation distribution, and (4) NEXRAD performs better at a daily temporal
scale than at an hourly temporal scale. These conclusions are derived based on the
analysis of two NEXRAD grids; their validity should be further examined using
high-quality data in other regions. Overall, the assessment of NEXRAD indicates
the need to remove bias of the NEXRAD precipitation product before its application.
The second case study examined the performance of three methods that use both
rain gauge and NEXRAD data for precipitation estimation in LREW. A visualization
process illustrated that substantial differences exist among the spatial precipitation
maps estimated by the different methods. On the average, although SKlm outperforms the other methods in terms of EB, R 2 , and EE, it performs the weakest in terms
of preserving variability of the spatial precipitation distribution. Further analysis
of the performance of different methods for daily spatial precipitation estimation
Multiscale Hydrologic Remote Sensing: Perspectives and Applications
of number of days that different methods performed best for the four evaluation coefficients (Table 12.9), BA, NEXRAD, SKlm, and RK perform best in terms of EB,
R 2 , EE, and RVar, respectively. Implementing multiple methods to estimate spatial
precipitation maps is a practical way of providing a more accurate spatial precipitation map.
12.6 CONCLUSIONS
NEXRAD has emerged as a valuable precipitation product. In this chapter, we
reviewed the current literature on accuracy evaluation of NEXRAD and calibrated
NEXRAD data using rain gauge observations. The review indicates that the development of user-friendly GIS-based NEXRAD processing and calibration software is
critical for the application of NEXRAD precipitation products in hydrology, ecology,
agriculture, and meteorology. We have also introduced recently developed GIS software (NEXRAD-VC) that can calibrate NEXRAD data with rain gauge observations
using geostatistical approaches and automatically process NEXRAD data for hydrologic and ecological models. NEXRAD-VC can (1) automatically read NEXRAD
data in NetCDF or XMRG format, transforming the projection of NEXRAD data to
match rain gauge observations, (2) apply different geostatistical approaches to calibrate NEXRAD data using rain gauge data, (3) evaluate the performance of different calibration methods using the leave-one-out cross-validation scheme, (4) output
spatial precipitation maps in ArcGIS grid format, and (5) calculate spatial average
precipitation for each spatial modeling unit used by hydrologic and ecological models. NEXRAD-VC is a public-domain software, which is expected to facilitate the
application of NEXRAD.
Two case studies on evaluating the accuracy of NEXRAD and calibrating
NEXRAD data with rain gauge observations were presented. The first case study
examined the performance of NEXRAD in a mountainous region versus the southern plains in the United States and during a cold season versus a warm season. In
addition, we explored the effect of subgrid variability and temporal scale. Results
from these comparisons indicate that (1) NEXRAD performs better in the plains
region than in a mountainous region with complex terrain, (2) NEXRAD performs
better in a warm season than in a cold season, (3) NEXRAD should be evaluated
using a dense rain gauge network to reduce the influence of subgrid heterogeneity
of precipitation distribution, and (4) NEXRAD performs better at a daily temporal
scale than at an hourly temporal scale. These conclusions are derived based on the
analysis of two NEXRAD grids; their validity should be further examined using
high-quality data in other regions. Overall, the assessment of NEXRAD indicates
the need to remove bias of the NEXRAD precipitation product before its application.
The second case study examined the performance of three methods that use both
rain gauge and NEXRAD data for precipitation estimation in LREW. A visualization
process illustrated that substantial differences exist among the spatial precipitation
maps estimated by the different methods. On the average, although SKlm outperforms the other methods in terms of EB, R 2 , and EE, it performs the weakest in terms
of preserving variability of the spatial precipitation distribution. Further analysis
of the performance of different methods for daily spatial precipitation estimation