Evaluating NEXRAD Rainfall Estimates with Rain Gauge Data

Abstract

Radar rainfall estimates are an alternative source of information when a rain gauge is unavailable in an area. However, radar rainfall data may over- or under-estimate rainfall compared to rain gauges. In this study, comparisons were made with rain gauge data to verify the accuracy of the radar rainfall in Puerto Rico. Rainfall can be estimated with remote sensing techniques, such as satellite sensors or radar (Harmsen et al., 2008). Satellite sensors may be active or passive. Active sensors, such as radar, transmit and receive the reflected energy. On the other hand, passive sensors measure radiant energy emitted or reflected from bodies, such as a cloud top. The Next Generation Radar (NEXRAD) estimates parameters such as rainfall and wind speed and direction. The Advanced Hydrologic Prediction Service (AHPS) of the National Oceanic and Atmospheric Administration (NOAA) provide daily cumulative bias corrected NEXRAD data (McEnery et al., 2004). <fig><graphic xlink:href=23520_files/23520-00.jpg id=ID_436f699a-f74e-4795-9c2a-8ce9daf0c60d></graphic></fig> The advantages of using radar rainfall data include low cost, easy accessibility, large coverage areas, and data in gridded format that can be input into hydrologic models. The disadvantages of radar rainfall data include rainfall depth underestimation due to the Earth's curvature at increasing distances from the radar signal (Torres-Molina, 2017) and its inability to characterize sub-pixel rainfall variation, which can be significant in Puerto Rico (Harmsen et al., 2008). In Puerto Rico, the NOAA Doppler radar is located in Cayey, providing coverage for the entire island and the US Virgin Islands, though in 2017 the radar was destroyed by Hurricane Maria (Weather.com, 2018). In this study, rain gauge data were combined with AHPS data from the website pragwater.com for corresponding days during 2022. A MATLAB computer program was developed to extract the AHPS rainfall data for the pixel corresponding to the locations of the rain gauges. Regression analysis including development of 1:1 plots, linear regression equations, and coefficients of determination (r²) to evaluate the gauge and radar data agreement for the rainfall data pairs. Figure 1 shows an example of a 1:1 plot for data collected at Maricao, PR, between March and August 2022. The plot shows reasonably good correlation between the gauge and radar data, however, the radar severely underestimated rainfall on August 11 (gauge 132.6 mm vs. AHPS 67.7 mm). In this poster, we will present comparisons between rain gauge and AHPS data for ten locations in Puerto Rico.