Deriving gridded hourly rainfall on Oahu by combining gauge network and radar rainfall

In Hawaii, flash flooding causes the highest amount of life and property damage among natural hazards and result in sediment erosion, subsequently impacting downstream and coastal ecosystem. Rainfall estimates at high temporal and spatial resolution are vital for the hydrological sciences, especially their applications to floods. However, historical gridded hourly rainfall is unavailable in Hawaii. Radar has been widely used to supplement the spatial rainfall distribution with point rainfall gauge data and has been applied to various hydrological research studies in many regions, while studies in Hawaii are impeded in part due to lack of confidence and validation in gridded hourly rainfall. In this study, we used radar reflectivity from the Next Generation Weather Radar (NEXRAD) on Molokai, which has full coverage of the island of Oahu. We derived radar rainfall by operating the Lidar Radar Open Software Environment (LROSE) for four different types of extreme events, including a front, an upper-level trough, a kona low, and a hurricane. Then, we combined the radar rainfall with gauge rainfall by applying kriging with external drift (KED). We applied leave-one-out cross validation (LOOCV) to evaluate the product performance and used multivariate analysis to understand what may lead to any biases or errors. The results show that the combination of radar and gauge rainfall provided better rainfall estimates than those from interpolation with rain gauges alone. We also examined variables (e.g., elevation, beam blockage fraction) that associate with the bias/error to adjust rainfall values. The newly derived hourly gridded rainfall dataset, obtained by incorporating gauge and radar information, will contribute to hydrological studies, e.g., flooding, erosion, and sediment transport analyses on Oahu. In addition, the established workflow can be the basis for deriving gridded hourly rainfall for other islands or other time periods.