Spatial and temporal resolution considerations in evaluating and utilizing radar quantitative precipitation estimates

Abstract

In recent years, Weather Surveillance Radar-1988 Doppler base data have become available at substantially finer spatial resolution—i.e., 0.25 km  0.5o (0.25 km  1o for use in precipitation processing) versus the legacy 1 km  1o. Quantitative precipitation estimation (QPE) at this higher resolution may yield greater accuracy in the discretization of precipitation to delineated stream and river basins, in turn resulting in operational benefit in improved performance of hydrologic models and forecasting tools. To assess this potential, 1-h radar QPEs were determined from two experimental, S-band radar systems across several spatial resolutions—starting with base reflectivity data at or near the newly available, fine resolution, and recombining them up to, and including, the legacy resolution. Next, 1-h QPE was calculated at these various resolutions (based on the traditional Z–R relationship) and matched against 1-h rain gauge accumulations from dense gauge networks. After performing several steps of manual quality control, the dataset contained over 9500 warm-season gauge–radar pairs; these were assessed in various sub-groups and configurations (including stratification by precipitation intensity and range from the radar) and in simulated small stream basins. In various statistical analyses, however, the error differences between the fine- and legacy-resolutions were not statistically significant. Several possible causes for this supposed counterintuitive result were investigated, including (1) increased sampling error/noisiness in smaller versus larger sample bins, (2) sub-beam factors acting on falling hydrometeors, and (3) the matter that temporal sampling frequencies employed operationally were not increased commensurately with the increase in spatial sampling resolution.