Effect of rainfall variability on hydrologic simulation using WSR-88D (NEXRAD) data

Abstract

Rainfall, one of the most important hydrologic model inputs, varies spatially and temporally. Simulations using rainfall data with a reasonable representation of this variability can be expected to produce better results for hydrologic system responses. Raingage rainfall data that have been used in hydrologic analyses involve several disadvantages: they are only point observations; their availability for subdaily time steps is very limited. Rainfall data from weather radars overcome these disadvantages. Effect of using spatially and temporally varied rainfall data on hydrologic simulation was analyzed in this study. Stage III WSR-88D radar rainfall data with sixteen square kilometer spatial resolution and one hour temporal resolution were used for this analysis. Soil and Water Assessment Tool, a distributed parameter hydrologic model, was used for hydrologic simulation in six watersheds located in Texas and New Mexico for the period 1994-99 without model parameter calibration. Results were compared with those obtained using raingage data. Rainfall data of the nearest raingage was assigned to each subbasin in this analysis. High spatial and temporal variability of rainfall was observed in the study area. In general, radar data underpredicted streamflow while raingage data resulted in overprediction. Evaluation of WSR-88D data quality is necessary before applying them in hydrologic modeling. Quality of Stage III WSR-88D rainfall data over the Texas-Gulf basin was evaluated using available raingage data. WSR-88D performance varied significantly over this basin. Radar performance was satisfactory at about 50% of the raingage locations. Bias adjustment of Stage III radar data was done using daily raingage data by calculating a daily Bias Adjustment Factor (BAF). Daily Stage III radar data were adjusted using BAF values. Hydrologic simulation efficiency improved with the use of bias-adjusted radar data. Selective bias adjustment for rainfall events with a flexible range for BAF may improve the results further. Analyses using temporally varying radar and raingage data suggested that consideration of rainfall temporal variability could improve the simulation results. High resolution Stage III WSR-88D rainfall data are useful for incorporating rainfall variability in hydrologic modeling and they will play an important role in short-term flood prediction and long-term waters resources planning studies.