Spatial Pattern of Bias in Areal Rainfall Estimations and Its Impact on Hydrological Modeling: A Comparative Analysis of Estimating Areal Rainfall Based on Radar and Weather Station Networks in South Korea

Abstract

Abstract Areal rainfall is routinely estimated based on the observed rainfall data using distributed point rainfall gauges. However, the data collected are sparse and cannot represent the continuous rainfall distribution (or field) over a large watershed due to the limitations of weather station networks. Recent improvements in remote-sensing-based rainfall estimation facilitate more accurate and effective hydrological modeling with a continuous spatial representation of rainfall over a watershed of interest. In this study, we conducted a systematic stepwise comparison of the areal rainfalls estimated by a synoptic weather station and radar station networks throughout South Korea. The bias in the areal rainfalls computed by the automated synoptic observing system and automatic weather system networks was analyzed on an hourly basis for the year 2021. The results showed that the bias increased significantly for hydrological analysis; more importantly, the identified bias exhibited a magnitude comparable to that of the low flow. This discrepancy could potentially mislead the overall rainfall-runoff modeling process. Moreover, the areal rainfall estimated by the radar-based approach significantly differed from that estimated by the existing Thiessen Weighting approach by 4–100%, indicating that areal rainfalls from a limited number of weather stations are problematic for hydrologic studies. Our case study demonstrated that the gauging station density must be within 10 km2 on average for accurate areal rainfall estimation. This study recommends the use of radar rainfall networks to reduce uncertainties in the measurement and prediction of areal rainfalls with a limited number of ground weather station networks.