Abstract
Albeit weather surveillance radar (WSR)-88D stage III radar rainfall (RR) data can generally capture the spatial variability of precipitation fields, its rainfall depth for cold seasons dominated by stratiform storms tends to be underestimated. This study proposed merging WSR-88D stage III data with rain gauge data using the Haar wavelet scheme and compared its with that merged by the statistical objective analysis (SOA) scheme. The idea is to exploit the strength of radar that better captures the spatial variability of rainfall and that of rain gauges that measure the rainfall depth more accurately. A Haar wavelet was used because of its simplicity and the appealing physical interpretation of its coefficients as directional gradients of rainfall, whose spatial correlation structure was accounted for through a polynomial function. From analysing 89 storms over the Blue River Basin (BRB), Oklahoma, during 1994–2000, the results show that the underestimation problem of WSR-88D RR was generally more pronounced during the cold season dominated by stratiform storms than warm season dominated by convective storms. The wavelet scheme was better than SOA in reducing the radar's underestimation of rainfall depths while maintaining the spatial variability of the original radar data, as shown by its merged rainfall patterns and the more accurate streamflow hydrographs simulated by a semi-distributed, physics-based rainfall-runoff model – semi-distributed physics-based hydrologic model using remote sensing (DPHM-RS) – driven by the merged data.
Acknowledgements
This research is partly funded by the NSERC of Canada, and the first author is also supported by the F.S. Chia scholarship of the University of Alberta. The Oklahoma Mesonet program provided the meteorological data while the NWS HL of USA provided the operational hourly WSR-88D RR data and hourly observed streamflow data through the DMIP.