Comparison of gage and multi-sensor precipitation estimates over a range of spatial and temporal scales in the Midwestern United States

Abstract

An intercomparison of radar-estimated precipitation and gage precipitation at a monthly time scale with a county spatial resolution was undertaken for a nine-state region of the Midwestern United States. Daily gage and radar-estimated precipitation data also were examined at the county and grid cell scale for several smaller regions. Precipitation data were collected from February 2002 to August 2005 from three sources: (1) gridded radar (stage II, RDR) and multi-sensor precipitation estimates (MPE) based on the stage III/IV algorithm developed by the Office of hydrology/NWS River Forecast Centers, (2) quality-controlled National Weather Service (NWS) cooperative gage (QC_Coop) data from the National Climatic Data Center (NCDC), and (3) gage data from three high density networks in Illinois. Both the QC_Coop and high density gage data were employed as the reference standard. Sixty-four percent of QC_Coop versus MPE county-averaged monthly precipitation estimates agreed to within ±25%, with a median difference of 5.6% (QC_Coop greater than MPE) for the Midwest region. The difference between gage and MPE monthly values decreased somewhat through the 41-month period of study, and the correlation between monthly estimates increased, averaging 0.80. Data from three regional gage networks indicated that on a daily basis, network-averaged MPE and gage data also agreed to within about ±25%, and the MPE values tended to be lower than gage amounts at higher precipitation values. When examining multiple gages within single MPE grid cells, it was found that the number of gages employed in computing the gage average did not appreciably affect the correspondence between MPE and gage precipitation amounts. This also was found examining monthly values at the county level. For daily precipitation at the grid cell scale, for daily networked-averaged precipitation for each of the regional networks, and for monthly county-averaged precipitation values across the Midwest, MPE values are often larger than gage values for lower gage precipitation totals, and as precipitation totals increase, MPE values are more likely to be the same or smaller than the gage value.