|Zhang, Yu - USEPA|
|Adams, Thomas - NATIONAL WEATHER SERVICE|
Submitted to: Journal of Hydrometeorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: March 15, 2007
Publication Date: December 1, 2007
Citation: Zhang, Y., Adams, T., Bonta, J.V. 2007. Sub-pixel scale rainfall variability and the effects on separation of radar and gauge rainfall errors. Journal of Hydrometeorology. 8(6):1348-1363. Interpretive Summary: Radar-based rainfall estimates and ground-level precipitation measurements often do not agree. One of the reasons is that rain-gage measurements are point measurements representing only the small area of a rain gauge, while a radar estimate is associated with a square area of about 1 km2. The differences are dependent on the time during a storm, storm characteristics, and the location on the landscape. This study evaluates an extension of a method to quantify the variability in small scale ground-level rainfall measurements and the impact of this variability on errors in radar-based estimates of rainfall. Eleven rain gauges located within a 1-km radius at the North Appalachian Experimental Watershed at Coshocton Ohio were used in this study. The 11 gauges lie within 4 square radar elements superimposed on the land surface (pixels). To represent different seasonal rainfall characteristics, July and October data were used. The extension of the method for quantifying errors shows a dependence of the estimated contribution of area-point rainfall differences on the representation of the within-radar-pixel correlation, which in turn depends on storm characteristics. The study is useful for improving our understanding of errors in spatial radar rainfall estimates, and can be used by researchers investigating the use of radar for watershed models. Other implications for the research are presented in the paper.
Technical Abstract: As long recognized, one of the primary sources of the discrepancies in the radar-based rainfall estimates and rain gauge measurements is the point-area difference, i.e., the intrinsic difference in the spatial dimensions of the rainfall fields that radar rainfall estimates and gauge measurements are meant to represent (i.e., ~1 km2 for the former and 1m2 for the latter). The point-area difference, which in turn determined by the small-scale rainfall variability, can manifest itself as the difference in gauge- and radar-based areal rainfall estimates at various spatial and temporal scales. This study presents an extension of the error separation method (ESM) that would allow the utilization of the latter’s formulation of the point-areal difference to quantify the impacts of small-scale rainfall variability on the difference in the radar- and gauge-based representations of areal rainfall. The methodology, herein referred to as EESM, is implemented for a study site in the USDA North Appalachian Experimental Watershed (NAEW) near Coshocton, OH, where a dense network of 11 rain gauges are clustered within an approximately 1-km radius and are interspersed within four Hydrologic Rainfall Analysis Project (HRAP) pixels (4 km by 4 km). For each pixel, areal rainfall estimates based on NEXRAD Stage III data and corresponding gauge records are used jointly via EESM to determine the relative contribution of point-area difference to the observed difference in radar and gauge-based areal rainfall estimates. To help illuminate the connection between EESM results and precipitation climate, July and October 2001, two months of contrasting rainfall regime, are designated as the temporal settings of this study. The results of the EESM point to a tangible dependence of the estimated contribution of area-point difference on the conceptualization of sub-pixel correlation structure, which, however, varies depending on the prevalent rainfall structure. The study suggests that accounting for the variations in the correlation structure within a pixel might be necessary for accurate delineation of the impacts of spatial rainfall variability and particularly during the summer months. The implications of the findings for applying EESM to field scale and watershed scale analysis are further discussed.