JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION VOL. 34, NO.1 AMERICAN WATER RESOURCES ASSOCIATION FEBRUARY 1998 EFFECTS OF IMPROVED PRECIPITATION ESTIMATES ON AUTOMATED RUNOFF MAPPING: EASTERN UNITED STATES' Gary D. Bishop, M. Robbins Church, and Christopher Daly2 ABSTRACT: We evaluated maps of runoff created by means of two automated procedures. We implemented each procedure using pre- cipitation estimates of both 5-km and 10-km resolution from PRISM (Parameter-elevation Regressions on Independent Slopes Model). Our goal was to determine if using the 5-km PRISM esti- mates would improve map accuracy. Visual inspection showed good general agreement among our runoff maps, as well as between our maps and one produced using a manual method. A quantitative uncertainty analysis comparing runoff interpolated from our maps with gage data that had been withheld showed slightly smaller actual and percentage interpolation errors for the 5-km PRISM- based maps. Our analyses suggest a modest region-wide improve. ment in runoff map accuracy with the use of PRISM-based precipitation estimates of 5-km (compared to 10-km) resolution. (KEY TERMS: runoff mapping; runoff-depth; runoffYprecipitation; PRISM precipitation; evapotranspiration.) INTRODUCTION Maps of runoff (i.e., runoff-depth) have proven to be useful tools in water resource planning (e.g., Solomon et al., 1968) and in regional scientific studies (e.g., Rochelle and Church, 1987). Traditionally these maps were produced with manual methods (e.g., Domokos and Sass, 1990; Krug et al., 1990). Manual maps can be relatively accurate, but are time consuming and expensive to create. They also lack reproducibility, i.e., from the same set of data, two groups of expert mappers might produce slightly different maps. Auto- mated methods have been developed (e.g., Solomon et al., 1968; Foyster, 1975) but they have not gained wide acceptance. We have developed several automated methods of producing maps of runoff (Bishop and Church, 1992, 1995) with the goals of simplicity and reproducibility. Two of our mapping methods, which we have given the shorthand names RPRIS and ETPRIS (Bishop and Church, 1995), make use of precipitation esti- mates of 10-km resolution from PRISM (Parameter- elevation Regressions on Independent Slopes Model; Daly et al., 1994) and gaged runoff data (Krug et al., 1990). A preliminary version of PRISM estimates, based partly on digital elevation model (DEM) data of 5-km resolution and employing an improved PRISM algorithm and additional gaged data, recently has become available. We wondered if the use of these finer-scale PRISM estimates would produce more accurate runoff maps. Although intuitively one would expect greater accuracy, the 5-km pixels might prove inappropriate for the generalized regional runoff maps produced with our methods. This article examines long-term runoff maps pro- duced with PRISM estimates of 5- and 10-km resolu- tion for the eastern United States, defined here as the areal extent of the manually produced runoff map (Figure 1) of Krug et al. (1990). Our goal was to deter- mine what, if any, improvements in accuracy could be derived from using the 5-km resolution estimates. We use two of the evaluation techniques employed by Rochelle et al. (1989), and Bishop and Church (1992, 1995) in similar studies: visual comparison and a quantitative uncertainty analysis using gaged sites withheld from the creation of any of the maps exam- ined. 1Paper No. 97018 of the Journal of the American Water Resources Association. Discussions are open until October 1, 1998. 2Respectively, Geographer/GIS Specialist, OAO Corporation, U.S. EPA NHEERL, 200 S.W. 35th Street, Corvallis, Oregon 97333; Research Environmental Scientist, U.S. EPA National Health and Environmental Effects Research Laboratory, 200 SW. 35th Street, Corvallis, Oregon 97333; and Assistant Professor, Department of Geosciences, Oregon State University, Corvallis, Oregon 97331 (e-m/Bishop: bishop@mail.cor.epa.gov). JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 159 JAWRA