A comparison of mapped estimates of long-term runoff in the northeast United States

We evaluated the relative accuracy of four methods of producing maps of long-term runoff for part of the northeast United States: MAN, a manual procedure that incorporates expert opinion in contour placement; RPRIS, an automated procedure based on water balance considerations; PnET-II, a physiologically based model of carbon/water balance in forests; and MAPSS (Mapped Atmosphere-Plant Soil System), a rule/process-based vegetation distribution/water balance model. Our goal was to confirm the accuracy of the modeling and mapping procedures, and to see if any improvements to the models and methods might be suggested. In our analyses, we compared contour maps derived from the four methods both qualitatively (visual inspection) and quantitatively (raster overlay and uncertainty analysis). The manual and automated (RPRIS) methods gave the best results. Our analyses suggest that methods directly integrating gaged runoff data (i.e. MAN and RPRIS) provide the best results under current climatic conditions. For predicting runoff under altered conditions, e.g. climate change, the existing models studied here (i.e. PnET-II and MAPSS) hold significant promise.