The Project for Intercomparison of Land-surface Parameterization Schemes PILPS phase 2 c Red–Arkansas River basin experiment: 3. Spatial and temporal analysis of water fluxes

(cid:14) . The water-balance components of 16 Soil–Vegetation Atmospheric Transfer SVAT schemes were evaluated by comparing predicted and observed streamflow, predicted evapotranspiration and evapotranspiration inferred from an (cid:14) . atmospheric moisture budget analysis, and soil moisture storage changes for a seven-year period 1980–1986 using data from the Red–Arkansas River basins of the Southern Great Plains of the USA. The evaluations support the following (cid:14) . suggestions: a The mean annual runoff of all models follows, at least generally, the strong climatic East–West gradient of (cid:14) . precipitation, although most models predict too much runoff in the dry part of the basin. b The mean monthly storage change tends to be underestimated, even though all models capture reasonably well the seasonality of the evapotranspiration. (cid:14) . c The wide range of conceptualizations used for generation of surface and subsurface runoff strongly affect runoff generation on seasonal, and shorter, time scales. Model responses to summer precipitation ranged from almost no summer (cid:14) . (cid:14) . runoff one model to the more common situation of persistent overprediction of summer runoff, especially in the driest (cid:14) . (cid:14) . part of the basin. d All models tended to underpredict evapotranspiration in summer and overpredict in winter. e Model-derived mean seasonal cycles of changes in soil moisture storage are qualitatively similar to those inferred from observations, but most models do not predict the decrease in April soil moisture storage and the increase in October that is inferred from observations. q 1998 Elsevier Science B.V. All rights reserved.