Implications of model selection: a comparison of publicly available, conterminous US-extent hydrologic component estimates

Abstract. Spatiotemporally continuous estimates of the hydrologic cycle are often generated through modeling, reanalysis, or remote sensing (RS) methods and commonly applied as a supplement to, substitute for, in situ measurements when observational data sparse unavailable. This study compares precipitation (P), actual evapotranspiration (ET), runoff (R), snow water equivalent (SWE), soil moisture (SM) from 87 unique sets by 47 models, reanalysis sets, products across conterminous United States (CONUS). Uncertainty between component was shown to be high western CONUS, with median uncertainty (measured coefficient variation) ranging 11 % 21 for P, 14 26 ET, 28 82 R, 76 84 SWE, 36 96 SM. lower eastern medians 5 13 22 53 63 42 83 Interannual trends 1982 2010 show common disagreement Correlating fluxes stores against remote-sensing-derived poor overall correlation CONUS ET SM estimates. Study results that can substantial, sometimes exceeding magnitude themselves. The authors conclude multimodel ensembles not only useful but fact necessity accurately representing research results. Spatial biases model values targeted efforts arid semiarid water-limited regions warranted, greatest emphasis on storage components, better describe complexities terrestrial system reconcile disagreement.