How well do terrestrial biosphere models simulate coarse-scale runoff in the contiguous United States?

Significant changes in the water cycle are expected under current global environmental change. Robust assessment of present-day water cycle dynamics at continental to global scales is confounded by shortcomings in the observed record. Modeled assessments also yield conflicting results which are linked to differences in model structure and simulation protocol. Here we compare simulated gridded (1° spatial resolution) runoff from six terrestrial biosphere models (TBMs), seven reanalysis products, and one gridded surface station product in the contiguous United States (CONUS) from 2001 to 2005. We evaluate the consistency of these 14 estimates with stream gauge data, both as depleted flow and corrected for net withdrawals (2005 only), at the CONUS and water resource region scale, as well as examining similarity across TBMs and reanalysis products at the grid cell scale. Mean runoff across all simulated products and regions varies widely (range: 71 to 356mmyr−1) relative to observed continental-scale runoff (209 or 280mmyr−1 when corrected for net withdrawals). Across all 14 products 8 exhibit Nash–Sutcliffe efficiency values in excess of 0.8 and three are within 10% of the observed value. Region-level mismatch exhibits a weak pattern of overestimation in western and underestimation in eastern regions—although two products are systematically biased across all regions—and largely scales with water use. Although gridded composite TBM and reanalysis runoff show some regional similarities, individual product values are highly variable. At the coarse scales used here we find that progress in better constraining simulated runoff requires standardized forcing data and the explicit incorporation of human effects (e.g., water withdrawals by source, fire, and land use change).