Comparison and evaluation of multiple land surface products for the water budget in the Yellow River Basin
Multiple global datasets simulated by different land surface models (LSMs) are useful for providing information to decision-makers on the water balance in different subareas of a river basin. As a large basin covering a vast area, the Yellow River Basin (YRB) represents a typical case to evaluate the water cycle components in LSM products. The main objective of this study is to evaluate and compare multiple LSM products with respect to precipitation, runoff, evapotranspiration (ET) and terrestrial water storage anomaly (TWSA). The datasets analyzed include different Global Land Data Assimilation System (GLDAS) products, ERA-Interim/Land datasets, and augmented Noah LSM products with multiple parameterization options (Noah-MP) driven by different GLDAS forcing datasets. The evaluation is conducted with reference to gauge-based precipitation datasets, reconstructed natural streamflow and Gravity Recovery and Climate Experiment (GRACE)-derived TWSA. Several statistical metrics and the partitioning of runoff and ET were emphasized. The results show that the products generated diverse spatial patterns for the water cycle components. The ERA-Interim/Land data present the best performance almost for all water cycle components. The three models (variable infiltration capacity (VIC), Mosaic and community land model (CLM)) driven by GLDAS version 1 generated different but large systematic biases, and the Noah model performs the best among the four models. The Noah simulation of GLDAS version 2.0 (GLDAS-2.0) and the Noah-MP simulation with the Princeton Global Forcing also have systematically large biases for runoff. The different Noah-MP simulations perform better for runoff (although overestimated) than other models used in GLDAS, but present the poorest correlations of TWSA to GRACE-derived TWSA. The ensemble averages of multiple products perform the best for precipitation and ET than any other individual product but not for runoff.