Hydrologic Model Calibration With Remote Sensing Data Products in Global Large Basins

Parameters of hydrologic models are typically calibrated using in situ runoff observations. This calibration strategy cannot be implemented when runoff observations are not available. Additionally, the high accuracy of runoff simulation does not guarantee high simulation accuracy of other hydrologic variables, such as evaporation (ET) and terrestrial water storage change (TWSC), leading to uncertainties in hydrologic predictions. To overcome the shortcomings of the runoff‐based calibration strategy, ET and TWSC products from remote sensing were used to design parameter calibration schemes in 59 large basins worldwide. Runoff observations were used to evaluate the effects of different calibration schemes. Results show that single‐objective calibration schemes using the ET or TWSC products for parameter calibrations cannot obtain acceptable runoff simulations. Multi‐objective calibration schemes using the combination of ET and TWSC products for parameter calibrations had better runoff simulation accuracy. The multi‐objective calibration scheme using the combination of ET product adjusted by the Budyko formula and TWSC product as the calibration objective had the best runoff simulation accuracy. It is necessary to adjust ET products before using them for parameter calibrations of hydrologic models. The Budyko formula can be used to reduce the bias in remote sensing ET products, thereby providing better constraints on hydrologic simulations. We also find that different precipitation products used as inputs to hydrologic models can significantly affect the accuracy of runoff simulations. The results are expected to promote the application of remote sensing products in hydrologic simulations and predictions. Plain Language Summary Traditionally, when we use a hydrologic model to simulate runoff in a basin, we typically use the historical runoff observations to optimize the parameters of the hydrologic model. This traditional strategy cannot be conducted when runoff observations are not available in some basins, especially in less developed countries. Over the past two decades, with the development of satellite observations, a large number of hydrologic variable products based on remote sensing have been generated. Parameter optimizations of hydrologic models based on these remote sensing products is a promising solution to overcome the shortcomings of the runoff‐based calibration strategy. In this study, we compared different combinations of remote sensing produ