Abstract Hydrological models are typically calibrated using historical ground‐based streamflow observations to constrain model uncertainty. However, such a calibration strategy can lead to unrealistic model parameters and is not applicable in data‐sparse regions where streamflow observations are unavailable. Motivated by this limitation, a novel model calibration approach that leverages remote sensing (RS) soil moisture retrievals has been recently developed based on the assumption of perfect rank correlation. It calibrates model parameters by maximizing the rank correlation between RS pre‐storm soil moisture and modeled storm‐scale runoff coefficient (i.e., the ratio of runoff to precipitation). However, this calibration approach has so far been limited to basin‐scale applications and evaluated only in terms of storm‐scale runoff coefficients rather than actual streamflow simulations. Here, we extend the calibration approach to a grid‐by‐grid parameter calibration framework within the Variable Infiltration Capacity (VIC) model and incorporate a routing scheme to enable streamflow simulation. The model simulations are evaluated against independent ground‐based streamflow observations and other hydrological variables, including ground‐based soil moisture and RS‐based terrestrial water storage (TWS) and evapotranspiration (ET). Results show that the RS‐based calibration approach produces VIC streamflow simulations comparable to the conventional calibration using ground‐based streamflow in semi‐humid and humid basins—achieving a mean Nash‐Sutcliffe coefficient above 0.68. In addition, the calibration method leads to improvements in both VIC TWS and ET estimates (with average correlation increments of 0.06 and 0.07, respectively). The study offers valuable insights for streamflow modeling in data‐sparse regions.
Feng et al. (2026) studied this question.