Identification of spatially distributed parameters of hydrological models using the dimension-adaptive key grid calibration strategy

Distributed hydrological models have been developed to describe temporal and spatial variations. However, it is difficult to calibrate the spatially heterogeneous parameters because of dimensional issues and spatial uncertainty. This study aims to propose a dimension-adaptive key grid (DAKG) calibration strategy to identify spatial parameters of distributed hydrological models. The method uses sensitivity analyses to find important model parameters, and significantly fewer grid points are then selected as representative grids to reduce the dimensionality of parameters. Apart from the observed outlet streamflow, grid actual evaporation, and grid runoff are used for the multi-objective calibration framework. The Xiangjiang basin, the Yuanjiang basin, and the source region of the Yellow River in China are used as case studies, where the Variable Infiltration Capacity (VIC) land surface models have been calibrated. The results indicate that: (1) The proposed DAKG calibration strategy improves performance in terms of accuracy and model complexity for hydrological simulation. (2) The spatial distributions of the shape parameter B and the thickness of the second soil layer Depth2 are easy to be identified, where the spatial distribution of Depth2 is the most positively correlated with that of elevation. (3) The fusion of the runoff and evaporation products helps to reduce the uncertainty of the distributed parameters. Overall, the DAKG calibration strategy is an effective technique for parameter identification in distributed hydrological models.

Automated summary

The study introduces a new calibration strategy, DAKG, to improve the accuracy and reduce the complexity of distributed hydrological models. Important model parameters are identified through sensitivity analysis, and fewer grid points are selected to reduce parameter dimensionality.