Calibration for Improving the Medium-Range Soil Forecast over Central Tibet: Effects of Objective Metrics’ Diversity

The high spatial complexities of soil temperature modeling over semiarid land have challenged the calibration–forecast framework, whose composited objective lacks comprehensive evaluation. Therefore, this study, based on the Noah land surface model and its full parameter table, utilizes two global searching algorithms and eight kinds of objectives with dimensional-varied metrics, combined with dense site soil moisture and temperature observations of central Tibet, to explore different metrics’ performances on the spatial heterogeneity and uncertainty of regional land surface parameters, calibration efficiency and effectiveness, and spatiotemporal complexities in surface forecasting. Results have shown that metrics’ diversity has shown greater influence on the calibration—predication framework than the global searching algorithm’s differences. The enhanced multi-objective metric (EMO) and the enhanced Kling–Gupta efficiency (EKGE) have their own advantages and disadvantages in simulations and parameters, respectively. In particular, the EMO composited with the four metrics of correlated coefficient, root mean square error, mean absolute error, and Nash–Sutcliffe efficiency has shown relatively balanced performance in surface soil temperature forecasting when compared to other metrics. In addition, the calibration–forecast framework that benefited from the EMO could greatly reduce the spatial complexities in surface soil modeling of semiarid land. In general, these findings could enhance the knowledge of metrics’ advantages in solving the complexities of the LSM’s parameters and simulations and promote the application of the calibration–forecast framework, thereby potentially improving regional surface forecasting over semiarid regions.