Assessing snow simulation performance of typical combination schemes within Noah-MP in northern Xinjiang, China

Abstract

The performance of the Noah land surface model (LSM) with multi-parameterization options (Noah-MP) in simulating snow depth was evaluated in northern Xinjiang, China. A total number of 13,824 Noah-MP physics-ensemble simulations were conducted at the Altay site by combining different parameterization schemes of physical processes while disregarding the uncertainties of forcing data and model parameters. The natural selection approach and Tukey’s test, which are two sensitivity analysis methods, were used to analyze the sensitivity of snow to parameterization schemes. Then, the uncertainty intervals of the ensemble simulation experiments were compared. According to the results of the sensitivity and uncertainty experiments, snow depth could be simulated by three typical combination schemes at the regional scale: the longest snow melting time scheme (LT), the shortest snow melting time scheme (ST) and the default combination scheme (DT). Observation data of snow depth from thirty-nine meteorological stations in northern Xinjiang were used to evaluate the snow simulation performance of typical combination schemes. The simulation performances of the three typical combination schemes were examined and compared in groups that were divided according to elevation and land cover. The results demonstrated that the simulation results of snow depth and snow water equivalent (SWE) were sensitive to four of the eleven physics options within Noah-MP. The exclusion of the parameterization schemes that notably reduced the simulation performance in the sensitive physical processes can significantly reduce uncertainty. Snow simulation performances of three typical combination schemes were diverse in northern Xinjiang, China; no single scheme performed best at all sites, but the length of the snow melting phase exhibited the best performance.