Application Of The Conditional Nonlinear Optimal Parameter Perturbation Method In Common Land Model

Common Land Model (CoLM) is the widely applied and advanced land model in the current world. Itcarefully considers the land ecological process and the hydrological process and preferably describes thesoil, the vegetation, the snow, the transmission of energy and moisture between the atmosphere. A largeamount of experiments show that CoLM has a good simulating ability in the areas of China. The extensionof conditional nonlinear optimal parameter perturbation method is used in a nonlinear dynamical system toobtain the parameter perturbation, which could make the objective function get the optimal value at theforecast moment with the given parameter perturbation restraint.In this paper, we choose sand, the percentage of sand in the soil, and clay, the percentage of clay in thesoil, as the optimal parameters and take the extension of conditional optimal parameter perturbation methodto optimize the target parameters in CoLM. In this paper, we choose35.23750N,118.1250E in theNorth China Plain and48.57050N,1200E in Inner Mongolia as the trial areas to study the impact ofthe percentage of sand and clay of the shallow soil in CoLM on the ability to simulate the shallow soilmoisture, with NMC Reanalysis6-hourly surface fluxes data (dataset Ⅰ) and NCEP/DOE AMIP-Ⅱ6-hourly Reanalysis Gaussian Grid data (dataset Ⅱ). For this purpose, we separately design experiment Ⅰ Ⅱ Ⅲ and Ⅳ to optimize the target parameters in CoLM and predict the shallow soil moisture in thefollowing month with the optimal parameters. The main numerical results are concluded as follows:(1) Whether the trial area is the North China Plain or the Hulunbuir Steppe, the optimal parameters,which are attained with the extension of conditional nonlinear optimal parameter perturbation method,could make CoLM simulate the shallow soil moisture better, and the optimal parameters attained by thedouble-parameter optimal experiment could make CoLM simulate the shallow soil moisture the best in theoptimization slot; predicting the shallow soil moisture in the prediction slot with the optimal parametersattained in the optimization slot, the results show that the optimal parameters could significantly improvethe prediction results of CoLM at the stage of prediction.(2) As this paper employs two different datasets, it is necessary to compare the corresponding numerical results. The results show that the optimal results attained with dataset Ⅱ are more reasonablethan dataset Ⅰ. The former results still show that the optimal parameters attained by the double-parameteroptimal experiment in the optimization slot could make CoLM simulate the shallow soil moisture the bestin the prediction slot, but the latter results do not. As dataset Ⅱ are more accurate than dataset Ⅰ, it isappropriate to say that the more accurate the atmospheric forcing data and observation data are, the moresignificant the results of optimization may be.