The Numerical Inversion Of Soil Hydraulic Parameters Based On PEST

The estimation of the hydraulic parameters is an important part of the application of the agricultural hydrological models.In recent years,numerical inversion methods have been widely used to estimate soil hydraulic parameters.Although the method can avoid the time consuming and laboriousness of the artificial trial and error method,there are some problems such as local convergence and "parameter equifinality",and the accuracy of the obtained result is still debated.Therefore,it has great significance for the application of numerical inversion methods and the improvement of model accuracy to study the numerical inversion of soil hydraulic parameters,and to evaluate the optimized parameters and simulation results in many aspects.In this paper,the soil hydrological model SWAP(Soil-Water-Atmosphere and Plant)is coupled with the numerical inversion program PEST.Based on the three soil profiles of Zhangjiawan Experimental Station in Tongzhou District of Beijing,the soil water content and actual evapotranspiration were obtained by the positive simulation of the SWAP model.These values are taken as observations for inversion simulation and are considered to be no errors.Secondly,the single objective optimization of numerical inversion has been explored.The uncertainties of the estimated parameters and simulation results were evaluated from the three aspects,including soil water content observation error,observation frequency and spatial arrangement of observation point.The results show that when the soil hydraulic parameters are estimated by soil moisture only,the soil water content observation error should be within 37.5%,and the uncertainty of soil hydraulic parameters increases with the increase of soil moisture observation error.The inversion of soil hydraulic parameters is not sensitive to the changes of the observation frequency.The observed frequency of soil water content is preferably not more than four weeks.The fitting precision of the soil water content has been improved by increasing the number of observations point.The simulation results are the best when each layer adds a point.But if the conditions are not allowed,the contribution for the improvement of inversion accuracy is greater when the observation point was added to the first layer.On this basis,considering the unreliability of a single variable,the multi-objective optimization of numerical inversion also has been discussed.The actual evapotranspiration(ETa)was added as another inversion variable based on the application of soil water content.The fitting results of inversion variables that obtained by inversion of soil water content(?),soil water content and actual evapotranspiration(ETa)and actual evapotranspiration(ETa)respectively were compared under different crop systems.Results indicate that under the assumption of realistic conditions(biweekly soil moisture observations with a noise level of 10 volumetric percent for one observation point of each soil layer),the results obtained by ETa inversion are the worst especially for the simulation of soil moisture.The addition of ETa does not improve the simulation accuracy of the inversion variables obviously,and there even has reverse effect in the rotation.This suggests that it is sufficient to use only soil moisture to estimate parameters in most cases,and ETa has little effect on the improvement of soil water content and ETa.