Analysis And Correction Of Sensitivity Parameters Of Cotton2k Model

The soil salinization of cotton fields in Xinjiang is usually high and often faces drought stress,especially in southern Xinjiang.Cotton2k model can better simulate cotton growth and production under arid and semi-arid conditions,and consider the effect of soil salinity on cotton growth.However,due to different regions,crop varieties,field management level and other factors,the accuracy of the model will be greatly reduced.Therefore,in the regional study of the model,in order to make the model better serve the local area,we must correct the model.Parameter sensitivity analysis can effectively screen out the parameters that have a greater impact on the output of the model,and only adjust these parameters during model calibration,and can effectively reduce the amount of data processing.Taking"Zhongmian 619"as the research object and Alar city as the research area,this paper used three different typical weather types(general year type,cold year type and hot year type)to simulate cotton growth,combined with Extended Fourier amplitude sensitivity test,EFAST)and multi-objective optimization algorithm were used to analyze,screen and optimize the sensitivity of 46 cotton variety parameters based on cotton observation data(leaf dry matter,stem dry matter,lint yield)in Alar city in two years.So as to evaluate the simulation accuracy of cotton2k model in the study area,and provide an effective basis for model regionalization.The results are as follows:(1)Based on the extended fast method,the sensitivity indexes of model parameters in different typical years to different output variables are calculated,and the differences between the sensitivity indexes are analyzed.The analysis of leaf dry matter showed that varpar01(the effect of density on growth)had the greatest effect on leaf dry matter in cold and hot years,while varpar35(fruit branch development)had the highest contribution rate in general years;The analysis of stem dry matter showed that the top three sensitive parameters were the same under different typical weather conditions,varpar01(the effect of density on growth)was the most sensitive parameter,followed by varpar15(stem growth after square boll spreading)and varpar12(stem growth after square boll spreading);The analysis of lint yield showed that the most sensitive parameters were varpar43(abscission strength,C stress),varpar49(possibility of Boll abscission)and varpar42(temperature effect,yield)under different typical weather conditions(general year,colder year and hotter year).(2)Considering the results of global sensitivity analysis,10 parameters with strong sensitivity are selected and optimized by multi-objective optimization algorithm.It is found that 4 parameters are improved and 6 parameters are decreased.The change of parameters is mainly related to cotton varieties,soil conditions,climate and environment.(3)The optimized parameters showed that the model was the best model for cotton yield simulation,and the relative root mean square error(n RMSE)was less than 10.04%,root mean square error(RMSE)was between 172.65 kg/hm~2 and 174.38 kg/hm~2.Under different irrigation methods,the relative error(ARE)was less than 13.78%;The results of stem dry matter simulation were the second,n RMSE were less than19.1%,RMSE was 276.65 kg/hm~2and 297.24 kg/hm~2,and were less than 25.8%under different irrigation methods;The results of the simulation of dry matter of leaves ARE the worst,and the reliability of the simulation results is of medium level.This may be because the model does not consider the effect of cotton pruning and topping.For the simulation of the date of the birth period,except for the error of 60%of the catchy period,the other simulation results of the date of birth are within a reasonable range,and the model can provide the main prediction of the growth period.The research provides a scientific theoretical basis for the regional application and model correction of the cotton2k model.