Experiments Of Summer Maize Yield Estimation Based On Data Assimilation In Shandong Province

Real-time and accurate crop estimation is the basic information to guarantee the stability of national economic construction,which is of great significance to national economic production.This paper mainly study the summer maize yield estimation in Shandong Province through assimilating MODIS LAI into crop growth model WOFOST,so as to provide research basis and intensification of agriculture yield estimation technique.First of all,we researched the mechanism of crop growth model WOFOST,the sensitivity test of crop parameters and soil parameters for the sensitivity analysis.Then,we added the ensemble kalman filter En KF module to crop growth model WOFOST.After calibration and verification,we used the field measured data in Yucheng agricultural comprehensive experiment station to revised MODIS LAI data.We did a series of meaningful exploration experiment,including the collection size,assimilation step length,assimilation of different interval,the experiment of observation error and forecasting error.The specific research work and conclusions of the paper are as follows :(1)Through sensitivity analysis of the parameters of the crop growth model,the sensitive crop parameters and soil parameters were found,which was based on the leaf area index and the yield results of the model simulation.Among them,the most sensitive crop parameters include TBASE,AMAXTB,TSUM1,SLATB and SPAN,which are the most sensitive to leaf area index and relative to yield.The most sensitive soil parameters include SM0 and WAV,which are the most sensitive to leaf area index and relative to yield.(2)In this paper,the optimal solution of the filter window size and polynomial order number is found by adjusting the size and fitting frequency of the filter window.Through the leaf area index curve contrast,the 7 filtering smoothing of summer maize LAI curve is consistent with the actual situation,which is peaked in early July to August.Since it can well reflect the variation of the growth period of special characteristics,this study selected the time window in the filter.According to theorder filtering of different polynomials,the curve of first-order polynomial is too flat,which can not reflect the actual change of leaf area index in summer maize.The second order polynomial is the most consistent with the change of leaf area index during the whole growth period of summer maize.Third order polynomial,fourth-order polynomial and five order polynomial are obviously rush type.According to the actual situation of summer maize leaf area index,this paper selects the second order polynomial.(3)The experimental results showed that the larger the assimilation step,the closer the leaf area index curve was to the original simulation curve.The smaller the assimilation step,the larger the leaf area index was compared with the original simulation value.The smaller the assimilation step,the smaller the index of the leaf area.There is no difference from the results of the simulation of the growth period.From the output of simulation results,the leaf area index is the smallest when the step size is 16 days.The error of output is also small when the step length is 8 days.In general,the smaller the step size,the closer the simulated leaf area index to the observed value.Through the contrast to the size from different En KF time-series LAI curve collection,we found that the impact of different set of size is the distance of assimilation LAI curve and the observation data.The results show that the greater the assimilation collection,the farther away from the observed data.From the results of simulated yield,the setting of size of 50 and 90 are close to the measured results.The setting of size of 50 is most close compared with the measured value.(4)The experimental results show that the result of assimilation of different interval was totally different.We found that the time-series LAI curve of assimilation is larger than simulated value from seedling stage to jointing stage,which is more close to satellite observation data.The change tendency of the curve after jointing stage are basically identical with the original analog value.The time sequence LAI curve of assimilation is significantly smaller than the simulation value from jointing stage to spinning stage,and is close to the time sequence LAI curve of the satellite observation data.The LAI curve during the period of spinning to maturity is smaller than the simulated value in this period,which is consistent with the actual change.Accordingto the results of yield,the effect of the jointing stage to the spinning stage is the best.(5)By setting up different observation error and prediction error,some meaningful results were found.When the prediction error increases,the simulation of leaf area index is close the field measured values.When observed error decreases,the simulated leaf area index is close to the field measured value.When the observation error and prediction error are set to 0.1,the error of simulated summer maize yield is usually small.(6)The summer maize yield of assimilation is better than the original simulation yield at most of the sites.They are Liaocheng,Jiyang,Taian,Zibo,Hanting,Gaomi,Jiaozhou,Heze agrometeorological station.But there are individual relative errors of assimilation of summer maize yield increase at some sites.They are Laiyang and Jining agricultural meteorological station.The original simulation error is relatively small,and the relative error is increased after assimilation at these sites,but within an acceptable range.After assimilation,in general,the relative error of summer maize yield decreases at most of the sites,which means summer maize yield estimation precision is improved after assimilation,and the sites of large simulated error improved significantly.