| Crop growth models have become one of the most powerful tools in agricultural research because they can dynamic reflect mechanism of the crop growth processes. However, the description of some processes for agro-ecosystem in most crop growth models are oversimplified, which results in many more parameters and directly reduces its simulation precision. Assimilating multivariate data of remote sensing and ground observation data provides an effective way to improve simulation ability of crop growth model.In this paper, the assimilation method of observation data and crop growth model was mainly developed based on the Price and the Downhill-Simplex algorithm for summer maize in Gucheng, Hebei Province and Zhengzhou, Henan Province. The sensitivity of model parameters to state variables in WOFOST was firstly analyzed and the constraint of remote sensing and ground observation data to sensitive parameters was investigated. It was then compared that the assimilation effect of observation data and crop growth model based on the Price method and the Downhill-Simplex method. The assimilation effect of remote sensing information in different developmental stages and crop growth model was finally discussed. The main outcomes in this study are as follows:(1) Assimilation method of crop growth model and observation data was established based on determination of the parameters to be optimized by using the Price and the Downhill-Simplex algorithm. Correctness validating for assimilation method of crop growth model and observation data was firstly carried out. Sensitive parameters and initial value of the state variables in crop growth models were then selected by using the Price and the Downhill-Simplex algorithm. Subsequently, constrained parameters of different observed data were obtained through constrained analysis based on the Price algorithm, which was defined according to the relation of goodness of fit (QT) and optimization results of parameters. Finally, constrained parameters can be viewed as the parameters to be optimized for the Price algorithm and the parameters to be optimized were selected according to sensitivity, physiological meanings and the principle of as few as possible for the Downhill-Simplex algorithm. The optimal values of each parameter were got by the two optimization algorithms and the crop model was validated by measured data. So assimilation method of crop growth model and observation data was achieved.(2) Sensitive parameters of different state variables and constrained parameters of different observation data were obtained in WOFOST based on the Price and the Downhill-Simplex algorithm. Sensitivity analysis showed that both WSO and TAGP were the most sensitive to Specific leaf area at jointing stage to tasseling stage, relative maintenance respiration rate of storage organs and maximum leaf CO2assimilation rate at later stage of milking maturity, while the most sensitive parameters of the LAI were the initial total crop dry weight, initial specific leaf area and initial maximum leaf CO2assimilation rate under potential production level. The sensitive parameters of each state variable under water stress production level revolved ones related to soil water balance. Such as maximum daily increase in rooting depth, initial amount of available soil water in the root zone soil and maximum moisture content in topsoil as well.Constrained analysis based on The Price algorithm showed that observational data of summer maize can mainly constrain initial total crop dry weight, specific leaf area at different development stages, initial maximum leaf CO2assimilation rate,life span of leaves growing at35Celsius, initial amount of available water in total root zone, maximum daily increase in rooting depth, etc. Parameters constrained by LAI, WSO, TAGP and soil moisture content(SM) were not exactly the same each other.(3) The parameters to be optimized were obtained based on sensitivity analysis and constrained analysis. The optimal values of those parameters were estimated and the model was validated by using observed data. Thus the assimilation of ground observation data and crop growth model was achieved based on the Price and the Downhill-Simplex algorithm. Moreover, the simulated errors of the most state variables were smaller in assimilation of observation data and crop growth models by using the Price algorithm. Meanwhile, the process of determining the parameters to be optimized based on the Price algorithm was more theoretical. Therefore, the Price algorithm was determined as suitable optimization method to assimilating observation data and crop growth model, which provides the basis for the assimilation of remote sensing data and crop growth model.The values of crop and soil parameters in Gucheng and Zhengzhou were obtained based on assimilation of observation data and WOFOST model. The simulation test showed that WOFOST can reflect the process of summer maize development, growth and yield formation.(4) The LAI retrieved by remote sensing data (1km*1km) was calibrated by using measured LAI. The parameters to be optimized were determined by using the sensitivity of constrained parameters and the optimization effect. Assimilation of remote sensing information and crop growth model was then achieved in filed scale.The assimilation of remote sensing information in different development stages and crop growth model showed that the error of the simulated LAI was reduced with the increase of the number of remote sensing image. Meanwhile, remote sensing information in different development stages produced different assimilation effect. Remote sensing data during milking maturity stage to maturity stage was more critical for the simulated LAI. However, the data during sprout of seedling to the initial stage of jointing was more significant for the simulated WSO and TAGP and the data during the initial stage of jointing to tasseling stage is more important for the simulated WLV and WST.In this study, stepwise selection method based on sensitivity analysis and constrained analysis results in the fewer and more suitable parameters to be optimized. Meanwhile, the comprehensive action of several kinds of observation data was considered in assimilation study. Thus assimilation of measured data and crop growth model achieved better results. This study laid a foundation to build regional model of crop growth model of summer maize based on remote sensing data and to develop the quantitative evaluation of crop growth. |
PDF Full Text Request