Design Of Remote Sensing Assimilation Strategy For Pear Growth Simulation

Pears play a significant role in rural economic development as an important fruit industry.This paper establishes a set of pear growth simulation and remote sensing assimilation methodology under ideal growing conditions based on the research background of fragrant pear(Pyrus nivalis)in southern Xinjiang.It provides objective quantitative analysis tools for pear planting strategies,agricultural management,and yield evaluation.In terms of experimental data,7-year-old slender spindle-shaped pear trees at the full-fruit stage in southern Xinjiang were selected as the research objects.The schema design for agricultural management(pollination,fertilization,pruning,irrigation,rotary tillage,pesticides,fruit picking)is based on two aspects:data selection and measurement scheme(phenological stages,leaves,fruits,branches,soil,remote sensing images,meteorology).Simultaneously,the measured data is preprocessed appropriately.Two automated leaf area measuring and statistical assessment systems for fruit tree vegetation were proposed.The data collected provided a basic guarantee for the following process.In terms of remote sensing inversion,LAI was selected as the inversion object.DWT and MSC were used to calibrate the canopy spectra acquired by ASD.It is demonstrated that the vegetation index estimated by UAV fulfills the accuracy standards by comparing the calibrated spectrum with UAV remote sensing data.The ROI data set is then retrieved experimentally after clipping by using a threshold method to preprocess the remote sensing images and reduce the disruption of environmental factors.Finally,NDVI was used as a model variable in four periods by PLSR,SVR,RF,and XGBoost to construct LAI regression models.For the above methods,XGBoost had a better fitting degree for LAI in all four periods,with an average R~2 of 0.743 and an average RMSE of 0.163.The LAIs from May to August were 1.51,2.45,3.39,and 3.05,respectively,by using XGBoost to invert the clipped remote sensing image area.In terms of crop model,the pear growth model was constructed based on Wofost.The DVS and BBCH scales were combined to create a phenological stages scale appropriate for pear trees,and TDWI was re-defined as the dry weight of the original roots and branches.The sensitivity of state variables TSUM1,TSUM2,TSUMEM,SPAN,Q10,TDWI,and output variables DVS,LAI,TWSO,TWLV,TWST,TAGP,was analyzed.According to the findings of the sensitivity analysis,previous knowledge,and actual measurement data,the model was constructed using phenological stages,energy flux,and calibration sequence of model initialization.The results showed that the simulated phenological stages,vegetation index,and aboveground biomass conformed to the actual growth and development of the pear.For data assimilation of remote sensing inversion LAI and Wofost growth model,the forcing technique,En KF updating method,and SUBPLEX calibration method were used,and the results were compared.The result showed that the aforesaid assimilation methods improved the fitting degree of partial actual growth when compared to the original model simulation results.For three assimilation methods,when the forced technique is applied,the results are less suited to the real measurement data.When there are more observation sites,SUBPLEX is better.When the number of observation sites is reduced,En KF becomes more accurate.In conclusion,this paper used the assimilation method as a bridge.The benefits of the crop model’s strong biological mechanism and long time series were merged with the advantages of multispectral UAV’s great phenotypic parameter of plants estimate accuracy.The accuracy of the pear tree growth simulation model under ideal conditions was improved,according to the results.