Study On Grassland Biomass Retrieval Method Based On Cooperative Multi-Source Satellite Remote Sensing Data

As an important part of the natural ecosystem in Qinghai Province,grassland provides forage resources needed by herdsmen for animal husbandry,and occupies an important position in local animal husbandry.In recent decades,with the development and expansion of local animal husbandry economy,the stability of grassland ecology has been continuously damaged,which hinders the sustainable development of animal husbandry.Remote sensing technology has advanced space-time monitoring ability.How to use this technology to accurately estimate grassland biomass and timely learn the changes of grassland yield is of great significance to the scientific planning and management of animal husbandry,which is an important topic to be studied urgently.Taking the grassland of Qinghai Province as the research object,this study explores the feasibility of biomass estimation methods of different satellites and models based on Multi-source Satellite remote sensing data,and cooperates with multi-source remote sensing satellites to extend the biomass estimation method to the grassland of Qinghai Province.Finally,data assimilation technology is used to integrate remote sensing data and crop growth model to realize the dynamic estimation of total grassland biomass.The research contents and conclusions of this thesis are as follows:(1)According to satellite remote sensing data and ground measured data,this thesis explores the applicability of different remote sensing inversion methods in multi-source satellite remote sensing data.Firstly,the correlation between vegetation index and measured biomass data is analyzed,so as to select the appropriate vegetation index for different satellite remote sensing data.According to the selected vegetation index,the grassland biomass is retrieved by remote sensing using empirical model and physical model respectively,and the accuracy of the inversion results of empirical model(multiple linear regression and random forest regression)and physical model(PROSAILH model)under Multi-source Satellite remote sensing data is verified.The results show that PROSAILH model is more suitable for grassland biomass inversion in a large area under long-time series than empirical model.(2)According to the research results of the previous stage,the remote sensing inversion of biomass is realized in the grassland area of Qinghai Province by using the surface reflectance products of Landsat 8 satellite and Sentinel-2 satellite,PROSAILH radiation transfer model and look-up table algorithm,so as to produce the monthly grassland biomass products from September 2019 to August 2020,and verify the accuracy of temporal and spatial distribution of biomass.In order to solve the problem of low inversion efficiency caused by too large image data in the process of regional mapping,the image and model parameters are reconstructed and transformed into data matrix to improve the efficiency of batch inversion algorithm.Finally,the correlation between meteorological factors and biomass is analyzed by using the meteorological statistical data of Qinghai Province(temperature,precipitation and sunshine hours).(3)Taking Qinghai Lake Basin and Wutumieren grassland as the study area,based on data assimilation method,remote sensing data and crop growth model are integrated to realize the dynamic estimation of total grassland biomass.Firstly,the leaf area index(LAI)is retrieved by prosailh model as the observed LAI data in the subsequent data assimilation method.Secondly,the sensitivity analysis and model localization of crop growth model are carried out,and the observed LAI data are assimilated into WOFOST(World Food Studies)model by using ensemble Kalman filter(En KF)data assimilation algorithm.Finally,the accuracy changes of the model in Estimating Grassland total biomass before and after data assimilation were analyzed.The results show that there is a good correlation between the model simulation value and the measured data,and the comparative analysis of the model accuracy before and after assimilation(R~2 increased from 0.76 to 0.83,RMSE decreased from 0.19 kg/m~2 to 0.15 kg/m~2).The integration of remote sensing data can effectively solve the influence of crop growth model error and further improve the accuracy of the total biomass simulated by the model.