Study On The Inversion Method Of Leaf Area Index Of Summer Maize At Different Growth Stages Based On GF-2 Satellite

Leaf area index(LAI)is an important indicator of crop growth,which widely used in agriculture and forestry and has the important research significance in the construction of ecosystem and crop growth model and agricultural environmental monitoring.In view of the satellite payload of GF-2 autonomous satellite in China,the research of LAI inversion method has an important value for developing the potential of satellite application and promoting the application of autonomous satellite.In this study,Luancheng County of Shijiazhuang city in Hebei was used as the experimental base.LAI inversion method for GF-2 satellite was studyed based on the empirical model and the PROSAIL model in the the jointing and heading stage of summer maize,and then hybrid model combining physical model with machine learning algorithm was explored.the optimal LAI inversion model of different growth period is determined,and a perfect GF-2satellite LAI inversion technological process system is formed,comparing the inversion precision of the three models.The major several research conclusions were as follows:(1)In the empirical model,the correlation coefficients r between nine vegetation index and LAI were higher than 0.58,which can reflect the growth and changes of LAI in the jointing and heading stage of maize;exponential model of OSAVI and binomial model of ARVI obtained the highest accuracy of model building at the jointing and heading stage respectively in single variable regression model.It is found that modeling accuracy of multivariate regression model was higher than single variable regression model,but the accuracy verification shows that its inversion ability is lower than that of single variable model,which is related to the over fitting of multivariate variables and the number of samples.(2)In the physical model,The sensitivity analysis of PROSAIL model parameters shows that leaf area index and chlorophyll are most sensitive to the model in the near infrared and visible wavelengths.The results of LAI inversion based on PROSAIL model are basically consistent with the growth change at jointing and heading stage of maize leaf area index.For the PROSAIL model,R~2 of the estimated and measured LAI values of jointing and heading stage were 0.563,0.689,in contrast to the RMSE of 0.321,0.275,respectively.The PROSAIL model is 7.7%and 11.6%higher than the empirical model in the two growth period.(3)In the mixed model,the RF-PROSAIL model has the highest inversion accuracy.The determination coefficient R~2 of the estimated and measured LAI values of jointing and heading stage were 0.631,0.791,in contrast to the RMSE of 0.297,0.254,respectively.The result of BP-PROSAIL inversion is not ideal due to the image quality,but in the heading stage,the inversion effect is better,and the inversion accuracy is higher than that of the empirical model,but it is lower than the PROSAIL model.(4)In general,the LAI distribution of three kinds of inversion methods was as follows:The LAI inversion results of the empirical model and the BP-PROSAIL model were mainly distributed between 2-4,and the results of the inversion of the PROSAIL model and the RF-PROSAIL model are mainly between 1-2 in the jointing stage,and in heading stage,the inversion results of empirical model are mainly between 3.5-7.0,and the other methods results were 3-6.During the two growth periods,the LAI value is around 0 for towns,roads and buildings.To sum up,focusing on the GF-2 satellite data,the best effect of to retrieve LAI was RF-PROSAIL model in the jointing stage and heading stage of summer maize.In addition,the vegetation coverage is low,and the soil background is greatly affected in jointing stage and the empirical model(OSAVI vegetation index)is a better method when the parameter acquisition conditions of the mixed model and the PROSAIL model are not satisfied.At heading stage,with the growing vegetation,the ground coverage gradually increased and the physical model provides an effective inversion method for LAI inversion because of its strong radiation transmission mechanism.