Estimation Of Tobacco Chlorophyll Content Based On Sentinel-2 Data And Vegetation Index

Tobacco is one of the important economic crops in our country.Scientific and refined management methods are essential to improve the quality of tobacco.The growth and quality supervision of tobacco are mainly based on tobacco biochemical parameters.As one of the main biochemical parameters of tobacco,chlorophyll is closely related to photosynthesis.To this end,this paper takes tobacco in Xuanzhou District,Xuancheng City,Anhui Province as the research object,and uses near-ground hyperspectral data,Sentinel-2 satellite images and tobacco chlorophyll content as basic data to established a tobacco chlorophyll content estimation model,realized the estimation of tobacco chlorophyll content in different growth periods on a regional scale.The main research contents and conclusions are as follows:(1)The canopy reflectance spectra of different chlorophyll contents were simulated by PROSAIL5B.According to the spectral characteristics of reflectance spectra on chlorophyll content,the normalized triangular vegetation index(BTVI),ratio vegetation index(NTVI),short-near difference vegetation index(NVI),composite vegetation index(REVI)were constructed.The correlation between BTVI,NTVI,NVI,REVI and chlorophyll content was analyzed,and the correlation coefficients(R)between BTVI,NTVI,NVI,REVI and chlorophyll content were-0.9636,-0.9203,0.9249,0.9919,all reaching 0.001 Very significant correlation level,which can be used to estimate the chlorophyll content.(2)Firstly,the method of out-of-bag Elastic Net(OOBEN),out-of-bag Principal Component Analysis(OOBPCA)was used to optimize the sensitive factors.Then,the tobacco chlorophyll content estimation models OOBENRF and OOBPCARF were constructed with the sensitivity factors optimized by the OOBEN and OOBPCA methods.The R2 of model validation set was 0.7316 and 0.5190 respectively,and RMSE was 1.3369 and 1.7897 respectively.Compared with the OOBPCARF model R2,OOBENRF increased by 40.96%,than that of OOBPCARF,indicating that the optimal sensitive factor of the OOBEN method was the appropriate factor for estimating the chlorophyll content of tobacco.In order to further analyze the performance of the newly constructed vegetation index to estimate the tobacco chlorophyll content,OOBENI was used as the sensitive factor(OOBEN preferred sensitive factor to remove the newly constructed vegetation index REVI,BTVI,NVI),and the tobacco chlorophyll content estimation model OOBENRF1 was constructed.The results showed that compared with OOBENRF1 model R2,OOBENRF increased by 23.52%,and RMSE decreased by 18.86%.The integration of new indexes REVI,BTVI,and NVI improved the accuracy of tobacco chlorophyll content estimation.(3)Based on sentinel-2 image,the estimation models M1,M2,M3 and M4 of tobacco chlorophyll content were established by using suitable factors,including prosperous period(may 2019),mature period(July 2019),prosperous period(may 2020)and three periods of data(may 2019,July 2019 and may 2020),the R2 of the model was 0.8013,0.7126,0.8067 and 0.8338 respectively,which indicates that the method was feasible for estimating chlorophyll content.In M1,M2 and M3 models,the accuracy of M1 and M3 models was higher than that of M2,which indicated that the accuracy of the models built in prosperous period(may 2019 and may 2020)was higher than that in maturity period(July 2019).The reason for this situation may be that the tobacco leaves gradually turned yellow and the spectral characteristics of chlorophyll were weakened in maturity period.Based on the constructed M1,M2 and M3 models,the spatial distribution map of tobacco chlorophyll content in different growth stages was drawn by using sentinel-2 image,and the changes of tobacco chlorophyll content in different growth stages were analyzed.It was concluded that the estimated chlorophyll content was basically consistent with the chlorophyll content in the actual growth process of tobacco.