Inversion Of Rice Leaf Chlorophyll Content From Sentinel-2 Satellite Data

With the increasing demand for food production in human society,the introduction of modern technology to increase food production has become very important.Rice is an important food crop in my country and even the world,and its output greatly affects food security.Chlorophyll content is an important indicator for evaluating crop health,productivity and environmental stress.Real-time,rapid and accurate acquisition of rice leaf chlorophyll content is of great significance for rice growth monitoring and field precise management.The traditional method of obtaining leaf chlorophyll requires destructive sampling and is timeconsuming and labor-intensive.With the development of technology,remote sensing has gradually become an effective way to obtain regional and global crop leaf chlorophyll content.The inversion method based on canopy radiation transmission model can quantitatively describe the response of vegetation canopy or leaf reflection spectrum to leaf chlorophyll.It is an important tool for chlorophyll remote sensing inversion.However,the inversion accuracy of this method is susceptible to interference from the selected spectral index and the underlying surface background.To this end,this paper uses Sentinel-2 remote sensing satellite images as the data source,combined with the observation data of typical rice fields,and uses the PROSAIL radiation transmission model to establish a rice leaf chlorophyll content-rice canopy reflectivity lookup table.The leaf chlorophyll was inverted through the cost function constructed by a variety of spectral indexes,the accuracy difference of different inversion results was compared and analyzed,the most suitable spectral index for the inversion of rice leaf chlorophyll content was explored,and the influence of the change of background reflectance on the inversion was clarified,and the G index(Greenness index)is introduced to reduce the influence of background interference on the inversion of leaf green content.Research indicates:(1)There are differences in the accuracy of leaf chlorophyll content inversion based on the spectral index constructed in different bands.the R2 of the Chlorophyll Index(CI),Simple ratio Index(SR)and Normalized Vegetation Index(NDVI)inversion results without the red-edge bands are 0.58,0.49 and 0.42,and the RMSE are 13.52,11.81 and 20.00μg cm-2,respectively.However,the inversion accuracy of the spectral index constructed by replacing the red band and the near-infrared band in CI,SR and NDVI with the red-edges band is significantly improved.the R2 of the CI740、CI783、RSR、ZM、NDVI740、NDVI783、Macc and MTCI inversion results are 0.68、0.64、0.73、0.70、0.68、0.71、0.73 and 0.70,and the RMSE are 11.91、11.42、12.02、11.22、11.82、11.38、8.14 and 10.15μg cm-2。(2)Using the PROSAIL model to build a lookup table to retrieve rice leaf chlorophyll will be affected by the background reflectivity of the water surface.This is mainly because the PROSAIL model cannot accurately describe the background reflectivity of the water surface.Especially in the early stage of rice growth,the inversion results of some spectral indices that are sensitive to changes in background reflectance are obviously low due to the large water surface background interference.When LAI < 3,the mean relative errors(MRE)of the inversion results of CI740,CI783,RSR,ZM,NDVI740 and NDVI783 are:-35.22%,-33.80%,-34.29%,-33.91%,-33.36% and-34.71 %.Compared with LAI> 3,the MRE increased by-22.37%,-26.29%,-19.31%,-20.77%,-22.29%,-20.14%,respectively.(3)The introduction of G index can effectively eliminate the influence of paddy field water surface background on leaf chlorophyll inversion results.The spectral indexes constructed by introducing the G index improves the retrieval accuracy of the rice leaf chlorophyll content.The specific results are as follows: CI740/G(R2=0.82,RMSE=5.84μg cm-2,MRE=3.53%),CI783/G(R2=0.77,RMSE=7.15 μg cm-2,MRE=8.61%),RSR/G(R2=0.80,RMSE=6.11 μg cm-2,MRE=2.23%),ZM/G(R2=0.81,RMSE=7.14 μg cm-2,MRE=7.95%),NDVI740/G(R2=0.75,RMSE=6.87 μg cm-2,MRE=4.16%)and NDVI783/G(R2=0.72,RMSE=7.37 μg cm-2,MRE=2.01%).After the spectral index is combined with the G index,it is no longer sensitive to changes in the rice background reflectance,and effectively eliminates the influence of the rice field water surface background on the leaf chlorophyll inversion results.