Retrieval Of Conifer Canopy Leaf Water Content With Sentinel-1B And Landsat 8 OLI Data

Water is an indispensable element in photosynthesis and respiration of vegetation,and it plays a vital role in the life activities of vegetation.Vegetation Water Content(VWC)is of great significance for crop monitoring,early warning of grassland drought and desertification,and early warning of forest fires.This paper takes Sentinel-1B data,Landsat8 OLI data,and ground surface measured vegetation canopy leaf water content as data sources,and takes the coniferous forest of the moon lake national forest park in Changchun as the research object.Firstly,Landsat8 OLI data is used to extract multiple vegetation indices.Then,the correlations between the remote sensing image band and vegetation index and canopy leaf water content were analyzed,and the most relevant factor was selected as the optimal factor,and the principal component was extracted.The principal component and canopy were included.The amount of water builds a variety of models,using the highest precision model to invert the canopy leaf water content.The main research contents and results are as follows:(1)In this paper,through the analysis of the typical correlation between ground survey data and Sentinel-1B data and Landsat8 OLI data,the band and band combinations that have a large correlation with the canopy leaf water content are selected.Among them canopy leaf water content and VV polarization,VH/VV polarization ratio of Sentinel-1B,and Shortwave Infrared Red 1(SWIR1),Shortwave Infrared Red 2(SWIR2)bands of Landsat8 OLI Relatively high.(2)In this paper,the normalized vegetation index(NDVI),normalized difference water index(NDWI)and Ratio Vegetation Index(RVI)were extracted using Landsat8 OLI data.Correlation analysis with canopy leaf water content showed that the correlation between NDVI and canopy leaf water content was low,while the correlation between NDWI and RVI and canopy leaf water content was high,and there was a high correlation between NDWI and RVI.(3)In this paper,the principal components of Landsat8 OLI data(SWIR1,SWIR2,NDWI,RVI)and Sentinel-1B combined with Landsat8 OLI data(VV,VH/VV,SWIR1,SWIR2,NDWI,and RVI)are analyzed to extract the first principal component.The linear,quadratic polynomial,cubic polynomial,and exponential function models were established with the canopy leaf water content.Accuracy analysis showed that the cubic polynomial model had the highest precision(R2 = 0.6299,RMSE = 0.0358),followed by the second-order polynomial(R2 = 0.6214.RMSE=0.0362),and the R2 and RMSE of the model after binding Sentinel-1B data(VV,VH/VV)are better than before.(4)In this paper,a BP neural network model is constructed by the first principal component extracted from Sentinel-1B combined with Landsat8 OLI data(VV,VH/VV,SWIR1,SWIR2,NDWI,RVI)and canopy leaf water content.After the training,the inversion accuracy of the model(R2=0.896,RMSE=0.0282)is better than the quadratic polynomial.In summary,for the moon lake national forest park,the precision of synthetic aperture radar(SAR)image combined with optical image inverting canopy leaf water content is higher than that of using only optical image,and the inversion accuracy of BP neural network model is higher than that of the second-order polynomial.This shows that SAR combined with optical images can improve the accuracy of vegetation canopy leaf water content inversion.