Research On Leaf Area Index Retrieve Methods Based On The Red Edge Bands From Multi-platform Remote Sensing Data

Leaf area index(LAI)is one of the most important vegetation structure parameter,which is one of the most important agronomic indices for crop condition monitoring,yield estimation as well as fertilization and irrigation management.The emergence of remote sensing technology provides an approach for timely estimation of LAI at a large scale.The red edge region of vegetation spectra is the basis of quantitative remote sensing analysis,because it plays an important role for nutrition and health status monitoring,vegetation recognition and biophysical and biochemical variable retrieval.There are prons and cons for each remote sensing platform data when applied for LAI retrieval,therefore,the features of different remote sensing data with red edge bands should be well explored during LAI retrieval.For example,hyperslectral data is equiped with plenty narrow bands within red edge region,but facing the problem of data redundancy due to significant relevancy among the bands;As for multispentral data with single red edge band,the band width is wider than that of hyperspectral data,hence loosing spectra details;As for multispectral data with multi red edge bands,it provides spectra information at different red edge wavelengths,which are discrepant from each other because the reflectance varies sharply within the red edge region,hereby,the red edge bands should be reasonably chosen for LAI estimation.With the objective of using red edge bands for LAI estimation considering the features of different remote sensing data,the main research contents and conclusions are as follows:(1)LAI estimation based on the red edge bands from ground and airborne hyperspectal data.The field collected LAI data and corresponding ground and airborne hyperspectral data in our study area were used to evaluation the capability of hyperspetral red edge retion for winter wheat LAI estimaiton.Firstly,LAI estimation models were built using hyperspectral vegetation indices(VIs),then three band combination approaches were investigated including red edge approach,conventional approach and band-by-band approach.The results showed that VIs based on bands between 680 nm and 750 nm were highly related to LAI.Finally,as the characteristic spectra related to leaf area index(LAI)varies among winter wheat under different nitrogen and irrigation conditions,universal and accurate LAI estimation models have been built with the MSAVI(Modified Soil-Adjusted Vegetation Index),NDVI(Normalized Difference Vegetation Index)and MTVI2(Modified Triangular Vegetation Index 2),based on the airborne and ground hyperspectral data,as well as in situ data.(2)Crop LAI estimation based on multispectral satellite data with single red edge band.Many vegetation indices(VIs)were improved with red band replaced by red edge band,and were used for LAI estimation at single stage or for single crop type,whereby the chlorophyll content variation was not properly considered.In this context,three newly improved VIs,i.e.NDVIred&RE(Red-edge Normalized Difference Vegetation Index),MSRred&RE(Red-edge Modified Simple Ratio Index)and CIred&RE(Red-edge Chlorophyll Index),were established by combining the red and red edge bands together to replace the red band.LAI estimation models were built using field collected LAI of wheat,barley,alfalfa and maize at different growth stages,as well as synchronous RapidEye satellite images.The estimation results were validated against the in situ data,showing that the newly improved indices(NDVIred&RE,MSRred&RE,CIred&RE)were resistant to chlorophyll content change and could efectively immprove the LAI estimation of multi crop types at multi stages,with the coefficient of determination(R2)improved by 10% of the R2 yielded by the VIs with red band replaced by red edge band.(3)Crop LAI estimation based on multispectral satellite data with multi red edge band.Facing the fact that the newly launched multispectral satellites equiped with more than one red edge bands are not sufficiently explored for crop variables retrieval,this paper takes the two red edge bands satellite Sentinel-2 as a paradigm for winter wheat LAI estimation.Three classical retrieving methods were used including the look-up table,neural network and vegetation index,and the discrepancy between the two red edge bands were investigated in order to make reasonable choice between them for LAI estimation.As comparison,the Landsat 8 satellite data was also used to retrieve LAI.Using the satellite data as well as agronimic information and field collected data in Shunyi District,Beijing,winter whear LAI estimation models were built.The results demonstrated that,with high temporal-spacial-spectral resolution,the Sentinel-2 data yielded higher LAI estimation accuracy than Landsat 8 data.And the results proved that the two Sentinel-2 red edge bands centered at 705 nm and 740 nm could provide more informaiton than the single red edge band from satellites like RapidEye,moreover,the two red edge bands allow the calculation of vegetation indices highly related to LAI based on 705 nm and 750 nm reflectance.This study lauches a pilot research for the application of multispetral satellites with multi red edge bands in quantitative remote sensing of vegetation.The results of this dissertation demonstrate that for hyperspectral data,VIs based on bands between 680 nm and 750 nm were highly related to LAI;For single red edge band multispectral data,the newly improved VIs based on the combination of red and red edge bands could resist the interference of chlorophyll content change and improve the LAI estimation accuracy;Regarding multispectral data with multi red edge bands,the variation among the red edge bands provides more information than the single red edge band multispectral data,which enables the usage of more VIs and helps to improve LAI estimation accuracy.These results could provide reference for the application of hyperspectral data,multispectral data with single or multi red edge bands in LAI estimation,and further offer reliable information for crop growth monitoring and field management decisions.In addition,the two red edge bands of Sentinel-2 centered at 705 nm and 740 nm were proved to have great potential in LAI estimation,which is a reference for future multispectral sensor design.