Research On Leaf Area Index Retrieval Of Different Crops

Leaf area index（LAI）is a key input in models describing biosphere processes and has widely been used in monitoring crop growth and yield estimation.Multi-platform and multi-sensor remote sensing technology has become an important means of crop growth monitoring to achieve refined agricultural management,which is fast,accurate as well as economical.At present,the use of remote sensing means for the acquisition of the leaf area index has been relatively mature,and domestic as well as foreign scholars have established a set of leaf area index inversion system based on multi-platform（ground,drone and satellite data）.However,the current research is mainly aimed at the LAI inversion of single crops,which has certain limitations on the research area with wide planting area and complex planting structure.Although the remote estimation for leaf area index has been studied for decades and various model including empirical methods and physical methods were developed and achieved high accuracy in selected experimental areas,the robustness of the algorithms are restricted among crop types,growth phenology,and field managements,especially for small-size croplands planted in developing countries and therefore,no routine and universal model is applicable for a wide range of operational applications and for all crop types.The general objective of this study is to assess different reflectance band based vegetation indices for the estimation of leaf area index（LAI）of four crop types（maize,soybean,oilseed,and rice）with contrasting canopy architectures and leaf structures.The main research contents include as followings:（1）The accuracy of the inversion of LAI by different methods（vegetation index method,machine learning classic algorithm including support vector machine,K-nearest neighbor method）is compared.The experimental results show that the machine learning method can self-learn the complex correlation between features and achieved better performance of the leaf area index inversion.However,compared with the machine learning and neural network methods,the vegetation index based on the red edge band can achieve simple and efficient leaf area index inversion results for crops,and has physical meaning interpretability;（2）Among the indices using different reflectance band tested in terms of ratio vegetation index and normalized difference vegetation index,red edge band based vegetation index exhibited strong and significant close-linear relationship with LAI,indicating being insensitive to crop type.Additionally,compared with CI vegetation index,the normalized vegetation index exhibited better accuracy of the estimation of LAI.The vegetation index experiment was designed using hyperspectral datasets to explore the optimal bands for inverting the LAI of the four crops.The experimental results show that the high spectral reflectance data with a center wavelength of 720 nm is the best band for the LAI inversion of the four crops,which means if the red edge band around 720 nm is available,as for hyperspectral radiometers such as Ocean Optics,ASD and hyperspectral imaging sensors such as AISA and HYPERION,720 nm centered red edge band reflectance based index VI_Rededge was recommended as non-species-specific index for estimating leaf area index in both wheat,soybean,oilseed and rice.（3）In order to evaluate the influence of chlorophyll content difference in LAI inversion,the widely used canopy radiation transport model PROSAIL was used to simulate the spectral differences under different chlorophyll content.Based on this,the sensitivity of chlorophyll to PROSAIL model was analyzed.It is concluded that the sensitivity of the long-wave red edge band is much lower than that of the short-wave red edge,and then the short-wave red edge is replaced by a certain ratio of the two bands,based which the vegetation index is improved,and the LAI inversion accuracy is increased.（4）The bias of absolute integral method was proposed to find the best vegetation index for LAI estimation of different crops combinations,and the red edge was proved to be the best reflectance band to form VIs for LAI estimation of crop combination with different canopy and leaf structure.