| Leaf area index (LAI) is defined as the single-side leaf area per unit ground area, which is an important biological parameter reflecting crop growing, providing structured quantitative information for describing the procedure of matter and energy exchange on plant canopy surface, thus playing a key role in quantitative remote sensing inversion of vegetation, carbon cycling, gross primary production and energy homeostasis of the interaction between vegetation, soil and atmosphere. LAI is a significant input parameter for many ecological and land surface model and is also one of the most important agronomy indicators for evaluating crop growth status and forecasting its yield. Conventional LAI measurement methods are based on direct measuring on the ground, while remote sensing technology provides an effective approach for the large-scale, rapid monitoring of vegetation LAI and becomes the trend of monitoring LAI. This paper is targeted at the problems in LAI remote sensing inversion such as discontinuity of LAI time series production, low inverting accuracy and poor universality. The main research contents and the experimental results are as follows:(1) In order to solve the problem of low estimation accuracy based on single data resource, multiple platform data from the airborne and ground access was used in comparative study on different inversion methods for estimating winter wheat LAI. Compared with wavelet transform and principle component analysis, SVM method has stronger ability in estimating winter wheat LAI and shows higher accuracy, effectively suppressing the problem of overestimation and underestimation. Therefore SVM is more suitable for winter wheat LAI inversion study.(2) Proposed the study of the universality of SVM (support vector machines) method on estimating winter wheat leaf area index under different plant types, different period, different nitrogenous fertilizer and different water conditions. LS-SVM method performs better than NDVI method in every winter wheat LAI inversion model, that is to say, LS-SVM method has universality in estimating winter wheat leaf area index under different plant types, different period, different nitrogenous fertilizer and different water conditions.(3) LS-SVM (least squares support vector machines) method was used to estimate MODIS (Moderate-resolution Imaging Spectroradiometer) LAI time series production, solved the discontinuity problem in moderate resolution LAI inversion production. SVM and LS-SVM method have certain ability in estimating MODIS LAI time series production in the study area of Naqu. Primary precision verification is based on MODIS LAI production and further verification is based on ground data, both show that LS-SVM is the better method. The result of this experiment provides a theory for improving remote sensing production data quality. |
