Hyperspectral estimation of leaf area index of winter wheat based on Akaike’s information criterion

Abstract

Leaf Area Index (LAI) is an important parameter for assessing the crop growth and winter wheat yield prediction. The objectives of this study were(1) to establish and verify a model for the LAI of winter wheat, where the regression models, extended the Grey Relational Analysis (GRA), Akaike’s Information Criterion (AIC), Least Squares Support Vector Machine (LSSVM) and (ii) to compare the performance of proposed models GRA-LSSVM-AIC. Spectral reflectance of leaves and concurrent LAI parameters of samples were acquired in Tongzhou and Shunyi districts, Beijing city, China, during 2008/2009 and 2009/2010 winter wheat growth seasons. In the combined model, GRA was used to analyse the correlation between vegetation index and LAI, LSSVM was used to conduct regression analysis according to the GRA for different vegetation index order of the number of independent variables, AIC was used to select the optimal models in LSSVM models. Our results indicated that GRA-LSSVM-AIC optimal models came out robust LAI evaluation (R = 0.81 and 0.80, RMSE = 0.765 and 0.733, individually). The GRA-LSSVM-AIC had higher applicability between different years and achieved prediction of LAI estimation of winter wheat between regional and annual levels, and had a wide range of potential applications.