Estimation of Leaf Water Content in Winter Wheat Using Grey Relational Analysis–Partial Least Squares Modeling with Hyperspectral Data

Abstract

Water availability limits wheat (Triticum aestivum L.) production in many areas of the world. It may be possible to use crop reflectance to estimate leaf water content (LWC). The objectives of this study were (i) to develop and test a model for assessing water stress in winter wheat, where the regression models, grey relational analysis–partial least squares (GRA‐PLS), the optimal band ratio normalized difference, and three bands algorithms were tested, and (ii) to compare the performance of the proposed models using GRA‐PLS and the three‐band algorithm. Spectral variables and concurrent LWC parameters of samples were acquired in Tongzhou and Shunyi districts, Beijing, China, during 2008 and 2009 winter wheat growth seasons. The GRA‐PLS model was established to estimate LWC. In the combined model, GRA was used to select sensitive spectral variables and PLS was used to perform regression analysis. The widely used three‐band algorithm was applied for LWC estimation to compare model performances. Both the three‐band algorithm and GRA‐PLS resulted in robust LWC estimation (R2 = 0.60 and 0.74, RMSE = 13.15 and 9.82%, respectively). Our results indicated that the GRA‐PLS model has great potential for LWC estimation in winter wheat; however, the three‐band algorithm also has merit, particularly when taking into consideration the simplicity of its application. This method may provide a guideline for improving the estimation of wheat LWC at the regional scale by combining different algorithms. We will further simplify the GRA‐PLS model to apply widely different satellite data in the future.