Machine Learning on the estimation of Leaf Area Index

Abstract

The Leaf Area Index (LAI) is an important indicator in agriculture that can be considered a reliable plant growth parameter. The objective of this study is to make use of two different machine learning algorithms including Support Vector Machine (SVM), and Random Forest (RF) to improve the estimation of leaf area index using multispectral, thermal, and hyperspectral data. The results showed that RF was the best model to improve the accuracy of the LAI estimation compared to the simple linear regression (previous study) and SVM (R2 = 0.91 for RF and R2 = 0.87 for SVM). To evaluate the effects of spectral portions on LAI estimation without calculating the spectral indices, (SI) we inputted each pair of spectral bands for training and testing both RF and SVM. It was found that the best correlation was lower compared to use SIs. However, R2 variations were more homogeneous across the whole spectrum, which suggests that even by using multispectral broadband bands in RF and SVM, a good correlation will be achieved.