Chapter 14 - Prediction of plant pigments for phytosanitary and yield estimation

Abstract

This chapter reviews the research on monitoring of crop phytosanitary conditions, as well as a novel approach to calculating wheat yield based on predicted endmember abundances attributable to photosynthetic pigments, which is based on Landsat 8 images obtained during maximum greenness. The case study demonstrates an unmixing technique that is used to find endmembers within the wheat pixels, which are then further adjusted to maximize their abundances' predictive power for yields. The endmembers' spectral fingerprints are similar to those of photosynthetic pigments, and their predictive power suggests that they are related to the pigments. Artificial intelligence methods are used to estimate wheat yields using abundances and other their relevant data. For performance evaluation, harvester records from 142 fields are used as the ground truth. When the Random Forest technique is used with critical parameters, the yields are calculated with 82% accuracy (RMSE=22.51kg/da).