CA-Markov model application to predict crop yield using remote sensing indices

Abstract

Drought and related water scarcity have a significant impact on crop production. The purpose of this study was to predict the yield of pomegranate trees and palm trees in southern Iran based on the probability of future drought. We propose a novel meteorological drought-based approach that can predict yield of two crops in 2040 by using Cellular Automata (CA)-Markov chains. From these data in 2000, 2010, and 2020, the regression analysis of yield determination was done with the most important effective indicators that were identified by principal component analysis (PCA), thus leads to highly accuracy. The modelling results of remote-sensing indices (Standardized Precipitation Index-SPI, Standardized Precipitation Evapotranspiration Index-SPEI, Precipitation Condition Index-PCI, Vegetation Condition Index-VCI, Normalized Difference Vegetation Index-NDVI, and Temperature Condition Index-TCI) depicted the expansion of drought areas in southern parts than others regions, and the decreasing yield of two crops in 2000–2020. Additionally, the results of PCA showed that NDVI, PCI, and VCI indices were the most effective drought indices in determining palm’ yield, while SPEI and TCI indices were most effective in determining pomegranate’ yield. According to the results of the CA-Markov chain and regression, approximately 50–60% of the region will have low pomegranate and palm yields in 2040. The approach provides a framework to predict what the decreasing of crop yield is due to the drought effects, and for supporting the optimal decision-making on sustainable horticultural management.