Unlocking drought in pistachio orchards: monitoring and forecasting using landsat time series and machine learning techniques

Agricultural drought leads to reduced water resources, vegetation changes, and, in turn, to durable impacts on agricultural productivity. Rafsanjan, a major hub for pistachio production in Iran, has faced significant challenges due to drought, resulting in reduced cultivation and orchard abandonment, which poses serious issues for farmers. This study aimed to monitor and predict drought conditions in pistachio orchards via remote sensing and to examine their relationships with groundwater resources. Landsat time series data and three classification algorithms, including random forest, support vector machine (SVM), and artificial neural network (ANN), were used to classify pistachio orchards from 2008 to 2022. We also compared the effectiveness of the two data-driven models for forecasting drought. Vegetation health index (VHI) values were forecasted via ANN and SVM over one- and three-year lead times, revealing a declining trend in pistachio orchards linked to groundwater resources. Monitoring results revealed that the SVM outperformed the other methods, achieving an average overall accuracy of 95% and a kappa coefficient of 0.90. Additionally, the ANN excelled in forecasting the VHI, with root-mean-square errors of 0.024 and 0.043, mean absolute errors of 0.019 and 0.034, and R2 values of 0.85 and 0.78 for one- and three-year lead times, respectively. This study underscores the necessity of simultaneously examining the factors contributing to drought in arid regions and determining the relationships and contributions of these factors to drought levels in horticultural key crops.