Drought is one of the most uncertain and perilous natural disasters which is caused due to the rapid climate changes that is dampening and worsening with each passing day in terms of its intensity and frequency. In this context, drought modelling is of immense importance, keeping in view its highly negative impact on the affected community globally. Drought has four stages, which are, drought, normal (conditions), pre-drought and after-drought. In this study, drought stages are predicted employing a two-part strategy. The first part of strategy is the forecasting of the remote sensing-based water and vegetation indexes (Enhanced Vegetation Index (EVI), Normalized Difference Vegetation Index (NDVI), and Normalized Difference Water Index (NDWI)), while in the second part, drought stages are predicted using EVI, NDVI and NDWI. Unfortunately, Pakistan is one of the most drought-prone regions with a densely populated desert called, “Thar” in Tharpakar, Sindh, Pakistan. Therefore, in this study, the dataset is collected from one of the most drought effected district (Tharpakar) of Pakistan. In this study, ARIMA (Auto-regressive Integrated Moving Average) is used the forecasting of water and vegetation indexes, and, Multiclass Decision Forest (MDF), Multiclass Decision Jungle (MDJ), Multiclass Logistic Regression (MLR) and Multiclass Neural Network (MNN) are employed for the classification of drought stages. Based on the experiments and various assessments performed in this research, the RMSE for EVI ranges between 0.035 and 0.088, for NDVI, it falls between 0.034 and 0.075, and for NDWI, it ranges between 0.032 and 0.075. The best performing algorithm for drought stage prediction is Multi-class Decision Forest with an accuracy of 97.35%.