AbstractDroughts are one of the most disastrous natural hazards, primarily due to their persistence and spatial distribution. Drought prediction is one of the key challenges for effective drought management and to do so, studies often involve the use of station-based data which are effective only in regions with high-gauge density. Therefore, there is growing interest in the use of interpolated climatic grids to predict droughts. In recent decades, drought conditions have been aggravated by climate change and for that reason the use of climatic variables is important to accurately predict droughts. The analysis of any aspect of drought can be affected by the choice of data and drought index. Therefore, this study aims to identify the most suitable dataset and drought index for the New South Wales (NSW) region of Australia. The present study evaluates various precipitation datasets (Climate Research Unit (CRU), ERA-5, and Scientific Information for Land Owners (SILO)) and their corresponding variations on the Standardised Precipitation Index (SPI) at different time scales. Based on the findings, CRU was used to predict meteorological drought using machine learning techniques. The different machine learning models are Support Vector Regression, Random Forest and Artificial Neural Networks. The results suggest SVM to be the best performing model among these models for predicting SPI at short time scales (1 month and 3 month) and ANN to be the best performing model for long-term scales (6 months and 12 months). Such findings depict the capabilities of different models in examining drought characteristics and confirming the use of interpolated climatic grids thereby assisting in regional drought management.