Accurately estimating saffron yield is crucial for agricultural planning and managing this valuable product. The objectives of this study encompass developing a robust model utilizing meteorological and satellite data to predict saffron yield, assessing the integration of Sentinel-2 and Landsat-8 imagery, establishing a predictive framework, selecting suitable machine learning algorithms, and computing various vegetation indices. The methodology employed involves calculating vegetation indices (such as NDVI, and EVI) from satellite images (Sentinel-2, Landsat 7 and 8), merging them with meteorological data, and employing machine learning algorithms for predictive analysis. Through correlation analysis, the most influential factors affecting saffron performance were identified. Evaluation of the models generated during the testing phase revealed that the coefficient of determination (R2) values indicate the model\\\'s accuracy. By analyzing the top-performing models, it was determined that the most effective saffron yield estimation model incorporates three datasets: vegetation indices, geographical information, and meteorological data, leading to a more precise and reliable estimate. This model serves as a comprehensive biomass and yield predictor for saffron, identifying key variables influencing yield and offering a dependable forecast for arid regions with limited meteorological data.

Keywords