Agricultural droughts are a periodic phenomenon in many regions of the word. Reducing damages to crops from drought events requires increasing the precision of agricultural drought mapping and making forecasts. However, there are very few studies that employ cutting-edge ensemble learning methods to estimate the risk of agricultural drought. In order to strengthen predictions of agricultural drought risk in a systematically explicit manner, we propose and evaluate ensemble models grounded on the credal decision tree (CDT) with Decorate (CDT-Decorate) supervised learning approaches as a case study in Iran. In the Esfahan Province of Iran, a thorough evaluation of the possibility of agricultural drought was conducted. This assessment coupled the five risk components—vulnerability, hazard, exposure, and mitigation—with the parameters that were suitable. Eighteen drought conditioning factors were identified and used to build both the training and validation datasets. A number of evaluation measures that showed the ensemble model's capacity to explain the underlying spatial pattern of agricultural drought events within the research area and forecast the likelihood of future drought phenomena that were used to validate the models. Area Under the Receiver Operating Characteristic Curve (AUCROC) showed that the ensemble CDT-Decorate model was better than the CDT model (AUCROC = 0.755) because it had an AUCROC value of 0.962. When assessing the risk of agricultural drought, the most important elements are population density, land usage, land cover, and distance to the river. The center region, with its intermediate risk (17 %), has a significant disruption of human agricultural activities; the southern region, with its very high risk (16 %), should receive the greatest attention due to its high susceptibility, significant hazardousness, and limited mitigation capability. An analysis of the models' e performances revealed that the ensemble model offered a trustworthy assessment of the risks associated with agricultural drought, and the risk maps it produced are suitable for drought mitigation techniques in the agricultural sector and could be applied in other drought-prone areas.