Explainable Environmental Decision Support System for Drought Monitoring with Decision-Tree Knowledge, Fuzzy Reasoning, and Independent SMAP Evidence in Indonesia

The increasing need for explainable environmental decision support requires drought monitoring systems to move beyond categorical classification toward traceable and operationally meaningful recommendations. This study proposes an Explainable Environmental Decision Support System (EEDSS) that integrates a Decision Tree, fuzzy reasoning, operational decision mapping, and independent SMAP soil-moisture evidence for drought monitoring in Indonesia. NASA POWER meteorological observations from 470 spatial locations covering 2020-2024 were used, with SPEI-3 serving as the reference for nine dry-to-wet environmental states. A Decision Tree was developed using 2020-2023 data and evaluated on an independent 2024 temporal test set. The selected balanced Decision Tree achieved an accuracy of 0.721, balanced accuracy of 0.704, and Macro-F1 of 0.681, indicating moderate classification performance while providing an interpretable structure for extracting 110 meteorological rules. These rules were transformed through fuzzy reasoning into continuous environmental evidence and five operational categories: NORMAL, WATCH, URGENT, WET WATCH, and URGENT WET. All 27,723 observations received an operational recommendation. Independent SMAP Level-4 observations provided valid SSM and RZSM evidence for 98.72% of the observations. SSM demonstrated the expected dry-to-normal-to-wet directional consistency, whereas RZSM did not. Both variables showed statistically significant distributional differences across the nine environmental states based on the Kruskal-Wallis test (p 0.05). Overall, the findings indicate that EEDSS provides a transparent decision-support pipeline with partial independent environmental consistency, particularly at the surface soil-moisture level. The system is therefore positioned as an explainable environmental decision-support framework, rather than solely as a high-accuracy drought classification model