<![CDATA[Global climate change has significantly disrupted rainfall patterns in tropical regions, posing major challenges to agricultural productivity and food security. Accurate rainfall prediction has become a critical component of data-driven agricultural management. This study conducts a systematic literature review (SLR) following the PRISMA 2020 guidelines to analyze the evolution of AI-based rainfall prediction systems and their multidimensional impacts on tropical agricultural management over the period 2008–2026. Data were sourced from Scopus using three Boolean search strings, yielding 239 records, of which 235 articles were retained after duplicate removal and quality assessment using the Mixed Methods Appraisal Tool (MMAT) with a threshold score of ≥5. Bibliometric analysis was conducted using VOSviewer and Bibliometrix (R), while thematic narrative synthesis was performed using NVivo 14. Results reveal a clear four-phase technological evolution: conventional methods (2008–2015), machine learning adoption (2016–2020), deep learning and IoT integration (2021–2023), and multimodal and large language model era (2024–2026). Technical impacts dominated the corpus (accuracy improvements of 18–35%), while social and economic impact studies remain critically underrepresented (2.6% and 0.9%, respectively). Key research gaps identified include poor model interpretability (black-box problem), limited integration with decision support systems (DSS), inadequate tropical-specific model development, and the near-total absence of longitudinal impact evaluations. This study contributes a holistic synthesis integrating technological evolution with multidimensional impact analysis, offering strategic recommendations for developing more adaptive, transparent, and equitable AI rainfall prediction systems aligned with SDG 2, SDG 13, and SDG 15]]>
