Page 108 - Demo

1st Int. Transborder Conf. of the Timor Island: Timor %u2013 Science without borderDili, 7-8 May 2025107Harnessing Machine Learning to Address Climate Change in Southeast Asia: A PRISMA-Guided Systematic Literature ReviewShannon M. D. ViegasUniversidade Cat%u00f3lica Timorense (UCT), Office of Research, Post-graduate & Cooperation, Dili, Timor-Lestee-mail: shanviegas@gmail.comAbstractSoutheast Asia (SEA) is one of the globe's most climate-exposed areas, with escalating risks from sea-level rise, weather extremes, and land degradation. To combat these, machine learning (ML) has proven a valuable resource to support climate change studies and decision-making. This systematic literature review combines findings from 19 peerreviewed articles from 2014 to 2025 that apply ML to address climate concerns in SEA. Articles were found through keyword searching on SCOPUS academic database, focusing on empirical studies with a specified ML component. The reviews analyzed employ diverse ML techniques%u2014random forests, support vector machines, neural networks, self-organizing maps, fuzzy c-means, and genetic programming. Most of the articles (74%) apply remote sensing or satellite data, i.e., Landsat, Sentinel-2, and TRMM, with applications varying from fire prediction, mangrove and crop mapping, carbon stock estimation, and drought monitoring. Thematically, studies addressed climate change mitigation (e.g., forest carbon sequestration), adaptation (e.g., planning rainfed agriculture), and impact assessment (e.g., Natech risk analysis). Geographically, studies are mainly concentrated in Indonesia, the Philippines, and Malaysia, whereas Laos, Cambodia, Timor-Leste and Myanmar are considerably underrepresented. Despite advancements, the significant challenges remain. Data scarcity and heterogeneity limit model generalizability, especially in rural or transboundary settings. Methodologically, supervised ML is most common in most studies, with hybrid and unsupervised models not yet being fully investigated. A minority of studies integrate ML with domain-specific tools such as hydrological models or socio-economic data. Moreover, only two studies directly engage with policy stakeholders or local communities. This review stresses the growing maturity of ML applications in climate science in SEA but underscores the need for growing methodological diversity, interdisciplinary integration, and regional equity in research. Future studies should prioritize open data sharing, hybrid modeling approaches, and co-designed research with policy actors to enhance real-world climate resilience.Keywords: Machine Learning, Southeast Asia, Climate Change, Remote Sensing, Adaptation, Mitigation.