Forest and land fires substantially contribute to greenhouse gas emissions, posing a major challenge to Indonesia's climate commitments. Fire prevention is a critical component of the FOLU Net Sink 2030 policy, which targets net carbon sequestration of 140 million tons CO₂e by 2030 as part of the country's Nationally Determined Contribution. Although threat-based messages are widely used in environmental communication, their effectiveness in forest fire prevention remains understudied. This study evaluated the comparative effectiveness of weak versus strong threat-based messages among community-based farmers. A quasi-experimental design (non-random sampling) with a non-equivalent control group was conducted involving 21 farmers in the Desa Makmur Peduli Alam (DMPA) program in Sri Gading Village, Ogan Komering Ilir Regency, South Sumatra Province. Farmers were allocated into weak threat-based messages, strong threat-based messages, and no treatment of fear appeal group. Intervention messages were designed using the Extended Parallel Process Model (EPPM), McGuire's persuasion stages and applied Venus's ten communication principles. The weak threat-based messages group achieved significantly higher participation scores (16.29 ± 2.36) than the strong threat-based messages and no treatment groups (9.86 ± 4.49 and 9.86 ± 3.34; p-value = 0.004). Discriminant analysis achieved 95.2 % accuracy in classifying participation level (as behavioral intention encompassing cognitive and affective dimensions), demonstrating that the territorial map effectively links demographic, attitudinal, and spatial variables to systematic patterns of community engagement in forest fire prevention. Community-based forest fire prevention can be strengthened by integrating weak threat-based messages with participatory platforms to support Indonesia's FOLU Net Sink 2030 target. © 2025