Annual burning events in Indonesia release massive amounts of carbon, averaging 49 TgCyr-1 in Sumatra and are exacerbated during El Niño events. The recent 2015 fires were greatly affected by an extended drought period associated with a strong El Niño, resulting in an extreme haze event with higher carbon emissions than the record breaking 1997 fires, as well as enormous social and environmental damages. This study aims to empirically evaluate the socio-ecological causes of fire occurrences within fire-prone regions in Indonesia (specifically Riau, Jambi and South Sumatra), by integrating geospatial and social analyses techniques. Specifically, we are interested to know what were the major contributing socio-ecological factors to fire activity in the eastern provinces of Sumatra in 2015 at two spatial scales (i) a regency administrative level and (ii) a pixel (1km) level. We built upon past conceptual frameworks on how mega-fires are the consequence of predisposing conditions in the landscape, which define the vulnerability of the landscape to fires, and fire ignitions, which are largely from anthropogenic sources. We developed indicators for plausible social and ecological drivers of fire activity which could contribute to mega-fires as a predisposing condition or as a source of fire ignitions. We used generalized linear mixed models (GLMM) and boosted regression trees to model fire activity as a result of our list of social and ecological factors at the regency- and pixel-level respectively.
Results/Conclusions
We showed that for fire activity at a regency-level, our most parsimonious GLMM model had a combination of both social and ecological factors. These factors included the proportion of small plantations in a regency, the proportion of villages who reported burning for agriculture, the mean slope of a regency, and the total length of roads in a regency. The preliminary results of our pixel-level analysis showed that rainfall and population density were most important in predicting fires. Due to the spatially explicit nature of our pixel-level analysis, we expect it to be more reliable than our regency-level analysis. We will present comparisons of our results with published studies on drivers of fire activity in Indonesia and provide suggestions on future fire management and prevention strategies for a more resilient socio-ecological landscape in Sumatra.