Land use as driver of vegetation fires in Indonesia : a spacial modelling approach to identify and quantify causes of fire

The large forests and other land cover fires in 1997/98 that ravaged throughout Indonesia drew international attention to the problem of fires as an important agent of deforestation and a global source of environmental pollution and CO2. It was estimated that over 9 million ha were burned, resulting in a smog cloud over 4 million km² and elevated CO2 concentrations of 30% over Southeast Asia. Observations by satellite present an unparalleled means for analysis of fire occurrence and impacts and capturing their variation from year-to-year on local, regional and global scales. Satellite data on fire are derived, via different techniques, from coarse resolution heat or optical light sensors, using day- and/or night-time values resulting in active fire points. A comparison of eighth different active fire datasets in Sumatra and Kalimantan (Indonesia) showed that each dataset detects different types of fires, depending on fire regime and detection method. However, some datasets identify the general trend of fire occurrence but with a large variation in actual number of fires detected. Fires are the result of predisposing conditions (e.g. drought) and ignition sources (human intend). With the active fire datasets and data on land use it is therefore possible to determine the probability of fire occurrence as a function of human action in land use. The causal relation between land use and active fires are quantified by a spacio-statistical model in one province for one year and then applied on a larger scale in different years. Results confirmed the hypothesis that fires were determined by multiple interacting, predisposing conditions (mostly climate, elevation and suitability for specific tree crops) and human-related causes (presence of transmigration projects and land allocation to specific land uses) and that these interactions are not the same in all provinces. National policies, responsible for the allocation of land, however are identified as a major driving force of fires through land allocation. Accessibility (by roads) is only an important determinant of fires in forests. Few fires seem to be accidental. Three land use pathways that increase fire probability were identified that are valid in all provinces and years, and further concluded was that areas not yet used by large-holders are more prone to fire, indicating a serious threat to the remaining forest in this region. The spacio-statistical model identified determinants through correlation of landscape variables and active fire points. To validate these results they are compared to results from detailed analysis of fieldwork on fire activity. In order to be able to compare these studies both are being analysed by a hierarchical cause-consequence chain method. The results indicated that in both studies similar pathways of fire were found, although detailed site-specific pathways were found too. The spacio-statistical model thus in general captures the human dimension in burning well. The findings suggest that fire occurrence was not a pure natural phenomenon but driven by land-use change and as such part of government policies and a serious threat to the remaining forest. Hence governments have a responsibility in fire occurrence and thus could also be able to take measures against fire occurrence.Doctorat en sciences (géographie) -- UCL, 200