During periods of drought in Indonesia, fires used for land clearance on peatland spread uncontrollably, with sometimes millions of hectares of peat carbon stocks affected along with significant forest loss. Poor and/or unregulated land management, including peatland drainage, deforestation and forest degradation, and installation of industrial monoculture crops (e.g. palm oil) exacerbate the problem. Peat is especially carbon rich, and when burned releases large amounts of greenhouse gases (GHGs) per unit mass burned. Climate change targets require reduced carbon emissions, and if these are to be met confidently it is of significant importance to yield precise estimates of GHG emissions from these fires.
Airborne LiDAR has been an important tool used to help quantify atmospheric emissions from peatland fires, but they are not commonly available. This thesis firstly focuses on integrating pre-burn peat surface measurements made with airborne LiDAR with a novel, low-cost UAV-based photographic methodology for measuring post-burn surface topography. Using structure-from-motion photogrammetry the latter produces accurate, high spatial resolution digital terrain models (DTMs) from RGB photography, and combined with the pre-burn LiDAR the two datasets can be used to calculate overall peatland depth of burn. The DoB data show peatland fires burn deepest around the roots of trees, to a depth of up to >1 m in the case of the fires close to Jambi (Sumatra) studied here. Mean (±1) depth of burn is found to be 0.23 ± 0.19, meaning an equivalent Mean (±1) fuel consumption per unit area of 134 ±29 tC ha-1 which supports previous findings that extremely high carbon emissions per unit area come from tropical peatland forest fires.
Comparisons of the pre-burn LiDAR data and the higher spatial resolution post-burn UAV-derived DTM model, conducted in areas that did not burn, highlighted some anomalies in the pre-burn topographic measurements. These were considered potentially related to pre-burn vegetation cover, and an investigation was conducted into the vertical structure of the overlying vegetation and its effect on LiDAR-derived DTMs accuracy. A field study using highly accurate global navigation satellite system derived x,y,z positioning data and a total station survey was conducted, along with LiDAR measurements during both leaf-on and leaf-off conditions in a UK deciduous woodland. It was found that surface DTM accuracy was significantly decreased in the presence of dense undergrowth vegetation (ferns, bramble), and less-so by the presence of tall canopy trees. This suggests that pre-burn DTM measurements in degraded forests are subject to large biases where LiDAR pulses are returned from near-surface vegetation, rather than the ground (soil) surface itself. Both the first two research chapters of the thesis have been published as papers in the journal Remote Sensing.
Along with peatland depth of burn, the factor which most controls GHG emissions estimates from burning peatlands is the burned area. Global fire emissions inventories use automated burned area products such as MODIS MCD64A1 to map this parameter. However, the coverage and accuracy of the MCD64A1 burned area estimates has not previously been assessed in detail in a tropical peatland environment. Here this product is compared with Landsat-derived burned indices, and as both burned area products’ coverage is hindered by persistent cloud cover, their relative coverage is compared to an independent dataset recording active fire locations (MCD14ML) which has a ~ 4 times per day temporal resolution. Comparisons are made across different land cover types, and the accuracy of the Landsat-derived burned area is compared with both Synthetic Aperture Radar (Sentinel-1) burned area and aerial orthophotos and airborne LiDAR-derived burned are maps as a reference.
Finally, the above results are used to calculate new emissions estimates for the 2015 fires that occurred in the Berbak region, Jambi Province (Sumatra) for the target period (1989-2015), along with their uncertainties. These estimates are compared to those of the most widely used fire emissions database (GFED) for the survey area.
Methods for improving greenhouse gas emissions estimates from peatland fires in indonesia
Simpson, J. E. (Author). 2018
Student thesis: Doctoral Thesis › Doctor of Philosophy