Development of indicators of burning efficiency based on time series of SPOT VEGETATION data

Accurate estimates of the aerial extent and burning efficiency are essential parameters for understanding and monitoring the impact of fires on the atmospheric and terrestrial ecosystems. Such inputs are important for global and regional carbon cycle models and provide a more comprehensive and accurate basis for understanding and monitoring vegetation recovery. Time series of satellite imagery offer the potential to quantify these parameters with spatial and temporal accuracy. Several studies (GLOBSCAR project of European Space Agency (ESA), GBA2000 project of Joint Research Centre (JRC, Italy)) have been developed to detect burned areas from satellite imagery, using the evolution of the spectral characteristics of a burn through time. The resulting data sets are available for the globe for the year 2000. Current research investigates the potential of time series of SPOT Vegetation (SPOT-VGT) S10 data (1998-2004) to quantify burning efficiency of the burns detected in the framework of GLOBSCAR and GBA2000. Burning efficiency is defined as the percentage biomass of a pixel that is burned. General constants have up to now been assigned per biome to define burning efficiency. The objective of this study was to provide quantitative estimates of the spatio-temporal variation of burning efficiency based on remote sensing indicators. Two different techniques were tested to quantify the burning efficiency of every detected fire pixel by means of consistent indicators. Southern Africa was used as a pilot study area due to the availability of both ground and satellite data. Firstly, a technique based on time series analysis of sub-pixel fractions was developed. A linear spectral mixture analysis (SMA) technique was applied to assess the fraction vegetation and non-vegetation components for every pixel through time. Endmember analysis was performed to extract appropriate pure endmembers for the SMA. Detailed vegetation geo-datasets and fire scar records were used to validate the endmembers. The results of the SMA provided a fractional variable for every pixel. These fractional variables were subsequently subjected to an image differencing change detection algorithm to quantify the changes in the vegetation component as a measure of burning efficiency. Secondly, a technique based on temporal variations of vegetation indices was developed. The temporal difference of several vegetation indices (Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Global Environment Monitoring Index (GEMI), and Normalized Burn Ratio (NBR)) was used to quantify the changes in the time trajectory. This difference is hypothesized to represent the changes in vegetation content and burned area and to provide a scaled index of the magnitude of change caused by fire, hence the burning efficiency. The results of both techniques were compared and validated with field data containing burn severity and severity indices based on Landsat imagery with higher spatial resolution. Statistical regression techniques were used to assess the performance of both techniques and the resulting quantitative indicators of burning efficiency. The effects of vegetation structure and fire regime features (timing of the fire, spatial pattern of fire) on the burning efficiency were subsequently analyzed. Statistical analyses were conducted to assess whether the indicators for burning efficiency were different for different vegetation types, fractions of treegrass-soil, stages of the burning season, and fragmented or continuous burns. Results show that the developed indicators of burning efficiency can provide useful information for understanding and monitoring the impact of fires. The temporal changes in fractional variables and vegetation indices provide consistent measures of burning efficiency. Finally, this study shows that indicators for burning efficiency can provide a useful input to global and regional carbon cycle models to decrease uncertainties in estimates of emissions by fire. They can also provide an accurate basis for understanding and monitoring the vegetation recovery, since the rate of recovery after a fire depends partly on burning efficiency.