Feasibility of Burned Forest Classification and Biomass Burning Emission Estimation Using ICESat-2 Data
Abstract
Climate change induces environmental concerns such as rising temperatures, increasing droughts, and more fires. The risk of large fires has increased over the past decades, causing threats to forest growth and biosphere sustainability. Rapid release of carbon dioxide during fire events contributes to atmospheric carbon accumulation and global warming, especially when combined with deforestation. To evaluate the effect of fire on forest ecosystems and climate, it is crucial to accurately classify burned regions and estimate the amount of fire induced carbon emissions. Remote sensing data have been employed extensively in landscape monitoring and Earth observation. The newly launched mission, Ice, Cloud, and land Elevation Satellite-2 (ICESat-2), provides photon counting light detection and ranging (LiDAR) data with high spatial resolution and dense samples along-track, offering new opportunities for measuring three-dimensional (3D) forest structure and detecting fire disturbance.
This research focuses on analyzing the feasibility of using ICESat-2 data to classify burned forests and evaluate biomass burning emissions of large wildfires. The main objectives are: 1) investigating the feasibility of classifying burned forest with ICESat-2 data; 2) analyzing the scale effects of segment length on burned forest classification when using ICESat-2 data; 3) estimating fire caused carbon emissions by integrating ICESat-2, Landsat 8, and Sentinel-1 data. Based on the analysis, the accuracy of burned forest classification reached 83% with the Random Forest model using ICESat-2 Land and Vegetation Height product (ATL08). Segment length was found to significantly influence the canopy structure measurements. Additionally, the classification accuracy of burned forest increases along with coarser spatial resolution and saturates at 100m segment length. For carbon emissions, a hierarchy model was established and employed to estimate consumed biomass as well as biomass burning emissions. Forests have the highest biomass burning emissions, 30.59 Mg ha-1, which is approximately twice and six times the emissions from shrubs and grasses, respectively. This study highlights the merit of adopting ICESat-2 photon counting data for fire mapping and biomass burning emission estimation, opening an avenue for accurate forest disturbance and carbon dynamics monitoring as well as climate change mitigation.
Citation
Liu, Meng (2022). Feasibility of Burned Forest Classification and Biomass Burning Emission Estimation Using ICESat-2 Data. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /197211.