Silvilaser 2019 - Poster Presentations »
Estimating Burned Ecosystem Properties with NEONs Airborne Observation Platform
Wildfires represent a critical component of the global carbon budget and generate atmospheric pollutants that have negative short and long-term human and environmental consequences. Characterizing the full impact of a wildfire requires observation of the fire emissions as well as the affected ecosystem. In the summer of 2018, the BB-FLUX wildfire emission observation campaign was supported by the National Ecological Observatory Network (NEON) Airborne Observation Platform (AOP) to collect aerial waveform lidar and hyperspectral data over the burn scars of four wildfires in the Western United States. The flights also encompassed some unburned forested area surrounding the burn scars, which were used as a proxy for the pre-burned conditions. The remote sensing observations were used to estimate burned ecosystem parameters of area burned and above ground biomass in order to better constrain the amount of available fuel. We present an algorithm for predicting live biomass using a random forest supervised machine-learning model incorporating lidar and hyperspectral data, and ground-based measurements at nearby NEON sites for training data. Since NEON has collected coincident remote sensing data and ground-based measurements over a number of eco-climactic domains throughout North America, it provides a rich dataset for assessing biomass model performance across variable ecological regimes. We conducted exploratory analysis to determine optimal model parameters for predicting biomass in differing ecosystems. Results of this analysis are shown, along with lessons learned on developing a model that works across large spatial scales and in dissimilar forest types.