Estimation of Plot-Level Burn Severity Using Terrestrial Laser Scanning

Monitoring wildland fire burn severity is important for assessing ecological outcomes of and their spatial patterning as well guiding efforts to mitigate or restore areas where are negative. Burn mapping products typically created using satellite reflectance data but must be calibrated field derive meaning. The composite index (CBI) the most widely used field-based method calibrate satellite-based limitations this approach have yet resolved. objective study was focused on predicting CBI from point cloud visible-spectrum camera (RGB) metrics derived single-scan terrestrial laser scanning (TLS) datasets determine viability TLS an alternative estimating in field. In our approach, we considered predictive potential post-scan-only metrics, differenced pre- post-scan RGB all three together predict evaluated these with candidate algorithms (i.e., linear model, random forest (RF), support vector machines (SVM) two evaluation criteria (R-squared root mean square error (RMSE)). congruence strata-based observations calculate CBI, approaches at strata level plot 70 10 independent variables that generated data. Machine learning successfully predicted total strata-specific CBI; however, accuracy predictions varied among by algorithm. improved when conjunction variables, alone proved a poor predictor below canopy. Although results highlight TLS-based methods quantifying can improvement over many ways because repeatable, quantitative, faster, requires less field-expertise, more flexible phenological variation biomass change understory prescribed effects pronounced. We also out leveraged inform other monitoring needs beyond those specific fire, representing additional efficiency approach.