A novel post-fire method to estimate individual tree crown scorch height and volume using simple RPAS-derived data

BACKGROUND: An accurate understanding of wildfire impacts is critical to the success of any post-fire management framework. Fire severity maps are typically created from satellite-derived imagery that are capable of mapping fires across large spatial extents, but cannot detect damage to individual trees. In recent years, higher resolution fire severity maps have been created from orthomosaics collected from remotely piloted aerial systems (RPAS). Digital aerial photogrammetric (DAP) point clouds can be derived from these same systems, allowing for spectral and structural features to be collected concurrently. In this note, a methodology was developed to analyze fire impacts within individual trees using these two synergistic data types. The novel methodology presented here uses RPAS-acquired orthomosaics to classify trees based on a binary presence of fire damage. Crown scorch heights and volumes are then extracted from fire-damaged trees using RPAS-acquired DAP point clouds. Such an analysis allows for crown scorch heights and volumes to be estimated across much broader spatial scales than is possible from field data. RESULTS: There was a distinct difference in the spectral values for burned and unburned trees, which allowed the developed methodology to correctly classify 92.1% of trees as either burned or unburned. Following a correct classification, the crown scorch heights of burned trees were extracted at high accuracies that when regressed against field-measured heights yielded a slope of 0.85, an R-squared value of 0.78, and an RMSE value of 2.2 m. When converted to crown volume scorched, 83.3% of the DAP-derived values were within ± 10% of field-measured values. CONCLUSION: This research presents a novel post-fire methodology that utilizes cost-effective RPAS-acquired data to accurately characterize individual tree-level fire severity through an estimation of crown scorch heights and volumes. Though the results were favorable, improvements can be made. Specifically, through the addition of processing steps that would remove shadows and better calibrate the spectral data used in this study. Additionally, the utility of this approach would be made more apparent through a detailed cost analysis comparing these methods with more conventional field-based approaches.