Multitemporal lidar captures heterogeneity in fuel loads and consumption on the Kaibab Plateau

BACKGROUND: Characterization of physical fuel distributions across heterogeneous landscapes is needed to understand fire behavior, account for smoke emissions, and manage for ecosystem resilience. Remote sensing measurements at various scales inform fuel maps for improved fire and smoke models. Airborne lidar that directly senses variation in vegetation height and density has proven to be especially useful for landscape-scale fuel load and consumption mapping. Here we predicted field-observed fuel loads from airborne lidar and Landsat-derived fire history metrics with random forest (RF) modeling. RF models were then applied across multiple lidar acquisitions (years 2012, 2019, 2020) to create fuel maps across our study area on the Kaibab Plateau in northern Arizona, USA. We estimated consumption across the 2019 Castle and Ikes Fires by subtracting 2020 fuel load maps from 2019 fuel load maps and examined the relationship between mapped surface fuels and years since fire, as recorded in the Monitoring Trends in Burn Severity (MTBS) database. RESULTS: R-squared correlations between predicted and ground-observed fuels were 50, 39, 59, and 48% for available canopy fuel, 1- to 1000-h fuels, litter and duff, and total surface fuel (sum of 1- to 1000-h, litter and duff fuels), respectively. Lidar metrics describing overstory distribution and density, understory density, Landsat fire history metrics, and elevation were important predictors. Mapped surface fuel loads were positively and nonlinearly related to time since fire, with asymptotes to stable fuel loads at 10–15 years post fire. Surface fuel consumption averaged 16.1 and 14.0 Mg ha(− 1) for the Castle and Ikes Fires, respectively, and was positively correlated with the differenced Normalized Burn Ratio (dNBR). We estimated surface fuel consumption to be 125.3 ± 54.6 Gg for the Castle Fire and 27.6 ± 12.0 Gg for the portion of the Ikes Fire (42%) where pre- and post-fire airborne lidar were available. CONCLUSIONS: We demonstrated and reinforced that canopy and surface fuels can be predicted and mapped with moderate accuracy using airborne lidar data. Landsat-derived fire history helped account for spatial and temporal variation in surface fuel loads and allowed us to describe temporal trends in surface fuel loads. Our fuel load and consumption maps and methods have utility for land managers and researchers who need landscape-wide estimates of fuel loads and emissions. Fuel load maps based on active remote sensing can be used to inform fuel management decisions and assess fuel structure goals, thereby promoting ecosystem resilience. Multitemporal lidar-based consumption estimates can inform emissions estimates and provide independent validation of conventional fire emission inventories. Our methods also provide a remote sensing framework that could be applied in other areas where airborne lidar is available for quantifying relationships between fuels and time since fire across landscapes.