Predicting Surface Fuel Models and Fuel Metrics Using Lidar and CIR Imagery in a Dense, Mountainous Forest

We compared the ability of several classification and regression algorithms to predict forest stand structure metrics and standard surface fuel models. Our study area spans a dense, topographically complex Sierra Nevada mixed-conifer forest. We used clustering, regression trees, and support vector machine algorithms to analyze high density (average 9 pulses/m(2)), discrete return, small-footprint lidar data, along with multispectral imagery. Stand structure metric predictions generally decreased with increased canopy penetration. For example, from the top of canopy, we predicted canopy height (r(2) = 0.87), canopy cover (r(2) = 0.83), basal area (r(2) = 0.82), shrub cover (r(2) = 0.62), shrub height (r(2) = 0.59), combined fuel loads (r(2) = 0.48), and fuel bed depth (r(2) = 0.35). While the general fuel types were predicted accurately, specific surface fuel model predictions were poor (76 percent and <50 percent correct classification, respectively) using all algorithms. These fuel components are critical inputs for wildfire behavior modeling, which ultimately support forest management decisions. This comprehensive examination of the relative utility of lidar and optical imagery will be useful for forest science and management.