Abstract
Elevated soil temperatures resulting from reintroduction of prescribed fire into long unburnt stands have been associated with unintended tree mortality. Several models exist to predict soil temperatures resulting from soil heating by fire; however, data to develop and validate these models are limited. A model to predict soil temperature at depths up to 25 inches (0.63 m) was developed from a data set from 46 prescribed burns in coniferous forests in national forests and parks in Arizona and California. Soil temperature was less than 140 °F (60 °C) at depths greater than 6 inches (0.15 m) and constant below 10 inches (0.25 m). Using a Bayesian generalized nonlinear additive model, nine models formed from combinations of soil and humus moisture contents, fuel consumption and tree species were fit to soil temperature data for Pinus ponderosa, P. lambertiana , and Sequoiadendron gigantea . Bayesian R 2 for the full model and the reduced model containing tree species and fuel consumption was 0.70 and 0.67, respectively. The Bayesian model predicted higher maximum temperatures than two soil heating models in the First Order Fire Effects Model. Based on parsimony, the model using fuel consumption and tree species is recommended for use.