•Predictive models of southern and western pine consumption are presented.•Fuel consumption modeling is an important step in emissions prediction and prescribed burn planning.•Models fall within the range of observed biomass and consumption in independent datasets.•Southern pine models of herbs, shrub and fine wood have close model fit with narrow prediction intervals.•Models of coarse wood, litter and duff have greater uncertainty, particularly at high pre-burn biomass. Reliable estimates of pre-burn biomass and fuel consumption are important to estimate wildland fire emissions and assist in prescribed burn planning. We present empirical models for predicting fuel consumption in natural fuels from 60 prescribed fires in ponderosa pine-dominated forests in the western US and 60 prescribed fires in long-needle pine forests in the southeastern US. There was high variability across sites, but total surface fuel biomass was generally much lower on southern sites (23.0±11.6Mgha−1) than western sites (61.5±35.8Mgha−1). Differences in surface fuel composition, pre-burn loading and fuel consumption between the southern and western pine consumption datasets justify the development of regional models for predicting fuel consumption. Southern pine models of herb, shrub and 1-h consumption have close model fit with narrow prediction intervals across the range of sampled values. Relationships between 10-h and 100-h pre-burn loading and consumption produced models with reasonable fit but with no significant correlation with fuel moisture. Model fit of litter and duff consumption models was relatively poor compared to the other southern fuel categories. Western models were developed for 1-h, 10-h and 100-h fine wood, sound coarse wood, rotten coarse wood, litter and duff. All western models had high coefficients of variability, and model residuals indicate higher uncertainty with increasing pre-burn biomass. Although empirical models are widely used, they have limitations in that they are constrained by burning conditions and ranges of predictor variables.