Loosely packed biomass dusts may self-heat and spontaneously ignite, especially when kept under elevated temperatures as in power plant mills. In this work, a mechanistic model was developed to predict self-ignition of biomass under such conditions. The model takes temperature- and gas phase species gradients into consideration. Reaction kinetic parameters were taken from a previous work, while material properties of biomass where found in the open literature. The model therefore did not require any parameter fitting. Model equations were discretized in one spatial dimension (here: for a cylinder). A series of lab scale experiments (10–40 g biomass dust) with beech, pine, sunflower husk pellets and wheat straw were used to validate the model. Predicted ignition temperatures were in good agreement (≈5% error) with experimental data. A sensitivity analysis showed the model to be most sensitive to reaction kinetic parameters, and to a lesser degree towards parameters influencing heat transfer. Mass transfer limitations (oxygen diffusion) did not appear to have a significant effect on the predicted onset of ignition. A scaling study showed sample size to have a larger influence on ignition temperatures than bulk density of the sample or oxygen availability. This study demonstrates that models based on chemical kinetics, heat- and mass transfer phenomena (as opposed to numerical correlations based on the Frank-Kamenetskii method) can yield accurate predictions of self-ignition temperatures. It also underlines the importance of finding realistic reaction kinetic parameters for low temperatures.