A statistical model linking Siberian forest fire scars with early summer rainfall anomalies

[1] Forest fires in Siberia have a significant effect on the global carbon balance. It is therefore of interest to study the environmental factors that may be responsible for their variability. Here we examine variability in the annual number of forest fire scars at a spatial scale of 2.5°. This is decomposed statistically into a spatio–temporal component correlated with low summer rainfall, a spatial component correlated with population density and a temporal component correlated with the Arctic Oscillation. Data come from ten years of satellite–derived data, incorporating both the number of forest fire scars and monthly rainfall. The expected number of fire scars halves for each additional 0.35 mm per day of rainfall in the period April–July. Our findings may prove useful in parameterising both fire models within climate simulations and fire warning systems based on numerical weather predictions of regional dry anomalies.