Patterns of long-term regeneration of forest fire slopes in the Northern European Alps – a logistic regression approach

We report on vegetation recovery after stand-replacing wildfires on steep slopes on limestone and dolostone bedrock. The sites burnt in the nineteenth and twentieth century and some have remained almost bare of vegetation. We mapped and analysed 26 burned slopes to determine site-specific factors influencing regeneration. We followed a multinomial-logistic regression approach to explain the patterns of grassland, shrub and tree vegetation in 10 × 10 m cells. We considered eight potential controlling parameters (time since fire, elevation, aspect, slope, topographical index, distance to forest edge, land cover of neighbouring pixels and lithology). The pseudo R² values for the multinomial-logistic regression point to a reasonably good model fit, even if some potentially important parameters (e.g. post-fire management) could not be incorporated. Tree regeneration is impeded by higher elevation and by steep slope angles while shrubs are unaffected by inclination. Although the probability of pixels with grasslands is higher in proximity to neighbouring grassland pixels, we found no such effect for trees or shrubs. There is no statistical correlation between the type of vegetation and time elapsed since the fire. This means that under adverse conditions, the time required for forest recovery is longer than the time period captured by our investigation (200–300 yr).