Skip to main content
Dryad logo

Fire refugia facilitate forest and savanna co‐existence as alternative stable states

Citation

Beckett, Heath; Bond, William J. (2019), Fire refugia facilitate forest and savanna co‐existence as alternative stable states, Dryad, Dataset, https://doi.org/10.5061/dryad.hmgqnk9bs

Abstract

Aim: Does complex topography facilitate the establishment and persistence of fire‐ sensitive (forest) vegetation in a fire‐prone landscape? We test the prediction that fire‐sensitive vegetation will establish and persist in areas where the fire return interval is lower due to a topographic hindrance on fire spread. Location: Hluhluwe–iMfolozi Park, KwaZulu Natal, South Africa.

Methods: Using aerial photographs from six time periods between 1937 and 2013, we mapped vegetation changes in Hluhluwe–iMfolozi Park (HiP). Using a general‐ ized additive model (GAM), we built a habitat suitability index (HSI) map for forest vegetation based on topographic variables related to fire behaviour and the vegeta‐ tion distribution maps from each time period. We investigated transitions between vegetation types between time periods based on the HSI map, as well as the effects of neighbourhood vegetation on transition probabilities.

Results: Forest cover has increased overall from 1937 to 2013, however, this has not been a linear increase with a peak in extent in the early 1990s. The HSI, using topo‐ graphic predictors associated with fire behaviour, correlates with areas of expansion and contraction of forest vegetation. The patterns of expansion and contraction are, however, more nuanced, with the in situ vegetation neighbourhood playing a large role.

Main Conclusions: Forest distributions in HiP have not remained static over time and have expanded into areas that were once savannas. This is a dynamic system where both forest and savanna boundaries can change considerably. Fire refugia are impor‐ tant for the long‐term persistence of forests in fire‐prone landscapes.

Usage Notes

Vegetation maps and elevation rasters