Spatiotemporal Dynamics of Afrotemperate Forests in KwaZulu-Natal, South Africa

Abstract

This study investigated Afrotemperate forests’ extent and composition from 1940 to 2022 in KwaZulu-Natal, South Africa. Thirty patches were examined (10 small patches (~one hectare), 10 medium patches (~ between 1 and 1000 hectares ) and 10 large patches (~1000 hectares)) over three time periods (1940s, 1980s and 2020s). Aerial photographs, with their high spatial resolution and long temporal extent, were used to estimate forest extent, composition (indigenous versus non-indigenous) and the surrounding land cover. Georeferencing and digitising of aerial photographs was done in QGIS. Fire extent and frequency from 1990-2024 was calculated seasonally using the difference in Normalised Burn Ratio (dNBR) on Landsat and Sentinel images in Google Earth Engine and ArcGIS Pro. Generalised Linear Mixed Models (GLMMs) were run to determine the relationship between forest extent, surrounding land cover and fire. The results of this study indicated that changes in forest extent were patch size specific, with smaller and larger patches increasing in extent and medium patches decreasing in extent. Smaller forests’ expansion in extent was due to the invasion of non-indigenous vegetation. The invasion of non-indigenous vegetation was aided by the high fire extent and frequency that the small forests experienced. Medium sized patches had the highest amount of non-indigenous vegetation in their surrounding matrix which can contribute to the reduction of available water and the degradation of indigenous forests. Larger forest patches are able to buffer the effects of changes in fire and land cover patterns which can aid in indigenous forest expansion. The surrounding matrix of the larger forests had the highest amount of abandoned agricultural land which can promote forest expansion and bush encroachment. Fire extent was highest for the smaller forests possibly due to the high amounts of non-indigenous vegetation found within them. Fire frequency was highest in medium forests’ matrices possibly due to the medium forests having the highest levels of non-indigenous vegetation in the surrounding matrix and the lowest reduction in herbaceous vegetation in the surrounding matrix. Statistically fire extent and land cover were found to impact forest extent independently. The results of this study indicate that small and medium sized forests are at a higher risk of degradation in comparison to larger forests. Therefore the results of this study can be used to inform decisions around where to focus prioritisation efforts for forest conservation.