Semi-arid African savanna habitat suitability for two Vachellia species

The habitat suitability of Vachellia stuhlmannii and Vachellia tortilis was assessed using Maximum Entropy model. Location data was collected and nineteen bioclimatic variables data for 1950 to 2000 downloaded from the WorldClim database. Soils were analyzed to investigate their influence on habitat suitability of the two species. MaxEnt effectively predicted current habitat suitability with an average test Area Under Curve value of 0.936 for V. stuhlmannii and 0.689 for V. tortilis. Key influential variables were BIO-1 (Annual mean temperature) with 71.7% highest gain for V. stuhlmannii and BIO-17 (Precipitation of driest quarter) with 65.3% highest gain for V. tortilis. However, BIO-14 (Precipitation of driest month) exhibited limited influence in predicting habitat suitability for both species. V. stuhlmannii occupies twenty-nine hectares and V. tortilis covers 102 hectares. Both species predominantly occur on Prismacutanic/pedocutanic B-horizons. V. tortilis also thrive on Glenrosa and Mispah soils, which are crucial for diverse plant and animal life. These findings have practical implications for conservation efforts aimed at protecting both species. Identifying suitable habitats and preserving soil types is crucial as soil affects microclimate conditions and influences moisture retention. Overall, this study provides insights for the conservation of V. stuhlmannii and V. tortilis in the semi-arid African savanna.

Methods
Data collection
Records of presence for V. stuhlmannii and V. tortilis were derived from Geographic Positioning System (GPS) locations’ data collected in June 2018 (Nkosi et al., 2022a). Different bioclimatic variables are key biological variables that determine species’ environmental niches (Yi, Cheng, Yang, & Zhang., 2016). Historical climatic data of 19 bioclimatic variables (Table 1) were downloaded from the WorldClim database at approximately 1 km2 (30 arc-seconds) spatial resolution (Fick & Hijmans, 2017) from the website: https://www.worldclim.org/. This is climatic data for 1970 – 2000, which was released in January 2020. The data include monthly climate data for minimum, mean and maximum temperatures, precipitation, solar radiation, wind speed, water vapour pressure and total precipitation (Hijmans, Cameron, Parra, Jones, & Jarvis, 2005). 

Data analysis
To build the model for each of the two Vachellia species, partitions were created by randomly selecting 70% of the presence localities as training data with the remaining 30% as test data (Phillips et al., 2006). This resulted in eighty-two training localities and twenty-eight test localities for V. stuhlmannii, and V. tortilis with fifty-six training localities and nineteen test localities. The MaxEnt software version 3.4.4 was downloaded from https://biodiversityinformatics.amnh.org/open_source/maxent and applied in this study. The bioclimatic data were converted to “ASCII” file format in Quantum-GIS (QGIS Development Team, 2012) for compatibility with MaxEnt. The GPS location data of the two species (V. stuhlmannii and V. tortilis) were converted into “CSV” format and the converted file was used as input to MaxEnt. MaxEnt generated output results that predict the suitability of a habitat for each species using a scale of 0-1, where the lowest suitability areas are represented by zero, while the highest suitability areas are represented by 1 (Sharma et al., 2018). Response curves for each predictor variable were generated from MaxEnt. To assess the performance of the model, the metric used was the Area under the ROC (receiver operating characteristic) curve (AUC), which ranges from 0 to 1 and is a measure of the model’s effectiveness (Vanagas, 2004). According to Swets (1988), the closer the AUC value is to one, the better the model’s performance. MaxEnt also uses the jack-knife method to highlight the relative influence of each predictor variable (Khanum, Mumtaz & Kumar, 2013). To produce the potential habitat maps of the two species, the VLNR boundary map was overlaid on the grid file generated by MaxEnt in QGIS.

Soil types (Mucina & Rutherford, 2006) were incorporated into the suitability maps to assess the habitat suitability of the two species in relation to soil characteristics. This is because soil type and structure can impact microclimate conditions by affecting moisture retention (Rost, Gerten, Hoff, Lucht, Falkenmark, & Rockström, 2009). The distribution of habitats could be influenced by the interplay between soil resources and herbivore impact (Skarpe et al., 2004). Moreover, other factors such as variations in soil types could potentially impact the differences observed in the attributes of woody vegetation (Gandiwa et al., 2011). Geology, topography, soil, and water are critical environmental factors that directly affect the structure and composition of vegetation at various scales, ranging from individual plants to landscape levels (Sankaran et al., 2005).