**5. Conclusions**

Savannas are an extremely important socio-economic landscape in Africa. These landscapes are inherently multi-functional, balancing the needs of pastoral societies with conservation of these dynamic ecosystems. Global trends of savanna to shrubland conversion [6,7] will have important ecological and economic consequences. Here we investigated the impact of regional scale environmental drivers (a scale that is more relevant to governments and land managers across Africa and beyond) on woody vegetation diversity and abundance. Data on over 44 species was collected over a two-year period at fifteen sites along the Kalahari transect. At each site, six 100 m transects recorded diversity ranging from one species to thirteen species, and abundance ranging from two individuals to 62 individuals. A mixture of Poisson and Tobit regression models identified that rainfall was the most important environmental variable when all species were considered equally, corroborating previous research conducted at continental [11,16] and garden [17,19] scales. Interestingly, bipinnate species abundance decreased with increasing boreholes and livestock. These results contradict the consensus that borehole density and grazing increase the presence of such species, and sugges<sup>t</sup> that by rotating cattle between boreholes, the impact of trampling and grazing on the soil is reduced and savanna landscapes are maintained. The deconstruction of species into different morphological groups provided better insights into the differences in the ways woody vegetation responds to environmental factors, and this deconstruction could aid in reconciling the divergent hypotheses surrounding woody vegetation patterns in savanna ecosystems, as all variables had a significant relationship with richness and abundance across all morphological groups. The results of this research should support land managers, governments and researchers working in transitional savanna landscapes worldwide.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-445X/8/8/122/s1, Data.

**Author Contributions:** Conceptualization, T.M., P.H., J.A.M., T.B.C., P.D.O., G.S.O.; Data Collection, T.M., P.D.O., G.S.O.; Data Analysis, T.M., P.H., J.A.M., T.B.C.; writing—original draft preparation, T.M., P.H., J.A.M., T.B.C., P.D.O., G.S.O.; writing—review and editing, T.M., P.H., J.A.M.

**Funding:** This research was funded by National Science Foundation, gran<sup>t</sup> number DEB-0717448 supported T.M., P.D.O., and G.S.O., and gran<sup>t</sup> number 0962198 supported P.H., and J.A.M.

**Acknowledgments:** We would like to thank the numerous volunteers and field assistants who spent months measuring vegetation under difficult field conditions. We would like to offer a special thank you for support in the field to Kelley A Crews-Meyer and our field assistant Gully. Finally, we would like to thank editors and the anonymous reviewers for their comments and suggestions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **Appendix A Derived Burned Area Product**

The MODIS MCD64A1 (LP DAAC 2010b) burned area product as described by Giglio et al. (2009) was used to determine the fire locations and fire frequency across Botswana from year 2001 to 2011. Fire frequency was calculated on a per pixel basis based on MODIS derived fire events, with frequency ranging from no fire to a maximum of eleven fires recorded for the northern part of the country along the northern border with Namibia. Due to fire frequency having more influence on vegetation than time since last fire, the estimated uncertainty in date of burn was not incorporated in the fire frequency product as this does not impact the output. Figure A1 shows that most areas that burned have a fire frequency between one fire and three fires for the observed time frame. Most of these areas are located in the northern and central parts of the region. Across all field sites, fire frequency ranged from 0–6, with a total of eight sites having been affected at least once during the time period. Return intervals for fire occurrences seem to be higher at the northern site locations (sites 1–6) while all central to southern sites (sites 7–15) with the exception the Bere and Tshane South (site 10 and 12) were unaffected by fire.

**Figure A1.** Fire frequency and burned area across Botswana from 2001 to 2011.
