The unsustainable exploitation of savanna woodland resources in Botswana threatens the capabilities of the ecosystem to provide goods and services that are vital to rural household livelihoods, hence affecting the ecosystem sustainability and rural livelihoods. Therefore, this study documented the levels of rural households’ dependence on and socio-economic factors contributing to woodland resource dependency in Foley and Makomoto localities. Data was obtained in 92 sampled households using semi-structured questionnaires. The probit model was used to predict factors contributing to woodland resource dependency. Results established that rural household adapted various portfolios of livelihood activities which contribute to household income. Furthermore, woodland resources have a great significance to rural households’ well-being. The results show that the income derived from the woodland resources constitutes 58% in Makomoto and 43% in Foley of the total household income. The econometric results revealed that education, place of residence, non-agricultural income and household size significantly influenced the dependence on woodland resources . Woodland dependency decreased with educational level and non-agricultural income. Hence, a need to promote education and non-farming enterprises in rural communities. To combat unsustainable exploitation of the woodland resources, this study recommends the introduction of a community monitoring tool to rural communities.

Savanna woodlands make a significant contribution to rural households’ livelihoods, providing a wide range of resources as well as generating income. However, the sustainability of the woodland ecosystem is generally affected by the human harvesting intensities. This study aimed at quantifying the woodland resource extraction rates and its effects on woodland structure and density. Data were collected from households using a semi-structured questionnaire, key informant interviews, and from 128 rectangular plots measuring 30 m × 30 m. Vegetation parameters, such as tree density, tree height, diameter at breast height, species, stump density, and stump diameter, were recorded. The results of the study revealed that a total of 649 woody stumps were recorded, with an average of 56 trees removed per hectare. The findings showed that the intensities of stumps varied widely, with a high number of stumps recorded at a distance of 10–15 km from the settlements, where major land-use activities are fuelwood extraction, cultivation, and livestock farming. The results also showed that the increased commercial fuelwood production led to cutting of large mopane trees, which is an indicator of unsustainable harvesting. Based on the harvest rates provided by the respondents, results showed that large quantities of fuelwood were harvested for trading, which could put pressure on the woodland ecosystem, consequently resulting in woodland degradation. This study provides forest and range resource managers with valuable information on the quantities of stumps as an indication of tree removals and could be useful in developing effective monitoring strategies and promoting sustainable forest and woodland management.

Land use and land cover (LULC) dynamics are increasingly shaped by environmental and socio-economic activities, which has modified natural habitats and altered ecosystem functions across many African cities. However, comprehensive spatiotemporal assessments of land use and land cover (LULC) change in relation to population growth remain limited. Remote sensing combined with machine learning and deep learning algorithms provides an effective means of capturing spatiotemporal environmental changes associated with land use and land cover dynamics. This study examined land use and land cover (LULC) dynamics in Gaborone, Botswana, from 1991 to 2021 using the random forest classification algorithm. The study also predicted LULC changes for the year 2031 using One-dimensional convolutional neural networks (1D-CNN). Furthermore, statistical analysis (Pearson correlation and regression) was used to examine the relationship between population dynamics and LULC change. The Random Forest model produced more accurate land cover maps at ten-year intervals between 1991 and 2021. The LULC prediction for 2031 also produced acceptable values, with overall accuracy values of 0.88, 0.85, 0.82 and 0.84, respectively. The findings revealed a notable increase in the built-up area from 15% in 1991 to 31% in 2021, bare ground increased from 2% to 7%, while vegetation declined from 57% to 41% and agriculture decreased 5% to 3%. 1D-CNN projections for 2031 indicated that built-up areas will expand to 41%, with a decrease in vegetation and agricultural land. Statistical analysis revealed a positive correlation between population growth and the expansion of built-up areas and negatively correlated with the vegetation category. This study provides insights on future land use and land cover transformation, which can be used to achieve sustainable urban development.