Joni Koskikala

Spatially explicit, evidence-based and regionally contextualized data on biophysical landscape characteristics is an essential basis for regionally sustainable landscape management schemes. In many regions of the Global South, the availability of such information is poor, especially at the subnational level, and the spatial management is often based on generic and outdated information, leading to severe threats for land sustainability. We have developed a biophysical regionalization of the Southern Highlands area of Tanzania. The map is based on open-source global datasets depicting climate, soil, topography and vegetation. Through replicable statistical and geospatial analyses, we have identified 7 regions and 18 subsections with biophysically similar and spatially distinctive environmental conditions. The regions provide spatially contextualized support for understanding and managing the landscapes of the Southern Highlands. The applications for such data sets are numerous, from screening suitability areas for e.g. afforestation schemes to evaluating the distinctiveness and vulnerability of landscapes to degradation.

Global terrestrial biodiversity hotspots (GBH) represent areas featuring exceptional concentrations of endemism and habitat loss in the world. Unfortunately, geospatial data of natural habitats of the GBHs are often outdated, imprecise, and coarse, and need updating for improved management and protection actions. Recent developments in satellite image availability, combined with enhanced machine learning algorithms and computing capacity, enable cost-efficient updating of geospatial information of these already severely fragmented habitats. This study aimed to develop a more accurate method for mapping closed canopy evergreen natural forest (CCEF) of the Eastern Arc Mountains (EAM) ecoregion in Tanzania and Kenya, and to update the knowledge on its spatial extent, level of fragmentation, and conservation status. We tested 1023 model possibilities stemming from a combination of Sentinel-1 (S1) and Sentinel-2 (S2) satellite imagery, spatial texture of S1 and S2, seasonality derived from Landsat-8 time series, and topographic information, using random forest modelling approach. We compared the best CCEF model with existing spatial forest products from the EAM through independent accuracy assessment. Finally, the CCEF model was used to estimate the fragmentation and conservation coverage of the EAM. The CCEF model has moderate accuracy measured in True Skill Statistic (0.57), and it clearly outperforms other similar products from the region. Based on this model, there are about 296,000 ha of Eastern Arc Forests (EAF) left. Furthermore, acknowledging small forest fragments (1–10 ha) implies that the EAFs are more fragmented than previously considered. Currently, the official protection of EAFs is disproportionally targeting well-studied mountain blocks, while less known areas and small fragments are underrepresented in the protected area network. Thus, the generated CCEF model should be used to design updates and more informed and detailed conservation allocation plans to balance this situation. The results highlight that spatial texture of S2, seasonality, and topography are the most important variables describing the EAFs, while spatial texture of S1 increases the model performance slightly. All in all, our work demonstrates that recent developments in Earth observation allows significant enhancements in mapping, which should be utilized in areas with outstanding biodiversity values for better forest and conservation planning.

Remotely sensed assisted forest inventory has emerged in the past decade as a robust and cost efficient method for generating accurate information on forest biophysical parameters. The launching and public access of ALOS PALSAR-2, Sentinel-1 (SAR), and Sentinel-2 together with the associated open-source software, has further increased the opportunity for application of remotely sensed data in forest inventories. In this study, we evaluated the ability of ALOS PALSAR-2, Sentinel-1 (SAR) and Sentinel-2 and their combinations to predict growing stock volume in small-scale forest plantations of Tanzania. The effects of two variable extraction approaches (i.e., centroid and weighted mean), seasonality (i.e., rainy and dry), and tree species on the prediction accuracy of growing stock volume when using each of the three remotely sensed data were also investigated. Statistical models relating growing stock volume and remotely sensed predictor variables at the plot-level were fitted using multiple linear regression. The models were evaluated using the k-fold cross validation and judged based on the relative root mean square error values (RMSEr). The results showed that: Sentinel-2 (RMSEr = 42.03% and pseudo − R² = 0.63) and the combination of Sentinel-1 and Sentinel-2 (RMSEr = 46.98% and pseudo − R² = 0.52), had better performance in predicting growing stock volume, as compared to Sentinel-1 (RMSEr = 59.48% and pseudo − R² = 0.18) alone. Models fitted with variables extracted from the weighted mean approach, turned out to have relatively lower RMSEr % values, as compared to centroid approaches. Sentinel-2 rainy season based models had slightly smaller RMSEr values, as compared to dry season based models. Dense time series (i.e., annual) data resulted to the models with relatively lower RMSEr values, as compared to seasonal based models when using variables extracted from the weighted mean approach. For the centroid approach there was no notable difference between the models fitted using dense time series versus rain season based predictor variables. Stratifications based on tree species resulted into lower RMSEr values for Pinus patula tree species, as compared to other tree species. Finally, our study concluded that combination of Sentinel-1&2 as well as the use Sentinel-2 alone can be considered for remote-sensing assisted forest inventory in the small-scale plantation forests of Tanzania. Further studies on the effect of field plot size, stratification and statistical methods on the prediction accuracy are recommended.