Abstract
Dry forests in Sub-Saharan Africa are of critical importance for the livelihood of the local population given their strong dependence on forest products. Yet these forests are threatened due to rapid population growth and predicted changes in rainfall patterns. As such, large-scale woody cover monitoring of tropical dry forests is urgently required. Although promising, remote sensing-based estimation of woody cover in tropical dry forest ecosystems is challenging due to the heterogeneous woody and herbaceous vegetation structure and the large intra-annual variability in the vegetation due to the seasonal rainfall. To test the capability of Sentinel-2 satellite imagery for producing accurate woody cover estimations, two contrasting study sites in Ethiopia and Tanzania were used. The estimation accuracy of a linear regression model using the Normalised Difference Vegetation Index (NDVI), a Partial Least Squares Regression (PLSR), and a Random Forest regression model using both single-date and multi-temporal Sentinel-2 images were compared. Additionally, the robustness and site transferability of these methods were tested. Overall, the multi-temporal PLSR model achieved the most accurate and transferable estimations (R2 = 0.70, RMSE = 4.12%). This model was then used to monitor the potential increase in woody coverage within several reforestation projects in the Degua Tembien district. In six of these projects, a significant increase in woody cover could be measured since the start of the project, which could be linked to their initial vegetation, location and shape. It can be concluded that a PLSR model combined with Sentinel-2 satellite imagery is capable of monitoring woody cover in these tropical dry forest regions, which can be used in support of reforestation efforts.
Highlights
In Sub-Saharan Africa, forests are crucial for the local population as they provide a range of ecosystem services such as the supply of firewood, food, and pharmaceuticals, maintenance of biodiversity, and management of water and carbon cycles in the soil [1,2]
In order to test the ability of the different methods to quantify the woody cover using Sentinel-2 satellite imagery at each site individually, they were first trained and tested with data from the same site only
The Partial Least Squares Regression (PLSR) and the Random Forest regression were run using three different combinations of predictor variables, one including only the monthly Sentinel-2 bands, one including both the monthly Normalised Difference Vegetation Index (NDVI) value and Sentinel-2 bands and one including the NDVI values of the whole year (Tables 5 and 6). For both the PLSR and the Random Forest regression models, the best scoring months shifted and achieved better results when NDVI was added as an extra predictor variable to the Sentinel-2 bands
Summary
In Sub-Saharan Africa, forests are crucial for the local population as they provide a range of ecosystem services such as the supply of firewood, food, and pharmaceuticals, maintenance of biodiversity, and management of water and carbon cycles in the soil [1,2]. Predicted changes in rainfall patterns caused by climate change are expected to affect the dry woodlands and forests [4] Within this context, there is a growing focus on more sustainable forest management and reforestation projects in tropical forest areas, dry forests and savannas are still under-represented [1,5]. Canopy structural parameters are measured in situ as indicators of forest stand development and they are needed to assess primary productivity, water cycles, and carbon stocks [6,7]. One important structural measure is woody cover It is defined by Gonsamo et al [6] as “the sum of the vertical projection areas of the tree crowns divided by the horizontal area on which the trees are growing”. Woody cover is a proxy for the spatial heterogeneity and fragmentation of an area, and can be related directly to species competition and diversity, why it is one of the most used structural parameters [6,8,9]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
