International Journal of Applied Earth Observation and Geoinformation · 2020 · 62 citations · 44 references
Natural ResourcesForestryForest ConservationForest ManagementRemote SensingDeforestationForest Change DetectionForest Disturbance DetectionForest Disturbance DynamicsVegetation Change DetectionVegetation MonitoringDeforestation DetectionDeforestation DynamicsDeforestation MappingDeforestation MonitoringForest MonitoringLong-term Forest MonitoringMontane ForestsSpatiotemporal Forest Monitoring
Naturally isolated montane forests in East Africa are hotspots of biodiversity, often characterised by high species endemism, and are fundamental contributors to water services. However, they are located in areas highly suitable for agriculture, making them a prime target for agricultural activities. The Eastern Arc Mountains (EAM) in Eastern Tanzania are within the target regions for agricultural development under the Southern Agricultural Growth Corridor of Tanzania (SAGCOT). However, forest monitoring initiatives that track long-term forest dynamics and the ecological impact of current agricultural development policies on forests, are lacking. Here, we use the STEF (Space-Time Extremes and Features) algorithm and Landsat time series to track forest disturbances (deforestation and degradation) and forest gains (regeneration) as spatio-temporal events over seventeen years (2001–2017) in the montane forests of the Mvomero District in Tanzania. We found that 27 % (∼ 20 487 ha) of montane forests were disturbed between 2001 and 2017, mainly led by deforestation (70 %). Small-scale crop farms with maize, banana, and cassava crops, were the most planted on deforested areas. Most disturbances occurred at lower elevation (lowland montane), but there was an increasing shift to higher elevations in recent years (2011–2017). Forest disturbances exclusively occurred at small spatial scales, a pattern similar to other forest montane landscapes in Africa, which lowers detection capabilities in global forest loss products. Our locally calibrated and validated deforestation map (Producer's accuracy = 80 %; User’s accuracy = 78 %) shows a gross underestimation of forest cover loss (>10 000 ha) by global forest loss products in these mountainous forest landscapes. Overall, we found few areas undergoing forest regeneration, with only 9 % of the disturbed forest regenerating over 17 years. Long-term conversion to cropland prevented regeneration in the lowlands, with regeneration mainly happening at higher elevations. However, the shift of deforestation and forest degradation to higher elevations may challenge high elevation regeneration trends, leaving the remaining blocks of montane forest in the Mvomero District at a risk of degradation and disappearance. Without effective forest conservation measures, market-driven agricultural development is likely to trigger an expansion of cropland at the expense of forests to meet the increased demand for the agricultural products promoted, with negative impact on biodiversity, carbon sequestration and water services.
44
Leo Breiman · Machine Learning · 2001 · 126.2K citations · Full text
Biodiversity hotspots for conservation priorities
Norman Myers, Russell A. Mittermeier, Cristina G. Mittermeier et al. · Nature · 2000 · 31.2K citations · Full text
Natural Resources, Biodiversity Conservation, Conservation Biology +13
Classification and Regression by randomForest
Andy Liaw, Matthew C. Wiener · 2007 · 18.4K citations
High-Resolution Global Maps of 21st-Century Forest Cover Change
Matthew C. Hansen, Peter Potapov, Rebecca Moore et al. · Science · 2013 · 11.6K citations