Long-Term Land Use/Land Cover Change Assessment of the Kilombero Catchment in Tanzania Using Random Forest Classification and Robust Change Vector Analysis

摘要: Information about land use/land cover (LULC) and their changes is useful for different stakeholders to assess future pathways of sustainable use food production as well nature conservation. In this study, we LULC in the Kilombero catchment Tanzania, an important area recent development East Africa. change assessed two ways: first, post-classification comparison (PCC) which allows us directly from one class another, second, spectral detection. We perform classification by applying random forests (RF) on sets multitemporal metrics that account seasonal within-class dynamics. For detection, make robust vector analysis (RCVA) determine those do not necessarily lead another class. The combination approaches enables distinguish areas show (a) only PCC changes, (b) affect a pixel, (c) both types change, or (d) no at all. Our results reveal one-quarter has experienced any change. One-third shows both, conversion. Changes detected with methods predominantly occur major regions, West catchment, floodplain. Both regions are economic Tanzania. floodplain Ramsar protected area, half was converted agricultural past decades. Therefore, monitoring required support management. Relatively poor performances revealed several challenges during process. combined approach RCVA detect spatial patterns distinct dimensions intensities. With assessment additional classifier output, namely class-specific per-pixel probabilities derived parameters, uncertainty across space. overlay reliability provide thorough picture taking place catchment.