Current socio-economic drivers of land-use change associated with globalization are producing two contrasting land-use trends. On the one hand, the continuing increase in global food demand accelerates the conversion of natural ecosystems into agricultural lands resulting in loss of biodiversity and ecosystem goods and services. On the other hand, the concentration of modern agriculture on the most productive soils has favoured the abandonment of marginal agricultural land and recovery of forested land trough natural regeneration or afforestation. This process of forest transition, i.e. the change from decreasing to expanding national forest areas, is currently taking place in several countries. These transitions are often associated with positive feedbacks on ecosystem goods and services. Forest recovery occurs on marginal lands, such as mountain environments. Assessing the rate, spatial patterns and ecosystem impacts of forest cover change in these areas is challenging given the ruggedness and inaccessibility of mountains. Remote sensing methods are the privileged tool, and yet suffer from methodological challenges due to topographical and shadowing effects. Recent techniques have been developed to correct high and very high resolution imagery for radiometric and geometric, and illumination and shade effects on steep surfaces. These sophisticated correction methods are not only highly labour intensive, but also demand site-specific calibration which makes them particularly difficult to apply in streamlined processing schemes. At present, it is not clear what the added value of complex pre-processing techniques is compared to relative simple empirical methods and to what extent more sophisticated processing enhances results of subsequent analyses. This project will specifically address this methodological research question, and will develop an optimal pre-processing chain to be used for semi-automatic analyses of high resolution satellite data on mountainous terrain. The project aims at a better understanding of the impact of pre-processing techniques on the detection accuracy of forest transitions and the mapping accuracy of ecosystem services. In this study, remote sensing images will be used to derive spatial proxies of ecosystem goods and services. The Ecuadorian Andes was selected as study site, as this region is characterized by intense land use changes with major environmental threats. Two major ecosystem services will be evaluated: carbon storage in aboveground biomass and natural hazard regulation (flooding, landsliding and soil erosion).  Regarding natural hazard regulation, landslides will be detected based on high resolution remote sensing data. For this purpose, multi-spectral images have proven to be effective for landslide mapping, particularly using band combinations, fusion techniques and vegetation indexes. In a first step, semi-automated procedures, based on 15m resolution ASTER images but also on instability factors such as topographic characteristics, are being validated by very-high resolution remote sensing imagery used for ground-truthing. Secondly, these regional landslide inventories will be used to calibrate and validate a slope stability model. After this model adjustment, and based on forest cover maps, we will run the model in a dynamic mode to create slope stability maps as an indicator of natural hazard regulation. Concerning runoff quantity and quality, hydrological data will be used to calibrate semi-distributed hydrological models and distributed erosion models in order to obtain indicators of flood control and erosion regulation. There is no direct methodology to measure forest carbon stocks across a landscape. Therefore, we will indirectly estimate carbon storage in aboveground biomass by combining information from existing ground based forest inventories with very high-resolution aerial imagery. Forest attributes (diameter at breast height, tree height) will then be converted into estimates of forest carbon stocks using empirical equations. Changes in carbon stocks (emission or sequestration) will then be estimated from the satellite-based forest cover change map and the average carbon stock value for different forest types and densities. This quantitative approach will allow us to quantify changes in different environmental services after forest cover change.