Migrations have been a feature of human existence for centuries. The migrations of migrants who risk their lives to reach Europe via the Mediterranean have increased dramatically in recent times. Their journey often ends in refugee camps where they are housed in poor conditions. In this dissertation we use remote sensing data in the context of the broader issue of migration. We focus on the detection of migrant vessels at sea and the environmental impact that increased migration can have around refugee centers.
In order to obtain data on the movement of migrants at sea, we investigate the possibilities of using a variety of (very) high-resolution optical satellite images. We have developed a method for automatic vessel detection consisting of four consecutive steps: sea-land separation, candidate detection, vessel discrimination and vessel classification. The results show that the developed algorithm more accurately detects vessels from images with more favourable weather conditions. False positives are a greater
problem in detection than undetected vessels. To detect changes in the vicinity of the refugee camps, we used the time series analysis BFAST (Breaks For Additive Season and Trend) Monitor, which monitors disturbances in time series based on a model for stable historical behaviour. The analysis was applied on Sentinel-2 images in areas of refugee camps on the Mediterranean islands that have been experiencing long term influxes of migrants. We observed
negatively detected changes in the NDVI (normalised difference vegetation index) in 2019. Sentinel-2 data proved to be suitable for time series analysis due to their dense time series. The assessment of potential environmental impacts in the vicinity of refugee camps are further determined on the basis of probability classes defined according to the negative magnitude of the observed changes.