Remote sensing and geoinformatics applied to forests. Works with radar and optical satellite data, Sentinel-1 coherence time series, Sentinel-2 spectral indices, and embeddings from Earth observation foundation models, to detect where forests have been disturbed, when the change happened, and what is growing back. His interest is in making these signals accurate and repeatable enough that forest managers, not only researchers, can act on them.
Education
Professional Experience
Where has the forest changed, when did the change happen, and what is standing there now? Satellites pass over Central Europe every few days and record the answer whether or not anyone is looking. Prosper´s work is turning that record into something usable.
He combines Sentinel-1 radar, which sees through cloud and works at night, with Sentinel-2 optical imagery to map disturbance and harvesting across large areas. Radar coherence is particularly good at catching the moment a stand is felled or salvaged, which is difficult to see in optical data alone. For working out what grows back, Prosper increasingly uses embeddings from Earth observation foundation models: instead of reducing a year of imagery to a handful of indices, these models compress the full seasonal signal of each pixel into a learned representation, which separates species that look nearly identical in any single image.
Knowing not only how much forest was lost but which species are replacing it is what makes the difference between describing a disturbance and understanding what it leaves behind.
I run, which is where most of my thinking about a problem actually happens, and I play rugby, a useful reminder that some things cannot be solved alone or from a desk.
