The observational capabilities provided by Sentinel-1 SAR (S1) and Sentinel-2 multi-spectral have considerable potential to enhance information retrievals related to time series approaches.
Cloud coverage often severely limits the utility of multi-spectral time-series, while SAR data remains largely unaffected by clouds. At the same time, many biophysical parameters cannot be derived directly from SAR data alone.
Combining and integrating coinciding SAR and optical observations through data fusion is therefore of great relevance to overcome these shortcomings. Data fusion methodologies have benefited from recent advances in new machine learning (ML) approaches, offering new opportunities for efficient SAR-optical data fusion at the data and parameter level.
This activity seeks to develop a data fusion framework for S1 SAR and S2 multi-spectral data, taking advantage of new ML approaches, enabling the creation of continuous fused data streams. This will be expanded by advanced time-series modelling and analytic approaches that allow enhanced characterisation of various land surface dynamics.
The developed framework will be assessed and evaluated in an experimental set-up regarding its performance to detect and predict (near-term) change in different components of the land environment, with a special emphasis on agriculture and food systems. Finally the developed framework will be deployed as a cloud-based public EO service offering, enabling scalable data fusion and advanced time series analytics as a service.
Learn more about this Invitation To Tender on the esa-star Publication page.