06-12, 15:00–15:15 (Europe/London), Sala Biblioteca @ PoliBa
Copernicus Sentinel-2 satellite constellation allows to sense Earth surface at high spatial and spectral resolution and its high revisit frequency foster new advances for land monitoring capacity. Sentinel-2 MSI data exhibit variable geolocation spatial accuracy, resulting in a weak spatial coherence that significantly affect time series consistency at pixel level. Despite evolving Sentinel-2 MSI processing baselines aims, among other objectives, to improve image co-registration with respect to a Global Reference Image (GRI), geospatial accuracy is not yet adequate for detailed time series analysis. Many methodologies to quantify image shifts, developed in the past years, require a significant computational effort to effectively co-register satellite acquisition time series.
To undertake operational image co-registration, Sentinel-2 Shift DataBase (S2SDB) has been established. The S2SDB contains information about horizontal linear local shifts, that can be easily applied to any Sentinel-2 MSI spectral band or derived spatially explicit products, using various image processing software solutions. The DataBase, by releasing simple but relevant information with an open access data policy, can contribute to reduce time and computational effort required to significantly improve Sentinel-2 MSI imagery spatial coherence and time series consistency. Improved co-registration may also contribute to strengthen satellite sensor interoperability, producing denser time series to improve Earth observation land monitoring for a wide range of applications.
S2SDB is freely accessible from the open access data repository available at link https://github.com/ffilipponi/S2SDB.
Improvements in time series consistency at pixel level using the S2SDB is demonstrated for selected case studies, related to monitoring of forest disturbances for logging identification and to the use of time series analysis for the estimation of phenological metrics at Italian national scale.
Filipponi Federico is a research scientist with 10 year relevant experience.
Research interests focus on the ecological processes investigated by use of Remote Sensing and spatio-temporal statistical analysis. Research topics are the investigation of spatial and temporal patterns of physical and ecological processes linked to the development of environmental processes at different ecological scales. Main activities concern remote sensing data processing, software code development, in-situ data collection, geostatistical analysis, environmental characterization.