Faculty of Mathematics and Natural Sciences - Earth Observation Lab

 

Island2VAP – Integrating Sentinel-2 and Landsat-8 to systematically generate value added products at high resolution

Project period: 04/2015 - 03/2017

 

Quality and quantity of current high resolution optical earth observation data is unprecedented and provides an opportunity to advance remote sensing land system analyses. However, cloud coverage and a lack of gridded higher level products still hampers the widespread usability of the data. This research addresses these shortcomings by developing toolsets to combine data streams from Sentinel-2 and Landsat-8 and that allow for the systematic (i.e. weekly, monthly, seasonal-) generation of composited reflectance and subsequently value-added products (e.g. percent cover estimates, annual phenology metrics). This suite of generated products will be synergetically exploited in order to address higher level land-use science questions that cannot readily be answered using spectral data only. While methods are developed to be capable of working with most ecosystems, a specific focus is on improving agricultural mapping and analyses.

Click on the grafic to enjoy enlarged viewing

3 animated Remote Sensing pictures - described in the Fig1 caption

 

Fig1: 10-day time series of reflectance composites generated from data acquired by Sentinel-2a, Landsat-7 and -8 and Spot-5. Black areas indicate pixel for which no cloud free observation was available in the temporal window. The sensor flag image (middle) indicated which sensor observed which pixel [Red:Spot5, Green:Landsat-7, Blue:Landsat-8, Magenta:Sentinel2a]. The image on right shows the data coverage after performing a bandwise linear interpolation of adjacent cloud free pixel values.

 

NDVI graph for different satellites and remote sensing image ships - picture explained in the caption

 

Fig2: Temporal detail obtained from the time series of 10-day interval composites. Example shows a corn crop for which key stages of phenological development (e.g. green-up, length of season, etc.) can be determined at 20m spatial detail. Symbols denote which sensor observed the pixel at which temporal day-of-year interval. Corresponding image chips are shown below [RGB=NIR, SWIR, Red].


Lab members involved
Funding

This project is funded by the ESA - Living Planet Fellowship