Insight into Sentinel-2 Global Mosaic products

  • Posted on: 15 March 2019
  • By: S2GM

Insight into Sentinel-2 Global Mosaic products

The Sentinel-2 Global Mosaic Service allows users to choose between multiple compositing options relating to the spatial resolution, the spectral bands, and the period from which observations are considered for compositing. In this blog post, several examples of the resulting mosaic products are presented, and their properties and characteristics are discussed to support users in adapting the compositing parameters to their specific needs. we like to show you some of these features and give you a deeper insight to better adapt the parameters to your needs.

If you are not interested in intra-annual surface changes (vegetation cycles) and want to get a spatially homogeneous composite product, we recommend requesting an annual product is clearly your best choice. As the cloud screening of the Sentinel-2 Level 2A products, which are the inputs to S2GM, does not always work perfect, the input products may suffer from residual clouds. With an increase in available input data, the occurrence of such undesired artifacts typically decreases. Particularly the Medoid algorithm, which is applied if more than four input data are available, is robust against outliers like residual clouds.

Figures 1 & 2 show an annual composite over southern France. It is apparently free of remaining clouds and spatially consistent.

These annual composites could thus be used to make interannual change detection, e.g. for urban areas.

If you are more interested in seasonal changes, a quarterly composite will better suit your needs, as shown in figures 3 & 4. Quarterly products are derived from observations within fixed 3-month periods (Jan-Mar, Apr-Jun, Jul-Sep, and Oct-Dec). 

Detecting phenological features requires shorter compositing periods. For such use cases, S2GM offers monthly, 10-day or daily products, which, however, come with limitations caused by the number of input data.

Figure 5 shows a time series of monthly products over Lunel in southern France. The monthly composites nicely show the two vegetation cycles in this agricultural area.

With the high temporal resolution of 10-day products, you can monitor even small changes in the vegetation. Due to the low availability of input products, composites may suffer from residual cloud contaminations or lack pixel values due to unavailability of input data.

Figure 6 shows a time series of 10-days products over Lunel in southern France between June and September 2018. For the first 10-days period, no data were available for compositing due to cloud coverage. Therefore, the composite shows no reflectance values. You can clearly see the gradual change in vegetation over time and track the harvest on individual agricultural fields during each period. Nevertheless, cloud contaminations, as occurring in the second 10-days period, will become more likely with decreasing length of the compositing period.