With the onset of the this years User Conference, Esri unveiled various updates to its existing services and apps, one of which is the Sentinel-2 Land Cover Explorer announced in last February. You can also find this in the Living Atlas. Apart of the the visual updates, the biggest change this time is that the changes over the year are now more accurate.
A fundamental aspect of global LULC maps is the ability to detect and assess land cover changes over time. Improving on an already accurate set of annual maps, Impact Observatory has incorporated new features and methodologies in their proprietary deep learning classification model, resulting in better temporal consistency across the entire time series.
Change between two LULC maps can potentially signify an important and developing change in an area of interest. However, in some cases, classification results may vary from one annual cycle to the next due to modeling insufficiencies, variability in seasonal observations, and/or class ambiguity at 10-meter resolution. Such cases can lead to false or spurious change results when conducting temporal change analysis. With the improvements to the temporal consistency, users assessing temporal change across the time series can be confident that what they are seeing represents the natural world.
Link - https://livingatlas.arcgis.com/landcoverexplorer/
Source
https://www.esri.com/arcgis-blog/products/arcgis-living-atlas/imagery/global-land-cover-updates/