We recently came across this interesting article by National Geographic about a recent study of land/water changes over the last 30 years. The study is by researchers at the Deltares Research Institute, who used Google Earth Engine to gather and process the data. The data itself comes from Landsat imagery.
Sadly, we were not able to figure out a way to view the data in Google Earth. Google Maps Engine (not to be confused with Google Earth Engine), which was shut down earlier this year, was notable for being able to easily display maps in Google Earth. However, it appears Google Earth Engine does not have any such features and is mostly focused on displaying data in Google Maps style 2D tiles. This is a pity, because we find Google Earth a much better platform for exploring this kind of data.
The researchers published the full description and analysis of the project in the journal Nature Climate Change, which is subscription based. However, the data itself is published as a publicly available 2D map.
What we were not able to determine were the details of how the water bodies were measured. Most inland water bodies are quite seasonal, so we wonder how the researchers corrected for that. The Landsat imagery typically covers each spot on earth once every 16 days, but a fairly high percentage of the images are obscured by cloud. This often means only a few good images per year. For the global mosaic used in Google Earth, many different images over multiple years are combined together to get the cloud-free image. However, this can result in some weird effects where water is concerned, because water is not just seasonable, but can vary considerably from year to year (more on this in a future post). Droughts or floods can, for example, be one-in-a-hundred-years events. This doesn’t apply to coastal land reclamation, which tends to be permanent.
Land reclamation near Seoul, South Korea, as seen in Aqua Monitor.
We have previously created animations of land reclamation and artificial islands being built, but we restricted ourselves to the time-frame visible in Google Earth historical imagery, rather than the 30 years used in the above study.