Google recently refreshed Google Earth’s global mosaic with newer, sharper imagery. So far, we like it very much and think it is definitely an improvement. However, we will have a look around and see if we can find any flaws or interesting aspects to the new imagery.
Landsat 7 stripes
We already pointed out yesterday that although Landsat 8 imagery was used in the new mosaic, it is not entirely free of Landsat 7 imagery with its characteristic stripes. They typically show up in hard to photograph places, such as those that have near year round snow cover or cloud cover, but we think we even saw some in the Sahara.
There are a few locations where non-Landsat imagery has been included in the mosaic. This includes a number of islands, such as Svalbard and the islands in the South China Sea.
Below we can see a particularly noticeable strip across Smith Island, which is part of the Andaman and Nicobar Islands in the Indian Ocean. The image is actually a DigitalGlobe image from 2011 which disappears as you zoom in.
We believe the reason for this is that there simply aren’t any good quality, snow-free and cloud-free Landsat images of the locations in question. Islands, it would appear, are cloud magnets.
Overall, the contrast in the imagery is noticeably higher and features you may have never noticed before stand out. Lakes generally seem to be greener or browner than before.
It is important to note that the ocean floors are actually a different data set and have not, as far as we know, been updated at this time. However, they did receive a significant update in January this year. We have come across some oddities in the margin between land and sea. For example, along the coast of Vietnam there is a thin border of brown where the Landsat mosaic ends, but between that and the sea floor data is some other imagery which includes some clouds.
We saw this same effect in a number of other locations around the world.
We also found that if you zoom in on Chandler Sound, which is part of the Mississippi delta in the Gulf of Mexico, Google Earth shows this strange pattern:
We are not sure if this has anything to do with the global mosaic update.
Dating the imagery
The imagery is a mosaic collected from parts of images from the Landsat archive going back many years, so it is impossible to put a date on the whole mosaic. However, there are specific places where it is possible to determine the approximate date of the imagery used. The best locations to do this would be large lakes or inland seas that are shrinking or growing over time. We haven’t yet done this for any such lakes, but we did check the Nansen Ice-shelf in Antarctica and determined that the imagery there has not been changed from the previous mosaic. The imagery is from 2003 as we determined when watching the ice sheet crack. We also checked Bento Rodrigues in Brazil and are fairly sure that the imagery is from before the disaster that took place there in December last year
Landsat imagery has a resolution of only about 30m per pixel and we suggested yesterday that Google consider using Sentinel imagery, which is higher resolution. However, after some consideration we have realised that for the global mosaic, the important factors are consistent colouring and good global coverage. As you zoom in, Google Earth transitions to higher resolution imagery where available so greater resolution of the global mosaic is not necessary. It is, however, the case that there are some parts of the world where no higher resolution imagery exists and the Landsat imagery is used even when you zoom in and only for these locations does Google need to seek alternative sources. For much of the globe they have already used medium resolution imagery from Spot Image. For more on what image sets are used where, see our series on Google Maps API Maximum Zoom.
To see the locations featured in this post in Google Earth download this KML file.