Google Earth Blog

The amazing things about Google Earth


landsat

Exploring the Chile wildfires with Landsat and Sentinel-2 imagery

Yesterday we had a look at some DigitalGlobe imagery of Chile showing the damage caused by wildfires in late January and early February this year.

We thought it would be interesting to explore the region with Landsat and Sentinel-2 imagery.


Landsat imagery. Credit: USGS/NASA Landsat


Sentinel-2 Imagery. Credit: Copernicus Sentinel data, 2017.

The wildfires extended far beyond the area covered by DigitalGlobe imagery (red outlines in the first two images) and almost certainly there were more fires outside the areas shown above, but we did not have time to gather imagery for the whole country. The southernmost Sentinel-2 image was captured earlier than the others and shows at least two wildfires still in progress. Both Landsat and Sentinel-2 imagery clearly show the areas affected by the wildfires and we have outlined them below.

Unfortunately the imagery above is too large for us to share as a KML on this blog.

To obtain the imagery, we started with our KMLs that show previews of Landsat and Sentinel-2 imagery. Once we have selected a suitable image that covers the region of interest in the right date range and without excessive cloud cover, we download the key files from Amazon AWS. For Landsat imagery you need bands 2, 3, 4 and 8 and for Sentinel-2 just bands 2, 3 and 4. We then used GeoSage’s ‘Spectral Discovery’ tools (Landsat Sentinel-2) to process the imagery. An alternative method is to use GIMP or any other suitable image processing tool. GeoSage’s tools certainly make the process much easier, but one downside is the colour processing is automatic, which resulted in significantly different colours for the three Sentinel images.

Obtaining and processing Landsat and Sentinel-2 imagery has certainly become easier over time, with both AWS and Google Cloud hosting the imagery archives. However, what is sorely needed is for a provider to pre-process the imagery and even better would be to then provide it in tiled format. Doing so would change the two public sources of imagery from a niche product mostly used by GIS professionals to a public resource readily available to all.

Moving sand in Landsat animations

Last week Google added global mosaics in historical imagery for each year going back to 1985. The mosaics are created from mostly Landsat imagery with a bit of Sentinel 2 imagery for the last couple of years (the Sentinel 2 satellite is new). The mosaics are created by gathering all the Landsat/Sentinel 2 imagery for a given year and looking for cloud-free and snow-free pixels then combining the images to create a single global mosaic for the year. Although it is fantastic for viewing long term change, the overall result is that we actually miss out on short term changes.

Back in July we wrote a post about watching sand dunes move with Google Earth imagery. Today we are looking at a similar concept, but with Landsat imagery instead. In August we created a KML file that automatically creates animations using Landsat imagery with thumbnails from Amazon Web Services. We used that tool to create the two GIF animations below:


Sand blowing in the Sahara (southern Libya). Explore original animation.


Sand blowing and irrigation circles in Oman. Note especially south west of the centre of the animation there are wisps of sand moving north west. Explore original animation.

What we found interesting is that established dunes hardly move at all over the three years covered by the animations, but loose sand can clearly be seen moving. It looks no different than sand being blown on the beach. Yet the scales involved in the animations are tens of kilometres and a three year period.

You will also notice in the second animation the dunes to the north east appear to pulse sharper, then softer again. This is because the angle of the sun affects the shadows during the course of the year.

Animating the new Landsat/Sentinel global mosaics with a dynamic tour

Earlier this week Google added to Google Earth global mosaics based on Landsat and Sentinel 2 data from 1985 to 2016. Google Earth Engine provides animations for the data using a web based interface. We promised to release a way to animate the data in Google Earth.

Simply download this KML file and open it in Google Earth. Then switch to ‘historical imagery’ and find a location of interest then play the ‘Animation’ tour found in the KML. If you wish to move to a different location, close the tour, move to the new location, wait a moment, then open it again. Depending on your internet speed, either let the tour run through a couple of times to load all the imagery, or you can manually go through each year first to make sure it is loaded, then run the tour to see it animated.


The growth of Kuala Lumpur, Malaysia, over 30 years (1985-2016) using the new Google Earth global mosaics.

The imagery covers a period of just over 30 years and there is almost no place on the planet that hasn’t changed significantly in that time, so get exploring!

Remember these are long term changes, not month-to-month changes.

We chose a timing of one second per image. We would have liked to go faster, but Google Earth cannot handle faster tours and starts to skip images if you increase the speed.

More about the new Landsat / Sentinel 2 data in Google Earth

Yesterday Google surprised us by adding global mosaics created from Landsat and Sentinel 2 data to the Google Earth’s ‘historical imagery’. The data comes to Google Earth via Google Earth Engine, which has long had this time-lapse feature that has just been updated to include imagery up to 2016. The timelapse page also tells us more about the imagery:

Using Earth Engine, we combined over 5 million satellite images acquired over the past three decades by 5 different satellites. The majority of the images come from Landsat, a joint USGS/NASA Earth observation program that has observed the Earth since the 1970s. For 2015 and 2016, we combined Landsat 8 imagery with imagery from Sentinel-2A, part of the European Commission and European Space Agency’s Copernicus Earth observation program.”

Google Earth Engine has also published this post about the update.

On Friday we will release a KML file that will allow you to view similar time-lapses from directly within Google Earth.

Today we are making a few observations about the imagery.

The first observation comes from GEB reader ‘haflaa’ who points out that although the mosaics are global for every year between 1985 and 2016, there are some locations, such as the Maldives, where the imagery is identical for a number of years. In the case of the Maldives, the imagery is identical from 1985 to 1999.

We also note that the Google Earth Engine animation starts in 1984, whereas the Google Earth data starts with December 1985 (which we presume represents the data for the whole of 1985).

The path of the satellite is often noticeable in the imagery. For more on the orbit of the Landsat satellites see this post and this post.

In many locations you will notice narrow stripes in the imagery. This is because of a faulty part on the Landsat 7 satellite. Learn more about it in this post.

The mosaics are created by trying to select imagery from throughout a given year then selecting cloud and snow free pixels where possible. However, there are a few locations on earth that are almost always cloudy, a problem we discussed in this post.


This location in Cameroon features both Landsat 7 stripes and clouds that just could not be eliminated.

Google Earth now has global Landsat and Sentinel 2 data from 1985 to 2016

Thank you for GEB reader Jonah for being the first to let us know. Google has added yearly global mosaics using Landsat and Sentinel 2 data going back to 1985. Google has long had these mosaics on Google Earth Engine which also provides a ‘Timelapse’ tool. As of this writing, the Google Earth Engine timelapse only goes up to 2012, but that will likely be updated too.

To see the mosaics, simply zoom out a bit and switch to ‘historical imagery’.

We plan to create tools to animate the new data, but that may take a few days.

Here are a few ‘before and afters’ showing changes between 1985 and 2016.

before
after

The Amazon, 1985 vs 2016

before
after

Dubai, 1985 vs 2016

before
after

The region around Shenzhen, China, 1985 vs 2016

Note that the new data is created by blending all Landsat/Sentinel 2 data for a whole year to remove clouds and snow cover. The result is that changes that happen on timescales less than a year, such as seasonal changes, will not be visible.

Google has not (as of this writing) updated the historical imagery layer with all the high resolution imagery that has been added since June 2016.

There is one disadvantage to the new data – it is much harder to find historical imagery, as you need to zoom in very close to the earth to see the high resolution historical imagery.