historical imagery density

How much historical imagery is there in Google Earth?

April 7, 2016

This is the third post in a series looking at how much data is in Google Earth. We have already looked at how much data per unit area 3D imagery requires and how much data different types of 2D imagery require. Today we are looking at historical imagery.

Our first test was to pick a location that we know has a lot of historical imagery. We chose Rio de Janeiro, Brazil. We prepared the view we were interested in and cleared the Google Earth cache then allowed the default layer to load. Just the default layer, without moving the view at all, filled the cache to 4 MB. We then switched to historical imagery and cycled through all the historical imagery, again without moving the view. The cache was now at 580 MB. The reason for the enormous size is that Rio has approximately 220 unique images for the location we chose. On average, each image added about 2.6 MB to the cache. This is less than the default view takes, because many of the historical images do not cover the full area in the view, with some of them being only barely visible at the edge of the screen.

The area we chose in Rio de Janeiro, Brazil.

So how much historical imagery is there globally? We have tried to answer that question before with our historical imagery density map. Since then, we have created a higher resolution version, although it is not quite complete.

In the standard heat map, we used a logarithmic scale to help make the patterns in the historical imagery stand out. To get a better idea of just how much historical imagery there is overall, we have a simplified version shown below. It shows areas with 10 to 24 images in green, areas with 25 to 49 images in yellow and areas with 50 and over images in red. Most of the world has less than 10 images in a given location and in fact most of the oceans and much of the polar regions and deserts have no historical imagery whatsoever. There is also a bug in the way Google Earth shows historical imagery on the timeline that causes it to incorrectly report the number of images along coastlines or other regions that have no historical imagery, but are near to areas with a lot of historical imagery. This is what causes all the large patches around the edges of the continents in the screenshot below. This makes it impossible to directly do calculations on the data. However, by rough approximation we believe that historical imagery, if uniformly distributed, could cover the worlds land masses at least 5 times over.

To see the historical imagery density map in Google Earth download this KML file.

Historical imagery density in Google Earth: Part 2

August 26, 2015

Last week we started on a project to try and map out the density of historical imagery in Google Earth using the Google Earth plugin.

We created a short piece of JavaScript that queries the Google Earth plugin for the number of images available at a given latitude and longitude. We did this every half degree latitude and longitude between 60° N and 60° S and cut it down to every degree longitude between 60° and 80° towards the poles, pausing one second between each one to give Google Earth time to retrieve the data. The result is just over 200,000 points of data, which took several days to generate, running 24 hours a day.

This really gave us a new appreciation of just how big the earth is. DigitalGlobe images cover about a ninth of the size of each square we generated, so at a rough estimate capturing one image per second, would take three weeks to photograph the whole globe. Landsat 8 actually manages to image the whole globe in just 16 days, but takes much lower resolution images, which cover a larger area for each image.

We had discovered with our ‘Chinese map offsets’ map that Google Earth cannot handle very large numbers of icons. However, it seems to have no great difficult with Placemarks that have no icons. You can load the whole dataset without significantly affecting Google Earth’s performance. Download the data here as a KMZ file. You have to zoom in a bit before you see the numbers – which probably helps the performance. If it looks too clustered try changing Google Earth label size to ‘Small’ in Tools->Options->3D View.

The next step was to try and make a heatmap effect. Google Fusion Tables that we used last week has a limit of 1000 points. So we decided to try and create our own in KML. You can download the result here. Again, we are quite impressed by Google Earth’s performance. We did have to optimize it a bit, as our initial attempts did make Google Earth very sluggish. If you have a slow computer you may find even the optimized version is a bit too much when zoomed out. However, performance should improve when you zoom in.

You can make out US state boundaries where aerial imagery sets overlap, and see the Amazon river.

As we noted last week, there are some odd effects in the data, such as high counts just off the coast in many places and some interesting bands in the data. There are also remarkably large numbers in some places. These could be related to how we collected the data or how the Google Earth plugin reports the data.

Historical imagery density in Google Earth: Part 1

August 21, 2015

Earlier this week when creating the map of Chinese map offsets it reminded us of a project we have long wanted to work on – mapping the density of historical imagery in Google Earth.

When you are in historical imagery and you look at the timeline, you can tell roughly how many historical images there are for the region you are viewing by the number of light blue bands on the timeline.

This location (Rio de Janeiro, Brazil) has a lot of historical imagery.

This location in rural Brazil, has very little historical imagery.

The only way we know of for automating the counting process is with the Google Earth API, and with the deadline for its end of life coming up this December we had better make use of it while we still can.

It will take a while to get the whole world done and decide on the best way to collect the data and the best format to show the results in, but we thought we would share some of our preliminary findings.

We have created a historical imagery density map for Spain and Portugal as seen below.

Strangely, over the oceans Google Earth sometimes reports no historical imagery but in other locations reports unusually high numbers. We are not entirely sure why this is, but it appears to be a bug in the way Google Earth reports historical imagery on the timeline. You can download the above KML file here.

We discovered that Google Fusion Tables has a remarkably easy way to create a heat map from a suitable KML file, as seen above. However the problem with high figures over the oceans drowns out the differences over the land. Nevertheless, there is a clearly more imagery over centres of population.