Continuing the theme started in May’s Geospatial Tip of the Month (GTM), this month we debunk more satellite imagery myths that are focused on DigitalGlobe’s most recently launched bird, WorldView-3. WorldView-3 is without question the world’s most advanced commercial satellite every launched with resolution down to 30-cm; but still the news stories about its capabilities are not even close to the truth. Before we jump into this GTM, you might check out more about WorldView-3’s technical specifications here as well as our sample images.

Myth 1 – WorldView-3 Offers Resolution Better Than 12-inches

One common myth associated with satellite imagery is the resolution they can offer and thus the objects they can “see” on the ground. For example, take this statement published by the US news site, NBC News:

The resolution is so ultra-high, in fact, that the federal government isn’t allowing DigitalGlobe to release the pictures at their best just yet. Due to regulatory restrictions, they have to be downgraded from a resolution of 12 inches per pixel to a fuzzier 15 inches per pixel (30 to 40 centimeters [cm] per pixel) until Feb. 21, 2015.

When this article was published on August 26, 2014, this statement was partially true but still its essence was completely wrong. The truth in this statement related to US government regulations which capped the maximum resolution for satellite imagery to 50-cm right up until WorldView-3 data was made commercially available. When WorldView-3 data was first sold, the maximum resolution you could obtain was 40-cm. And it is also true that by February 21^st the US government eased regulations on resolution again so that 30-cm data could be sold.

But what is far from the truth is the statement that WorldView-3 imagery is downgraded to 30-cm resolution per pixel. This is clearly the driving theme of this short paragraph on resolution. However, in actuality, the maximum resolution that WorldView-3 collects is 31-cm. And in fact, that is only when the satellite is pointing directly down at the ground (also called off-nadir). As off-nadir increases, the resolution decreases so that the maximum resolution at 20 degrees off-nadir is only 34-cm. So the reality is that WorldView-3 does not collect 12 inch imagery; and further that most of its data, even if delivered with 30-cm pixels, has native resolution of 31-cm or higher.

Myth 2 – Daily Revisits to Anywhere and Spotting Small Objects

Our second myth, well actually two of them in one, comes from an online article published by the UK news group, The Register. In the article, this claim is made about seeing small objects in satellite imagery:

A little sharpening of the images will mean it is possible to spot objects just 31cm long, almost exactly the length of your correspondent’s size 13 flip flops, from space.

This is actually a myth we debunked in last month’s edition as well. In that article, a claim was made that you could see objects on a desk. In this case, while yes the at-nadir resolution of WorldView-3 is 31-cm, this does not mean you can truly see an object 31-cm in size. First keep in mind that detecting an object’s presence and seeing it are two different things. While flip flops sitting on a piece of grass (as long as they are not green in color too) might cause this specific pixel to look black surrounded by a sea of green, it would only be a single pixel. Determining if this single pixel is a flip flop or just a small bag on the grass would require actually seeing the object with more detail than one pixel – myth debunked!

This same article goes on to make this claim:

The satellite is also exciting geogeeks thanks to specs that promise it will be able to revisit any corner of the globe every 24 hours.

And again, this statement could be called partially true. In latitudes above and below 40 degrees, it is possible to see every piece of ground every day; but it does not mean it is possible to image every piece of ground every day given the limited storage capacity of WorldView-3. Second, it does not mean that if you have a request in for a new collection of data with WorldView-3 that you are guaranteed a collection every day, it can take days, weeks or even months to receive data over a specific polygon you requested. Third, if you were to pay the price required to get daily (or near daily) images over your specific polygon, in many cases the imagery would only have 1-m resolution given the off-nadir of collection. And finally, as you move closer to the Equator, the Earth’s landmass increases in area (consider the Earth is a sphere and as you move to the center of a sphere it bulges out) so that even at 1-m resolution, daily collections are simply not physically possible.

This graphic can help to illustrate my point on revisit time better than words alone. At the Equator, there is about 25,000 miles of land mass that WorldView-3 (or any satellite for that matter) can possibly see. And then from 50 to 60 degrees North or South latitude, this same satellite can see about 18,500 miles. This means that WorldView-3 can see the same location at 50-60 degrees latitude more often than at the Equator as it has less land to potentially image as it moves to the Poles, but the same number of physical rotations around the planet. (Graphic Source: University of California Berkeley)

This graphic can help to illustrate my point on revisit time better than words alone. At the Equator, there is about 25,000 miles of land mass that WorldView-3 (or any satellite for that matter) can possibly see. And then from 50 to 60 degrees North or South latitude, this same satellite can see about 18,500 miles. This means that WorldView-3 can see the same location at 50-60 degrees latitude more often than at the Equator as it has less land to potentially image as it moves to the Poles, but the same number of physical rotations around the planet. (Graphic Source: University of California Berkeley)

Myth 3 – Features can be Detected Automatically in WorldView-3 Data

The final myth we will bust in this GTM was published August 13, 2014 on Yahoo! Finance:

Our automated algorithms … are able to identify all the potential football fields in Colorado. We can do it on Colorado state, we can do it on the United States, we can do it anywhere across the globe. Automated data extraction on a global scale is now possible. For the changing planet, we can create a living inventory of various things.

Again, you might call this statement half true as it is 100% correct to say that features like football fields can be extracted automatically from satellite imagery. But the real question to me is how accurate are these results? When information is extracted automatically from imagery is typically done by looking at both spatial and spectral characteristics – in this example, we would be looking for long rectangles (i.e. spatial) covered with living green grass or Astroturf (i.e. spectral). So while it may be possible to find some of the football fields in Colorado, those covered by a dome or by lots of trees would likely be missed; and then it would likely mis-ID large park areas which are rectangular in shape as football fields. This same story with regards to false and missing identifications using automated classification techniques can be extended to any feature such as homes, sidewalks and roads. In my experience with automated techniques, they are generally 70 to 80% accurate. How many businesses would rely upon data that is only 70 to 80% accurate? Another myth busted!

Do you have an idea for a future GTM? If so, let me know by email at brock@apollomapping.com.

Brock Adam McCarty

Map Wizard
(720) 470-7988
brock@apollomapping.com