Needles in a haystack – finding a plane in the Indian Ocean

In the weeks following the disappearance of Malaysia Airlines Flight 370 there has been a huge international effort to locate the plane. People keep talking about finding a needle in a haystack; somehow I think the needle is easier to find than this plane. It’s a big ocean, and there’s not a lot of traffic in that part of it.

Analysts use Worldview-2 imagery to search for the missing plane. Here’s how they are probably doing it. If you look at the imagery as a standard RGB composite, it’s hard to quickly scan an image and see where a plane might be. Metal and water are pretty similar to our eyes. However, if you start looking at the infrared bands, metal reflects strongly whereas water absorbs that energy. So it’s easier to quickly look at an image and figure out if there is anything that warrants closer inspection. If you can identify all of the areas where the plane cannot be, then you greatly limit the areas where it might be. On top of the contrast between water and metal, you have less of an issue with sun glint as you move further down the electromagnetic spectrum.

Here’s an RGB composite image over Singapore that illustrates what I’m talking about. You can make out the container vessels, which are several times larger than the plane that went missing.

RGB Composite, plane cannot be seen

Nothing to see here, right?

Here’s the Near Infrared-2 band which shows the greatest contrast between metal and water.

NIR2 band; plane is visible as bright pixels

Or, maybe there is … ?

And if I zoom into the area in question (it was total luck that I found this by the way, it’s not like they tell you what’s in the image before you open it) you can see that tiny little speck is actually a plane flying over the ocean.

plane spotted in near infrared

If there was a debris field, you would essentially be looking for just a few pixels that look this bright. It’s kind of weird to wrap your head around, but you’re looking for bright pixels, not features.

Speaking of tiny specs, here’s the extent of that one Worldview-2 scene in relation to the rest of the search area.

One WV2 scene in the ocean

Landsat would actually be the best option for narrowing down the potential areas because it allows you to quickly scan a larger area. You have shortwave infrared bands which show even more contrast between water and metal than the near infrared that you see on Worldview-2. Metal will reflect so strongly that even if the amount of metal does not cover the entire pixel (30 meters for Landsat) it can still change the value of the pixel. You won’t know exactly what you’re looking at, but you’ll know that there is something other than ocean in those brighter pixels. Unfortunately, as the name implies, Landsat is not turned on over the deep ocean.

The biggest challenge with using satellite imagery isn’t the size of the ocean or its spectral characteristics. It’s the clouds. Check out this global image from taken yesterday. Click to open the map in a new window.

MODIS clouds over the Indian Ocean

Lots of clouds over that ocean.

This entry was posted in ArcGIS Online, Imagery, Oceans & Maritime, Uncategorized and tagged , . Bookmark the permalink.

Leave a Reply


  1. ernestocd says:

    Interesting article! It makes me wonder why crowsourced efforts such as tomnod ( uses RGB approaches rather than the one described here. Wouldn’t we want to narrow down the area of interest? By the way, I believe they are doing a great job by deploying this tool.

    • kevin_butler says:

      Thanks! We’ve had some interesting remote sensing conversation here at Esri about this. I just took a look at tomnod, and I think they’re using a mixture of NIR and RGB, probably just using whatever data they can get their hands on I imagine. I’ve been hearing a lot about sighting from a Thai satellite recently, but I haven’t been able to find anything on its specs.

  2. brutusbuckeye says:

    Does anybody have the new search coordinates for the large area due west of Perth?

  3. brutusbuckeye says:

    And does anybody know why Digital Globes Image Finder tool does not bring up more images than they do for the old MH370 search areas? I used a center corrdinate pair of -44 41 00 by 90 31 00 with a 1000 Km box size yesterday plus a 1/1/2013 to 3/27/2014 time window. I set all the sensor parameters to min or max depending so it would open up the widest search. All I got were 126 images ALL acquired since 3/16/2014 and ALL were in the NE quadrant of the box!!