Thursday, May 22, 2014

The Search for MH370 - A Cautionary Tale

Today I was looking at high-resolution satellite images of the ocean and ran across something that reminded me of the search for the still-missing Malaysian Airlines flight MH370.  You might recall that several nations published satellite images that supposedly showed hundreds of pieces of floating debris in the Indian Ocean.  Twice I was booked by CNN to appear on the Jake Tapper show to talk about what we thought of this imagery; both times they canceled.  Maybe in part because I couldn't be very confident about what I was seeing, and that's exactly what I told the producers.  Without access to the raw, full-resolution image data, I had no idea how the images had been processed.  The Thai satellite images in particular were contrast-enhanced so severely that all shades of gray had been reduced to simple black-and-white, eliminating many of the visual clues that could help an experienced analyst.  Talking-head "experts" on cable TV repeatedly claimed with absolute certainty that these images showed debris. 

If you ever stare at the ocean from the window of a jet at 35,000 feet, you'll notice white patches of various size and shape that randomly seemed to appear out of nowhere, slowly get brighter, then slowly fade away.  It's mesmerizing, and kind of beautiful. You're looking at sunglint as the faces of waves momentarily line up to send a mirror-like reflection of the sun your way; and possibly you're seeing whitecaps, generated by the wind occasionally blowing the tops off the steepest waves.  The rougher the sea-state, the more of these bright patches you'll see.

Now imagine taking a snapshot of that to capture an instant in time.  What would it look like?  Right. A random bunch of white spots of various size and shape and brightness, that could easily be mistaken for floating debris if that's what you desperately want to see.  This is what gave me pause, as I said in a Washington Post article on March 20 and in an email to a Post reporter on March 28.  That same day a CIA image analyst suggested the same thing:
Stephen Wood, a former CIA analyst and satellite imagery expert, said the satellites could be seeing something as simple as whitecaps, which he said can look deceptively like solid objects.
So let's look at some images from Google Earth, taken just east of the island nation of Palau in the Pacific Ocean way back in  June 2004.  I'm confident these images show nothing but sunglint and whitecaps; you can see how big the waves and swell appear.  It was a very rough day for these latitudes, but not at all uncommon for the Roaring 40's in the southern Indian Ocean where the search for MH370 was focused:


Zooming in on the image above:


Zooming in even more on a strangely artificial-looking triangular patch of whitecap and/or sunglint.  Note the measurement: it's about 70 feet across (21 meters).  Compare with the "possible object" shown in this Australian satellite image.  The straight edges in this case are an artifact, a function of the spatial resolution of the imaging system as it interacts with the target; a commonly signal-processing phenomenon known as "aliasing." 


 Let's jump back up to the first image.  Here's how it looks if you push the contrast so hard that all shades of gray are reduced to either black or white; an image processing function called thresholding.  Compare with this Thaicote image showing hundreds of white spots described as "pieces of debris."


I don't mean to say neener-neener to my fellow image analysts in Australia, Thailand and China. I still don't have the images they had, so who knows -- some of that stuff really might have been wreckage, floating trash, discarded fishing gear, or something else.  There's a lot going on out there in the big ocean.  We rushed to our own conclusions in the heat of the moment with the low-resolution imagery that was publicly available (here and here).  And who knows what careful caveats and cautions those analysts provided with their work, that got tossed aside by their superiors or the media in the furious grinding rush to be The First to find the missing airplane?

But I think this is a tragic example of how easy it is to see whatever we want to see, and hear whatever we want to hear, when we're confronted with a confusing mashup of noise, inconclusive data, and life-or-death urgency.  My sincere condolences to the friends and family of those who are still missing.

1 comment:

  1. We are fortunate to live in an area chock-full of remote sensing and image analysis experts. Recently we've been in touch with the NGA Alumni Association, a group of folks who honed their skills on the intelligence side of this business. I've asked them what they think of this blog post and some of the claims made using satellite imagery during the hunt for MH370, and gotten these two responses so far:

    "I think John Amos is spot on. I always liked to say that a digital image is like a statistic -- you can manipulate it as much as you want to come to the conclusion you want. Unfortunately!!"

    " feel free to tell Mr. Amos that at least one NGA retiree is in his corner. (45+ yrs as an IA, Naval specialist)"

    ReplyDelete