Thoughts on the New ATSB Report on MH370 — UPDATED

The ATSB has just issued a lengthy and detailed new report explaining its latest thinking regarding the underwater search for MH370 in the southern Indian Ocean, available here. For the most part the media are reporting that its basic point is that the current search area is the right one. While that’s true, there are some more interesting points buried within in it, and within its companion volume from Australia’s Defence Science and Technology Group (DSTG) which explores the logic in further depth.

Here are my bullet points—I invite readers to add their own, or to correct or elaborate upon my points.

— One of the most jaw-dropping revelations in this report is that after 18:01:49 there was just a single radar return point. Note only does this contradict data shown to passenger family members soon after the disappearance (see Victor’s note below, and the image above), it also raises questions about the reliability of that piece of radar data. Since it was detected at the far limits of the radar equipment, it is relatively inaccurate, and as a stand-alone piece of data it is much more likely to be anomalous.

— The report reiterates that the only way to deliberately depower the SDU is by pulling circuit breakers in the E/E or isolating the left AC bus from the cockpit, but offers no explanation of why this might have happened prior to 18:25.

— It turns out that the time to recycle SDU is not 2.5 min but only 60 seconds. This is particularly important when it comes to laying out an end-of-flight scenario that presumes fuel exhaustion.

— The report says that after fuel exhaustion ditching not possible, with or without a conscious pilot. There has been a great deal of debate about the possibility of ditching in this forum, and I hope (but doubt) that the report will lay the issue to rest.

—At last, we know the cost index for the initial portion of the flight: it was set at 52. Of course there is no reason to assume that the later portion of the flight was conducted at this setting but it helps us calibrate likely flight modes.

— Overall point: With these two documents, the Australian authorities have shed a commendable quantity of light onto the subject of how they have determined the likely flight paths that MH370 took after it disappeared. It is heartening to note that they have greatly shrunk the length along the 7th arc along which the plane might plausibly lie: by my reckoning, from about 630 nm to 380 nm. And the “fried egg” of maximum probability is smaller still, only about 150 nm long. However, I find it baffling that, given the incredible level of effort poured into figuring out how the plane might have traveled prior to fuel exhaustion, there seems to have been basically zero time spent figuring out how the plane would likely have traveled after the fuel ran out. Frankly, I was expecting a lot of analysis along the lines of Brock McEwen’s work on this topic. As it is, it seems that instead of examining flight modes they took a guesstimate from accident-investigation experts and added a fudge factor. The result is that, while this latest analysis shrinks the search’s target area on one dimension, it makes it fatter on the other. In current southern search area, they’ve looked about 18nm inside, 30nm outside 7th arc. According to the new report, they should expand the search box to a width of 80 nm, symmetric around the 7th arc. This is not progress, and I think the ATSB can do much better (and hopefully will in a future report.)

Some additional points from Victor Iannello

Before I could post the above thoughts, Victor emailed me some observations of his own, which I include here unedited:

“There are some very strange results reported starting on page 17. Here are some comments related to that and the BFO bias:
1. As the attached graphic illustrates, if we are to believe that primary radar data exists every 10 seconds up to 18:02 and then only a single capture at 18:22, the slide presented to the NOK on Mar 21 at the LIDO hotel is false or includes data not used in the report.
2. There is a statement that the ground speed observed by the radar prior to 18:02 is relatively high and implies the aircraft would be at low altitude. While this would be allowed from Ma number and available thrust considerations, the indicated air speed would be extremely high, the airframe would be stressed, and the fuel efficiency would be incredibly wasteful. This is not consistent with the fuel calculations after 18:22.
3. The groundspeeds they calculate from the radar data have tremendous variability, even after the Kalman filter is applied. I estimate the peaks to be about 550 kn (!) We need the raw radar data to see what the hell they are doing.
4. The bias term was observed to be time-varying and modeled with a SD of 25 Hz. But there is also a statement that “Substantial effort was made to characterise this structured bias. It was found to have a geographic dependency but it has not been possible to determine a quantitative function to compensate for this change in bias.” This implies the drift might not be a simple OCXO drift issue. In fact, perhaps it is the correction term that has a geographic dependency rather than the oscillator drift.

Overall, I am not sure their work adds much value over the deterministic approach that the IG and others has followed. The PDFs and assumptions on heading/speed changes of previous flights are practically irrelevant.”

He later added:

“Their approach of randomly spaced turns, accelerations/decelerations, and climb/descents will always favor straighter, more constant speed paths that fit the BTO data. Since commercial flights are relatively straight to conserve fuel, it predicts those flights relatively well. If MH370 flew relatively straight, it should work there, too. But there are no guarantees the flight was near straight. For instance, if the flight flew south, a circle “loiter” above Sumatra would be ranked low, as would a curved path that followed the coast of Sumatra. There is bias in their model that is not acknowledged. Also, fuel calculations are only indirectly included by limiting the range of speeds. The model seems overly complicated for the value of what is produced. It seems developed more to impress than to enlighten.”

UPDATE: 12/4/15

Reader Paul Smithson asked “how much of the newly-defined priority area has already been searched?” In the image below I’ve outlined in black the area already searched (via Richard Cole) on the “fried egg” map. (Click to enlarge) The original 120,000 sq km search area is outlined in red; the new 120,000 sq km search area is outlined in purple. As you can see, almost all of the high-probability area has already been scanned. As more and more is searched, the probability density of the area being scanned will decrease, so that search becomes ever less fruitful. The effort expended between now and the end of the scheduled search will, by my seat-of-the-pants estimate, increase the probability distribution coverage from aroun 85 percent to 90 percent.