Part 2: The Analysis of Southern Airways flight 932
This is a continuation from my post last week on the crash of Southern Airways flight 932. In that post, I stepped through the sequence of events which led to the crash of the charter flight to Huntington Tri-State airport in a McDonnell Douglas DC-9 which killed all 75 on board.
The Chairman of the NTSB said of the accident: “If it hadn’t been for those trees, he probably would have made it. It was that close.”
The captain of that flight was 47 years old who had been hired by Southern in 1949. He held his ATPC with ratings in the DC-3, DC-4, DC-9 and M-202/404. He also held a flight instructor certificate with airplane and instrument ratings. He had 18,557 total flying hours, of which 2,1904 were on the DC-9. His first officer that day was less experienced. Hired in 1965, he held his commercial pilot certificate and had 5,872 hours, of which 1,196 were on the DC-9.
There were no known issues with the aircraft. Although the captain referred to issues at the performance of the autopilot, the descent was stable throughout and there was nothing to imply of a loss of control or a serious autopilot malfunction during the approach.
The airport lighting system, which consisted of high intensity approach lights, sequence flashers and high intensity runway lights, were all in operation and properly set at the time of the accident.
The report describes the conditions that night when the DC-9 was on approach as follows:
The terrain under the approach path was irregular with numerous hills of varying heights. There were few lights along the approach path excepting those of the refinery which were to the right of the inbound track. The lower clouds were ragged and the restrictions to visibility included darkness, rain, fog, and smoke. The pilot had his barometric altimeter, vertical speed indicator, airspeed indicator, and radio altimeter to aid him in establishing the desired descent profile. However, the pilot had little, if any, information instantly available to him regarding the elevation and character of the terrain below the aircraft or the flight path related thereto.
First Officer: A thousand feet above the ground, rate and speed good.
First Officer: Speed a little fast, looks good. Got bug and twelve.
Captain: See something?
First Officer: No, not yet. It’s beginning to lighten up a little bit on the ground here at ah… seven hundred feet.
First Officer: Bug and five.
First Officer: We’re two hundred above.
Charter Coordinator: Bet it’ll be a missed approach.
This comment by the charter coordinator could mean that he could see that the flight was approaching the Minimum Descent Altitude (MDA) and the runway was still not in sight. Other than that, though, the conversation doesn’t really help us to know what they were thinking or why they continued the approach past the MDA.
There’s clear issues with the cockpit resource management, as has been pointed out in the comments. There’s no clear checklist or challenge-and-response interaction between the captain and the first officer. The first officer didn’t call out five hundred, which was required, and never called out their rate of descent, even when the aircraft was descending faster than expected. The landing gear was lowered but no statement was made by either crew member in the cockpit.
When the first officer called out four hundred feet, they may well have been over the river valley, which was 500 feet above mean sea level. So if the 400 feet was from the radio altimeter, then they were only 900 feet above mean sea level, even though the minimum descent altitude for the approach was 1,240 feet above mean sea level. The runway was at 828 feet.
The swath cut through the trees, ground witness statements and the flight recorder altitude trace all show that the descent was stopped, power was added, and a gradual climb was commenced. Furthermore, the first officer called out airspeed in terms of a number (“Hundred and twenty-six”) instead of a reference speed, which is indicative of a go-around rather than a continuing approach.
This quote from the accident report is particularly interesting because later, the report concedes that it isn’t at all clear that the first officer was calling out an airspeed here. The board’s conclusion leaves open the question as to whether that final call of “hundred” meant the height above the ground or the current airspeed, as they were in the process of going around.
The captain responded to ask, “Is that the approach?” The minimum descent altitude of 1,240 feet would be 400 feet above the runway (what Rudy and I would call the MDH). So it’s reasonable to think that the captain presumed that they had reached the MDA for that approach. In that case it could simply be sloppiness that he hadn’t slowed his descent in preparation, so the DC-9 did not immediately level out.
However, I want to focus on the first officer’s monitoring of the approach. He made at least four call outs that are definitively height or altitude. If he was reporting the height above the ground, then there’s two ways that the first officer could have been making them:
1) He subtracted the airfield elevation (828 feet) from the altitude given on the barometric altimeter and called out the result, which would give them the correct height (over the runway, not over the terrain) to remain “400 feet above” until they had the runway in sight.
2) He erroneously used the radio altimeter, thus calling out the height over the current terrain rather than over the runway.
Here is a table from the report which shows:
- the call-out,
- the altitude as recorded by the FDR at the time of the callout
- the terrain elevation for the terrain that the aircraft was flying over at the time of the callout
- the height over the terrain, by subtracting the elevation from the altitude,
- and the height over the runway, by subtracting the runway elevation (828) from the flight recorder altitude.
|Call-out||Flight Recorder||Terrain Elevation (msl)||Height over Terrain||Height over Runway|
|ay, ah, 700||1,330||500||780||502|
As you can see, the first call certainly appears to be the height over the runway elevation. The second and third calls are closer to the height over the terrain, that is, those could have been done using the radio altimeter. And the fourth call, the important one, to show that they’d reached their MDA, doesn’t match up however we look at it.
Now, this isn’t an exact science. There’s a certain amount of variance here: the timestamps on the CVR and the FDR might not have been correctly aligned and even if they were, the exact location of the aircraft is affected by airspeed and wind, so the third column may not be quite correct. The first officer was clearly rounding so he may have called out shortly before or shortly after he believed that they had achieved those heights, adding another unknown factor.
It is certainly possible that all of the calls after the first one were made using the radio altimeter. Further, the last two calls, if they were indeed meant to reflect altitude, seem to be in line with the height above the ground. The hill with the trees reached a height of 990 feet above sea level, so about 150 feet higher than the runway that they were flying towards. Presuming that the terrain was rising again, they could have been 126 feet above the higher ground. The call of “hundred!” shortly before impact could either be the the height above the ground and the aircraft’s airspeed at that moment.
On the other hand, the investigators tracked the theoretical rates of descent between the calls. The calculated rates of descent do not match those recorded by the flight recorder, which tells us at the very least that some of these numbers must not be reliable.
The report’s main argument against this explanation is that a qualified pilot would not make this mistake, knowing they were flying over uneven terrain. Although the chart didn’t give detailed terrain information, Rudy shows in the comments of the previous post that it certainly made the high points clear. The investigators also struggled with the idea that the captain never glanced at the altitude information to notice that the call outs from his first officer were made from the radio altimeter.
It’s possible that he, like the first officer, was relying on his radio altimeter. A second possibility is that he was not using his barometric or radio altimeter, but rather was relying solely on the first officer for altitude information. Finally, he may have been including his barometric altimeter in his instrument scan, but was concerned with other items during the final stages of the approach to such an extent that he did not notice any variations.
That said, I have to admit, what with the outsider in the cockpit and his distraction with the autopilot, I don’t find it all that surprising that he didn’t notice.
However, there was another odd detail. When the barometric altimeters were recovered from the wreckage, both were correctly set at 29.67; however the first officer’s altimeter was out by 300 feet. Now if we look at the callouts in the table again, we can see that each is about 200 feet higher than the FDR reference to the actual altitude of the aircraft. If the altimeter was showing an altitude that was 200-300 feet high, then that would explain why the flight crew stopped their descent at about 300 feet below the MDA.
On the other hand, when the investigators dropped the barometric altimeter from a height, the indicated altitude increased by about 1,000 feet. So the altimeter could have gained those 300 feet in the impact. The report stated that they stopped testing the results of dropping the instrument because they didn’t want to break it; something that I’m sure would not stop the NTSB of today.
That’s an important point in itself. I mentioned that in 1970, cockpit resource management was nothing like as important an aspect of aviation as it is now. There’s also the point that neither the crew nor the charter coordinator thought it was a big deal that the approach plate was out of date. There was no sterile cockpit and the chatter from the back should have been cut off rather than half-heartedly ignored but again, clearly the coordinator was accustomed to being able to discuss fuel and other issues during the final approach. Sure, the captain didn’t handle the situation but based on the coordinator’s commentary, neither had any other captain.
That’s not the only thing that’s changed over the past fifty years. Reading the report, there are some unexpected omissions in how the NTSB handled the investigation.
The report mentions that Southern’s training program distinguished between the use of radio and barometric altimeters over level and irregular terrain but that Southern’s DC-9 Operating Manual “accentuated [the radio altimeters] use for all instrument approaches”. However, no one appears to have followed up on this, for example speaking to pilots at Southern to test their understanding of the radio altimeter’s weaknesses or interviews from the flight crew’s colleagues to discover if these actions during the descent were typical.
Similarly, the use of out-of-date charts, laughed off by the crew as normal, wasn’t followed up on in the recommendations, other than a vague statement that the FAA should keep a closer eye on airlines.
And finally, the issue with the barometric altimeter showing 300 feet out was left as a theoretical exercise, without the rigorous testing and simulations that the NTSB would have access to today.
Then there’s the following argument against the radio altimeter being incorrectly used during the approach:
The theory also assumes an unlikely dual human failure in that the captain was either also using his radio altimeter or did not recognize the differences between the barometric altimeter and the altitude information called by the first officer and was relying on the latter.
Now I’m not sure “dual human failure” is all that unlikely but more importantly, if there was an issue with the first officer’s barometric altimeter being 300 feet out, then we still have the same question: why didn’t the Captain notice? If that seemed unlikely in the first scenario, then why wasn’t it mentioned as equally an issue in the second?
Meanwhile, the investigators were not able to find any physical evidence of an issue with the static system or ports. And if there were a defect in the static system, this would also have given them inaccurate airspeed indications, which there was no reference to, unless there were an error in the pitot system which offset the error in the static system.
There is no known static system error which could cause a 200-foot altimeter error along with a corresponding air-speed error of -17.5 knots; without evidence of a phenomenon which could cause these errors, the Board cannot conclude that a static system error is supported by sufficient evidence to be termed a causal factor in this accident.
So both theories were dismissed as not having sufficient evidence to be termed a causal factor in this accident; a frustrating conclusion which could have been mitigated by more interviews and more tests. The probable cause was cited as the descent below Minimum Descent Altitude. I find this only slightly better than an explanation that the plane crash was caused by impact with the trees, to be honest.
However, we can certainly see steps towards modern aviation when looking at the conclusion and the recommendations.
Although the Safety Board has been unable to determine the probable reason for the unrecognized descent below MDA in this instance, the Board wishes to reiterate its concern with the general problem of landing and approach accidents and to re-emphasize its interest in the various preventive measures which might proof useful in reducing the rate of these kinds of accidents. There is a need for all segments of the aviation industry to continue to focus attention on the unique demands for crew coordination and vigilance during nonprecision approaches. Area navigation systems, now in the final proving stages of development, will apparently provide descent guidance capability within the aircraft and should be standard equipment on all future transport category aircraft.
Three recommendations were made:
- that the aviation industry continue to focus on the unique demands for crew coordination and vigilance during non-precision approaches, particularly with emphasis on developing area navigation systems with vertical guidance capability and heads-up display.
- That the FAA evaluate the need for the installation and use of ground proximity warning devices on air carrier aircraft
- That the FAA continue to emphasize the importance of Part 121.445 [US based airlines, regional air carriers, and cargo operators] carriers in its surveillance and inspection of flight operations under Part 121.
The NTSB had previously recommended the installation of altitude alerting devices (Ground Proximity Warning Systems) in reports released in 1968, 1969, 1971 and 1972 (the Southern flight 932 report was released in late 1972). However it was this high-profile crash which is largely credited for the focus on GPWS in the early 1970s and in 1974, the FAA required all large aircraft to install GPWS equipment.
In view of the apparent success of the nonstandard glide slope at Huntington, it is unfortunate that such an installation was not made sooner. However, the experience gained with this installation should provide a basis for possible application to other airports where standard installation criteria cannot be met without major construction.
If the DC-9 had been following this glide slope, then it would have been about 2,500 feet closer to the hill when it reached the minimum descent altitude. Only if it were pitched nose-down by 10° (an unusually steep angle) could it have crashed into the trees.
This student made a video about Southern Airways flight 932 as his “National History Day Project on Tragedy and Triumph” (which, to be honest, I watched with the sound off). I felt the images and newspaper headlines did a good job of showing the impact of the crash.