Airline Safety: Realistic Improvements

Observations of a SWA Line Captain

ATIS PDF  Jump to Runway Incursions | Pressurization | Obvious Knobs | NOTAMS


Airline safety is a continual defense against multiple threats: interruptions, weather, apathy, complacency, fatigue, mechanical failures, difficult personalities, stress, poor system design and more. We build armor in levels: automation and standardization are two of the most effective methods at breaking deadly chains wrought by attacks of circumstance.

It is up to those who design the system to build improvements at all levels and to be realistic in their expectations of the frail humanity putting them to practice.

The Cost of Safety

The best practices preserve as much safety as humanly possible within reasonable cost. Our reality is that we must weigh safety measures by their cost and benefit. We could cut the number of allowed pilot hours in half and safety would be increased by some amount. We could require airbags at every seat, double the number of flight attendants, install less flammable interiors, lighten the allowable weights for airliners (carry fewer people) to improve runway margins and many, many other things. Added together they would price the airline out of business. Even if all the airlines complied, overall traveling safety would decrease by sending people to their far, far more dangerous personal cars.

But there are some things that cost almost nothing (in relative terms) and have some measurable benefit. Not dramatic, to be sure, but a value seemingly well worth their expense.

Cost Per Life Saved

It doesn't sound appealing but it's a necessary calculation. If a million dollar improvement will save 1 life than the cost of the improvement is a million dollars per life. If a 100 million dollar improvement saves 1000 lives then the cost is $100,000 per life saved. The more expensive option turns out to be the better deal.

What if a billion dollar improvement saves 10 lives? That's 100 million per life saved. Since ticket prices must increase to cover the extra cost, some passengers will revert to less expensive and more dangerous travel modes—overall transportation safety will decrease.

Sometimes safety improvements come from very expensive systems and other times they are merely procedural. Just because a system is expensive doesn't mean it will be expensive per life saved. If it's very effective then it may be worthwhile.

Affordable Armor

Enhanced Ground Proximity Warning System (EGPWS) typifies an expensive technology that cumulatively cost billions but dramatically improved safety. Several "saves" have already been scored that kept flesh and aluminum intact. That alone probably paid for the fleet installation and maintenance of the gear for 20 years, not to mention the anguish avoided. 

Other examples abound. Southwest Airlines recently undertook a project that sought the reality of line operations and set to improve it. Seeing the process firsthand was refreshing. They avoided saying "we've always done it this way" but rather made improvements to procedures that worked—not just what sounded good on paper. They should be applauded. The changes cost money by delaying initial taxi. Accidents are too rare for trend analysis and so they concentrated on other metrics, fixing what they could. Although the process was expensive, most of the changes were not. They continue this practice. 

But there are certainly other areas with room for improvement.

1. Taxiing: 10-9 Airport Diagrams

Here is a page dedicated to improvements of these important charts.

It has been well recognized that attention must be concentrated outside the cockpit while moving. The more attention spent inside the cockpit, the more room for error. We've reduced how much time the captain taxis solo while the F/O is buried in cockpit chores, but one risk remains: pilots have to periodically look the airport diagram while taxiing. Jeppesen's 10-9 page it could use a lot of improvement. Obviously the airline doesn't produce these charts but it's a major customer that should implore Jeppesen to make these improvements.

The goal must be to minimize how long a pilot looks at the chart to decipher instructions. There are a number of visual tools that should be employed in the design of these charts and I have used them in the modified example. It's a given that pilots should try to understand taxi instructions before moving, but that is not always the reality.

Below is the current FLL 10-9 page and below that is a modified one. Look at the two diagrams and see which one is quicker to understand the sample instructions. Remember, a captain may be taxiing a moving airplane when getting instructions (normally he's not yet moving on the initial call but may get a revised routing while underway).

Southwest flight 1 is parked on the far east side at terminal one. A crew is getting ready to taxi and gets this:

“Southwest 1, taxi to runway 9L via T, T5, S, hold short of taxiway Q”.


Modified (below)


Now consider this one. The flight just landed on runway 13.

Southwest 1, turn left when able, left on Delta. Taxi to your gate using Sierra, Bravo, enter the ramp at T2.

Look at both charts to see which are easier to use.

The pilot is going to take much less time looking at the modified chart to figure (or re-figure if given new instructions while moving) his route. Given the FAA's appropriate emphasis on minimizing runway incursions, that is an important outcome. 

Other elements for improvement:

1. Increase the airport diagram size by eliminating surrounding airport features. How many pilots use these charts to know about items such as trees, roads, etc., near the airport? The answer is very few (none that I've talked to). We use them to taxi. The current full chart  gives up 20% of it's width to airport surroundings—valuable chart real estate. Bordering information may be valuable for studying the chart beforehand, but gives up clarity in its primary function, understanding taxi routings especially in a dimly lit cockpit.

2. Use visual icons for the most common taxiway notes. Icons can be overused but, in this case, they can help. The max wingspan notes are the best example since pilots will recognize quicker the need to check if they can use a taxiway. The most common values are max weight and max wingspan (or aircraft model). There's never a minimum wingspan so, if the pilot sees a wing icon with 112' in it, he knows immediately whether or not the note applies to him and can go on to read the rest of it. Importantly, this reduces clutter, too.

3. Further reduce clutter by putting runway numbers in the runway and terminal descriptions in the terminal using white on black.

4. Further reduce clutter by avoiding word descriptions of obvious things wherever possible such as VOR's. Every pilot knows the standard VOR symbol so the word "VOR" is superfluous and adds to clutter. Clarity is improved by having secondary information, such as buildings in a lighter shade of gray.

It's true that charts fall outside the control of the airline but we are, after all, the customer. And we pay dearly for this service—we can have some input. In all likelihood, Jeppesen (Boeing) wants to have a good product but they are like any other business, not wanting to spend money on changes unless the customer really wants them to.

2. Reducing Potential ATIS Errors

Anything we can to reduce pilot workload in dynamic situations, including taxiing, will improve safety. One area is how ATIS (Automatic Terminal Information System) information is recorded and presented to the pilot. Pilots should always be able to look in the same place every time for information. Milliseconds are meaningful when taxiing an aircraft. It's not enough for management to tell the captain to "just ask." The fact is that captains will occasionally look at the ATIS sheet for various information from closures to gate to altimeter setting.  Do we want them slurping up precious seconds hunting for that info?

Another reason for standardizing placement of the information is that it reduces the possibility of confusing data. That has happened to me before.  Most at risk is confusing the temperature/due point, altimeter setting and winds.

The solution is fortunately incredibly cheap and simple—a standard form. The form, used by most airlines, puts the data is in the same place every time and uses human factors layout concerns to make the data elements obviously different, even in low light. Pilots don't spend valuable attention trying to decipher it. While the form (see below) is certainly best, just standardizing the layout of the data would accomplish a lot. 

Consider this example. The left one was pulled off an airplane that I took over, the right one done as an example. This seems clear enough. But notice how the winds, temp/dewpoint and altimeter setting were written—with a dash. And this is common. 12 - 02 and 29 - 70. 

Now consider when the temperature is 30C and dewpoint 26C, not uncommon in the summer. The pilot, looking at such a napkin, would see 30 - 26 (temperature/dewpoint) which looks a lot like the altimeter setting of 30.26. It was exactly this error that caused a 500' deviation on one flight (see STL example above) and motivated me to develop the ATIS Sheets in the first place. That napkin suckered two reasonable pilots into misreading it. It will happen again.

This form is designed to separate the elements in very clear ways using shapes 
that are easily discernible even in low light conditions.

While the amount of extra time it saves is small, it frequently comes during critical flight phases that may be the difference between catching something or not. Having a dedicated form like this is ideal, but the least that should be done is putting data in standardized places to minimize potential confusion. For example, if napkins are used, the altimeter setting should always be in one corner, the same corner, and winds in the  opposing corner. It must be standardized across the airline in order to be effective.

Yes, the benefit is small, but so too is the effort/expense. It may take a billion flights to score a "save," but with the miniscule expense, it would be an easy chink to have in the armor.

This improvement has been adopted by my airline in an ingenious fashion. I wish I'd have thought of it. They are printed on an unused section of our weather packet. It requires no extra paper meaning the cost is almost incalculably small (toner).

3. Pavlov and Pressurization | Helios Crash

737Cockpit.jpg (111028 bytes)In many ways, humans behave quite predictably. If you respond to a stimulus often enough, the response becomes automatic and the relationship very powerful. We've all heard of Pavlov's famous dog which exemplifies the principle (actually, it goes a bit further). Human factors design takes this stimulus/reaction, along with many other factors, into account when creating warnings, switches, procedures and so forth. 

Amazingly, there is one critical warning system in the Boeing 737 that is doomed to be mis-interpreted. It completely ignores human factors. It is the cabin altitude warning system.

As a crew prepares for every single flight, they perform a check of the takeoff warning horn—an important warning intended to prevent attempting takeoff without all the controls properly set. The pilots move the throttles forward and listen for it's distinctive "" Every flight—just before leaving the gate, push a throttle up and listen for that buzzing. So every time hear they "" it's related to the takeoff configuration warning. Every time. While taxiing out, they move the throttle lever up to make sure they do not hear that distinctive warning while glancing at the flaps to insure they're set.

When cruising at altitude, losing pressurization can render a crew unconscious in seconds. While a rapid depressurization would be obvious from popping ears, hissing noises and probably fogging, a slow loss might be missed. Guess what the first indication the pilots will likely get of low cabin pressure? Yup, the "" of a takeoff warning horn. Boeing uses the exact same sound to warn of a cabin pressure loss.

You'd like to think that trained pilots would know the difference. You'd be wrong. Remember, the pilots connect this noise on every single flight with the throttles/takeoff warning that they test before pushing back and while taxiing out. 

It turns out, numerous pilots have gotten the warning during climb and did not realize what it was for. Upon hearing the "" they immediately suspected an issue with the takeoff warning. In fact, they didn't realize the pressurization was failing until the oxygen masks dropped (which occurs after the cabin has depressurized further). 

Given the dire consequences of an unconscious crew, this situation is truly unfathomable. This error has been happening on 737's since before I was flying them and is truly tragic that the system has not been modified. An operational change could help by acknowledging that the horn being tested is for both cabin altitude and takeoff warning. But the real fix is —a unique warning, preferably a voice that clearly states "cabin altitude". This warning is too important to save money by using the same buzzer.

It appears that this same exact failing caused the loss of aircraft and occupants in the Greek 737 accident where they crashed with an unconscious flight crew.


Thankfully, action has been taken to mitigate this risk in the last few years. Boeing and airlines have made changes to better prepare crews and to improve the warning system with a light that clearly specifies which problem the horn is sounding for. Apparently, it costs an inordinate amount of money to change the warning given how much better that would be. The proof would be in how many times airline crews actually allow the cabin to climb so high the masks drop. I haven't seen that data but haven't heard about it, either, so I suspect that overall it's now pretty rare.

4. Obvious Knobs

mcp300.jpg (38903 bytes)

This panel (above - click to expand) is very well engineered. The primary knobs used to control the plane's flight path are uniquely shaped so they are less likely to confuse, even if the pilot is looking elsewhere when he uses them. The new below uses the same shape knobs for everything. A distinct step backwards.

The Mode Control Panel (MCP) is how you tell the autopilot what to do, setting airspeed, heading and altitude. When I first used the 737-300's, I was impressed with how the knobs were fashioned, using shape to clearly differentiate them. It was hard to twist the wrong thing. If you put your hand on the cone-shaped airspeed knob it was nigh impossible to confuse it with the mushroom-shaped heading knob which was obviously different from the cylinder-shaped altitude knob. Brilliant! These were designers who obviously understand human-factors (read that frailties).

Then came the Boeing 737-700 with its new MCP. Amazingly, those obvious shapes were missing, replaced with the same size cylinder knobs that varied only by their texture--a far less obvious clue. That, in my opinion, was a significant step backwards. 

Practice has shown that indeed I have twisted the wrong thing a couple times. I've always caught it by looking up but why add the risk? The fix is cheap and easy—go back to the previous and more obvious method.

5. Noticing Notams 

It's not missing the forest for the trees, but rather picking out the one tree that matters in a forest of like-looking siblings. This almost bit me once on a runway who's last half was closed, a fact that I missed while perusing 2 pages of nearly irrelevant notices to airman (NOTAMS).

NOTAM1.gif (27350 bytes)NOTAM2.gif (20346 bytes)

The top NOTAM is one of several dozen that a pilot gets for LAS. The bottom one is the same NOTAM formatted in a way that pilots can read easily. While it may use more paper, the whole exercise is pointless if pilots are ignoring them. All NOTAMs should be formatted like the lower one to improve clarity and dramatically increase the use of this important resource.

NOTAMs, of course, explain special circumstances that may be important to pilots. There are several significant problems with the system and these problems have contributed to accidents and close calls. The tragedy is how easy these problems could be corrected.

1. The part of the airport affected gets lost in the time stamp. Yes the time stamp is important, but why wade through all that clutter only to determine it doesn't apply to you? One or two seconds per NOTAM item may not seem like a lot until you get to an airport with 20 or 30 NOTAMS and you've got 7 minutes to push. Of course you should just slow down but the reality of humans is that most pilots will justify skipping the novella by saying they'll look at it in flight which means they'll probably be skipped.

2. Put the most relevant NOTAM's at the top. Start with runway closures then go to Taxiway closures and, if your airline has the capability, prioritize only those taxiways you're actually allowed to use.

3. Make them obvious by how they're indented. This can improve the speed of interpretation by 10 fold since pilots could very quickly eliminate notices that don't affect them. You could almost scan down by the patterns alone and get to those you care about.

4. Only use abbreviations that are known by 90% of the pilots. That's a relatively short list. While it may be ideal for all pilots to know all likely abbreviations, that is not the reality. I've had numerous times when neither I nor my first officer knew what an abbreviation meant. In nearly every case a call to the dispatcher revealed he didn't know what it meant either! Some of them I knew back in various training stages or college, but years of filling my brain with more commonly appearing material has faded those memories. 

Lets Do What We Can

These are all things that could be done relatively cheaply, even on a per-life-saved basis. My hope is to encourage those who are in a position to enact change. We've done so much already, lets keep going. Lets at least do the easy stuff to continue making our operation the safest it can be.

Please send questions or comments to:

Thanks for listening.

Jeff Goin
Captain, 737 

This is a blank where the existing taxiway designators have been removed before creating a new chart. I have to go through this, Boeing/Jeppesen does not; they can just turn off layers in the source documents.

© 2016 Jeff Goin & Tim Kaiser   Remember: If there's air there, it should be flown in!