Putting Atlantis At Risk

Warning: A long story that has a lot of technical details – and a moral.

Reflecting on my career as a Space Shuttle Flight Director, there were many difficult decisions to make and sometimes some terrifying results.  Putting peoples’ lives at risk and potentially destroying complex and expensive vehicles paid for by the US taxpayers was a daily possibility for every decision made.  Sometimes the consequences were only clear in hindsight.

Fifteen years in the NASA Space Shuttle Flight Director’s office, I supported missions during the on-orbit phases which was always fun, sometimes frustrating, but only rarely worrisome. 

Prior to launch, during the countdown, all the problems really belonged to the Firing Room team and the Launch Director down in Florida.  They had to prep the vehicle and make sure every little piece was working as it should or the system would scrub for the day.  I had only to think about the weather at the abort landing sites – an interesting problem but almost always clear cut:  Go or No-Go. 

To prepare for the flying part of a shuttle launches, my team and I practiced very hard.  The training team would throw problem after problem into every launch simulation.  (See my earlier post ‘Nexus of Evil’ https://blogs.nasa.gov/waynehalesblog/2010/02/16/post_1266353065166/).  When we could proficiently handle each individual problem, the trainers would throw in combinations of problems, more and more and more.  Until we cried ‘uncle’ and past that.  Until we could respond almost automatically to any number of combinations of multiple situations.  Frequently the training days ended with sweat and pounding hearts as if we had been at the gym instead of sitting comfortably under the air conditioning; the work was mental and sometimes psychological. 

During the launch phase (a bare 8 ½ minutes) there was precious little time to think, the responses had to be conditioned in, automatic almost.  In the end, every one of ‘my’ launches in reality was uneventful with no significant problems.  At the end of a real flight day, I sometimes felt that our training had been wasted.  But we were prepared. 

On the other hand, being the Flight Director for the Entry phase of a Space Shuttle flight was the most nerve-wracking job that I ever had.  Takeoffs are optional, landings are mandatory.  With the shuttle in flight entry and landing was inevitable.  Planning and evaluation of potential opportunities, starting before flight, went on the entire mission.  While the three-team rotation of on-orbit flight controllers worked to make sure that the payloads were deployed, the science experiments were performed, etc., etc., the Entry team and its Flight director would come into MCC every day to plan, evaluate, discuss, and digest all the possibilities. 

Evaluate the situation if anything on the orbiter was broken – rarely.  Talk with the landing site folks at the primary and backup sites to gauge the readiness of equipment and personnel at the runways – regularly.  Spend a lot of time in the weather office looking at forecasts – always.  Did you know that meteorology is an inexact science?  We had the best forecasters in the world, but when the landing was more than a few days away pondering forecasts was an exercise in futility – they always changed. 

As the day of landing grew closer, the forecasts were scrutinized to an increasingly minute degree.  All the senior NASA management felt that they were amateur meteorologist and felt free to provide opinions.  But only one person had to make the decision.  Pressure grew daily.  A lot of time was spent at the coffee pot in the hallway outside mission control debating the options with various astronauts, managers, schedulers, and other interested parties. 

About 3 days before landing, we had to commit to send the convoy team to one of the three sites in the US we guessed to be best.  If Florida, we could use the Shuttle Landing Facility at Kennedy Space Center with its single long runway.  If California, the Edwards AFB/Dryden Flight Research Center where they similarly had a single long paved runway but many other options on Roger’s dry lakebed.  The least used option was New Mexico with the hard-as-concrete gypsum dry lakebed runways at White Sands Space Harbor aka Northrop strip.  We only landing a shuttle at WSSH once and it was always the lowest priority in our thinking, fewest number of personnel, least facilities.

Edwards, like the SLF, had one concrete runway that was twice as wide as any normal big airport featured, twice as long as the typical airport runway, and specially constructed to bear the weight of extremely heavy aircraft landings.  But the concrete runways were singletons:  they pointed one direction.  The problem was that winds could come from any direction; not just straight down the runway but across it.  A crosswind could sometimes exceed the maximum allowable for the shuttle: 12 (later 15) knots (nautical miles per hour).  Without a runway in a different direction, well, the landing site was unusable.  At the SLF, surrounded by the swamp, you had to have the winds just right.

Edwards famously had Rogers Dry Lakebed.  The Air Force annually scribed out a large number of ‘runways’ on the hard surface of the lakebed.  They ‘painted’ with asphalt side stripes, threshold markings, and – for the shuttle – aimpoint markings.  To be considered for a normal landing (not emergency) we also required various indicator lights to be set up on a runway and, for the best change at landing, microwave beam scanning landing system (MSBLS) huts that gave precise angle information to an approaching shuttle.  Unfortunately, at Edwards, there were far more runway options than we had equipment.  Generally, only a couple of the lakebed runways were ‘instrumented’ which put them on the candidate list for places to land. 

Oh, and just one more thing.  Rogers Dry Lake sometimes was not dry.  In most winters the rains filled the lake, not very deep, but enough.  When wet – or even damp – the lakebed surface was too soft to land a shuttle.  The tires would dig in, perhaps snap off, maybe make the vehicle tumble.  So, among the criterion for using runways on the lakebed, the first was they had to be dry and hard.  The Air force provided us with the services of their geologist who use various instruments to test the load bearing characteristics of the lakebed. 

Who would have thought?

Runways are designated by the compass direction they point:  for example, runway 33/15 runs northwest to southeast.  An aircraft on the runway would show its compass heading as 330 degrees (northwest) or if pointed the other way at 150 degrees (southeast). 

STS-37 flew in early April 1991 and Rogers Lakebed was just drying out.  Since the prevailing winds in California blew from west to east, the waters had been pushed to the eastern side of the lakebed and the western side started drying out earlier.  We had some confidence that the most western of the lakebed runways 33/15 might be usable, but the geologist had to do his testing.

Just as an additional consideration, the shuttle program had never used lakebed runway 33/15.  The pilot astronauts took turns in the Shuttle Training Aircraft – which flew the final approach like an orbiter – and spend many hours – days – weeks – practicing landing at the runways likely to be used.  Edwards 33/15 was never on the list.  Nobody had ever trained to land on that runway.  Think about that as a potential problem.  It was not shown in the Flight Maps and Charts book that the commander on orbit could consult and become familiar with. 

As the day of landing approached, it was clear that weather was going to be a problem, as usual.  KSC had No-Go forecasts of rain, low clouds, and wind for several days in a row.  This forecast turned out to be accurate. WSSH similarly had unacceptable landing weather.  Edwards looked really with clear skies and no thunderstorms or rain of any kind in the forecast.  But strong north winds blew almost directly across the concrete runway situated east-west (22/04), far exceeding the crosswind limit.  Only the lakebed runway 33/15 pointed in the right direction, if it was dry and hard enough to use. 

The first or normally planned landing day, when we could not consider using the lakebed runways, I had to make the call to wave off for 24 hours – it was just too bad everywhere.  This was not a hard call.  Its easy when its clear; the hard part comes when its marginal.  Crews generally appreciated a wave off if it was called early – they got a day in orbit to rest from a hard paced mission, look out the windows and take pictures, and generally enjoy being in space.

There were limits on how long the Space Shuttle could stay in orbit:  oxygen, water, food, and the Lithium Hydroxide canisters which removed Carbon Dioxide from the air:  all were limited and being used at a substantial rate.  We needed to land soon.

On the +1 day as we called it, the stakes were higher. 

During the wave off day, the geologist completed his evaluation and reported that the 33/15 runway was hard enough to bear the weight of a shuttle landing.  The ground crews scrambled to move approach lights and other navigational aids – but not the MSBLS – to the end of the runway we might use:  the southern end which would put the wind right on Atlantis’ nose.  They were NOT able to respray the asphalt markings.  Those markings, vital cues for a pilot to make a good landing, had been dimmed by the winter rains.  We sent a seasoned astronaut commander to fly the STA and evaluate – all in all the 33-runway was deemed acceptable. 

Remember, we had never used runway 33 before.  And, as it turned out, never would again. 

No pressure on the Flight Director.  No sirree, no pressure.  Never let them see you sweat. 

On deorbit day April 11 I ordered the crew wake up early enough to try to land at KSC in case the weather cleared.  No such luck.  An hour and a half later would be the first opportunity at Edwards. I spent my time between sitting on the console, pacing back and forth, getting briefings on the STA runs, visiting the weather office in person for updates, and having those ad hoc meetings at the coffee pot. 

Too much coffee was consumed. 

There was only one runway in the continental United States that would be go for a shuttle landing that day.  The wind was right down the runway 12 gust to 18 knots, no crosswind. 

But wait.  It gets more complicated. 

Every hour or so a weather balloon is launched at the landing site and tracked it measure the winds aloft as high as 50,000 ft.  On that day the upper winds were ugly, high and shifting.  Worse, there was a tremendous shear in windspeed and direction just above 10,000 ft.  The measured winds aloft exceeded anything previously analyzed for a shuttle landing.  

No pressure, right.  Don’t let my cardiologist know. 

The Space Shuttle Orbiter is a 100-ton glider that has only one shot at making a successful landing.  After the deorbit burn which occurs an hour before there was no way to delay, no go-around possibility, no second try.  Theoretically possible to ‘redesignate’ to another runway, there wasn’t even an emergency runway with acceptable conditions anywhere near Edwards.  To say that the shuttle would be committed was an understatement. 

During shuttle landings, the autopilot is in control from the start of reentry down to about 70,000 ft.  At about the point the shuttle drops below the speed of sound, the commander takes manual control and flies the rest of the way down to landing – which is only 2 ½ minutes.  Very shortly after taking control, the commander starts a turn to align with the runway.  An imaginary cone in the sky – the Heading Alignment Cone – is mathematically displayed to the commander.  On this approach on April 11, we planned for him to make a right turn of ¾ of a circle – 271 degrees.  Commanders sit on the left of the cockpit so they prefer a left hand turn to see the runway as they turn; we planned for a right hand turn to conserve energy.  This meant the commander he could only look at the instruments for flying cues. 

Every time we got balloon data, the mission control team ran a simulation of an automated landing.  We used the autopilot model because we had no way to predict what a real man-in-the-loop pilot would do.  Use of the autopilot for the final part of entry was not allowed because it had certain well-known limitations, and it would not perform well without a MSBLS – which we did not have installed on runway 33. 

But doing a computer simulation using the autopilot all the way to landing gave us critical information about the affects of the upper air winds on the one and only landing attempt that the commander could make. 

After rolling off the HAC and onto the final approach at about 12,000 feet, the shuttle dives to gain speed, using one of the ‘aimpoints’ painted on the lakebed the commander puts the nose down to a glideslope of 19 degrees – 6 times steeper than commercial airlines.  Gaining speed to about 290 knots until the commander starts his pull up at about 2,000 feet to set up for landing.    

Normally the commander is diving at the ‘nominal aimpoint’ – in this case a large black triangle painted in asphalt on the lakebed surface about 7,500 feet from the runway threshold.  Speed – which is synonymous with energy at this point – is controlled by the rudder/speedbrake on the tail of the orbiter.  The automated system checks to see if there is enough energy to make the landing at the criteria desired – 2500 feet past the threshold at 195 kts.  The automated system will open the speedbrake if the energy is too high and close it down if the energy is low:  15% is the minimum setting if the energy is low; it is desirable to approach with the ‘boards’ open to about 20 to 30%. 

If the energy is evaluated to be low from the balloon data, the commander will change to the ‘close in aimpoint’ which is 6,500 feet from the threshold, thus picking up 1000 feet of margin to touchdown.  In even lower energy situations, the commander can stretch the landing by going as low as 185 kts – which gains another 1000 feet in equivalent touchdown distance.  All this to try to get to our normal touchdown point 2,500 feet past the threshold.

The simulations predicted that the landing would be ‘low energy’.  Using the forecast winds the simulation needed the close-in aimpoint and the speedbrakes were full closed all the way down yielding a touchdown distance of 2380 feet past the threshold at 195 knots.  All ‘go’ by the flight rules but with yellow flags indicating caution all around. 

The only options left for such a low energy day would be allowing the touchdown distance to go as short as 1,100 feet past the threshold – considered the minimum for safety – and holding off landing for another 10 knots of airspeed to 185 knots.  Those were it.  The coin of the realm in these situations is distance past the threshold and our purse was almost empty. 

Using the data from the first balloon – at 4 hours prior to touchdown – the situation had deteriorated.  Again, using close in aimpoint, closed boards, the touchdown distance was predicted to be 1800 feet at 195 knots.  Still ‘go’ but getting a brighter shade of yellow. 

My blood pressure ticked up. 

At Edwards the Shuttle training aircraft was flying mimicking a shuttle landing.  The first approach the aircraft would not go into simulated shuttle mode so no touchdown data came in.  The experienced shuttle commander at the controls reported that the turbulence was not an issue and it appeared to be a good day to land. 

Using the L-2.5-hour balloon data the touchdown prediction had deteriorated to 1300 feet past the threshold – still ‘go’ but close to the 1100 foot minimum required by flight rules. We reported this result to the crew.

On his last dive at the runway, the STA pilot reported simulated touchdown at 1600 feet past the threshold which gave me some very slight sense of relief. 

Time to decide.  It always gets really quiet in Mission Control and all the senior managers had stopped giving advice.  It was my call. 

Knowing that the only runway on the continent where the shuttle could safely land had marginal energy conditions, I gave a “Go for Deorbit” to the Capcom to pass to the crew.  I knew that we could have safely waved off for a second day, but tomorrow’s weather forecasts were about the same.

A few minutes later the crew executed the deorbit burn and we were committed.  COMMITTED. 

While the shuttle was on its way down, we got our last update from the L-1 hour balloon data, touchdown 1800 feet past the threshold. Better. I began breathing again.

As the shuttle approached the HAC, the commander delayed by a fraction of a second turning onto the cone, something that was not at all uncommon.  However, flying outside the HAC increased the distance the glider would have to fly and thus made the energy situation even worse. 

Unbeknownst to us there was another phenomenon at play; the wind shear altitude had dropped from 13,000 feet to just under 10,000 feet.  This change in the wind speed and direction effectively robbed the orbiter of some energy. 

In the shuttle software, at 10,000 feet, there is a check of speed and distance to go which, if it is good, the guidance scheme transitions to what is called ‘approach and land’.  When the wind shear was at 13,000 feet the software detected the energy difference and directed the commander to fly more aggressively to make up energy and the A/L transition occurred on time.  With the wind shear below 10,000 feet, the A/L transition criteria was not acceptable and the guidance did not direct the commander to fly as aggressively. Rolling out on final the wind shear delayed the software transition which then did not indicate to the commander he should fly a more aggressive high lift/drag profile.

Coming down the energy situation just got worse.  The commander could see it, but we in MCC could not.  He pulled up at the right point, lowered the gear at the right point, but it was a long way to the threshold.  Too long.  The speedbrakes were closed and at the point the only option at that point was to hold off landing by holding the nose higher as the speed decreased.  Touchdown finally occurred at 166 knots – the second slowest in shuttle history.  The main gear slapped the lakebed surface some 623 feet short of the threshold of the runway.

Disaster?

We got lucky.  Every runway the shuttle was allowed to use has a 1000-foot underrun – safe space before the threshold.  On the lakebed where the stripes were rather arbitrary there were miles of safe landing space. 

It was a safe landing. 

I was unaware at how bad it had been.

It was several hours later when I got the call.  The first apology I made was to the commander for putting him in that position. 

We had been lucky when they paid us to be smart. 

We spent the next six weeks going painstakingly through the situation. 

I thought I was going to get fired for putting the crew at risk.  But the senior folks decided I had learned a valuable lesson. We had all learned a valuable lesson. 

We made many changes to the entry planning process.  One was to make sure the STA pilot reported if the A/L transition was delayed (it had been delayed during his dives but it was not reported to MCC).  Another was to steepen the glideslope by an additional 2 degrees to increase maximum speed before pullout to 300 kts to give more energy.

We trained the pilots to fly the HAC turn closer, although that had mixed success; that is partly what you get with a human in the loop.

But most important was to re-emphasize to Entry Flight Directors to use extreme caution when committing the orbiter to a landing.

It is not a job for the fainthearted. 

The next time we fly a winged vehicle down from orbit, I hope they don’t have to learn that lesson again.