CRFI - Know what you're getting!

[Vsplat]

I continue to encounter pilots who are either struggling with CRFI or are in flight operations organizations that do not understand its use. I thought another thread on this topic might help.

First, some background. http://www.tc.gc.ca/civilaviation/commerce...lars/AC0164.htm

CRFI is a modelled value that provides a theoretical prediction of the distance needed from 50' to a full stop under the conditions resulting in a given CRFI value. The number is a raw, unfactored value, valid 95 times out of 100 when the actual runway conditions match those in the RSC report. The safety margins that TC notes are assumed conservatism in their model. Cold comfort if you live one of the 5 out of 100 discarded data points or learn the hard way that the averaging in the CRFI masked a really low friction spot.

CRFI is NOT a predictive value. It is a forensic one, and a fragile one at that.

Why do I say this? Because CRFI is measured at several points along a runway, then processed, then reported to the tower or FSS. This all takes time. As precipitation continues to fall, the temperature changes, or the runway is plowed, swept, or treated, the reported CRFI value falls out of sync with reality. Last week I was inbound to YYZ for the better part of a day and at one point had CRFI reports for 16 hours! Most of these were of course nonesense. Numerous runway maintenance events had taken place, yet the CRFI values prior to the maintenance actions were not pulled.

There are a number of folks, some of them at TC, who believe CRFI would make a great dispatch tool. Of course, this relies on the predictive value, reference comments above. It also relies on basic stuff like temperature, which is not in aviation forecasts. So 5 hours from now, will the runway be wet, slushy, or ice covered?

When used within its context and with the caveat that it has to be fresh, CRFI can really help round out what the QRH has to say about your upcoming arrival. I use it frequently and like its usefulness with crosswinds. Used like a silver bullet to override the AFM prediction, CRFI can and has caused crews to land when the AFM predicted the overrun that occurred.

So, ask yourself - on the 'day' when you face the lawyer, do you want to be holding up the QRH for your real airplane or the CRFI model extrapolated from other machines? Me- I'm going to be holding both. And I will never, ever accept a runway when CRFI says I can but my QRH says I can't.

Cheers

Vs

[Specs]

One of the questions that nags at me it is that if the manufacturer's numbers were not reliable then there would be a whole whack of lawsuits against them and regulatory action to change their numbers. There hasn't been.

Misapplication of CRFI data though has resulted in overruns.

Why can't crew just rely on the manufacturer's numbers for actual landing performance?

[J.O.]

Hi Vs:

Your knowledge is extensive and I certainly appreciate the comments you've shared with us on the subject. I sent some of your previous comments to our flight ops managers. They were quite interested in it and may use it in the development of guidance material. Can I ask where you gained your knowledge? Have you been involved in performance engineering or something similar?

Jeff

[J.O.]

Why can't crew just rely on the manufacturer's numbers for actual landing performance?

Vs can probably provide more informed comment, but I believe that one of the reasons for the development of CRFI was because some of the manufacturers guidance material was very poor, and in some cases non-existent. In addition, it was an attempt to put a more consistent system in place than the old RSC system which was wide open to errors and misleading information.

[cp fa]

Good Day Vsplat,

Jennifer brought this article to my attention, knowing I am interested in contaminated runway data...I actually put down my beer to come and read it.

Thanks much for the link to the TC information. I had not seen it before.

I can't agree entirely with your comments regarding using CRFI for dispatch. I have used it on occasion when deciding whether or not to dispatch. In all cases it supported a decison not to go. IE: time stamped CRFI+FT=a runway condition that cannot possibly be acceptable for our arrival...

I think guys should use them like SAs...a snapshot. However, I don't think I would exceed CRFI guidlines for crosswind limitations without a more up to date PIREP.

I can only speak to contaminated runway QRH data for the B-767, and if this applies to you, I would be pleased to discuss it via email.

Thanks again for the link.

Cheers

Dave Thomson

[Vsplat]

Good morning all. Glad to see some discussion on this topic.

Specs, one of the reasons JBI, the forerunner of CRFI, was first developed was to address a lack of manufacturer information in this regard. It wasn't required by certification and runway friction measurement was an emerging science. Even today, manufacturers tend to relate to runway friction using mu (can't put the right symbol in).

Manufacturer data has steadily improved to the point where the certification standards for new (not derivative) aircraft require this data. Still, the way in which the data is gathered in certification, like most certification data points, varies from real life. The surface is assumed to be 100% covered with the same depth and type of contaminant, rarely found on the line. Even with that, manufacturers aren't interested in destroying flight test articles, so the bulk of the data they provide is itself modelled. This is important - CRFI is one model, the manufacturer's another.

When both CRFI and QRH speak to the same runway state, I will calculate both and weigh the more conservative view very heavily. Does my manufacturer require me to use thrust reverse? Where does that leave me if directional tracking requires me to cancel reverse to regain the centreline? Stuff like that.

In line operations, CRFI is pretty good in places where the runway is prone to shallow contamination like patchy frost or highly compact snow. CRFI is useless in deeper contamination with high fluid content and in fact is not supposed to be reported when conditions are not valid for measurement. But how many times do you see a current RSC report sent out with the last reported CRFI which was several hours old? This leads to reports where the runway is slush covered and apparently reporting a CRFI.

J.O. thanks for your kind words. Yes, I do this stuff for money and have been trained in performance engineering. I also conduct training. That said, I would not recommend cutting and pasting any of my posts for use in a company setting. One of the reasons I use a handle is so that folks specifically don't take my word for anything. My posts are intended to promote discussion and thought, and, if you feel more is necessary, provide the motivation to consult an official source and find out for yourself if what I say here is true and then decide what you want to do about it. Some of what I do is used to dodge rock while in cloud, so liability is a big thing to me.

Dave, I have also used CRFI as you describe, for short haul operations, say YOW-YUL or times when the forecast is for more of the same and the airport is already reporting trouble. The use you describe has some subtle yet critical differences from using CRFI for dispatch.

My objection to use for dispatch was in response to a recent reguatory initiative that would have mandated CRFI consideration for dispatch, every time. The TC goal was to have us guess what the runway condition would be like based on a forecast that did not have a temperature, assign a CRFI value to this guess, then base our required landing field length on that CRFI distance. Real easy to do when leaving LHR or PVG - not.

Contrast this with what I understand your use to be. I support the use of CRFI to get a picture of what the airport is like right now. That's what it's made for. I would suggest that, once aware of what is going on right now at the airport and recent weather, you would apply the rest of the flight planning package to predict whether or not the airport is likely to afford a safe arrival. For example, if I'm sitting in YUL flight planning, on my way to YYZ and they are down to one runway in heavy snow at -10 and RSC shows 1/2" of snow and CRFI of .2-ish with a forecast of continued heavy snow changing to IP and ZR, I would strongly suspect that the only thing that runway is going to do is get worse, as snow clearing and sweeping are going to drop the CRFI - the chemicals may or may not be effective by the time I get there. I'd pick up the phone and have a meaningful discussion with Dispatch.

I hope that clarifies things a bit.

All the best

Vs

[IFG]

Thanks for the link, Vsplat. It's definitely an area of differing interpretations. Your efforts toward clarity are appreciated.

.... The number is a raw, unfactored value, valid 95 times out of 100 when the actual runway conditions match those in the RSC report.  The safety margins that TC notes are assumed conservatism in their model.  Cold comfort if you live one of the 5 out of 100 discarded data points or learn the hard way that the averaging in the CRFI masked a really low friction spot.

....  Used like a silver bullet to override the AFM prediction, CRFI can and has caused crews to land when the AFM predicted the overrun that occurred .... do you want to be holding up the QRH for your real airplane or the CRFI model extrapolated from other machines?  Me- I'm going to be holding both.  And I will never, ever accept a runway when CRFI says I can but my QRH says I can't.

Re: "CRFI can and has caused crews to land when the AFM predicted the overrun that occurred" - Have there been any incidents in which inadequate CRFI data was specifically cited as contributing? The problem is parsing out contributing issues like incorrect #'s, landing profile/technique, equipment problems etc.

Re: "silver bullet to override the AFM" - the CBAAC you linked seems quite clear:

The onus for the production of information, guidance or advice on the operation of aircraft on a wet and contaminated runway still rests with the aircraft manufacturer. The information published in the AIP Canada on the CRFI and updated by this CBAAC does not change, create any additional, authorize changes in, or permit deviations from regulatory requirements.

This of course offers little backup to those of us operating A/C without any AFM/QRH guidance for landing on contaminated runways (Bingo, J.O. ). Experience bears out that CRFI #'s are generally ‘conservative', and regardless they are the only slippery runway guidance for many of us, but if no 60% factor is applied, one does wonder how we quantify:

The new CRFI table in the AIP Canada and the newer table in this CBAAC include an implicit safety factor such that for 19 out of 20 landings the stated distances will be conservative. This is what is meant by a 95% level of confidence. Or put more simply, if a page contains 100 data points, we have ignored the results from the worst 5 data points in determining the deceleration models.

.... [&] ....

The factors take into account screen heights greater than 50 feet, glidepath angles less than those used by the manufacturer in establishing the dry AFM distance, speed greater than threshold speed (Vref) at 50 feet, touchdown rate of descent less than that used by the manufacturer in establishing the dry AFM distance, pilot and/or system delays in selecting the full braking configuration, less than full braking effort by the pilot, adverse tailwind gusts, effects of temperatures above International Standard Atmosphere (ISA) and/or downhill runway slope, and lastly, variation in the tire to ground maximum friction coefficient

.

(Seems to echo that polling disclaimer: accurate 19 times out of 20 ... & if the values are "conservative" 19/20, how would the other 1 of 20 be characterized? Of course the pollsters happily disavow their exclusions altogether )

Except for the most egregious deviations, that list of considerations "account"-ed for is more than covered under dry conditions by the 60% factor, but experience and intuition will indicate far less buffer in the CRFI chart #'s. Is there any quantitative data available on their “account”-ing?

I can't agree entirely with your comments regarding using CRFI for dispatch. I have used it on occasion when deciding whether or not to dispatch. In all cases it supported a decision not to go.

I also wonder how adding CRFI considerations to the dispatch requirements has a downside, Vsplat, since it wouldn't seem to enable any departure that would currently be disallowed. As you say, the last CRFI carries all the predictive value of the last METAR, but it can only add another hurdle which, cleared, leaves us ultimately no worse off than we are now. IAC, would the alternate alternative (CAR_705.60(3)) still prevail under the TC folks' proposal? Right now we can depart with abominable current destination RSC’s, with both an acceptable alternate, and anticipated improvement at destination, e.g. when a wind shift dictates moving to a different runway. Would they close that option?

In the meantime, until we can quantify the “conservatism” in the CRFI models, I guess one can only encourage the treatment of all lower-CRFI landings as the 'abnormal' landing that they are (as one might, e.g., a tailwind or flapless landing), i.e: absolutely stabilized on speed, on profile, firm touchdown, and in the low-CRFI case, maximum effort stopping - certainly no time for any complacency.

Cheers, IFG

[Vsplat]

IFG I'll try to answer your questions in sequence. I may have to come back later for some.

Have their been incidents caused by trusting CRFI instead of manufacturer's data? Yes. Have there been incidents caused by trusting stale CRFI? Yes. Check the TSB and Aviation Safety letter online for them. One occured in Ottawa while the Working Group considering CRFI was meeting.

Not sure what you mean in your comments regarding the discarded data points. Points were 'discarded' statistically - that is, if they fell outside the reduction model, either on the high or low side, they were dropped. I would draw no assurance that standard dispatch factor makes up for shortcomings in CRFI.

How can CFRI hurt at dispatch? A whole bunch of ways. Company culture is a funny thing. I do not believe for a minute that, once CRFI was incorporated, only unfavorable CRFI reports would count at dispatch. Competitive pressures would eventually affect the operations - the PIC says it looks like a bad idea, but there is the CRFI and forecast saying it will be fine. But, that's not the main reason it bugs me.

When the whole discussion of CRFI and dispatch began, all of the safety cases were clearly slanted to short-haul operations - 2 hours or less. In these circumstances, it was argued, the tool would work. Ignore the fact that there were ongoing overruns on segments of that length already. But what happens when you go out to transcon or overseas operations? The forecasts are worse than meaningless. The saying "it takes fuel to carry fuel" is true. So, if the forecast for your destination 14 hours away means a CRFI of .2 is possible, and the first alternate where you can land is 500nm from the destination, you are going to have to board a whole hockey-sock full of fuel to ensure you have alternate and reserves overhead destination.

Now, apart from the economics of boarding a whole lot of gas just to blow more out the pipes, consider that the accelerate-stop, net takeoff flight path, enroute cruise, driftdown, in fact every phase of flight but landing at the alternate, will be negatively affected. But wait a minute, wasn't the aim to improve safety at destination? Consider that dispatch factors don't apply after takeoff, and you end up with the same aircraft at the same destination with the same lousy conditions, only now, because of your CRFI-driven alternate, you are tonnes heavier trying to stop on that slick runway with a CRFI report of .4 that is 2 hours old. Your risk of an overrun is actually higher.

I am also not paid to kill off the airlines. The costs of implementing this ineffective option would have been huge. Watining for the forecast to improve or landing short? Try that with a duty-time limited crew, or with a slot-limited airspace restriction. Very quickly the operations would descend into IRROPS, and that kind of chaos exacts its own safety toll.

Like so many other performance discussions, CRFI starts out as a numbers exercise, but in reality affects safety through human factors. So to see all of its aspects we have to step way back and look at the pilots, the runway maintainers, the ATC folks handling the data and the dispatchers that tie it all together. When something is going to go through that many hands, you need a bulletproof system to make sure that its accuracy does not degrade as it moves. Aviation weather is such a product - CRFI is simply not on the same level in terms of data accuracy or shelf life. Maybe one day it will be.

Cheers

Vs

[IFG]

Vsplat - I know well there have been over-runs (acquainted as I am with pilots who have suffered them), but I'd be most interested in any report that indicates a failure of the CRFI model itself, whether used "instead of manufacturer's data", or in the absence of any other altogether; i.e., not stale, or misused in any other way, hence: "The problem is parsing out contributing issues [to any incidents] like incorrect #'s, landing profile/technique, equipment problems etc." I'm aware of the TSB site, having enjoyed a few cumbersome searches thru' it. I did check back about 7 years (when the CBAAC came out), and the over-runs reported all contained other issues, so I guess I was slothfully hoping for a little DF steer (A/C type maybe, or month/year?)

Please understand I'm not trying to argue, but I am approaching this as a user of the CRFI charts (they're all I've got), who has some questions on theoretical grounds, but whose own experience so far (fairly numerous, continous short-haul since the first publications of CRFI's) has rather supported a claimed conservatism in the charted values, notwithstanding any reservations about the whole thing.

Re: the discarded data points - I'm not sure what my comment was that you don't understand . But I do have questions, and have drawn no assurances. The stats courses I took have faded somewhat, but I still have some handle on the principles, altho' terminology is ropey. My question/concern in the CRFI case is that I understand discarding extreme data points at both ends, in calculating averages, mean deviations etc, but the CBAAC says: ".... if a page contains 100 data points, we have ignored the results from the worst 5 data points in determining the deceleration models." Seems to my layman's eye to bias the results.

On the dispatch issue, I see your reservations and agree that unintended consequences have to be considered. But the perfect can be the enemy of modest improvement. I don't see how providing a means for mandated cancellation in some instances affects a departure in any other instance, that would be going anyway under the current protocols, particularly in an aggressive go-oriented 'culture'. The long-haul points are well-taken (again the law of unintended consequences rears its head), but again, what would be different in your hypothetical case under current rules? Or is it that you advocate a higher CRFI threshold for departure?

Cheers, IFG

[Vsplat]

IFG, I'll see what I can find for specific report numbers. It may take some time due to other time constraints.

I understand that you are looking for discrete cases where the CRFI report was timely and as accurate as possible, but the table itself predicted the wrong value. The nature of winter and the investigations themselves add so many confounding aspects that I doubt we would ever see such an isolated finding. Of course, we did see specific references to JBI - twice, at Terrace, B.C. if memory serves, but CRFI was intended to address the issues of JBI.

There is much internal debate on the statistical methods used to produce CRFI. Were they too conservative? Not conservative enough? All complicated by the fact that the CRFI model incorporates the air distance from 50 feet to the threshold - where friction is irrelevant but aircraft type differences have a measurable effect. Of all the things that make me scratch my head with CRFI, the decision to incorporate the air model in such a non-adjustable way is probably number 1.

In practice, we don't have the luxury of considering whether we are going to find a data point on the conservative side or not. We simply have to weigh the CRFI report in the context of our other data.

WRT the Dispatch case, we found that CRFI did NOT provide an improvement of any kind. To suggest that it is better than nothing is to reclaim the myth that CRFI has predictive value. Consider also that if CRFI expected BELOW a certain value will trigger a delay or cancellation, then the corollary must also hold - an expected CRFI ABOVE that value will support a departure. This is a wrong model from so many aspects, some of which we have already discussed. If we attach risk to the difference between the flight deck's understanding of a situation and what is actually taking place, then relying on CRFI for prediction adds to that difference, so adds risk.

As I posted above, in very short-haul operations where the CRFI may well be current throughout the flight time from departure to destination, there is benefit to the nowcasting of CRFI. But a CRFI of .3 in snow or IP tells you absolutely nothing about what the CRFI will be in two or three hours.

Of course, we have not yet had athe discussion of deciding between a runway that is reporting a poor CRFI and a runway that is not reporting CRFI at all. We have experienced diversions where the PIC turned down the airport with the poor CRFI, diverted to the alternate because they did not have a bad RSC report to dissuade the crew, but of course found the runway on arrival to be bad enough that they filed an safety report. We can extrapolate this to the dispatch case, where no news is good news, and now you have a conflict between the pilot and their company over a fictitous future value of CRFI, when in fact the forecasted conditions may be identical.

All said, I prefer to work with fact (as do you I'm certain). I consider even the best models to be educated guesses. There is a point beyond which I do not trust CRFI and will discard it as a reference tool because, like an IRS that is drifting, it is weakening the overall decision plot. But that is me.

A final disclaimer for all reading this thread, reference the post above about cutting and pasting this discussion for use elsewhere. I post anonymously for some very important reasons (important to me that is). One of the outcomes I expect is that no one would make an operational decision based on an anonymous post on an internet forum. If you disagree with me, fine. It is just an internet forum and we all have exactly the same value to our opinion here. If, on the other hand, you agree with me, let it be because you have verified my comments using independant sources. And here I mean stuff like the NRC reports, TC guidance material or your manufacturer's Ops Engineering reps. Don't settle for someone offering you their opinion as though it were gospel. Too much of that goes on and we all pay for it.

Here's a link you may find useful. It is for the Transport Canada Development Centre (TDC). This is the think tank for Transport. You will find a wealth of reports here, as well as the minutes of conferences where CRFI was on the menu.

http://www.tc.gc.ca/tdc/publication/listing.htm#air

Happy reading.

Vs

[handyman]

Vsplat, thanks for the link!

The CRFI seams like a logical reporting system. My problem is at last count, there are at least 10 systems in use around the world.

The ICAO friction co-efficient report and the US civil braking action appear similar but slightly different numbers using MU values. Just adding to all the confusion!