Trends in Cirrus Fatal Accident Rates - Pull early, pull often! - Safety and Training Programs - Cirrus Owners and Pilots Association

Trends in Cirrus Fatal Accident Rates

A couple of recent threads have highlighted the recent upsurge in Cirrus fatal accidents.  To help visualize what has been happening, let me post updates to the charts that I prepare for the plenary session of CPPP that discusses General Aviation and Cirrus accident statistics.

Cirrus Fatal Accidents By Quarter

First, the number of accidents over time.  For convenience, I use 3-month quarter-years.  This matches the GAMA production reports that are released each quarter.  It also helps to differentiate accidents that occur in the summer months (April to September) from the dark days of winter (October to March).  70% of Cirrus fatal accidents occur in the winter 50% of the year.  And you can see bunches of accidents that occur in those months with less daylight and worse weather.


Cumulative Cirrus Fatal Accidents

Next, the cumulative number of fatal accidents.  Of course, this trends upwards since there is no way to reduce the cumulative total.

Cirrus Fatal Accidents Per 1,000 Airplanes

What is worrisome is seeing the cumulative number of fatal accidents accelerate over time.  But so too is the number of airplanes delivered in the Cirrus fleet.  So that gives us the next chart, the ratio of fatal accidents per 1,000 airplanes delivered.

Now we see both upward and downward segments over time.  The downward trends occur when fewer accidents happen while the number of airplanes increases.  The upward trends occur when more accidents happen at a rate faster than new airplanes are delivered.  2001 was a good year with few accidents.  The time between 2003 and 2005 was good with the rate of accidents staying stable as the fleet grew.  However, late 2006 and late 2007 were horrific.  Again, it is those winter months.  The ratio grew from 6 to 8 fatals per 1,000 planes in 2006-07, and then grew again from 8 to 10 fatals per 1,000 planes in 2007-08.  So far in the first 2 months of the winter of 2008-09, we are holding even . . .

Cirrus Fatal Accident Rate Per 100,000 Flying Hours

Then there is the flying time rate.  This is the rate used by the FAA and NTSB to calculate the number of fatal accidents per 100,000 hours of flying time.  The GA fleet has about 500 accidents in 24,000,000 hours of flying time.  The GA rate has declined from 1.33 in 2002 to 1.19 in 2007.

Fortunately, for COPA, Cirrus Design gives us the flying hours for the SR2X fleet.  That enables us to plot this final chart.  The fatal accident rate is calculated for each quarter from the actual number of fatal accidents and interpolating the flying time for that quarter.

In the early years, a few accidents cause big swings in the rate.  The first single accident in 2001Q2 caused a jump to 4.5 accidents per 100,000 hours, but the fleet was only 200 airplanes that had flown only 25,000 hours.

The next jump to 4.5 was from 3 accidents in 3 months ending in January 2003 while the fleet was still small, less than 800 airplanes that had flown less than 200,000 hours.

Yet, the last year has seen 13 accidents and the rate has barely climbed to 1.8.  Now the fleet has over 4,000 airplanes that fly about 800,000 hours a year.  So that's stabilizing the rate.

That short green line?  If you take away from the GA accident rate all of the multiengine and turboprop/turbojet airplanes, I calculate the GA single-engine piston rate to be about 1.86 fatal accidents per 100,000 flying hours.

So that is a tutorial on accident rates based on the Cirrus fatal accident history.

Obviously, COPA wants to reduce the number of accidents by eliminating all of the pilot causes.  Just pulling the parachute more often would help.  But talking to other COPA Pilots about flight safety will also help.



  • Rick,

    your "Cirrus Fatal Accidents Per 1,000 Airplanes" is new and fascinating data.  I can imagine two competing trends in that graph.  One is we have seen from other data that the hours  flown per year goes down as aircraft age, so as the average fleet age increases one would expect average hours flown per aircraft to go down.  Less hours flow would presumably lead to fewer accidents and would make that graph trend down.

    A second factor is that many aircraft get resold, and the new owners often don't receive factory training.  One could postulate that could lead to higher accident rates.   Or even for aircraft that don't get resold, one could speculate that for many pilots their recurrent training regime may be less effective than the initial training.  Or that complacency sets in with long-term ownership.  In any event, it seems whatever upward factors exist are overwhelming the negative.  Figuring out what they are may give us clues on what to focus our safety programs on.

  • Ah, Gigi, good point.  I don't think the graph means that.  I believe it is the 'cumulative' graph, meaning does not suggest 'expected accidents per year', but some other strange datum such as 'expected accidents over the fleet life'.

    Now that I think about it, I'm not sure my previous comment is correct; even with a constant accident rate per 100,000 hours I believe this graph could trend up or down, depending on the changes in the production rate.  I don't think this graph of itself tells us much about the safety stats as it is conflated with the average fleet age.

  • Eric,

    This graph tell us several things :

    - First, Cirrus airplanes are very efficient killers!

    10% of the Cirri (41 over the 4000 produced) have killed (the average score is 2 people per Cirrus killer).

    - Second, the trend will decrease this quarter because a lot of Cirri are waiting an owner on Duluth apron : without a pilot a Cirrus cannot kill

    - Third,  on the contrary if CD stop the production, this graph will trend up even if the accident rate per 100,000 hours decreases. Cirri are potential killers if they find pilots.

    The story is : general aviation is dangerous because of the pilots.

    CD gave us planes and we, pilots, are killing us, our family and friends with them. Even if we have aboard Entegra, TKS, Stormscopes, TCAS, sofisticated AP etc...  

    If you learn to deal with the weather (stay away from thunderstorms and ice, don't fly above mountains if no blue sky and sun shining!), if you learn to hand fly your plane in IMC without any help from AP, if you learn to navigate only with a watch and a compass, if you learn what is a DA, a MDA and a MSA your Cirrus will not be a killer Cirrus...

  • of course you corrected the lapsus calami : 1% of Cirri are Cirrus killer...

  • Rick -

     As I recall, time in type is an excellent prediction of accident rates.  Do you have enough data to put together a rate vs. time-in-type graph?


  • Rate vs. time-in-type graph is an interesting idea if that is possible - the fleet has been growing rapidly, so has the numbers of pilots operating with limited time on type - such that the Cirrus accident rate rea have been that bad? .Just a hypothesis. Would data from the first few years of a comparable type be available  - say Bonanza- demonstrate that?

    I remember a forum comment (attributed to an insurer) about the prolific Robinson helicopter fleet having a relatively high accident rate, but this being normal for the initial period of a highly successful (sales wise) new aircraft type.

    What are the factors in a 'fleet break-in' period? What happened to the Bonanza and Robinsons fleets to improve their rates post 'break-in'? Maybe it only time in type? Maybe the 'collective wisdom/learning' of the experience of that initial pilot base making its way further across the pool of later pilots of the type?

    If the latter is a factor maybe the COPA big pool of knowledge needs to makes it way across the Cirris Pilot community - beyond the Cirrus'pride and joy/lifestyle' pilots and owners that tend to be visible at COPA forums/CPPP's?

  • Careful folks, trends are not causes.  These are indicators, not predictors.

    So we have some work to do with potential causes.  I'll do some Pareto charts of NTSB causes and factors, along with John's suggestion of SR2X time-in-type.

    Thanks for the extensive interest and discussion.



  • just to say the time-in-type graph wa Chucks idea, rather than mine, cheers

    I do have another idea, the concept of "Time On Mission", but that's for another day.