How accurate is it?
The FAA published the cosmic radiation dose for 27 nonstop flights in Advisory Circular 120-61B, computed with its own CARI-7 software. Here is what this site produces for those same flights. The software to check it is free, and section 5 tells you how.
1. The headline, and why it is not the flattering number
Our vertical flight profile has one fitted parameter: the share of a long cruise flown one step below the top flight level. It was fitted against these 27 flights. Scoring it on the same flights would tell you how well it fits data it has already seen, which is not what you want to know. So the parameter is refitted on 26 flights and scored on the 27th, for each flight in turn. Those are the numbers below.
- Flight level and time supplied
- 13.3%
- Neither supplied
- 17%
- Published band, supplied
- ±25%
- Published band, estimated
- ±30%
For the record, fitted and scored on the same 27 flights the deviation is 10.9%. We quote the higher figure because it is the one that predicts a flight we have not seen.
2. Two numbers, because there are two paths
If you tell us the cruise altitude and the flight time, we use them. If you do not, we estimate both from the route length, and the answer is less good. Most people do not know their cruise altitude, so the second case is the common one and it would be misleading to publish only the first.
- You supplied both. Mean absolute deviation 13.3%, root mean square 15.4%, bias +0.6%, individual routes from -28.6% to +34.4%.
- We estimated both. Mean absolute deviation 17%, root mean square 23.1%, bias +1.6%, individual routes from -27.3% to +79.6%.
The band printed on your report is the smallest multiple of 5% that contained at least 24 of the 27 routes in that test: ±25% when you supply the flight level and time, ±30% when we estimate them. That is a coverage figure you can check against the table below, not a confidence interval.
3. The comparison
Effective dose in microsieverts, ICRP-103 weighting, for January 2000 under quiet geomagnetic conditions, which is the basis the FAA used. The geomagnetic field is interpolated to that date, as CARI-7 does. Green is within 10% of the FAA figure, red more than 25% away.
| Route | FL | Hours | FAA | Ours, supplied | Diff | Ours, estimated | Diff |
|---|---|---|---|---|---|---|---|
| IAH to AUS | FL200 | 0.5 | 0.19 | 0.17 | -8.7% | 0.34 | +79.6% |
| MIA to TPA | FL240 | 0.6 | 0.41 | 0.31 | -23.3% | 0.48 | +16.0% |
| STL to TUL | FL350 | 0.9 | 1.41 | 1.58 | +12.1% | 1.62 | +14.8% |
| TPA to STL | FL310 | 2 | 4.06 | 4.35 | +7.2% | 5.53 | +36.2% |
| MSY to SAT | FL390 | 1.2 | 2.71 | 2.83 | +4.4% | 2.07 | -23.6% |
| LAX to HNL | FL350 | 5.2 | 12.2 | 8.76 | -28.6% | 9.75 | -20.6% |
| JFK to SJU | FL370 | 3 | 8.12 | 7.42 | -9.2% | 8.29 | +1.7% |
| HNL to LAX | FL400 | 5.1 | 13.5 | 12.12 | -10.8% | 10 | -27.3% |
| LAX to NRT | FL400 | 11.7 | 35.1 | 39.51 | +17.0% | 32.55 | -7.7% |
| NRT to LAX | FL370 | 8.8 | 26.9 | 24.49 | -9.9% | 33.14 | +22.4% |
| IAD to LAX | FL350 | 4.7 | 15.1 | 12.86 | -15.5% | 14.42 | -5.7% |
| JFK to ORD | FL390 | 1.8 | 6.6 | 7.05 | +6.8% | 5.25 | -20.4% |
| LIS to JFK | FL390 | 6.5 | 22.4 | 23.94 | +12.9% | 23.98 | +6.3% |
| LHR to DFW | FL390 | 9.7 | 33.2 | 42.28 | +34.4% | 39.27 | +17.6% |
| SEA to IAD | FL370 | 4.1 | 16.4 | 14.4 | -12.8% | 16.28 | -2.2% |
| DFW to LHR | FL370 | 8.5 | 29.7 | 31.88 | +14.1% | 39.4 | +33.1% |
| ORD to SFO | FL390 | 3.8 | 14.7 | 13.99 | -5.4% | 11.94 | -19.9% |
| SEA to ANC | FL350 | 3.4 | 12.4 | 10.07 | -19.7% | 10.13 | -19.6% |
| SFO to ORD | FL410 | 3.8 | 15.6 | 15.96 | +7.2% | 12.07 | -23.8% |
| JFK to SEA | FL390 | 4.9 | 20.2 | 20.48 | +6.8% | 17.25 | -15.9% |
| LHR to JFK | FL370 | 6.8 | 27.3 | 25.13 | -8.9% | 28.38 | +3.5% |
| JFK to NRT | FL430 | 13 | 55.1 | 68.42 | +29.3% | 55.59 | +0.5% |
| NRT to JFK | FL410 | 12.2 | 50.7 | 58.2 | +21.0% | 56.93 | +11.6% |
| LHR to LAX | FL390 | 10.5 | 44.6 | 45.82 | +8.5% | 45.06 | +0.3% |
| ORD to LHR | FL370 | 7.3 | 30.8 | 27.46 | -11.6% | 33.1 | +7.0% |
| LHR to ORD | FL390 | 7.8 | 34 | 34.22 | +6.1% | 33.18 | -3.8% |
| ATH to JFK | FL410 | 9.4 | 44.6 | 42.14 | -6.1% | 37.07 | -18.2% |
4. What this comparison cannot tell you
- The reference is itself an older CARI-7. Run CARI-7 v4.1 on the FAA's own published flight profiles and it comes out about 6% above the doses printed in the Advisory Circular. So part of the spread above is the distance between two versions of the FAA's own software, not our error.
- These 27 flights are not the world. They are almost all US-based and northern hemisphere. A southern-hemisphere or equatorial route is not represented.
- It measures agreement with CARI-7, not with reality. For that the reference is Meier et al. (2018), who compared CARI-7A against instruments flown on real aircraft. Individual deviations ran from -8% to +11%, and the authors put the upper limit of the model's uncertainty at about 20%. Note they compared ambient dose equivalent and absorbed dose, which are measurable; the effective dose on your report is defined by ICRP-103 tissue weighting and can only be calculated. None of that is included in the bands above.
- Solar particle events are not modeled, by us or in this comparison.
5. Check it yourself
CARI-7 is free and the FAA distributes it, with source, from its radiobiology page. Every report we issue lists the date, top flight level, block time and profile for each flight, so you can enter the same inputs and compare. If you get a materially different answer for the same profile, tell us: corrections go on the corrections log.