NavList:
A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding
Re: Astronomical Refraction: Computational Method for All Zenith Angles
From: Frank Reed CT
Date: 2005 Aug 22, 18:26 EDT
From: Frank Reed CT
Date: 2005 Aug 22, 18:26 EDT
[note to list members: I promise, no code in this message. there are some conclusions relevant to navigation in the last paragraphs] Marcel, you wrote: "It seems that there is something wrong in my stratospheric part. May I ask you, Frank, to have a look at it? May be you can see what is going wrong there." I am not using the analytic model from the Auer-Standish article so I can't help you with your code. In my opinion, the authors' choice of an analytic atmosphere model from 1944 was misguided. First, as I noted previously, that model is simplified in a fashion that was determined by the calculational limitations of sixty years ago. It uses the number "5" for the polytropic index of the troposphere simply because it's an integer. Second, the very idea that we need to treat "polytropic" and "exponential" sections of the atmosphere separately is a limitation of the analytic treatment that obscures the fundamental equivalence of these different sections of the atmosphere. The whole atmosphere (assumed to be hydrostatic and spherically symmetric) can be specified by giving temperature as a function of altitude based on real empirical observations. If I say, 'temperature falls 7.2 degrees per kilometer up to 11km and then remains constant above that height,' I have given enough information to determine the atmospheric density profile and therefore the refraction completely (apart from very small differences arising from changing humidity at low altitudes and changing composition at very high altitudes). Equally easily, I could feed the model a detailed temperature profile taken from actual weather balloon observations with a hundred different lapse rates and altitude limits. That's the beauty of this approach: it's completely general. I was able to reproduce the sample refraction tables in the Auer-Standish article almost perfectly. After a little more experimentation, I found that the best results come from setting the lapse rate to -5.692 (degrees Celsius per kilometer) below 11km and zero above that. Here's the zero observer height table: Alt----ref"------Au-St.----diff" 0 2189.4 2189.4 1 1532.6 1532.6 2 1145.5 1145.5 3 899.2 899.2 4 732.7 732.8 0.1 5 614.5 614.6 0.1 10 330.5 330.5 15 221.5 221.5 30 104.0 104.0 45 60.1 60.2 0.1 In this table, the leftmost column is the altitude in degrees, the second column is the refraction in arcseconds as calculated in my code, the third column is the refraction listed in the Auer-Standish article and the third column shows the difference between the two in arcseconds (when different from zero). These differences are extremely small, and I consider them insignificant. When the observer altitude is 15,000m, the agreement is still excellent though the differences close to the geometric horizon are a little larger: Tfactor of 1.0 and Lrate=-5.692 for obs h=15000 m: Alt----ref"------Au-St.----diff" -3 2316.1 2316.4 0.3 -2 1188.4 1187.9 -0.5 -1 600.0 600.6 0.6 0 353.3 353.4 0.1 1 235.6 235.8 0.2 2 171.3 171.5 0.2 3 132.4 132.5 0.1 4 106.9 107.0 0.1 5 89.1 89.2 0.1 10 47.4 47.5 0.1 15 31.7 31.7 30 14.9 14.9 45 8.6 8.6 I suspect that the small variances (all still less than 1 arcsecond!) at the lowest altitudes are due to the small differences in the way the troposphere is modeled. I don't think they're anything to worry about. Next, the tables in the Nautical Almanac. Here we have two versions: pre-2004 and post 2004. They require different lapse rate models: Using a temperature factor of 273.15/283.15 and changing the lapse rate to -7.25 reproduces a refraction table very close to the standard (pre-2004) Nautical Almanac: Alt----ref'------N.A.----diff' 0.00 34.4 34.5 0.1 0.25 31.4 31.4 0.50 28.7 28.7 0.75 26.4 26.4 1.00 24.4 24.3 -0.1 1.25 22.6 22.5 -0.1 1.50 21.0 20.9 -0.1 1.75 19.6 19.5 -0.1 2.00 18.3 18.3 2.25 17.2 17.2 2.50 16.2 16.1 -0.1 2.75 15.2 15.2 3.00 14.4 14.4 3.25 13.7 13.7 3.50 13.0 13.0 3.75 12.3 12.3 4.00 11.8 11.8 4.25 11.2 11.2 4.50 10.7 10.7 4.75 10.3 10.3 5.00 9.9 9.9 6 8.5 8.5 7 7.4 7.4 8 6.6 6.6 9 5.9 5.9 10 5.3 5.3 11 4.9 4.9 12 4.5 4.5 13 4.1 4.1 14 3.8 3.8 15 3.6 3.6 The NEW Nautical Almanac refraction table: Using a temperature factor of 273.15/283.15 and and changing the lapse rate as follows: ht<3000: LRate= -9 300013000: LRate= 0 produces a refraction table very close to standard (post-2004) Nautical Almanac: Alt----ref'------N.A.----diff' 0.00 33.6 33.8 +0.2 0.25 30.8 30.9 +0.1 0.50 28.3 28.3 0.75 26.1 26.1 1.00 24.2 24.1 -0.1 1.25 22.4 22.3 -0.1 1.50 20.9 20.8 -0.1 1.75 19.5 19.4 -0.1 2.00 18.3 18.2 -0.1 2.25 17.1 17.1 2.50 16.1 16.1 2.75 15.2 15.2 3.00 14.4 14.3 -0.1 Note that the differences in these latter tables are on the order of 0.1 arcMINUTES. Of course since that's the precision limit of the table, it's hard to experiment with greater accuracy. So which of these various lapse rate models is correct? The short answer is 'all of them' and 'none of them'. You can find various diagrams on the net of "typical" temperature profiles and "standard" atmosphere profiles, but all of these are just long-term averages and idealizations. There's a lot of variability about what happens in the real world in the lower atmosphere, and there's no particular reason to prefer, for example, the temperature profile that regenerates the current almanac's refraction table. I really do not believe that there is any good reason to prefer the post-2004 Nautical Almanac refraction tables to the pre-2004 tables (though later evidence may yet persuade me!). The difference in the lapse rate profile does not appear to be significant. Is all lost then? Is there anything positive that comes out of this sort of analysis?? I think there is. By varying the temperature profiles in ways that reflect real variability in Nature, we can determine in an obective (albeit theoretical) fashion just how far we can trust the refraction tables. Throwing in temperature inversions, changes in the lapse rate, the location of the lower edge of the stratosphere, and all of that, I find that the refraction tables are 'rock solid' above about 3.0 degrees. Almost nothing that can happen in the weather is likely to make any meaningful change of 0.1 minutes or larger above that height. Even at 2.0 degrees altitude, the changes in refraction amount to only a few tenths of a minute. This suggests that refraction tables can be relied upon safely down to very low altitudes (on the other hand, under the same conditions that would yield unusual changes in refraction, the 'dip' of the horizon will be much more liable to error). I've read in various places before that the refraction above 10 degrees or so is almost completely insensitive to atmospheric structure, and the refraction close to the horizon is sensitive mostly to the rate of change with height of the temperature in the atmosphere. These numerical integrations confirm those general statements beautifully and even place stronger limits on the numbers. Refraction above 15 degrees is COMPLETELY insensitive to atmospheric structure (and for that matter, it can be calculated easily from k*tan(z)). Refraction between 3 and 15 degrees is nearly insensitive to atmospheric structure. And refraction below 3 degrees depends in detail on the temperature structure of the lowest layers of the troposphere. There is no 'correct' refraction at such low altitudes. [note that all of this applies to an observer at sea level]. By the way, if anyone would like to experiment with these models, let me know. I've been considering making a little web app out of it. -FER 42.0N 87.7W, or 41.4N 72.1W. www.HistoricalAtlas.com/lunars