Benchmark astro nodal corrections against known sources

[joeberkovitz]

In the general spirit of making sure, it may be a good idea to benchmark the nodal corrections for Schureman and/or IHO fundamentals against some known sources. I did this in a quick and dirty fashion for constituent M2 as a proof of concept, comparing the node-year correction factors and equilibrium arguments for midyear 2026 with those spit out by XTide's congen program. I may not have done it right, but the node factor was very close (.0001 different) and the equilibrium argument was a little off (56 degrees vs 66 degrees).

Maybe this was already done, I just didn't see it captured anywhere so I thought I'd make sure.

[joeberkovitz]

I made an error in my quick attempt to compare the numbers, omitting the V0 contribution. Adding this in, it looks as though the equilibrium arg is closer than I previously stated.

The numbers come out as follows for M2 on midnight for these dates:

UTC	speed (Xtide)	u (Xtide)	speed (iho)	u(iho)	speed (schur)	u(schur)
2026-01-01	.9674	66.36	.9654	66.05	0.9650	66.06
2026-07-02	.9674	66.36	0.9677	308.96	0.9674	308.96

The Xtide based corrections are binned per-year. So I think the right way to benchmark those for neaps purposes may be to combine Xtide's midyear speed correction (it minimizes error across the year bin) and the Jan. 1 equilibrium arg (since this is intended as a phase offset to be added to the constituent calculated in days from Jan. 1 00Z on the given year).

So the errors for Schureman fundamentals appear nonexistent for speed and slight (but maybe significant) phase.

[bkeepers]

I think this is an excellent idea. Might make sense to benchmark against pytides too.

I'm a bit of a noob when it comes to xtide. If you can share how you're generating those outputs I can write a benchmark script that compares all constituents.

[joeberkovitz]

XTide is a bit annoying to build and run, so I got these numbers from the MarineTides repo which claims they got them from running XTide's congen program. I believe those corrections to be pretty accurate because benchmarking my own hand-rolled predictor in R using these numbers came in very close to NOAA. (I was playing around with R for my own edification, not because I am planning to use MarineTides.)

This very indirect sourcing of corrections is not super helpful, I realize. Benchmarking against pytides might well be a better idea, and I didn't research it. If you want the MarineTides corrections I can extract them in R as a CSV. Note I did not look into their license since I am not planning to use that package.

NOAA has also published tables of nodal corrections in these reports:

• The original Schureman report which you must have seen: https://tidesandcurrents.noaa.gov/publications/SpecialPubNo98.pdf
• Its supplement Zetler (82) which extrapolates to 2025 https://tidesandcurrents.noaa.gov/publications/SpecialPubNo98Suppl.pdf

I guess for benchmarking purposes it doesn't matter what year, so one could do worse than use these tables :-)

[joeberkovitz]

Note that congen with standard definitions doesn’t support the full IHO constituent set.

[bkeepers]

If you want the MarineTides corrections I can extract them in R as a CSV.

That would be great! Thank you.

Note that congen with standard definitions doesn’t support the full IHO constituent set.

Ok. I thought I would just poke around at it a little and see what I could get.

[bkeepers]

NOAA has also published tables of nodal corrections in these reports:

• The original Schureman report which you must have seen: https://tidesandcurrents.noaa.gov/publications/SpecialPubNo98.pdf
• Its supplement Zetler (82) which extrapolates to 2025 https://tidesandcurrents.noaa.gov/publications/SpecialPubNo98Suppl.pdf

I guess for benchmarking purposes it doesn't matter what year, so one could do worse than use these tables :-)

Getting them into a usable format could be a pain, but could be valuable. I had 4 different AI agents extract/check/double check the IHO constituents from the PDF and I still found errors upon manual inspection.

[joeberkovitz]

I’ll send the CSV later on today.

WRT the machine-sort-of-readable tables in the reports, I wonder if it’s ok to begin by focusing on a few high amplitude constituents which yield decent coverage of the correction code. To avoid or at least defer a bunch of distracting tedium…

[joeberkovitz]

I don't think you need the CSV since you found the congen_output.txt file - they should be identical.