Unexpected Solar Declination Value for Vernal Equinox Date (March 20, 2024, 12:00 UTC) with Skyfield 1.53

[OfferLa]

Hello Skyfield Developers and Community,

I've been using Skyfield (version 1.53) to calculate solar declination and have encountered an unexpected result for March 20, 2024, at 12:00:00 UTC, which I've reproduced on both Google Colab and Kaggle notebook environments.

Issue:
For the date and time 2024-03-20 12:00:00 UTC, my Skyfield setup consistently calculates an apparent geocentric solar declination of approximately +0.012173 degrees.

This value seems unexpectedly low. Based on the known Vernal Equinox time for 2024 (approx. March 20, 03:06 UTC), the declination at 12:00 UTC (almost 9 hours later) should be significantly more positive. Other sources and manual estimations suggest a value closer to +0.1465 degrees.

Minimal Reproducible Code:
Here's the minimal Python code I'm using:

from skyfield.api import load, Loader
import pandas as pd
import skyfield # To print version

# Ensure fresh loads (though the issue persists even without aggressive cache clearing)
# loader = Loader('~/.skyfield-data', verbose=True) # Path for local, Colab/Kaggle might use /root/ or /content/
# ts = loader.timescale(builtin=False)
# planets = loader.planets('de421.bsp')
# Skyfield seems to handle downloads and caching well, so simpler load:
ts = load.timescale()
planets = load('de421.bsp') # Using standard DE421 ephemeris

earth, sun = planets['earth'], planets['sun']

# Test date
date_utc = pd.Timestamp('2024-03-20 12:00:00', tz='UTC')
t = ts.from_datetime(date_utc)

# Calculate apparent geocentric position
astrometric = earth.at(t).observe(sun)
ra, dec, distance = astrometric.apparent().radec()

print(f"Skyfield version: {skyfield.__version__}")
print(f"Pandas version: {pd.__version__}") # pd.__version__ if pandas is used
print(f"Target Date (UTC): {date_utc}")
print(f"Calculated Apparent Geocentric Declination (degrees): {dec.degrees}")

# For comparison, astrometric declination:
# ra_astrom, dec_astrom, dist_astrom = astrometric.radec()
# print(f"Calculated Astrometric Geocentric Declination (degrees): {dec_astrom.degrees}")

Environment Details (Consistent across Colab & Kaggle for my tests):

• Skyfield Version: 1.53
• Python Version: (e.g., 3.10.x - Colab/Kaggle usually have recent versions)
• Pandas Version: (e.g., 2.2.x)
• Ephemeris: de421.bsp (as loaded by Skyfield)
• Operating System: Linux (via Colab/Kaggle)

Expected Result:
Based on other astronomical sources (e.g., NOAA Solar Calculator for a rough check, other astronomical libraries, and calculations by other users), the apparent geocentric solar declination for 2024-03-20 12:00:00 UTC is expected to be approximately +0.1465 degrees.
For example, a local test by another user with Skyfield 1.48 and de421.bsp yielded 0.14652916406613006 degrees.

Question:
Could there be a reason why Skyfield 1.53, specifically in these cloud notebook environments (or perhaps generally), might be producing this lower declination value for this particular date/time? Am I potentially misusing a feature or is there an environmental factor I might be overlooking? I have tried clearing the Skyfield cache in Colab with shutil.rmtree on the ~/.skyfield-data directory and forcing re-downloads, but the result remains the same.

Any insights or suggestions would be greatly appreciated. Thank you for developing and maintaining this excellent library!

Best regards!

[brandon-rhodes]

The problem is that you have asked Skyfield for RA and declination in the J2000 ("ITRF") reference frame, but apparent RA and dec are traditionally expressed using "equinox-of-date" coordinates. Here is some documentation on the topic in case it's useful:

https://rhodesmill.org/skyfield/positions.html#apparent-right-ascension-and-declination

You will want to compute the coordinates like this:

ra, dec, distance = astrometric.apparent().radec('date')

Let me know if there's an adjustment I can make to Skyfield's documentation that would put this information where you would have found it more easily, and I will edit the docs accordingly!