Astropy 4.3+ Produces Inaccurate Range Predictions via AltAz Calc

Description

We regularly compare our radar Range measurements to ILRS ephems. Upon upgrading to Astropy 5.1, we started seeing very large residuals. We’ve tried different versions and found astropy 4.2 gives expected agreement, while astropy 4.3 and later give the unexpected behavior.

We believe that the astropy 4.3+ predictions are in error because we have internal residual analysis and comparison to other data sources (not using astropy) that do not show this issue.

Expected behavior

When we run it with astropy 4.2, we get nice small residuals like this:

(astropy5) mas@ip-192-168-4-110 scripts % python astropy_snippet_20220608.py
	Calculated measurement residual
		range: 0.1m
	Calculated measurement residual
		range: 12.4m
	Calculated measurement residual
		range: 17.8m

Actual behavior

When we run it with astropy 4.3, we get great big residuals like this:

(astropy5) mas@ip-192-168-4-110 scripts % python astropy_snippet_20220608.py
	Calculated measurement residual
		range: -112.2m
	Calculated measurement residual
		range: -99.9m
	Calculated measurement residual
		range: -94.4m

Steps to Reproduce

1. Get some radar Range measurements
2. Get ILRS Ephem predictions
3. Create an EarthLocation for the radar location.
4. Interpolate the ILRS ephem for every measurement timestamp.
5. Create a SkyCoord object for the interpolated position.
6. Do an AltAz transformation to get the Range prediction.

import numpy as np
import scipy.interpolate
import datetime
import dateutil.parser

import astropy.time, astropy.coordinates
import astropy.units as u
from astropy.utils import iers
iers.conf.iers_auto_url='https://maia.usno.navy.mil/ser7/finals2000A.all'

# Instrument:
inst = {
    'longitude': -85.479833,
    'latitude': 10.413333,
    'altitude': 23.0,
    }

measurements = [
    {
        'timestamp_measurement': dateutil.parser.parse('2022-06-08 08:18:45.103596+00'),
        'range_value': 1813723.24341158,
    },
    {
        'timestamp_measurement': dateutil.parser.parse('2022-06-08 08:18:44.983595+00'),
        'range_value':  1813538.3470108,
    },
    {
        'timestamp_measurement': dateutil.parser.parse('2022-06-08 08:18:44.926049+00'),
        'range_value':  1813449.3366306,
    },
]

ilrs_mjd = 59738
ilrs_secs = np.array([
    27900.00000, 28080.00000, 28260.00000, 28440.00000, 28620.00000,
    28800.00000, 28980.00000, 29160.00000, 29340.00000, 29520.00000,
    29700.00000, 29880.00000, 30060.00000, 30240.00000, 30420.00000,
    30600.00000, 30780.00000, 30960.00000, 31140.00000, 31320.00000,
    31500.00000, 31680.00000, 31860.00000, 32040.00000, 32220.00000,
    ])

ilrs_data = np.array([
    [-2741944.579,   -235689.755,  -7325387.829],
    [-2711657.122,  -1474364.598,  -7189654.207],
    [-2641154.833,  -2675083.522,  -6860757.255],
    [-2530855.876,  -3806852.586,  -6347386.337],
    [-2381441.653,  -4840435.015,  -5663181.834],
    [-2193903.565,  -5749124.122,  -4826398.133],
    [-1969601.044,  -6509460.381,  -3859440.304],
    [-1710326.586,  -7101871.185,  -2788282.056],
    [-1418372.400,  -7511212.894,  -1641776.504],
    [-1096593.106,  -7727194.734,   -450875.156],
    [ -748458.835,  -7744665.880,    752222.831],
    [ -378094.716,  -7563753.660,   1934963.517],
    [    9698.097,  -7189847.764,   3065344.786],
    [  409443.848,  -6633429.388,   4112807.145],
    [  815017.796,  -5909751.857,   5049077.091],
    [ 1219697.263,  -5038388.983,   5848936.440],
    [ 1616238.610,  -4042669.245,   6490897.504],
    [ 1996980.100,  -2949016.560,   6957766.909],
    [ 2353970.392,  -1786221.732,   7237086.985],
    [ 2679123.360,   -584665.510,   7321450.016],
    [ 2964396.995,    624488.873,   7208680.076],
    [ 3201992.320,   1810113.002,   6901878.965],
    [ 3384567.960,   2941919.905,   6409337.507],
    [ 3505465.893,   3991242.320,   5744316.504],
    [ 3558942.459,   4931767.385,   4924700.825],
    ])

ilrs_time = astropy.time.Time(ilrs_mjd + ilrs_secs / 86400, format='mjd', scale='utc')
x = ilrs_data[:,0]
y = ilrs_data[:,1]
z = ilrs_data[:,2]

# instrument location
topo_loc = astropy.coordinates.EarthLocation(
    lat=inst['latitude']*u.deg, lon=inst['longitude']*u.deg, height=inst['altitude']*u.m
)

for meas in measurements:

    # measurements
    time_meas = meas['timestamp_measurement']
    ttime = astropy.time.Time(time_meas, format='datetime', scale='utc')
    trange = meas['range_value']

    # find closest point in time
    imin = abs(ilrs_time - ttime).argmin()

    # interpolate, 9 point
    tinterp = range(imin - 4, imin + 5)
    t0 = ilrs_time[tinterp][0]
    tl = (ilrs_time[tinterp] - t0).sec  # ilrs times
    teval = (ttime - t0).sec + np.array([-0.05, 0, 0.05])  # times to evaluate at
    px = scipy.interpolate.lagrange(tl, x[tinterp])
    py = scipy.interpolate.lagrange(tl, y[tinterp])
    pz = scipy.interpolate.lagrange(tl, z[tinterp])
    posi = np.array([px(teval), py(teval), pz(teval)])

    # position in ITRS
    dpos = astropy.coordinates.SkyCoord(
        x=posi[0, :], y=posi[1, :], z=posi[2, :],
        representation_type='cartesian',
        frame="itrs", unit="meter"
    )

    # position relative to site
    dpos_altaz = dpos.transform_to(
        astropy.coordinates.AltAz(location=topo_loc)
    )
    rs = dpos_altaz.distance.m

    # residuals
    r_resid = trange - rs[1]

    print('\tCalculated measurement residual')
    print('\t\trange: %2.1fm' % r_resid)

System Details

Please note that we have reproduced this in several environments.

We were able to reproduce it with the following environment:

Python 3.8.13 (default, Mar 16 2022, 20:38:02) 
[Clang 13.0.0 (clang-1300.0.29.30)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import platform; print(platform.platform())
macOS-11.6-x86_64-i386-64bit
>>> import sys; print("Python", sys.version)
Python 3.8.13 (default, Mar 16 2022, 20:38:02) 
[Clang 13.0.0 (clang-1300.0.29.30)]
>>> import numpy; print("Numpy", numpy.__version__)
Numpy 1.22.4
>>> import erfa; print("pyerfa", erfa.__version__)
pyerfa 2.0.0.1
>>> import astropy; print("astropy", astropy.__version__)
astropy 4.3
>>> import scipy; print("Scipy", scipy.__version__)
Scipy 1.8.0
>>> import matplotlib; print("Matplotlib", matplotlib.__version__)
Matplotlib 3.5.2
>>> 

[mkbrewer]

Since this is a recurring issue, I will raise a separate issue to propose the above more straightforward approach. If I get the go ahead, I will submit a pull request that will solve this problem once and for all.

[mkbrewer]

Actually, it would be better to form the ITRS to AltAz matrix outside the loop.