error in plot_alignment

[miniufo]

Describe the bug
When I tried to understand the kwarg alignment using the plot_alignment, it raises an strange error.

Code Sample

Below is the example codes. When slicing with 8 values of both init and time, it raises an error ValueError: The truth value of a CFTimeIndex is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all(). Change to 10 is fine in this case.

from climpred import HindcastEnsemble
from climpred.tutorial import load_dataset
import matplotlib.pyplot as plt
import numpy as np
import xarray as xr

hind = load_dataset("CESM-DP-SST")["SST"].isel(init=slice(8))
verif = load_dataset("ERSST")["SST"].isel(time=slice(8))

hindcast = HindcastEnsemble(hind)
hindcast = hindcast.add_observations(verif)

print(hind.init.values)
print(verif.time.values)
hindcast.plot_alignment(edgecolor="w", figsize=(15, 5))

Expected behavior
No exception is raised when slicing with 8.

Output of climpred.show_versions()

Details

INSTALLED VERSIONS ------------------ commit: None python: 3.13.7 | packaged by Anaconda, Inc. | (main, Sep 9 2025, 19:59:03) [GCC 11.2.0] python-bits: 64 OS: Linux OS-release: 6.14.0-29-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: zh_CN.UTF-8 LOCALE: zh_CN.UTF-8

climpred: 2.4.0
xarray: 2025.4.0
pandas: 2.3.2
numpy: 2.2.5
scipy: 1.16.1
cftime: 1.6.4
netcdf4: None
nc_time_axis: 1.4.1
matplotlib: 3.10.6
cf_xarray: 0.10.9
xclim: 0.58.1
dask: 2025.9.0
distributed: 2025.9.0
setuptools: 72.1.0
pip: 25.2
conda: None
IPython: 9.1.0
sphinx: None
None

Screenshots

Additional context
No

[aaronspring]

Thanks for raising this question.

I suspect for 8, there isn't any overlap in valid_time from forecast and time from verification (for some lead). Could you please add the returns from both print statements?

I agree it would be great to surface that failure mode as an error message.

[miniufo]

Yes you are right. The maximum lead is 10. So a simple if xxx < max_lead would be better than this error.

I am a bit confused by this plot. The 2D init-lead plot shows some boxes, but when should a box be shaded? I am trying to understand this through your example codes.

[aaronspring]

Could you please post the figure here?

[miniufo]

Could you please add the returns from both print statements?

The output with slice(8) is

[1954. 1955. 1956. 1957. 1958. 1959. 1960. 1961.]
[1955 1956 1957 1958 1959 1960 1961 1962]

Could you please post the figure here?

When slice(10), the plot is:

[aaronspring]

Thanks. Confirming that the error message should be: "no overlap between hindcast valid_time and verification time for all leads. Please try plot_alignment with less leads or longer verification time."

[miniufo]

Could you explain a bit on the 2D init-lead plot shows some boxes? When should a box be shaded and when it shouldn't?

[aaronspring]

When a box is colored this forecast verification pair enters the metric calculation

[miniufo]

Thanks @aaronspring.

I've tried to estimate the prediction skill like:

This is the evolution of RMSE as a function of lead. Can I get a similar curve but as a function of inits? So I can see how the RMSE increases with time (in cases of no data assimilation, RMSE will increase with inits).

[aaronspring]

If you want to get skill over dimension/dim init, don't use init in dim argument. Use member or [] (empty list should also work for rmse)

[miniufo]

Great. I didn't realize the empty list is also a choice, as I didn't find any docs on demonstrating this (maybe I've missed that).

[aaronspring]

Not very obvious from the docs. Here example with spatial dims only https://climpred.readthedocs.io/en/stable/examples/decadal/verify_dim_implications.html