Thanks to @jklymak, there was an update in PR I created (https://github.com/pydata/xarray/pull/7553). @ghiggi with the code from this PR, your code shows identical plots (except for the first one, but that should be the case). Hopefully the PR can be merged somewhere soon.

@veenstrajelmer I am not sure I understand what you are saying. In the example I pass only norm to the plotting function ... As suggested by @kmuehlbauer the solution to this issue is to specify the extend argument in both the plotting call and cbar_kwargs. In the example, extend was only defined in cbar_kwargs (since is an argument also of mpl.Figure.colorbar). Likely we should align the arguments somewhere in the code with something like this:

if extend is None and cbar_kwargs.get("extend) is not None: extend = cbar_kwargs.get("extend) if extend is not None and "extend" not in cbar_kwargs cbar_kwargs["extend"] = extend

@ghiggi: If I understand it correctly, your issue/examplecode covers multiple issues. Since one subissue might be using norm instead levels, I would recommend trimming down your example code do only show the remaining vmax issue.

Thanks to all the people above that have started digging into the problem ! @veenstrajelmer : Adding the levels=levels argument (together with norm, ... or dropping norm) does not correct/change the output figure. Of course commenting #da1.data[da1.data>=norm.vmax] = norm.vmax - 1 "solves" the issue, but this line of code is what enables to show up the bug, which is occurring when the array does not contain any value equal to or higher than norm.vmax

The related issues https://github.com/pydata/xarray/issues/4061 and https://github.com/Deltares/xugrid/issues/49 are fixed by supplying levels=levels instead of norm=norm to ds.plot(), as suggested by https://github.com/pydata/xarray/pull/7553#discussion_r1117264787. However, I cannot judge if this also fixes the example from this issue, since the plots still all look different. However, commenting #da1.data[da1.data>=norm.vmax] = norm.vmax - 1 solves most of the differences. @ghiggi: could you check if this suggestion solves your issue indeed?

I just combined @Huite's suggestion with splitting the if-statement. This works for both solving the issue and keeping the testcases in test_plot.py green. if levels is not None: cmap, newnorm = _build_discrete_cmap(cmap, levels, extend, filled) norm = newnorm if norm is None else norm if isinstance(norm, mpl.colors.BoundaryNorm): cmap, norm = _build_discrete_cmap(cmap, levels, extend, filled) This could replace this code: https://github.com/pydata/xarray/blob/1de881e5f5d4d7697403f9ab7686849ab02953bd/xarray/plot/utils.py#L307

However, a bit up in the code there is a if isinstance(norm, mpl.colors.BoundaryNorm) statement, which I guess could be combined. https://github.com/pydata/xarray/blob/1de881e5f5d4d7697403f9ab7686849ab02953bd/xarray/plot/utils.py#L277 if isinstance(norm, mpl.colors.BoundaryNorm): levels = norm.boundaries

I think it is a potential solution nevertheless, but some help is appreciated with the last steps. Also since the case of @ghiggi seems not to be solved with this fix. It does solve https://github.com/pydata/xarray/issues/4061 though.

Debugging this, @headtr1ck points correctly to _determine_cmap_params:

python if levels is not None or isinstance(norm, mpl.colors.BoundaryNorm): cmap, newnorm = _build_discrete_cmap(cmap, levels, extend, filled) norm = newnorm if norm is None else norm

The problem lies in the second line. In _build_discrete_cmap, a new cmap is returned with a different number of levels. However, if the original norm is not None, you end up with a mismatch, as the old norm expects the old cmap.

What then happens, is that the norm calls into the cmap. Calling into cmap doesn't do any checks whether the value is larger than N, it just takes the highest available value. The examples in #4061 show this quite clearly, but to illustrate:

```python import matplotlib as mpl import matplotlib.pyplot as plt import numpy as np

data = np.arange(100).reshape((10, 10)) cmap = mpl.cm.get_cmap("viridis") print(cmap.N) # 256 boundaries = [0, 25, 50, 75, 100] norm = mpl.colors.BoundaryNorm(boundaries, cmap.N) fig, ax = plt.subplots() ax.imshow(data, norm=norm, cmap=cmap)

%%

colors = [cmap(i/255) for i in np.linspace(0, cmap.N, len(boundaries) - 1)] new_cmap, new_norm = mpl.colors.from_levels_and_colors(boundaries, colors) print(new_cmap.N) # 4 fig, ax = plt.subplots() ax.imshow(data, norm=new_norm, cmap=new_cmap)

%%

Mismatched

fig, ax = plt.subplots() ax.imshow(data, norm=norm, cmap=new_cmap)

%%

```

This is avoided here by removing the conditional in the second line, or just making sure both cmap and norm are replaced by their new values:

python if levels is not None or isinstance(norm, mpl.colors.BoundaryNorm): cmap, norm = _build_discrete_cmap(cmap, levels, extend, filled)

Then, the cmap and norm remain in sync.

However, when running the tests in test_plot.py, this gives a single questionable(?) failure:

```python def test_norm_sets_vmin_vmax(self) -> None: vmin = self.data.min() vmax = self.data.max()

    for norm, extend, levels in zip(
        [
            mpl.colors.Normalize(),
            mpl.colors.Normalize(),
            mpl.colors.Normalize(vmin + 0.1, vmax - 0.1),
            mpl.colors.Normalize(None, vmax - 0.1),
            mpl.colors.Normalize(vmin + 0.1, None),
        ],
        ["neither", "neither", "both", "max", "min"],
        [7, None, None, None, None],
    ):

        test_min = vmin if norm.vmin is None else norm.vmin
        test_max = vmax if norm.vmax is None else norm.vmax

        cmap_params = _determine_cmap_params(self.data, norm=norm, levels=levels)
        assert cmap_params["vmin"] is None
        assert cmap_params["vmax"] is None
        assert cmap_params["norm"].vmin == test_min
        assert cmap_params["norm"].vmax == test_max
        assert cmap_params["extend"] == extend

      assert cmap_params["norm"] == norm

E assert <matplotlib.colors.BoundaryNorm object at 0x000001C7A4CB07C0> == <matplotlib.colors.Normalize object at 0x000001C7A50C4760> ```

I don't understand why the conditional is there. At first sight, it doesn't make a lot of sense to create a new norm and cmap, but then take the original norm?

I'm still struggling to understand that whole workflow and it looks like this is no easy way of getting to the bottom of this.

But adding kwarg extend="max" to the calls to pcolormesh/imshow fixes this.

The docstring gives a hint:

extend : {'neither', 'both', 'min', 'max'}, optional How to draw arrows extending the colorbar beyond its limits. If not provided, ``extend`` is inferred from ``vmin``, ``vmax`` and the data limits.

So in first case everything works nice since extend is derived from vmin/vmax and data limits. In the second case, extend is obviously not set to the expected value (as in cbar_kwargs) as we've tampered with the maximum data values.

Not sure if aligning extend and cbar_kwargs-extend is already the solution.

Is there any update on this issue? I have been running into the same problem recently and am happy to see that this issue was already recognized by others.

It seems that the cmap gets replaced by xarray in https://github.com/pydata/xarray/blob/c0011e13b1edabcd541610d7fbef084bcb387183/xarray/plot/plot.py#L1182 which calls https://github.com/pydata/xarray/blob/c0011e13b1edabcd541610d7fbef084bcb387183/xarray/plot/utils.py#L854 which then calls https://github.com/pydata/xarray/blob/c0011e13b1edabcd541610d7fbef084bcb387183/xarray/plot/utils.py#L154 and this method has some special casing of BoundaryNorm that will create a new cmap

Not sure what all this code is supposed to do, though... But hopefully these findings can help someone with better understanding to debug this :)