Single-scale image writes are serialized as multiscale pyramids with ome-zarr==0.14.0

[ArneDefauw]

Writing a single-scale xarray.DataArray image with SpatialData.write() produces a multiscale OME-Zarr pyramid on disk when using ome-zarr 0.14.0, even though the input is not multiscale.

Expected behavior

A single-scale image should remain single-scale on disk unless pyramid generation was explicitly requested.

Roundtrip expectation:

input: xarray.DataArray
output after read_zarr(): xarray.DataArray
Actual behavior
The image is written as a pyramid (s0, s1, s2, ...), and read_zarr() returns a multiscale object on roundtrip.

Minimal reproducible example

import tempfile
from pathlib import Path
from numpy.random import default_rng

import dask.array as da

from spatialdata import SpatialData, read_zarr
from spatialdata.models import Image2DModel

RNG = default_rng(0)

data = da.from_array(
 RNG.random((3, 800, 1000)),
 chunks=((3,), (300, 200, 300), (512, 488)),
)

image = Image2DModel.parse(data, dims=("c", "y", "x"))
print(type(image)) # xarray.DataArray

sdata = SpatialData(images={"image": image})

tmpdir = Path(tempfile.mkdtemp())
path = tmpdir / "data.zarr"

sdata.write(path)
sdata2 = read_zarr(path)

print(type(sdata2["image"]))

On-disk result
The written image group contains multiple scales:

data.zarr/images/image/
s0/
s1/
s2/
s3/
s4/

Suspected cause

spatialdata writes single-scale images through ome_zarr.writer.write_image().

In current ome_zarr, write_image() defaults to building a pyramid via:

scale_factors=(2, 4, 8, 16)
So calling it without overriding scale_factors writes a multiscale pyramid even for a single-scale input.