added regridding

evaltools/eval.py

@@ -1,5 +1,9 @@
 import numpy as np
 import xarray as xr
+import cordex as cx
+import cf_xarray as cfxr
+import xesmf as xe
+from warnings import warn
 
 
 def regional_mean(ds, regions=None, weights=None):
@@ -100,17 +104,12 @@ def height_correction(height1, height2):
     return (height1 - height2) * 0.0065
 
 
-def seasonal_mean(da):
-    """
-    Calculate seasonal averages from a time series of monthly means.
-
-    Parameters:
-    da (xarray.DataArray): The DataArray to compute seasonal means for.
+def seasonal_mean(da, skipna=True, min_count=1):
+    """Calculate seasonal averages from time series of monthly means
 
-    Returns:
-    xarray.DataArray: The seasonal mean values.
+    based on: https://xarray.pydata.org/en/stable/examples/monthly-means.html
     """
-    # Get number of days for each month
+    # Get number od days for each month
     month_length = da.time.dt.days_in_month
     # Calculate the weights by grouping by 'time.season'.
     weights = (
@@ -121,79 +120,134 @@ def seasonal_mean(da):
     # np.testing.assert_allclose(weights.groupby("time.season").sum().values, np.ones(4))
 
     # Calculate the weighted average
-    return (da * weights).groupby("time.season").sum(dim="time")
+    return (
+        (da * weights)
+        .groupby("time.season")
+        .sum(dim="time", skipna=skipna, min_count=min_count)
+    )
+
+
+def add_bounds(ds):
+    if "longitude" not in ds.cf.bounds and "latitude" not in ds.cf.bounds:
+        ds = cx.transform_bounds(ds, trg_dims=("vertices_lon", "vertices_lat"))
+    ds = ds.assign_coords(
+        lon_b=cfxr.bounds_to_vertices(
+            ds.vertices_lon, bounds_dim="vertices", order="counterclockwise"
+        ),
+        lat_b=cfxr.bounds_to_vertices(
+            ds.vertices_lat, bounds_dim="vertices", order="counterclockwise"
+        ),
+    )
+    return ds
+
 
+def mask_with_sftlf(ds, sftlf=None):
+    if sftlf is None and "sftlf" in ds:
+        sftlf = ds["sftlf"]
+        for var in ds.data_vars:
+            if var != "sftlf":
+                ds[var] = ds[var].where(sftlf > 0)
+        ds["mask"] = sftlf > 0
+    else:
+        warn(f"sftlf not found in dataset: {ds.source_id}")
+    return ds
 
-def get_regridder(finer, coarser, method="bilinear", **kwargs):
+
+def create_cordex_grid(domain_id):
     """
-    Regrid data bilinearly to a coarser grid.
+    Creates a CORDEX grid for the specified domain.
 
-    Parameters:
-    finer (xarray.Dataset): The dataset to regrid.
-    coarser (xarray.Dataset): The target grid dataset.
-    method (str, optional): The regridding method to use. Defaults to "bilinear".
-    **kwargs: Additional keyword arguments to pass to the regridding function.
+    Parameters
+    ----------
+    domain_id : str
+        The domain ID for the CORDEX grid.
 
-    Returns:
-    xesmf.Regridder: The regridder object.
+    Returns
+    -------
+    xarray.Dataset
+        The CORDEX grid with assigned coordinates for longitude and latitude bounds.
     """
-    import xesmf as xe
+    grid = cx.domain(domain_id, bounds=True, mip_era="CMIP6")
+    lon_b = cfxr.bounds_to_vertices(
+        grid.vertices_lon, bounds_dim="vertices", order="counterclockwise"
+    )
+    lat_b = cfxr.bounds_to_vertices(
+        grid.vertices_lat, bounds_dim="vertices", order="counterclockwise"
+    )
+    return grid.assign_coords(lon_b=lon_b, lat_b=lat_b)
+
 
-    regridder = xe.Regridder(finer, coarser, method, **kwargs)
+def create_regridder(source, target, method="bilinear"):
+    """
+    Creates a regridder for regridding data from the source grid to the target grid.
+
+    Parameters
+    ----------
+    source : xarray.Dataset
+        The source dataset to be regridded.
+    target : xarray.Dataset
+        The target grid dataset.
+    method : str, optional
+        The regridding method to use. Default is "bilinear".
+
+    Returns
+    -------
+    xesmf.Regridder
+        The regridder object.
+    """
+    regridder = xe.Regridder(source, target, method=method)
     return regridder
 
 
-def compare_seasons(
-    ds1, ds2, regrid="ds1", do_height_correction=False, orog1=None, orog2=None
-):
-    """
-    Function to compare seasonal means of two datasets.
-
-    Paramters
-    ---------
-    ds1 : xarray.Dataset
-        First variable data for comparision. Temporal resolution has to be less than monthly.
-        ds1 is mainly model output data.
-        ds2 is subtracted from ds1.
-    ds2 : xarray.Dataset
-        Second variable data for comparision. Temporal resolution has to be less than monthly.
-        ds2 is mainly observational or reanalysis data.
-        ds2 is subtracted from ds1.
-    regrid : {"ds1", "ds2"}, optional
-        Denotes the dataset to be bilinearly regridded. Specify the dataset with the finer spatial resolution:
-
-        - "ds1": Regrid ds1 to ds2's grid with coarser spatial resolution.
-        - "ds2": Regrid ds2 to ds1's grid with coarser spatial resolution.
-    do_height_correction : bool, optional
-        If ``do_height_correction=True``, do a height correction on ds1 using two orography files orog1 and orog2.
-    orog1 : xarray.Dataset, optional
-        Use only if ``do_height_correction=True``.
-        Specify a orography file referring to ds1.
-    orog2 : xarray.Dataset, optional
-        Use only if ``do_height_correction=True``.
-        Specify a orography file referring to ds2.
+def regrid(ds, regridder):
+    """
+    Regrids the dataset using the specified regridder.
+
+    Parameters
+    ----------
+    ds : xarray.Dataset
+        The dataset to be regridded.
+    regridder : xesmf.Regridder
+        The regridder object.
+
+    Returns
+    -------
+    xarray.Dataset
+        The regridded dataset.
+    """
+    ds_regrid = regridder(ds)
+    for var in ds.data_vars:
+        if var not in ["mask", "sftlf"]:
+            ds_regrid[var] = ds_regrid[var].where(ds_regrid["mask"] > 0.0)
+    return ds_regrid
+
+
+def regrid_dsets(dsets, target_grid, method="bilinear"):
+    """
+    Regrids multiple datasets to the target grid.
+
+    Parameters
+    ----------
+    dsets : dict
+        A dictionary of datasets to be regridded, with keys as dataset IDs and values as xarray.Datasets.
+    target_grid : xarray.Dataset
+        The target grid dataset.
+    method : str, optional
+        The regridding method to use. Default is "bilinear".
 
     Returns
     -------
-    seasonal_comparision : xarray.Dataset
-        Spatial mean differences of two datasets
-
-    """
-    ds1 = ds1.copy()
-    ds2 = ds2.copy()
-    ds1_seasmean = seasonal_mean(ds1)
-    ds2_seasmean = seasonal_mean(ds2)
-    if regrid == "ds1":
-        regridder = get_regridder(ds1, ds2)
-        print(regridder)
-        ds1_seasmean = regridder(ds1_seasmean)
-    elif regrid == "ds2":
-        regridder = get_regridder(ds2, ds1)
-        print(regridder)
-        ds2_seasmean = regridder(ds2_seasmean)
-
-    if do_height_correction is True:
-        orog1 = regridder(orog1)
-        ds1_seasmean += height_correction(orog1, orog2)
-    return ds1_seasmean - ds2_seasmean
-    # return xr.where(ds1_seasmean.mask, ds2_seasmean - ds1_seasmean, np.nan)
+    dict
+        A dictionary of regridded datasets.
+    """
+    for dset_id, ds in dsets.items():
+        print(dset_id)
+        mapping = ds.cf["grid_mapping"].grid_mapping_name
+        if mapping == "rotated_latitude_longitude":
+            dsets[dset_id] = ds.cx.rewrite_coords(coords="all")
+        else:
+            print(f"regridding {dset_id} with grid_mapping: {mapping}")
+            regridder = create_regridder(ds, target_grid, method=method)
+            print(regridder)
+            dsets[dset_id] = regrid(ds, regridder)
+    return dsets

evaltools/source.py

@@ -3,7 +3,7 @@
 import pandas as pd
 from warnings import warn
 
-from .utils import iid_to_dict, dict_to_iid
+from .utils import iid_to_dict, dict_to_iid, mask_with_sftlf, add_bounds
 
 xarray_open_kwargs = {"use_cftime": True, "decode_coords": "all", "chunks": None}
 time_range_default = slice("1979", "2020")
@@ -119,3 +119,15 @@ def open_and_sort(catalog, merge=None, concat=False, time_range="auto"):
             join="override",
         )
     return dsets
+
+
+def open_datasets(variables, frequency="mon", mask=True, add_missing_bounds=True):
+    catalog = get_source_collection(variables, frequency, add_fx=["areacella", "sftlf"])
+    dsets = open_and_sort(catalog, merge=True)
+    if mask is True:
+        for ds in dsets.values():
+            mask_with_sftlf(ds)
+    if add_missing_bounds is True:
+        for dset_id, ds in dsets.items():
+            dsets[dset_id] = add_bounds(ds)
+    return dsets

evaltools/utils.py

@@ -1,7 +1,22 @@
 from collections import defaultdict
 
 
-def iid_to_dict(dset_id, attrs):
+default_attrs = [
+    "project_id",
+    "domain_id",
+    "institution_id",
+    "driving_source_id",
+    "driving_experiment_id",
+    "driving_variant_label",
+    "source_id",
+    "version_realization",
+    #'frequency',
+    #'variable_id',
+    "version",
+]
+
+
+def iid_to_dict(iid, attrs=None):
     """
     Convert a dataset ID and its attributes to a dictionary.
 
@@ -12,7 +27,9 @@ def iid_to_dict(dset_id, attrs):
     Returns:
     dict: The dataset ID and attributes as a dictionary.
     """
-    values = dset_id.split(".")
+    if attrs is None:
+        attrs = default_attrs
+    values = iid.split(".")
     return dict(zip(attrs, values))
 
 
@@ -31,6 +48,22 @@ def dict_to_iid(attrs, drop=None):
     return ".".join(v for k, v in attrs.items() if k not in drop)
 
 
+def short_iid(iid, attrs=None):
+    """
+    Convert a dataset ID to a short ID.
+
+    Parameters:
+    iid (str): The dataset ID.
+    attrs (dict): The dataset attributes.
+
+    Returns:
+    str: The short ID.
+    """
+    if attrs is None:
+        attrs = ["institution_id", "source_id", "driving_source_id", "experiment_id"]
+    return dict_to_iid({k: v for k, v in iid_to_dict(iid).items() if k in attrs})
+
+
 def sort_by_grid_mapping(dsets):
     """
     Sort the datasets by their grid mapping.