def interpolate_esmf(obs_file, mod_file, depth, cmor_var):
"""The 2d interpolation with ESMF.
Parameters
----------
obs_file: str
path to file with observations/climatology,
will be used to extract the grid to interpolate on to.
mod_file: str
path to the file with model data.
depth: int
depth to interpolate to. First the closest depth from the
observations will be selected and then.
"""
metadata_obs = load_meta(obs_file, fxpath=None)
metadata_mod = load_meta(mod_file, fxpath=None)
data_obs = metadata_obs['datafile'].variables[cmor_var][:]
data_model = metadata_mod['datafile'].variables[cmor_var][:]
# Select depth in climatology that is closest to the desired depth
target_depth, level_depth = closest_depth(metadata_obs['lev'], depth)
# climatology and model data on the level
data_onlev_obs = data_obs[0, level_depth, :, :]
data_onlev_mod = interpolate_vert(metadata_mod['lev'], target_depth,
data_model[0, :, :, :])
# prepear interpolation fields
distfield = define_esmf_field(obs_file, data_onlev_obs, 'OBS')
distfield.data[:] = 0.0
sourcefield = define_esmf_field(mod_file, data_onlev_mod, 'Model')
sourcefield.data[...] = data_onlev_mod.T
lonc, latc, data_onlev_obs_cyc, data_interpolated_cyc = esmf_regriding(
sourcefield, distfield, metadata_obs, data_onlev_obs)
return lonc, latc, target_depth, data_onlev_obs_cyc, data_interpolated_cyc
def plot2d_bias(cfg, plot_params):
"""Plot 2d maps of the bias relative to climatology.
Parameters
----------
model_filenames:OrderedDict
OrderedDict with model names as keys and input files as values.
cmor_var: str
name of the variable
depth: int
we will plot the data on the model level
that is closest to the `depth`.
diagworkdir: str
path to the working directory
diagplotdir: str
path to the plot directory
levels: tuple
values to be used for vmin and vmax in the form of (vmin, vmax)
dpi: int
the dpi values to save the figure
observations: str
name of the observations
projection: instance of cartopy projection (ccrs)
bbox: list
bounding box. It will be the input for cartopy `set_extent`.
ncols: int
number of columns.
Retuns
------
None
"""
# setupa a base figure
figure, axis = create_plot(plot_params['model_filenames'],
ncols=plot_params['ncols'],
projection=plot_params['projection'])
# get the filename of observations
ifilename_obs = genfilename(cfg['work_dir'],
plot_params['variable'],
plot_params['observations'],
data_type='timmean',
extension='.nc')
# get the metadata for observations (we just need a size)
metadata = load_meta(
datapath=plot_params['model_filenames'][plot_params['observations']],
fxpath=None)
lon2d = metadata['lon2d']
# Create an empty array to store the mean.
# One point larger along long to acount for cyclic point
model_mean = np.zeros((lon2d.shape[0], lon2d.shape[1] + 1))
print("MODEL MEAN SHAPE {}".format(model_mean.shape))
# delete observations from the model list
model_filenames = plot_params['model_filenames'].copy()
del model_filenames[plot_params['observations']]
# loop over models
index = None
for index, mmodel in enumerate(model_filenames):
logger.info("Plot plot2d_bias %s for %s", plot_params['variable'],
mmodel)
# get the filename with the mean generated by the `timemean`
ifilename = genfilename(cfg['work_dir'],
plot_params['variable'],
mmodel,
data_type='timmean',
extension='.nc')
# do the interpolation to the observation grid
# the output is
lonc, latc, target_depth, data_obs, interpolated = interpolate_esmf(
ifilename_obs, ifilename, plot_params['depth'],
plot_params['variable'])
# get the label and convert data if needed
cb_label, data_obs = label_and_conversion(plot_params['variable'],
data_obs)
cb_label, interpolated = label_and_conversion(plot_params['variable'],
interpolated)
# add to the mean model
model_mean = model_mean + interpolated
# set the map extent
left, right, down, upper = plot_params['bbox']
axis[index].set_extent([left, right, down, upper],
crs=ccrs.PlateCarree())
# Only pcolormesh is working for now with cartopy,
# contourf is failing to plot curvilinear meshes,
# let along the unstructures ones.
image = axis[index].contourf(
lonc,
latc,
interpolated - data_obs,
levels=plot_params['levels'],
extend='both',
# vmin=contours[0],
# vmax=contours[-1],
transform=ccrs.PlateCarree(),
cmap=plot_params['cmap'],
)
# fill continents
axis[index].add_feature(
cfeature.GSHHSFeature(levels=[1],
scale="low",
facecolor="lightgray"))
axis[index].set_title("{}, {} m".format(mmodel, int(target_depth)),
size=18)
axis[index].set_rasterization_zorder(-1)
# calculate the model mean and plot it
if index:
index = index
else:
index = 0
model_mean = model_mean / len(model_filenames)
axis[index + 1].set_extent([left, right, down, upper],
crs=ccrs.PlateCarree())
image = axis[index + 1].contourf(
lonc,
latc,
model_mean - data_obs,
levels=plot_params['levels'],
extend='both',
# vmin=contours[0],
# vmax=contours[-1],
transform=ccrs.PlateCarree(),
cmap=cmo.balance,
)
axis[index + 1].add_feature(
cfeature.GSHHSFeature(levels=[1], scale="low", facecolor="lightgray"))
axis[index + 1].set_title("Model mean bias, {} m".format(
int(target_depth)),
size=18)
axis[index + 1].set_rasterization_zorder(-1)
# delete the axis that are not needed
for delind in range(index + 2, len(axis)):
figure.delaxes(axis[delind])
# set common colorbar
colorbar = figure.colorbar(image,
orientation='horizontal',
ax=axis,
pad=0.01,
shrink=0.9)
colorbar.set_label(cb_label, rotation='horizontal', size=18)
colorbar.ax.tick_params(labelsize=18)
# save the picture
pltoutname = genfilename(cfg['plot_dir'],
plot_params['variable'],
"MULTIMODEL",
data_type='plot2d_bias_{}_level'.format(
str(int(target_depth))))
plot_params['basedir'] = cfg['plot_dir']
plot_params['ori_file'] = [ifilename]
plot_params['areacello'] = None
plot_params['mmodel'] = None
plot_params['region'] = "Global"
plt.savefig(pltoutname, dpi=plot_params['dpi'])
provenance_record = get_provenance_record(plot_params, 'plot2d_bias',
'png')
with ProvenanceLogger(cfg) as provenance_logger:
provenance_logger.log(pltoutname + '.png', provenance_record)
def plot2d_original_grid(cfg, plot_params):
"""Plot 2d maps on original grid using cartopy.
Parameters
----------
model_filenames:OrderedDict
OrderedDict with model names as keys and input files as values.
cmor_var: str
name of the variable
depth: int
we will plot the data on the model
level that is closest to the `depth`.
Ignored if explicit_depths is provided.
levels: tuple
values to be used for vmin and vmax in the form of (vmin, vmax)
diagworkdir: str
path to the working directory
diagplotdir: str
path to the plot directory
cmap: matplotlib colormap
colormap
dpi: int
the dpi values to save the figure
explicit_depths: dict
Output of the `aw_core` function.
It's a dictionary where for each model there is a maximum temperature,
depth level in the model, index of the depth level in the model.
If provided the `depth` parameter is excluded.
projection: instance of cartopy projection (ccrs)
bbox: list
bounding box. It will be the input for cartopy `set_extent`.
ncols: int
number of columns.
Retuns
------
None
"""
figure, axis = create_plot(plot_params['model_filenames'],
ncols=plot_params['ncols'],
projection=plot_params['projection'])
index = None
for index, mmodel in enumerate(plot_params['model_filenames']):
logger.info("Plot plot2d_original_grid %s for %s",
plot_params['variable'], mmodel)
ifilename = genfilename(cfg['work_dir'],
plot_params['variable'],
mmodel,
data_type='timmean',
extension='.nc')
metadata = load_meta(ifilename, fxpath=None)
datafile = metadata['datafile']
lon2d = metadata['lon2d']
lat2d = metadata['lat2d']
lev = metadata['lev']
if not plot_params['explicit_depths']:
depth_target, level_target = closest_depth(lev,
plot_params['depth'])
else:
level_target = plot_params['explicit_depths'][mmodel][
'maxvalue_index']
depth_target = lev[level_target]
if datafile.variables[plot_params['variable']].ndim < 4:
data = datafile.variables[plot_params['variable']][
level_target, :, :]
else:
data = datafile.variables[plot_params['variable']][
0, level_target, :, :]
cb_label, data = label_and_conversion(plot_params['variable'], data)
left, right, down, upper = plot_params['bbox']
axis[index].set_extent([left, right, down, upper],
crs=ccrs.PlateCarree())
# Only pcolormesh is working for now with cartopy,
# contourf is failing to plot curvilinear meshes,
# let along the unstructures ones.
image = axis[index].pcolormesh(
lon2d,
lat2d,
data,
vmin=plot_params['levels'][0],
vmax=plot_params['levels'][-1],
transform=ccrs.PlateCarree(),
cmap=plot_params['cmap'],
)
axis[index].add_feature(
cfeature.GSHHSFeature(levels=[1],
scale="low",
facecolor="lightgray"))
axis[index].set_title("{}, {} m".format(mmodel,
np.round(depth_target, 1)),
size=18)
axis[index].set_rasterization_zorder(-1)
# delete unused axis
for delind in range(index + 1, len(axis)):
figure.delaxes(axis[delind])
# set common colorbar
colorbar = figure.colorbar(image,
orientation='horizontal',
ax=axis,
pad=0.01,
shrink=0.9)
colorbar.set_label(cb_label, rotation='horizontal', size=18)
colorbar.ax.tick_params(labelsize=18)
if not plot_params['explicit_depths']:
plot_type = 'plot2d_{}_depth'.format(str(plot_params['depth']))
else:
plot_type = "plot2d_different_levels"
# save the figure
pltoutname = genfilename(cfg['plot_dir'],
plot_params['variable'],
"MULTIMODEL",
data_type=plot_type)
plot_params['basedir'] = cfg['plot_dir']
plot_params['ori_file'] = [ifilename]
plot_params['areacello'] = None
plot_params['mmodel'] = None
plot_params['region'] = "Global"
plt.savefig(pltoutname, dpi=plot_params['dpi'])
provenance_record = get_provenance_record(plot_params, 'plot2d', 'png')
with ProvenanceLogger(cfg) as provenance_logger:
provenance_logger.log(pltoutname + '.png', provenance_record)