def plot_drydown(sim_data, flux_data, met_data, name, date_range):
"""Plot behaviour during dry-downs.
Plots rainfall, as well as Qh and Qle for obs and simulations.
:sim_data: xarray dataset from a simulation
:flux_data: xarray dataset from a simulation
:met_data: xarray dataset from a simulation
:name: model name
:returns: plot filename
"""
year_range = [parse(d).year for d in date_range]
year_range[1] += 1
year_range = ['%s-01-01' % d for d in year_range]
site = pals_site_name(met_data)
sns.set_palette(
sns.color_palette(['#aa0000', '#ff4444', '#0000aa', '#4477ff']))
# Plot rainfall in mm
Rainf = (pals_xr_to_df(met_data.sel(time=slice(*year_range)),
['Rainf']).resample('1W', how='sum') * 1000).mean()
obs = (pals_xr_to_df(flux_data.sel(time=slice(*year_range)),
['Qh', 'Qle']).resample('1D')).mean()
sim = (pals_xr_to_df(sim_data.sel(time=slice(*year_range)),
['Qh', 'Qle']).resample('1D')).mean()
x_vals = Rainf.index.to_pydatetime()
fig, ax = pl.subplots()
ax.bar(x_vals, Rainf.values, width=7, color='lightblue', linewidth=0)
x_vals = obs.index.to_pydatetime()
for c in obs:
if c == 'Qle':
offset = 0
if c == 'Qh':
offset = 100
ax.plot(x_vals,
pd.rolling_mean(obs[c], 14) + offset,
label='Obs %s + %d' % (c, offset))
ax.plot(x_vals,
pd.rolling_mean(sim[c], 14) + offset,
label='Sim %s + %d' % (c, offset))
ax.axvspan(parse(date_range[0]),
parse(date_range[1]),
color='black',
alpha=0.1,
zorder=-100)
pl.legend(loc=0)
pl.title('{n}: drydown plot at {s}'.format(n=name, s=site))
filename = '{n}_{s}_drydown_timeseries_plot.png'.format(n=name, s=site)
return filename
def plot_drydown_daily_cycles(sim_data, flux_data, met_data, name, date_range):
"""Plot daily cycle behaviour during dry-downs.
Plots rainfall, as well as Qh and Qle for obs and simulations.
:sim_data: xarray dataset from a simulation
:flux_data: xarray dataset from a simulation
:met_data: xarray dataset from a simulation
:name: model name
:returns: plot filename
"""
d_range = [np.datetime64(parse(d)) for d in date_range]
del_t = np.timedelta64(7, 'D')
first_cycle = d_range[0] + [-del_t, del_t]
last_cycle = d_range[1] + [-del_t, del_t]
site = pals_site_name(met_data)
sns.set_palette(
sns.color_palette(['#aa0000', '#ff4444', '#0000aa', '#4477ff']))
fig, _ = pl.subplots(2, 1, sharey=True)
periods = {0: 'start', 1: 'end'}
flux_vars = list(
set(['Qh', 'Qle']).intersection(list(sim_data.data_vars)).intersection(
list(sim_data.data_vars)))
for i, dr in enumerate([first_cycle, last_cycle]):
obs_df = pals_xr_to_df(flux_data.sel(time=slice(*dr)), flux_vars)
obs = obs_df.groupby(obs_df.index.time).mean()
sim_df = pals_xr_to_df(sim_data.sel(time=slice(*dr)), flux_vars)
sim = sim_df.groupby(sim_df.index.time).mean()
x_vals = obs.index.values
ax = pl.subplot(2, 1, i + 1)
for c in obs:
if c == 'Qle':
offset = 0
if c == 'Qh':
offset = 100
ax.plot(x_vals, obs[c] + offset, label='Obs %s + %d' % (c, offset))
ax.plot(x_vals, sim[c] + offset, label='Sim %s + %d' % (c, offset))
pl.title('{n}: daily cycles for {p} of drydown at {s}'.format(
n=name, s=site, p=periods[i]))
pl.legend(loc=0)
fig.tight_layout()
filename = '{n}_{s}_drydown_daily_cycles_plot.png'.format(n=name, s=site)
return filename
def get_multimodel_data(site, names, variables):
"""Returns a DF with variable columns, time/model indices."""
data = []
for n in names:
path = "model_data/{n}/{n}_{s}.nc".format(n=n, s=site)
try:
with xr.open_dataset(path) as ds:
df = pals_xr_to_df(ds, variables)
df['name'] = n
# offset time indices
if n == 'COLASSiB.2.0':
df.index = df.index + pd.Timedelta('-30min')
if n == 'ORCHIDEE.trunk_r1401':
df.index = df.index + pd.Timedelta('-15min')
df.set_index('name', append=True, inplace=True)
data.append(df)
except (OSError, RuntimeError) as e:
raise Exception(path, e)
df = pd.concat(data)
df.columns.name = 'variable'
return df
def diagnostic_plots(sim_data, flux_data, name, site=None):
"""Plot standard diagnostic plots for a single site
:sim_data: xarray dataset
:flux_data: xarray dataset
:name: mode name
:returns: list of paths to plots
"""
if site is None:
site = pals_site_name(flux_data)
logger.info('Running standard plots for %s at %s' % (name, site))
base_path = 'source/models/{n}'.format(n=name)
rel_path = 'figures/{s}'.format(s=site)
fig_path = os.path.join(base_path, rel_path)
os.makedirs(fig_path, exist_ok=True)
files = []
# benchmark_names = ['1lin', '2lin', '3km27']
benchmark_names = ['S_lin', 'ST_lin', 'STH_km27']
try:
benchmarks = [get_benchmark(bname, site) for bname in benchmark_names]
except RuntimeError as e:
logger.warning("Benchmark(s) not available at {s}, skipping. {e}".format(s=site, e=e))
return
sns.set_palette(sns.color_palette(['red', 'pink', 'orange', 'black', 'blue']))
# Generalise if more multi-variable plots needed
metric_df = get_PLUMBER_metrics(name, site)
for plot in [p_plumber_metrics, p_metric_rank_counts]:
filename = plot(metric_df, name, site)
rel_plot_path = save_plot(base_path, rel_path, filename)
files.append(rel_plot_path)
flux_vars = ['Qle', 'Qh', 'NEE']
for var in flux_vars:
try:
data = pd.concat([pals_xr_to_df(ds, [var]) for ds in
benchmarks + [flux_data, sim_data]], axis=1)
except Exception as e:
logger.warning('Data missing for {v} at {s}, skipping. {e}'.format(v=var, s=site, e=e))
continue
data.columns = benchmark_names + ['observed', 'modelled']
for plot in DIAGNOSTIC_PLOTS.values():
filename = plot(data, name, var, site)
rel_plot_path = save_plot(base_path, rel_path, filename)
files.append(rel_plot_path)
return files
def get_test_data(site, met_vars, use_names, qc=False, fix_closure=True):
# We use gap-filled data for the testing period, or the model fails.
met_test_xr = get_met_data(site)[site]
met_test = pals_xr_to_df(met_test_xr,
variables=met_vars,
qc=qc,
name=use_names)
return dict(site=site,
met_vars=met_vars,
fix_closure=fix_closure,
met_test=met_test,
met_test_xr=met_test_xr)