def build_climate_cube(
tlim=(1970, 2099), variables=['tmean'], downscaling_method='quantile-mapping-v3'
):
""""""
gcms = [
('CanESM5-CanOE', 'r3i1p2f1'),
('MIROC-ES2L', 'r1i1p1f2'),
('ACCESS-CM2', 'r1i1p1f1'),
('ACCESS-ESM1-5', 'r10i1p1f1'),
('MRI-ESM2-0', 'r1i1p1f1'),
('MPI-ESM1-2-LR', 'r10i1p1f1'),
]
scenarios = ['ssp245', 'ssp370', 'ssp585']
all_scenarios = []
for scenario in scenarios:
all_gcms = []
for (gcm, ensemble_member) in gcms:
cmip = load.cmip(
store='az',
model=gcm,
tlim=tlim,
scenario=scenario,
historical=True,
member=ensemble_member,
method=downscaling_method,
sampling='annual',
variables=variables,
)
all_gcms.append(cmip)
concatted = xr.concat(all_gcms, dim='gcm')
concatted = concatted.assign_coords({'gcm': [gcm[0] for gcm in gcms]})
all_scenarios.append(concatted)
ds = xr.concat(all_scenarios, dim='scenario')
ds = ds.assign_coords({'scenario': scenarios})
return ds
def cmip_fia_long(cmip_table, method):
# generate long data w/ cmip
df_fia = load.fia(store='az', states=states, clean=False)
keep_vars = (['lat', 'lon', 'plot_cn'] +
[var + '_min' for var in variables] +
[var + '_mean'
for var in variables] + [var + '_max' for var in variables])
for i, row in cmip_table.iterrows():
if 'hist' in row.scenario:
targets = targets_historical
else:
targets = targets_future
for target in targets:
historical = target == '2015'
tlim = (str(int(target) - 5), str(int(target) + 4))
print(tlim, row)
df = load.cmip(
store='az',
tlim=(int(tlim[0]), int(tlim[1])),
variables=variables,
df=df_fia,
model=row.model,
scenario=row.scenario,
member=row.member,
historical=historical,
method=method,
sampling='annual',
)
df = df[keep_vars]
write_df(
df,
f'{method}/FIA-CMIP6-Long-{row.model}.{row.scenario}.{row.member}-{tlim[0]}.{tlim[1]}',
)
targets = list(map(lambda x: str(x), np.arange(2020, 2120, 20)))
cmip_models = [
'BCC-CSM2-MR', 'ACCESS-ESM1-5', 'CanESM5', 'MIROC6', 'MPI-ESM1-2-LR'
]
scenarios = ['ssp245', 'ssp370', 'ssp585']
for it in tqdm(range(len(targets))):
target = targets[it]
tlim = (str(int(target) - 5), str(int(target) + 4))
for cmip_model in cmip_models:
for scenario in scenarios:
key = cmip_model + '_' + scenario + '_' + target
df = load.cmip(
store=store,
tlim=(int(tlim[0]), int(tlim[1])),
data_vars=data_vars,
data_aggs=data_aggs,
model=cmip_model,
scenario=scenario,
annual=True,
df=df,
)
pf[key] = np.NaN
x, meta = prepare.insects(df, eval_only=True, duration=10)
x_z = utils.zscore_2d(x, x_mean, x_std)
for code in codes:
if code in models.keys():
model = models[code]
inds = df['type_code'] == code
pf.loc[inds, key] = model.predict(x=x_z[inds])
pf['r2'] = pf['type_code'].map(lambda k: models[k].train_r2
if k in models.keys() else np.NaN)
('ACCESS-CM2', 'r1i1p1f1'),
('ACCESS-ESM1-5', 'r10i1p1f1'),
('MRI-ESM2-0', 'r1i1p1f1'),
('MPI-ESM1-2-LR', 'r10i1p1f1'),
]
scenarios = ['ssp245', 'ssp370', 'ssp585']
for (cmip_model, member) in cmip_models:
for scenario in tqdm(scenarios):
results = []
climate = load.cmip(
store=store,
model=cmip_model,
coarsen=coarsen_predict,
method='quantile-mapping-v3',
scenario=scenario,
tlim=('1969', '2099'),
variables=data_vars,
sampling='monthly',
member=member,
historical=True,
mask=mask,
)
try:
for year in np.arange(1970, 2100, 10):
ds_future = xr.Dataset()
print('[fire] conducting prediction for {} {} {}'.format(
cmip_model, scenario, year))
prepend_time_slice = slice(str(year - 1), str(year - 1))
analysis_time_slice = slice(str(year), str(year + 9))
'ACCESS-ESM1-5',
'MRI-ESM2-0',
'MPI-ESM1-2-LR',
]
scenarios = ['ssp245', 'ssp370', 'ssp585']
for it in tqdm(range(len(targets))):
target = targets[it]
tlim = (str(int(target) - 10), str(int(target) + 9))
for cmip_model in cmip_models:
for scenario in scenarios:
key = cmip_model + '_' + scenario + '_' + target
df = load.cmip(
store=store,
tlim=(int(tlim[0]), int(tlim[1])),
variables=variables,
historical=True if int(tlim[0]) < 2015 else False,
model=cmip_model,
scenario=scenario,
sampling='annual',
df=df,
)
pf[key] = np.NaN
for code in type_codes:
if code in models.keys():
model = models[code]
inds = df['type_code'] == code
x = df[inds]['age']
year = df[inds]['year']
f = [df[inds]['tmean_mean'], df[inds]['ppt_mean']]
if it == 0:
pf.loc[inds, key] = model.predict(
np.maximum(x + (float(target) - year), 0), f