],
))
pf[scenario + '_' + target] = pf[keys].mean(axis=1)
print(f'[{dataset}] regridding predictions')
nlcd = load.nlcd(store=store, year=2016, classes=[41, 42, 43, 90])
final_mask = nlcd.sum('band')
final_mask.attrs['crs'] = nlcd.attrs['crs']
if 'biomass' in dataset:
final_mask.values = final_mask.values * (final_mask.values > 0.5)
else:
final_mask.values = final_mask.values > 0.5
ds['historical'] = fit.interp(pf, final_mask, var='historical')
for scenario in tqdm(scenarios):
results = []
for target in targets:
key = scenario + '_' + target
gridded = fit.interp(pf, final_mask, var=key)
results.append(gridded)
da = xr.concat(results, dim=xr.Variable('year', targets))
ds[scenario] = da
account_key = os.environ.get('BLOB_ACCOUNT_KEY')
path = utils.get_store('carbonplan-forests',
f'risks/results/web/{dataset}.zarr',
account_key=account_key)
ds.to_zarr(path, mode='w')
results_dict = {}
# select out bounding boxes
# select out bounding boxes
for impact in ['insects', 'drought', 'fire', 'tmean']:
results_dict[impact] = {}
# read in the temperature data from its different sources and create a datacube
# of the same specs as the risks
if impact == 'tmean':
ds = build_climate_cube()
# grab the risks data
else:
store_path = 'risks/results/web/{}_full.zarr'.format(impact)
ds = xr.open_zarr(
get_store(
'carbonplan-forests',
store_path,
account_key=account_key,
)
)
ds = ds.assign_coords({'year': np.arange(1980, 2100, 10)})
if impact in ['insects', 'drought']:
# restructure the insects/drought ones to align with the temp/fire
ds = repackage_drought_insects(ds)
# assign the coords for all of the data sources (this helps make sure that
# the masking works appropriately and coordinates aren't off by 0.00000001)
print(ds)
ds = ds.assign_coords(
{
"x": website_mask.x,
dataset = args[1]
if len(args) > 2:
coarsen = int(args[2])
savename = f'{dataset}_d{coarsen}'
res = 4000 * coarsen
else:
coarsen = 0
savename = dataset
res = 4000
print(f'[{dataset}] converting to geojson')
precision = 2
store = utils.get_store('carbonplan-forests',
f'risks/results/web/{dataset}.zarr')
ds = xr.open_zarr(store)
if coarsen > 0:
ds = ds.coarsen(x=coarsen, y=coarsen, boundary='trim').mean().compute()
# if dataset == 'fire':
# scenarios = ['ssp245', 'ssp370', 'ssp585']
# for scenario in scenarios:
# keys = list(
# filter(lambda a: a is not None, [k if scenario in k else None for k in ds.data_vars])
# )
# ds[scenario] = functools.reduce(lambda a, b: a + b, [ds[key] for key in keys]) / len(keys)
if 'fire' in dataset or 'biomass' in dataset:
scenarios = ['ssp245', 'ssp370', 'ssp585']
store = 'az'
df = pd.read_csv(
'https://carbonplan.blob.core.windows.net/carbonplan-scratch/from-bill-04-14-2021/Fig1D_DroughtModel_ModeledFIAlongEnsembleHistMort_FIAlong_04-14-2021.csv'
)
pf = pd.DataFrame()
pf['lat'] = df['V3']
pf['lon'] = df['V2']
pf['mortality'] = df['V6']
pf = pf.dropna().reset_index(drop=True)
ds = xr.Dataset()
nlcd = load.nlcd(store=store, year=2016, classes=[41, 42, 43, 90])
final_mask = nlcd.sum('band')
final_mask.attrs['crs'] = nlcd.attrs['crs']
final_mask.values = final_mask.values > 0.5
gridded = fit.interp(pf, final_mask, var='mortality')
ds['historical'] = gridded
account_key = os.environ.get('BLOB_ACCOUNT_KEY')
path = utils.get_store('carbonplan-forests',
'risks/results/web/drought.zarr',
account_key=account_key)
ds.to_zarr(path, mode='w')
website_mask = (load.nlcd(store="az", year=2016).sel(
band=[41, 42, 43, 90]).sum("band") > 0.5).astype("float")
gcms = [
("ACCESS-CM2", "r1i1p1f1"),
("ACCESS-ESM1-5", "r10i1p1f1"),
("MRI-ESM2-0", "r1i1p1f1"),
("MIROC-ES2L", "r1i1p1f2"),
("MPI-ESM1-2-LR", "r10i1p1f1"),
("CanESM5-CanOE", "r3i1p2f1"),
]
for impact in impacts_to_process:
ds = xr.open_zarr(
get_store(
'carbonplan-forests',
'risks/results/web/{}_cmip_high_res.zarr'.format(impact),
account_key=account_key,
))
ds = ds.assign_coords({
"x": website_mask.x,
"y": website_mask.y,
})
if impact == 'fire':
ds = ds.groupby('time.year').sum().coarsen(year=10).mean().compute()
ds = ds.assign_coords({'year': np.arange(1970, 2100, 10)})
ds = ds.rolling(year=2).mean().drop_sel(year=1970)
else:
ds = ds.assign_coords({'year': np.arange(1970, 2100, 10)})
ds = ds.rolling(year=2).mean().drop_sel(year=1970)
ds = ds.assign_coords(
yhat = model.predict(x_z)
prediction = collect.fire(yhat, climate.sel(time=analysis_time_slice))
ds['historical'] = (['time', 'y', 'x'], prediction['prediction'])
ds = ds.assign_coords({
'x': climate.x,
'y': climate.y,
'time': climate.sel(time=analysis_time_slice).time,
'lat': climate.lat,
'lon': climate.lon,
})
if store == 'local':
ds.to_zarr('data/fire_historical.zarr', mode='w')
elif store == 'az':
path = get_store(
'carbonplan-forests',
'risks/results/paper/fire_terraclimate_{}.zarr'.format(run_name),
account_key=account_key,
)
if year == 1984:
ds.to_zarr(path, consolidated=True, mode='w')
else:
ds.to_zarr(path, consolidated=True, mode='a', append_dim='time')
print('[fire] evaluating on future climate')
cmip_models = [
('CanESM5-CanOE', 'r3i1p2f1'),
('MIROC-ES2L', 'r1i1p1f2'),
('ACCESS-CM2', 'r1i1p1f1'),
('ACCESS-ESM1-5', 'r10i1p1f1'),
('MRI-ESM2-0', 'r1i1p1f1'),
('MPI-ESM1-2-LR', 'r10i1p1f1'),
]
from carbonplan_forest_risks import load
from carbonplan_forest_risks.utils import get_store
# flake8: noqa
account_key = os.environ.get('BLOB_ACCOUNT_KEY')
# this is only used to provide the x/y template for the insects/drought tifs
grid_template = (load.nlcd(store="az", year=2016).sel(
band=[41, 42, 43, 90]).sum("band") > 0.5).astype("float")
# # by passing mask as none we don't mask out any values
# # we'll pass a mask for when we do the webmap data prep
cmip_insect_url_template = "https://carbonplan.blob.core.windows.net/carbonplan-scratch/from-bill-05-03-2021/InsectProjections_Maps_5-5-21/InsectModelProjection_{}.{}.{}-{}.{}-v18climate_05-05-2021.tif"
da = load.impacts(cmip_insect_url_template, grid_template, mask=None) * 100
out_path = get_store('carbonplan-forests',
'risks/results/paper/insects_cmip_v5.zarr')
ds = xr.Dataset()
ds['probability'] = da.to_array(dim='vars').rename({'vars': 'gcm'})
ds.to_zarr(out_path, mode='w', consolidated=True)
cmip_drought_url_template = "https://carbonplan.blob.core.windows.net/carbonplan-scratch/from-bill-05-03-2021/DroughtProjections_Maps_5-5-21/DroughtModelProjection_{}.{}.{}-{}.{}-v18climate_05-05-2021.tif"
da = load.impacts(cmip_drought_url_template, grid_template, mask=None) * 100
out_path = get_store('carbonplan-forests',
'risks/results/paper/drought_cmip_v5.zarr')
ds = xr.Dataset()
ds['probability'] = da.to_array(dim='vars').rename({'vars': 'gcm'})
ds.to_zarr(out_path, mode='w', consolidated=True)
# load in historical runs to create drought_terraclimate and insects_terraclimate
terraclimate_insect_url_template = "https://carbonplan.blob.core.windows.net/carbonplan-scratch/from-bill-05-03-2021/Fig2_TerraClimateHistModels_4-22-21/InsectModel_ModeledTerraClimateFIAlong_{}-{}_04-22-2021.tif"