def count_mappings_for_regions(reference_da: xr.DataArray,
comparison_da: xr.DataArray,
regions: List[Region]) -> pd.DataFrame:
all_df = []
variable = reference_da.name
for region in regions:
# add error catching in case need to invert latlon
try:
region_reference_da = select_bounding_box(
reference_da.to_dataset(), region)[variable]
except AssertionError:
region_reference_da = select_bounding_box(
reference_da.to_dataset(), region, inverse_lat=True)[variable]
try:
region_comparison_da = select_bounding_box(
comparison_da.to_dataset(), region)[variable]
except AssertionError:
region_comparison_da = select_bounding_box(
comparison_da.to_dataset(), region, inverse_lat=True)[variable]
# count the pixels in each group
d = convert_counts_dict_to_dataframe(
count_matching_pixels(region_reference_da, region_comparison_da))
d["region"] = [region.name for _ in range(len(d))]
all_df.append(d)
all_df = pd.concat(all_df)
return all_df
def test_select_bounding_box_inversed(self):
"""Test that inversion works correctly
"""
size = (100, 100)
ds, (lonmin, lonmax), (latmin, latmax) = _make_dataset(size)
global_region = Region(name='global',
lonmin=lonmax,
lonmax=lonmin,
latmin=latmin,
latmax=latmax)
subset = select_bounding_box(ds, global_region, inverse_lon=True)
# add the time dimension
assert subset.VHI.values.shape[1:] == size, \
f'Expected output subset to have size {size}, got {subset.VHI.values.shape[1:]}'
def test_selection(self):
size = (100, 100)
ds, (lonmin, lonmax), (latmin, latmax) = _make_dataset(size)
mid = lonmin + ((lonmax - lonmin) / 2)
half_region = Region(name='half',
lonmin=lonmin,
lonmax=mid,
latmin=latmin,
latmax=latmax)
subset = select_bounding_box(ds, half_region)
assert subset.VHI.values.shape[1:] == (100, 50), \
f'Expected output subset to have size (50, 100), got {subset.VHI.values.shape}'
assert max(subset.lon.values) < 0, \
f'Got a longitude greater than 0, {max(subset.lon.values)}'
name='victoria',
lonmin=cluster_ds.isel(lon=0).lon.values,
lonmax=cluster_ds.isel(lon=7).lon.values,
latmin=cluster_ds.isel(lat=-28).lat.values,
latmax=cluster_ds.isel(lat=-18).lat.values,
)
turkana = Region(
name='turkana',
lonmin=cluster_ds.isel(lon=5).lon.values,
lonmax=cluster_ds.isel(lon=16).lon.values,
latmin=cluster_ds.isel(lat=-16).lat.values,
latmax=cluster_ds.isel(lat=-6).lat.values,
)
southern_highlands = Region(
name='southern_highlands',
lonmin=cluster_ds.isel(lon=3).lon.values,
lonmax=cluster_ds.isel(lon=13).lon.values,
latmin=cluster_ds.isel(lat=-41).lat.values,
latmax=cluster_ds.isel(lat=-31).lat.values,
)
coastal = Region(
name='coastal',
lonmin=cluster_ds.isel(lon=15).lon.values,
lonmax=cluster_ds.isel(lon=20).lon.values,
latmin=cluster_ds.isel(lat=-44).lat.values,
latmax=cluster_ds.isel(lat=-34).lat.values,
)
fig, ax = plt.subplots()
select_bounding_box(cluster_ds, southern_highlands).cluster_5.isel(time=0).plot()
"""!ls -U glofas_africa| head -4"""
"""NOTE: https://github.com/esowc/ml_drought for the `src` code"""
from pathlib import Path
import xarray as xr
import pprint
from src.preprocess.utils import select_bounding_box
from src.utils import region_lookup
region = region_lookup['africa']
base_data_dir = Path("/lustre/soge1/projects/crop_yield/hackathon")
data_dir = base_data_dir / "glofas_africa"
out_file = base_data_dir / "glofas/glofas_africa.nc"
in_files = [f for f in data_dir.glob('*.nc')]
# ds = xr.open_mfdataset(
# (data_dir / '*.nc').as_posix(), chunks={'time': 1}
# )
# ds.rename({'latitude': 'lat', 'longitude': 'lon'})
select_bounding_box()