def __init__(self, file, grid=None, time=None, monthbegin=False):
"""Open the file and try to understand it.
Parameters
----------
file: path to the netcdf file
grid: a Grid object. This will override the normal behavior of
GeoNetcdf, which is to try to understand the grid automatically.
time: a time array. This will override the normal behavior of
GeoNetcdf, which is to try to understand the time automatically.
monthbegin: set to true if you are sure that your data is monthly
and that the data provider decided to tag the date as the center of
the month (stupid)
"""
self._nc = netCDF4.Dataset(file)
self.variables = self._nc.variables
if grid is None:
grid = sio.grid_from_dataset(self._nc)
if time is None:
time = sio.netcdf_time(self._nc, monthbegin=monthbegin)
dn = self._nc.dimensions.keys()
self.x_dim = utils.str_in_list(dn, utils.valid_names['x_dim'])[0]
self.y_dim = utils.str_in_list(dn, utils.valid_names['y_dim'])[0]
dim = utils.str_in_list(dn, utils.valid_names['t_dim'])
self.t_dim = dim[0] if dim else None
dim = utils.str_in_list(dn, utils.valid_names['z_dim'])
self.z_dim = dim[0] if dim else None
GeoDataset.__init__(self, grid, time=time)
def _salem_grid_from_dataset(ds):
"""Seek for coordinates that Salem might have created.
Current convention: x_coord, y_coord, pyproj_srs as attribute
"""
# Projection
try:
proj = ds.pyproj_srs
except AttributeError:
proj = None
proj = gis.check_crs(proj)
if proj is None:
return None
# Do we have some standard names as variable?
vns = ds.variables.keys()
xc = utils.str_in_list(vns, utils.valid_names['x_dim'])
yc = utils.str_in_list(vns, utils.valid_names['y_dim'])
# Sometimes there are more than one coordinates, one of which might have
# more dims (e.g. lons in WRF files): take the first one with ndim = 1:
x = None
for xp in xc:
if len(ds.variables[xp].shape) == 1:
x = xp
y = None
for yp in yc:
if len(ds.variables[yp].shape) == 1:
y = yp
if (x is None) or (y is None):
return None
# OK, get it
x = ds.variables[x][:]
y = ds.variables[y][:]
# Make the grid
dx = x[1] - x[0]
dy = y[1] - y[0]
args = dict(nxny=(x.shape[0], y.shape[0]),
proj=proj,
dxdy=(dx, dy),
x0y0=(x[0], y[0]))
return gis.Grid(**args)
def _salem_grid_from_dataset(ds):
"""Seek for coordinates that Salem might have created.
Current convention: x_coord, y_coord, pyproj_srs as attribute
"""
# Projection
try:
proj = ds.pyproj_srs
except AttributeError:
proj = None
proj = gis.check_crs(proj)
if proj is None:
return None
# Do we have some standard names as variable?
vns = ds.variables.keys()
xc = utils.str_in_list(vns, utils.valid_names['x_dim'])
yc = utils.str_in_list(vns, utils.valid_names['y_dim'])
# Sometimes there are more than one coordinates, one of which might have
# more dims (e.g. lons in WRF files): take the first one with ndim = 1:
x = None
for xp in xc:
if len(ds.variables[xp].shape) == 1:
x = xp
y = None
for yp in yc:
if len(ds.variables[yp].shape) == 1:
y = yp
if (x is None) or (y is None):
return None
# OK, get it
x = ds.variables[x][:]
y = ds.variables[y][:]
# Make the grid
dx = x[1]-x[0]
dy = y[1]-y[0]
args = dict(nxny=(x.shape[0], y.shape[0]), proj=proj, dxdy=(dx, dy))
args['corner'] = (x[0], y[0])
return gis.Grid(**args)
def _lonlat_grid_from_dataset(ds):
"""Seek for longitude and latitude coordinates."""
# Do we have some standard names as variable?
vns = ds.variables.keys()
xc = utils.str_in_list(vns, utils.valid_names['x_dim'])
yc = utils.str_in_list(vns, utils.valid_names['y_dim'])
# Sometimes there are more than one coordinates, one of which might have
# more dims (e.g. lons in WRF files): take the first one with ndim = 1:
x = None
for xp in xc:
if len(ds.variables[xp].shape) == 1:
x = xp
y = None
for yp in yc:
if len(ds.variables[yp].shape) == 1:
y = yp
if (x is None) or (y is None):
return None
# OK, get it
lon = ds.variables[x][:]
lat = ds.variables[y][:]
# double check for dubious variables
if not utils.str_in_list([x], utils.valid_names['lon_var']) or \
not utils.str_in_list([y], utils.valid_names['lat_var']):
# name not usual. see if at least the range follows some conv
if (np.max(np.abs(lon)) > 360.1) or (np.max(np.abs(lat)) > 90.1):
return None
# Make the grid
dx = lon[1] - lon[0]
dy = lat[1] - lat[0]
args = dict(nxny=(lon.shape[0], lat.shape[0]),
proj=wgs84,
dxdy=(dx, dy),
x0y0=(lon[0], lat[0]))
return gis.Grid(**args)
def __init__(self, xarray_obj):
self._obj = xarray_obj
if isinstance(xarray_obj, xr.DataArray):
xarray_obj = xarray_obj.to_dataset(name='var')
try: # maybe there was already some georef
xarray_obj.attrs['pyproj_srs'] = xarray_obj['var'].pyproj_srs
except:
pass
self.grid = grid_from_dataset(xarray_obj)
if self.grid is None:
raise RuntimeError('dataset Grid not understood.')
dn = xarray_obj.dims.keys()
self.x_dim = utils.str_in_list(dn, utils.valid_names['x_dim'])[0]
self.y_dim = utils.str_in_list(dn, utils.valid_names['y_dim'])[0]
dim = utils.str_in_list(dn, utils.valid_names['t_dim'])
self.t_dim = dim[0] if dim else None
dim = utils.str_in_list(dn, utils.valid_names['z_dim'])
self.z_dim = dim[0] if dim else None
def __init__(self, xarray_obj):
self._obj = xarray_obj
if isinstance(xarray_obj, xr.DataArray):
xarray_obj = xarray_obj.to_dataset(name='var')
try: # maybe there was already some georef
xarray_obj.attrs['pyproj_srs'] = xarray_obj['var'].pyproj_srs
except:
pass
self.grid = grid_from_dataset(xarray_obj)
if self.grid is None:
raise RuntimeError('dataset Grid not understood.')
dn = xarray_obj.dims.keys()
self.x_dim = utils.str_in_list(dn, utils.valid_names['x_dim'])[0]
self.y_dim = utils.str_in_list(dn, utils.valid_names['y_dim'])[0]
dim = utils.str_in_list(dn, utils.valid_names['t_dim'])
self.t_dim = dim[0] if dim else None
dim = utils.str_in_list(dn, utils.valid_names['z_dim'])
self.z_dim = dim[0] if dim else None
def _lonlat_grid_from_dataset(ds):
"""Seek for longitude and latitude coordinates."""
# Do we have some standard names as variable?
vns = ds.variables.keys()
xc = utils.str_in_list(vns, utils.valid_names['x_dim'])
yc = utils.str_in_list(vns, utils.valid_names['y_dim'])
# Sometimes there are more than one coordinates, one of which might have
# more dims (e.g. lons in WRF files): take the first one with ndim = 1:
x = None
for xp in xc:
if len(ds.variables[xp].shape) == 1:
x = xp
y = None
for yp in yc:
if len(ds.variables[yp].shape) == 1:
y = yp
if (x is None) or (y is None):
return None
# OK, get it
lon = ds.variables[x][:]
lat = ds.variables[y][:]
# double check for dubious variables
if not utils.str_in_list([x], utils.valid_names['lon_var']) or \
not utils.str_in_list([y], utils.valid_names['lat_var']):
# name not usual. see if at least the range follows some conv
if (np.max(np.abs(lon)) > 360.1) or (np.max(np.abs(lat)) > 90.1):
return None
# Make the grid
dx = lon[1]-lon[0]
dy = lat[1]-lat[0]
args = dict(nxny=(lon.shape[0], lat.shape[0]), proj=wgs84, dxdy=(dx, dy))
args['corner'] = (lon[0], lat[0])
return gis.Grid(**args)
def netcdf_time(ncobj, monthbegin=False):
"""Check if the netcdf file contains a time that Salem understands."""
import pandas as pd
time = None
try:
vt = utils.str_in_list(ncobj.variables.keys(),
utils.valid_names['time_var'])[0]
except IndexError:
# no time variable
return None
if hasattr(ncobj, 'TITLE') and 'GEOGRID' in ncobj.TITLE:
# geogrid file
pass
elif ncobj[vt].dtype in ['|S1', '|S19']:
# WRF file
time = []
try:
stimes = ncobj.variables['Times'][:].values
except AttributeError:
stimes = ncobj.variables['Times'][:]
for t in stimes:
time.append(
pd.to_datetime(t.tobytes().decode(),
errors='raise',
format='%Y-%m-%d_%H:%M:%S'))
elif vt is not None:
# CF time
var = ncobj.variables[vt]
try:
# We want python times because pandas doesn't understand
# CFtime
time = cftime.num2date(var[:],
var.units,
only_use_cftime_datetimes=False,
only_use_python_datetimes=True)
except TypeError:
# Old versions of cftime did return python times when possible
time = cftime.num2date(var[:], var.units)
if monthbegin:
# sometimes monthly data is centered in the month (stupid)
time = [datetime(t.year, t.month, 1) for t in time]
return time
def netcdf_time(ncobj, monthbegin=False):
"""Check if the netcdf file contains a time that Salem understands."""
import pandas as pd
time = None
try:
vt = utils.str_in_list(ncobj.variables.keys(),
utils.valid_names['time_var'])[0]
except IndexError:
# no time variable
return None
if hasattr(ncobj, 'TITLE') and 'GEOGRID' in ncobj.TITLE:
# geogrid file
pass
elif ncobj[vt].dtype in ['|S1', '|S19']:
# WRF file
time = []
try:
stimes = ncobj.variables['Times'][:].values
except AttributeError:
stimes = ncobj.variables['Times'][:]
for t in stimes:
time.append(pd.to_datetime(t.tostring().decode(),
errors='raise',
format='%Y-%m-%d_%H:%M:%S'))
elif vt is not None:
# CF time
var = ncobj.variables[vt]
time = netCDF4.num2date(var[:], var.units)
if monthbegin:
# sometimes monthly data is centered in the month (stupid)
time = [datetime(t.year, t.month, 1) for t in time]
return time
def _apply_transform(self, transform, grid, other, return_lut=False):
"""Common transform mixin"""
was_dataarray = False
if not isinstance(other, xr.Dataset):
try:
other = other.to_dataset(name=other.name)
was_dataarray = True
except AttributeError:
# must be a ndarray
if return_lut:
rdata, lut = transform(other, grid=grid, return_lut=True)
else:
rdata = transform(other, grid=grid)
# let's guess
sh = rdata.shape
nd = len(sh)
if nd == 2:
dims = (self.y_dim, self.x_dim)
elif nd == 3:
newdim = 'new_dim'
if self.t_dim and sh[0] == self._obj.dims[self.t_dim]:
newdim = self.t_dim
if self.z_dim and sh[0] == self._obj.dims[self.z_dim]:
newdim = self.z_dim
dims = (newdim, self.y_dim, self.x_dim)
else:
raise NotImplementedError('more than 3 dims not ok yet.')
coords = {}
for d in dims:
if d in self._obj:
coords[d] = self._obj[d]
out = xr.DataArray(rdata, coords=coords, dims=dims)
out.attrs['pyproj_srs'] = self.grid.proj.srs
if return_lut:
return out, lut
else:
return out
# go
out = xr.Dataset()
for v in other.data_vars:
var = other[v]
if return_lut:
rdata, lut = transform(var, return_lut=True)
else:
rdata = transform(var)
# remove old coords
dims = [d for d in var.dims]
coords = {}
for c in var.coords:
n = utils.str_in_list([c], utils.valid_names['x_dim'])
if n:
dims = [self.x_dim if x in n else x for x in dims]
continue
n = utils.str_in_list([c], utils.valid_names['y_dim'])
if n:
dims = [self.y_dim if x in n else x for x in dims]
continue
coords[c] = var.coords[c]
# add new ones
coords[self.x_dim] = self._obj[self.x_dim]
coords[self.y_dim] = self._obj[self.y_dim]
rdata = xr.DataArray(rdata,
coords=coords,
attrs=var.attrs,
dims=dims)
rdata.attrs['pyproj_srs'] = self.grid.proj.srs
out[v] = rdata
if was_dataarray:
out = out[v]
else:
out.attrs['pyproj_srs'] = self.grid.proj.srs
if return_lut:
return out, lut
else:
return out
def transform(self, other, grid=None, interp='nearest', ks=3):
"""Reprojects an other Dataset or DataArray onto this grid.
Parameters
----------
other: Dataset, DataArray or ndarray
the data to project onto self
grid: salem.Grid
in case the input dataset does not carry georef info
interp : str
'nearest' (default), 'linear', or 'spline'
ks : int
Degree of the bivariate spline. Default is 3.
Returns
-------
a dataset or a dataarray
"""
was_dataarray = False
if not isinstance(other, xr.Dataset):
try:
other = other.to_dataset(name=other.name)
was_dataarray = True
except AttributeError:
# must be a ndarray
rdata = self.grid.map_gridded_data(other, grid=grid,
interp=interp, ks=ks)
# let's guess
sh = rdata.shape
nd = len(sh)
if nd == 2:
dims = (self.y_dim, self.x_dim)
elif nd == 3:
newdim = 'new_dim'
if self.t_dim and sh[0] == self._obj.dims[self.t_dim]:
newdim = self.t_dim
if self.z_dim and sh[0] == self._obj.dims[self.z_dim]:
newdim = self.z_dim
dims = (newdim, self.y_dim, self.x_dim)
else:
raise NotImplementedError('more than 3 dims not ok yet.')
coords = {}
for d in dims:
if d in self._obj:
coords[d] = self._obj[d]
out = xr.DataArray(rdata, coords=coords, dims=dims)
out.attrs['pyproj_srs'] = self.grid.proj.srs
return out
# go
out = xr.Dataset()
for v in other.data_vars:
var = other[v]
rdata = self.grid.map_gridded_data(var, interp=interp, ks=ks)
# remove old coords
dims = [d for d in var.dims]
coords = {}
for c in var.coords:
n = utils.str_in_list([c], utils.valid_names['x_dim'])
if n:
dims = [self.x_dim if x in n else x for x in dims]
continue
n = utils.str_in_list([c], utils.valid_names['y_dim'])
if n:
dims = [self.y_dim if x in n else x for x in dims]
continue
coords[c] = var.coords[c]
# add new ones
coords[self.x_dim] = self._obj[self.x_dim]
coords[self.y_dim] = self._obj[self.y_dim]
rdata = xr.DataArray(rdata, coords=coords, attrs=var.attrs, dims=dims)
rdata.attrs['pyproj_srs'] = self.grid.proj.srs
out[v] = rdata
if was_dataarray:
out = out[v]
else:
out.attrs['pyproj_srs'] = self.grid.proj.srs
return out
