def get_values_on_grid(self, var, lon, lat, **kwargs):
z = kwargs.get('z', None)
t = kwargs.get('t', None)
tinds = None
zinds = None
bbox = [np.min(np.min(lon)), np.min(np.min(lat)), np.max(np.max(lon)), np.max(np.max(lat))]
if z is None:
zbounds = z
else:
zbounds = (np.min(np.min(np.min(np.min(z)))), np.max(np.max(np.max(np.max(z)))),)
if t is None:
tbounds = t
else:
if True:
tbounds = None
timeinds = np.arange(t[0], t[-1]+1)
else:
tbounds = (t[0], t[-1])
method = kwargs.get('method', 'nearest')
raw_vals = self.get_values(var, zbounds=zbounds, bbox=bbox,
timeinds=tinds, zinds=zinds, timebounds=tbounds)
coords_struct = self.sub_coords(var, zbounds=zbounds, bbox=bbox,
timeinds=tinds, zinds=zinds, timebounds=tbounds)
interpolator = CfGeoInterpolator(raw_vals, coords_struct.x, coords_struct.y,
z=coords_struct.z, t=coords_struct.time, method=method)
return interpolator.interpgrid(lon, lat, t=t, z=z)
def test_interpolator_2d(self):
lonbounds = [-70, -60]
latbounds = [40, 50]
nx, ny = 50, 50
lon, lat = create_grid(lonbounds[0], lonbounds[1], latbounds[0], latbounds[1], nx=50, ny=50)
data = np.random.rand(50, 50)
i = CfGeoInterpolator(data, lon, lat, method='nearest')
data2 = i.interpgrid(lon, lat)
assert np.all(data==data2)
def test_interpolator_3d_3dz_1dll(self):
lonbounds = [-70, -60]
latbounds = [40, 50]
nx, ny = 50, 50
lon, lat = create_grid(lonbounds[0], lonbounds[1], latbounds[0], latbounds[1], nx=50, ny=50)
data = np.random.rand(10, 50, 50)
z, dummy, dummy2 = np.meshgrid(np.arange(10), lon, lat, indexing='ij')
i = CfGeoInterpolator(data, lon, lat, z=z, method='nearest')
data2 = i.interpgrid(lon, lat, z=z)
assert np.all(data==data2)
def test_interpolator_3d_1dt_1dz_1dll(self):
lonbounds = [-70, -60]
latbounds = [40, 50]
nx, ny = 50, 50
lon, lat = create_grid(lonbounds[0], lonbounds[1], latbounds[0], latbounds[1], nx=50, ny=50)
data = np.random.rand(9, 10, 50, 50)
z = np.arange(10)
t = np.arange(9)
i = CfGeoInterpolator(data, lon, lat, z=z, t=t, method='nearest')
data2 = i.interpgrid(lon, lat, z=z, t=t)
assert np.all(data==data2)
def test_interpolator_2d(self):
lonbounds = [-70, -60]
latbounds = [40, 50]
nx, ny = 50, 50
lon, lat = create_grid(lonbounds[0],
lonbounds[1],
latbounds[0],
latbounds[1],
nx=50,
ny=50)
data = np.random.rand(50, 50)
i = CfGeoInterpolator(data, lon, lat, method='nearest')
data2 = i.interpgrid(lon, lat)
assert np.all(data == data2)
def test_interpolator_3d_3dz_1dll(self):
lonbounds = [-70, -60]
latbounds = [40, 50]
nx, ny = 50, 50
lon, lat = create_grid(lonbounds[0],
lonbounds[1],
latbounds[0],
latbounds[1],
nx=50,
ny=50)
data = np.random.rand(10, 50, 50)
z, dummy, dummy2 = np.meshgrid(np.arange(10), lon, lat, indexing='ij')
i = CfGeoInterpolator(data, lon, lat, z=z, method='nearest')
data2 = i.interpgrid(lon, lat, z=z)
assert np.all(data == data2)
def test_interpolator_3d_1dt_1dz_1dll(self):
lonbounds = [-70, -60]
latbounds = [40, 50]
nx, ny = 50, 50
lon, lat = create_grid(lonbounds[0],
lonbounds[1],
latbounds[0],
latbounds[1],
nx=50,
ny=50)
data = np.random.rand(9, 10, 50, 50)
z = np.arange(10)
t = np.arange(9)
i = CfGeoInterpolator(data, lon, lat, z=z, t=t, method='nearest')
data2 = i.interpgrid(lon, lat, z=z, t=t)
assert np.all(data == data2)
def regrid_roms(newfile, filename, lon_new, lat_new, t=None, z=None):
'''Function to regrid the entire roms datasets (all values on non-rho coords)
onto an arbitrary grid to support regridding on to regular grids as well.
'''
with pw.new(newfile) as new:
with netCDF4.Dataset(filename) as nc:
for key in nc.variables:
try:
if "since" in nc.variables[key].units:
time = nc.variables[key][:]
except AttributeError:
pass
# Identify the rho coordinates, and get them
lon_rho = nc.variables["lon_rho"][:]
lat_rho = nc.variables["lat_rho"][:]
rho_y, rho_x = lat_rho.shape
depth_rho = nc.variables["s_rho"][:]
#depth_w = nc.variables["s_w"][:]
if t == None:
t = time
# Put dimensions into the new netcdf file, should only be for
# time, rhos, psis (not sure what to do about w yet)
if len(depth_rho.shape) == 4:
s_rho = depth_rho.shape[1]
else:
s_rho = depth_rho.shape[0]
if len(lon_new.shape) == 2 and len(lon_new.shape) == 2:
eta_new = lat_new.shape[0]
xi_new = lon_new.shape[1]
elif len(lon_new.shape) == 1 and len(lon_new.shape) == 1:
eta_new = lat_new.shape[0]
xi_new = lon_new.shape[0]
else:
raise ValueError("New lat and lon have invalid shapes or don't match in shape.")
if z == None:
z = depth_rho
s_new = z.shape[0]
time_new = t.shape[0]
# [pw.add_attribute(new, at, new.getncattr(at)) for at in new.ncattrs()]
pw.add_coordinates(new, OrderedDict([("time_new",time_new),("s_new",s_new),("eta_new",eta_new),("xi_new",xi_new)]))
# print("Coordinates " + str([("time_new",time_new),("s_new",s_new),("eta_new",eta_new),("xi_new",xi_new)]))
pw.add_variable(new, "ocean_time", t, ("time_new",))
pw.add_variable(new, "s_new", z, ("s_new",))
if len(lon_new.shape) == 2 and len(lat_new.shape) == 2:
pw.add_variable(new, "lat_new", lat_new, ("eta_new", "xi_new",))
pw.add_variable(new, "lon_new", lon_new, ("eta_new", "xi_new",))
elif len(lon_new.shape) == 1 and len(lat_new.shape) == 1:
pw.add_variable(new, "lat_new", lat_new, ("eta_new",))
pw.add_variable(new, "lon_new", lon_new, ("xi_new",))
# Identify variables that aren't coordinates, and their native grids,
# convert non-rho variables to rho with `average_adjacents`. Create
# sequence of rho-based variables in `roms_variables` object
for key in nc.variables:
var = nc.variables[key]
try:
if key == "w":
if z == None:
z = depth_w
grid_type = "rho"
else:
if "_psi" in var.coordinates:
grid_type = "psi"
elif "_u" in var.coordinates:
grid_type = "u"
elif "_v" in var.coordinates:
grid_type = "v"
else:
grid_type = "rho"
except AttributeError:
grid_type = None
if grid_type != None and grid_type != "psi":
if key in set([ "sustr", "bustr", "DU_avg1", "DU_avg2", "u", "w", "ubar", "mask_u", "mask_v", "mask_psi", "mask_rho" ]):
pass
elif grid_type == "v":
var_dimensionality = len(var.shape)
if "sv" in key or "bv" in key or "DV" in key or key == "vbar" or key == "v":
paired_vector = {"svstr":"sustr",
"bvstr":"bustr",
"DV_avg1":"DU_avg1",
"DV_avg2":"DU_avg2",
"vbar":"ubar",
"v":"u"}
for i in range(var.shape[0]):
if var_dimensionality == 3:
complex_uv = _uv_to_rho(nc.variables[paired_vector[key]][i,:,:], var[i,:,:], nc.variables["angle"][:], rho_x, rho_y)
uz, vz = complex_uv.real, complex_uv.imag
elif var_dimensionality == 4:
for j in range(var.shape[1]):
complex_uv = _uv_to_rho(nc.variables["u"][i,j,:,:], var[i,j,:,:], nc.variables["angle"][:], rho_x, rho_y)
if j == 0:
uz, vz = complex_uv.real[np.newaxis,:], complex_uv.imag[np.newaxis,:]
else:
uz = np.vstack((uz, complex_uv.real[np.newaxis,:]))
vz = np.vstack((vz, complex_uv.imag[np.newaxis,:]))
if i == 0:
u, v = uz[np.newaxis,:], vz[np.newaxis,:]
else:
u = np.vstack((u, uz[np.newaxis,:]))
v = np.vstack((v, vz[np.newaxis,:]))
values = {paired_vector[key]:u, key:v}
for new_key in values:
values_interp[:, np.where(nc.variables["mask_rho"]==0)] = np.nan
if var_dimensionality == 3:
interpolator = CfGeoInterpolator(values[new_key], lon_rho, lat_rho, t=time)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t)
coordtuple = ("time_new", "eta_new", "xi_new",)
coordattr = "ocean_time lat_new lon_new"
elif var_dimensionality == 4:
interpolator = CfGeoInterpolator(values[new_key], lon_rho, lat_rho, t=time, z=depth_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t, z=z)
coordtuple = ("time_new", "s_new", "eta_new", "xi_new",)
coordattr = "ocean_time s_new lat_new lon_new"
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, new_key, values_interp, coordtuple)
[pw.add_attribute(new, at, nc.variables[new_key].getncattr(at), var=new_key) for at in nc.variables[new_key].ncattrs()]
pw.add_attribute(new, "coordinates", coordattr, var=new_key)
else:
var_dimensionality = len(var.shape)
vartmp = var[:]
if var_dimensionality == 4:
vartmp[:,:, np.where(nc.variables["mask_rho"]==0)] = np.nan
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho, t=time, z=depth_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t, z=z)
pw.add_variable(new, key, values_interp, ("time_new", "s_new", "eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "ocean_time s_new lat_new lon_new", var=key)
elif var_dimensionality == 3:
vartmp[:, np.where(nc.variables["mask_rho"]==0)] = np.nan
if var.shape[0] == time.shape[0]:
try:
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho, t=time)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t)
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, key, values_interp, ("time_new", "eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "ocean_time lat_new lon_new", var=key)
except:
print(key, vartmp.shape, lon_rho.shape, lat_rho.shape, time.shape)
elif var.shape[0] == depth_rho.shape[0]:
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho, z=depth_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new, z=z)
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, key, values_interp, ("depth_new", "eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "s_new lat_new lon_new", var=key)
else:
raise ValueError("unsure about what dimension this variable varies with in addition to lat/lon.")
elif var_dimensionality == 2:
vartmp[np.where(nc.variables["mask_rho"]==0)] = np.nan
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new)
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, key, values_interp, ("eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "lat_new lon_new", var=key)
else:
# todo if 1-d check for which dimension it matches and interp based on that...
# if 5+ d, only interpolate to the 4d dimensions that we can specify i guess...
raise ValueError("sort of confused about the dimensionality of the variable i am attempting to regrid...")
elif grid_type == "psi":
pass
else:
if len(var.dimensions) == 0:
pw.add_scalar(new, key, var[:])
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
# Add time attributes
for key in nc.variables:
var = nc.variables[key]
if "time" in key:
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var="ocean_time") for at in nc.variables[key].ncattrs()]
# Add global attributes to the file
[pw.add_attribute(new, at, nc.getncattr(at)) for at in nc.ncattrs()]
try:
new.history = new.history + ", regridded by Python tool 'paegan' at " + str(datetime.datetime.now())
except:
new.history = "regridded by Python tool 'paegan' at " + str(datetime.datetime.now())
new.sync()
def regrid_roms(newfile, filename, lon_new, lat_new, t=None, z=None):
'''Function to regrid the entire roms datasets (all values on non-rho coords)
onto an arbitrary grid to support regridding on to regular grids as well.
'''
with pw.new(newfile) as new:
with netCDF4.Dataset(filename) as nc:
for key in nc.variables:
try:
if "since" in nc.variables[key].units:
time = nc.variables[key][:]
except AttributeError:
pass
# Identify the rho coordinates, and get them
lon_rho = nc.variables["lon_rho"][:]
lat_rho = nc.variables["lat_rho"][:]
rho_y, rho_x = lat_rho.shape
depth_rho = nc.variables["s_rho"][:]
#depth_w = nc.variables["s_w"][:]
if t == None:
t = time
# Put dimensions into the new netcdf file, should only be for
# time, rhos, psis (not sure what to do about w yet)
if len(depth_rho.shape) == 4:
s_rho = depth_rho.shape[1]
else:
s_rho = depth_rho.shape[0]
if len(lon_new.shape) == 2 and len(lon_new.shape) == 2:
eta_new = lat_new.shape[0]
xi_new = lon_new.shape[1]
elif len(lon_new.shape) == 1 and len(lon_new.shape) == 1:
eta_new = lat_new.shape[0]
xi_new = lon_new.shape[0]
else:
raise ValueError("New lat and lon have invalid shapes or don't match in shape.")
if z == None:
z = depth_rho
s_new = z.shape[0]
time_new = t.shape[0]
#[pw.add_attribute(new, at, new.getncattr(at)) for at in new.ncattrs()]
pw.add_coordinates(new, OrderedDict([("time_new",time_new),("s_new",s_new),("eta_new",eta_new),("xi_new",xi_new)]))
#print "Coordinates " + str([("time_new",time_new),("s_new",s_new),("eta_new",eta_new),("xi_new",xi_new)])
pw.add_variable(new, "ocean_time", t, ("time_new",))
pw.add_variable(new, "s_new", z, ("s_new",))
if len(lon_new.shape) == 2 and len(lat_new.shape) == 2:
pw.add_variable(new, "lat_new", lat_new, ("eta_new", "xi_new",))
pw.add_variable(new, "lon_new", lon_new, ("eta_new", "xi_new",))
elif len(lon_new.shape) == 1 and len(lat_new.shape) == 1:
pw.add_variable(new, "lat_new", lat_new, ("eta_new",))
pw.add_variable(new, "lon_new", lon_new, ("xi_new",))
# Identify variables that arn't coordinates, and their native grids,
# convert non-rho variables to rho with `average_adjacents`. Create
# sequence of rho-based variables in `roms_variables` object
for key in nc.variables:
var = nc.variables[key]
try:
if key == "w":
if z == None:
z = depth_w
grid_type = "rho"
else:
if "_psi" in var.coordinates:
grid_type = "psi"
elif "_u" in var.coordinates:
grid_type = "u"
elif "_v" in var.coordinates:
grid_type = "v"
else:
grid_type = "rho"
except AttributeError:
grid_type = None
if grid_type != None and grid_type != "psi":
if key in set([ "sustr", "bustr", "DU_avg1", "DU_avg2", "u", "w", "ubar", "mask_u", "mask_v", "mask_psi", "mask_rho" ]):
pass
elif grid_type == "v":
var_dimensionality = len(var.shape)
if "sv" in key or "bv" in key or "DV" in key or key == "vbar" or key == "v":
paired_vector = {"svstr":"sustr",
"bvstr":"bustr",
"DV_avg1":"DU_avg1",
"DV_avg2":"DU_avg2",
"vbar":"ubar",
"v":"u"}
for i in xrange(var.shape[0]):
if var_dimensionality == 3:
complex_uv = _uv_to_rho(nc.variables[paired_vector[key]][i,:,:], var[i,:,:], nc.variables["angle"][:], rho_x, rho_y)
uz, vz = complex_uv.real, complex_uv.imag
elif var_dimensionality == 4:
for j in xrange(var.shape[1]):
complex_uv = _uv_to_rho(nc.variables["u"][i,j,:,:], var[i,j,:,:], nc.variables["angle"][:], rho_x, rho_y)
if j == 0:
uz, vz = complex_uv.real[np.newaxis,:], complex_uv.imag[np.newaxis,:]
else:
uz = np.vstack((uz, complex_uv.real[np.newaxis,:]))
vz = np.vstack((vz, complex_uv.imag[np.newaxis,:]))
if i == 0:
u, v = uz[np.newaxis,:], vz[np.newaxis,:]
else:
u = np.vstack((u, uz[np.newaxis,:]))
v = np.vstack((v, vz[np.newaxis,:]))
values = {paired_vector[key]:u, key:v}
for new_key in values:
values_interp[:, np.where(nc.variables["mask_rho"]==0)] = np.nan
if var_dimensionality == 3:
interpolator = CfGeoInterpolator(values[new_key], lon_rho, lat_rho, t=time)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t)
coordtuple = ("time_new", "eta_new", "xi_new",)
coordattr = "ocean_time lat_new lon_new"
elif var_dimensionality == 4:
interpolator = CfGeoInterpolator(values[new_key], lon_rho, lat_rho, t=time, z=depth_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t, z=z)
coordtuple = ("time_new", "s_new", "eta_new", "xi_new",)
coordattr = "ocean_time s_new lat_new lon_new"
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, new_key, values_interp, coordtuple)
[pw.add_attribute(new, at, nc.variables[new_key].getncattr(at), var=new_key) for at in nc.variables[new_key].ncattrs()]
pw.add_attribute(new, "coordinates", coordattr, var=new_key)
else:
var_dimensionality = len(var.shape)
vartmp = var[:]
if var_dimensionality == 4:
vartmp[:,:, np.where(nc.variables["mask_rho"]==0)] = np.nan
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho, t=time, z=depth_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t, z=z)
pw.add_variable(new, key, values_interp, ("time_new", "s_new", "eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "ocean_time s_new lat_new lon_new", var=key)
elif var_dimensionality == 3:
vartmp[:, np.where(nc.variables["mask_rho"]==0)] = np.nan
if var.shape[0] == time.shape[0]:
try:
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho, t=time)
values_interp = interpolator.interpgrid(lon_new, lat_new, t=t)
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, key, values_interp, ("time_new", "eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "ocean_time lat_new lon_new", var=key)
except:
print key, vartmp.shape, lon_rho.shape, lat_rho.shape, time.shape
elif var.shape[0] == depth_rho.shape[0]:
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho, z=depth_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new, z=z)
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, key, values_interp, ("depth_new", "eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "s_new lat_new lon_new", var=key)
else:
raise valueerror("unsure about what dimension this varaible varies with in addition to lat/lon.")
elif var_dimensionality == 2:
vartmp[np.where(nc.variables["mask_rho"]==0)] = np.nan
interpolator = CfGeoInterpolator(vartmp, lon_rho, lat_rho)
values_interp = interpolator.interpgrid(lon_new, lat_new)
values_interp = np.ma.MaskedArray(values_interp, mask=values_interp==np.nan)
pw.add_variable(new, key, values_interp, ("eta_new", "xi_new",))
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
pw.add_attribute(new, "coordinates", "lat_new lon_new", var=key)
else:
# todo if 1-d check for which dimension it matches and interp based on that...
# if 5+ d, only interpolate to the 4d dimensions that we can specify i guess...
raise valueerror("sort of confused about the dimensionality of the variable i am attempting to regrid...")
elif grid_type == "psi":
pass
else:
if len(var.dimensions) == 0:
pw.add_scalar(new, key, var[:])
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var=key) for at in nc.variables[key].ncattrs()]
# Add time attributes
for key in nc.variables:
var = nc.variables[key]
if "time" in key:
[pw.add_attribute(new, at, nc.variables[key].getncattr(at), var="ocean_time") for at in nc.variables[key].ncattrs()]
# Add global attributes to the file
[pw.add_attribute(new, at, nc.getncattr(at)) for at in nc.ncattrs()]
try:
new.history = new.history + ", regridded by Python tool 'paegan' at " + str(datetime.datetime.now())
except:
new.history = "regridded by Python tool 'paegan' at " + str(datetime.datetime.now())
new.sync()
