def make_intialization_from_hrrr(Grid, file_path):
"""
This function will read an HRRR GRIB2 file and return initial guess
u, v, and w fields from the model
Parameters
----------
Grid: Py-ART Grid
The Py-ART Grid to use as the grid specification. The HRRR values
will be interpolated to the Grid's specficiation and added as a field.
file_path: string
The path to the GRIB2 file to load.
Returns
-------
Grid: Py-ART Grid
This returns the Py-ART grid with the HRRR u, and v fields added.
The shape will be the same shape as the fields in Grid and will
correspond to the same x, y, and z locations as in Grid.
"""
if(CFGRIB_AVAILABLE is False):
raise RuntimeError(("The cfgrib optional dependency needs to be " +
"installed for the HRRR integration feature."))
the_grib = cfgrib.open_file(
file_path, filter_by_keys={'typeOfLevel': 'isobaricInhPa'})
# Load the HRR data and tranform longitude coordinates
grb_u = the_grib.variables['u']
grb_v = the_grib.variables['v']
grb_w = the_grib.variables['w']
gh = the_grib.variables['gh']
lat = the_grib.variables['latitude'].data[:, :]
lon = the_grib.variables['longitude'].data[:, :]
lon[lon > 180] = lon[lon > 180] - 360
# Convert geometric height to geopotential height
EARTH_MEAN_RADIUS = 6.3781e6
gh = gh.data[:, :, :]
height = (EARTH_MEAN_RADIUS*gh)/(EARTH_MEAN_RADIUS-gh)
height = height - Grid.radar_altitude['data']
radar_grid_lat = Grid.point_latitude['data']
radar_grid_lon = Grid.point_longitude['data']
radar_grid_alt = Grid.point_z['data']
lat_min = radar_grid_lat.min()
lat_max = radar_grid_lat.max()
lon_min = radar_grid_lon.min()
lon_max = radar_grid_lon.max()
lon_r = np.tile(lon, (height.shape[0], 1, 1))
lat_r = np.tile(lat, (height.shape[0], 1, 1))
lon_flattened = lon_r.flatten()
lat_flattened = lat_r.flatten()
height_flattened = gh.flatten()
the_box = np.where(np.logical_and.reduce(
(lon_flattened >= lon_min,
lat_flattened >= lat_min,
lon_flattened <= lon_max,
lat_flattened <= lat_max)))[0]
lon_flattened = lon_flattened[the_box]
lat_flattened = lat_flattened[the_box]
height_flattened = height_flattened[the_box]
u_flattened = grb_u.data[:, :, :].flatten()
u_flattened = u_flattened[the_box]
u_interp = NearestNDInterpolator(
(height_flattened, lat_flattened, lon_flattened),
u_flattened, rescale=True)
u_new = u_interp(radar_grid_alt, radar_grid_lat, radar_grid_lon)
v_flattened = grb_v.data[:, :, :].flatten()
v_flattened = v_flattened[the_box]
v_interp = NearestNDInterpolator(
(height_flattened, lat_flattened, lon_flattened),
v_flattened, rescale=True)
v_new = v_interp(radar_grid_alt, radar_grid_lat, radar_grid_lon)
w_flattened = grb_v.data[:, :, :].flatten()
w_flattened = w_flattened[the_box]
w_interp = NearestNDInterpolator(
(height_flattened, lat_flattened, lon_flattened),
w_flattened, rescale=True)
w_new = w_interp(radar_grid_alt, radar_grid_lat, radar_grid_lon)
del grb_u, grb_v, grb_w, lat, lon
del the_grib
gc.collect()
return u_new, v_new, w_new
def __init__(self, filename, lock=None, **backend_kwargs):
import cfgrib
if lock is None:
lock = ECCODES_LOCK
self.lock = ensure_lock(lock)
self.ds = cfgrib.open_file(filename, **backend_kwargs)
def scan_grib(url, common_vars, storage_options, inline_threashold=100, skip=0, filter={}):
"""
Generate references for a GRIB2 file
Parameters
----------
url: str
File location
common_vars: list[str]
Names of variables that are common to multiple measurable (i.e., coordinates)
storage_options: dict
For accessing the data, passed to filesystem
inline_threashold: int
If given, store array data smaller than this value directly in the output
skip: int
If non-zero, stop processing the file after this many messages
filter: dict
cfgrib-style filter dictionary
Returns
-------
dict: references dict in Version 1 format.
"""
if filter:
assert "typeOfLevel" in filter
logger.debug(f"Open {url}")
store = {}
z = zarr.open_group(store, mode='w')
common = False
with fsspec.open(url, "rb", **storage_options) as f:
for fn, offset, size in _split_file(f, skip=skip):
logger.debug(f"File {fn}")
ds = cfgrib.open_file(fn)
if filter:
var = filter["typeOfLevel"]
if var not in ds.variables:
continue
if "level" in filter and ds.variables[var].data not in np.array(filter["level"]):
continue
attr = ds.variables[var].attributes or {}
attr['_ARRAY_DIMENSIONS'] = []
if var not in z:
_store_array(store, z, np.array(ds.variables[var].data), var, 100000, 0, 0,
attr)
if common is False:
# done for first valid message
logger.debug("Common variables")
z.attrs.update(ds.attributes)
for var in common_vars:
# assume grid, etc is the same across all messages
attr = ds.variables[var].attributes or {}
attr['_ARRAY_DIMENSIONS'] = ds.variables[var].dimensions
_store_array(store, z, ds.variables[var].data, var, inline_threashold, offset, size,
attr)
common = True
for var in ds.variables:
if (
var not in common_vars and getattr(ds.variables[var].data, "shape", None)
and var != filter.get("typeOfLevel", "")
):
attr = ds.variables[var].attributes or {}
if "(deprecated)" in attr.get("GRIB_name", ""):
continue
attr['_ARRAY_DIMENSIONS'] = ds.variables[var].dimensions
_store_array(store, z, ds.variables[var].data, var, inline_threashold, offset, size,
attr)
logger.debug("Done")
return {"version": 1,
"refs": {k: v.decode() if isinstance(v, bytes) else v for k, v in store.items()},
"templates": {"u": url}}
def __init__(self, filename, lock=None, **backend_kwargs):
if lock is None:
lock = ECCODES_LOCK
self.lock = ensure_lock(lock)
self.ds = cfgrib.open_file(filename, **backend_kwargs)
def QueryGrib(gribFile,LatLonTimeFile,\
filterKeys={'dataType':'an','numberOfPoints':65160},\
ignoreList=['number','time','step','valid_time','longitude','latitude','heightAboveGround'],\
DistanceLim=1,TimeDelayLim=3,QueryColumns=['time','lat','lon'],\
forecastInd=0):
try:
#Load GRIB:
#ds = xarray.open_dataset(gribFile, engine='cfgrib')
ds = cfgrib.open_file(gribFile, filter_by_keys=filterKeys)
#Load Query file (comma separated; time, lon, lat):
Queries = pd.read_csv(LatLonTimeFile, names=QueryColumns)
except:
print('Error loading GRIB or Query files, ')
return -1
else:
tic = time.time()
#Extract frame data:
Glat = list(ds.variables['latitude'].data) #list(ds.latitude.data)
Glon = list(ds.variables['longitude'].data -
180) #list(ds.longitude.data)
# list((ds.time.data).astype(float))/1e9 #we convert implicitly from np.datetime64 to timestamp!
Gtime = list(ds.variables['time'].data)
#Extract list of variables:
gribVariables = ds.variables.keys()
numActualVars = len(
[var for var in gribVariables if var not in ignoreList])
iVars = -1
FRM = defaultdict(list)
FRM['Time'] = Queries.time
FRM['Latitude'] = Queries.lat
FRM['Longitude'] = Queries.lon
#
dimsLogged = False
for key in gribVariables: #load the Grib Variables into memory one by one!
if key not in ignoreList: #Leaev open the possibility for a ignoreList to ignore specific keys
iVars += 1
try:
keyData = ds.variables[key].data # try loading the data:
except:
print('failed to load variable {}'.format(key))
else:
for i in range(len(
Queries)): # cycle through the number of queries!
#find the closest (here we find the closest neighbor in the three dimensions (latitude, longitude and time)!
print('{:02.1f}% complete, {:.1f} seconds \elapsed'.format((iVars/numActualVars + \
i/len(Queries)/numActualVars)*100,time.time()- tic), end="\r", flush=True)
iLat, dLat = find_nearest(Queries.lat[i], Glat)
iLon, dLon = find_nearest(Queries.lon[i], Glon)
iTime, dTime = find_nearest(float(Queries.time[i]),
Gtime)
#
NearingDist = np.sqrt(dLat * dLat + dLon * dLon)
#
if not dimsLogged:
FRM['DistFrmGrdPnt'].append(NearingDist)
FRM['TimeOffset'].append(dTime)
if NearingDist <= DistanceLim and dTime <= TimeDelayLim * 3600:
try:
if len(
ds.dimensions.keys()
) == 4: #we are dealing with a forecast array!
bufr = keyData[forecastInd, iTime, iLat,
iLon]
if type(bufr) == 'numpy.ndarray':
bufr[
bufr == keyData.
missing_value] = np.nan #remove any missing values!
elif type(bufr) == 'numpy.float32':
if bufr == keyData.missing_value:
bufr = np.nan
val = bufr.mean()
else:
val = keyData[iTime, iLat, iLon]
if val == 9999:
val = np.nan
except:
val = np.nan
else:
val = np.nan
#Assign value
FRM[key].append(val)
dimsLogged = True
return FRM
