statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# 10 sea level pressure observations, 3 hours after model is run. Accuracy is
# approx 0.01 in mercury, 33.8 Pa, rounded to 34
#Ketchikan
ob1 = Observation(
value=101320,
time=datetime(2017, 9, 6, 6),
lat=55.356,
lon=228.286, #101210
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=34)
#Seatac
ob2 = Observation(
value=101420,
time=datetime(2017, 9, 6, 6),
lat=47.445,
lon=237.686, #101280
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=34)
#Spokane
ob_lon = float(ob_split[2])
#get longitude positive-definite- ie -130 lon is 230 E
# if ob_lon < 0:
# ob_lon = ob_lon + 360
ob_elev = float(ob_split[3])
ob_time = mt.timestamp2utc(float(ob_split[4]))
ob_value = float(ob_split[5])
#call function to check elevation to see whether or not to assimilate ob
TorF = efa.check_elevation(lats, lons, elevs, ob, ob_lat,
ob_lon, ob_elev)
obser = Observation(value=ob_value,
time=ob_time,
lat=ob_lat,
lon=ob_lon,
obtype=ob_type,
localize_radius=1000,
assimilate_this=TorF,
error=1)
if TorF == True and ob_id != 'SHIP':
#create dictionary for station ID if it doesn't exist
ob_dict[ob_type][sfh][hour][ob_id] = ob_dict[ob_type][sfh][
hour].get(ob_id, {})
#create empty lists for the station ID if they don't exist yet
ob_dict[ob_type][sfh][hour][ob_id]['hx'] = ob_dict[
ob_type][sfh][hour][ob_id].get('hx', [])
ob_dict[ob_type][sfh][hour][ob_id]['ob_all'] = ob_dict[
ob_type][sfh][hour][ob_id].get('ob_all', [])
ob_dict[ob_type][sfh][hour][ob_id]['obs'] = ob_dict[
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# Creating 2 dummy obs to test EFA- if exact coordinates in the lat/lon,
# interpolate will fail.
#key west
ob1 = Observation(value=5900,
time=datetime(2017, 9, 6, 12),
lat=24.55,
lon=278.21,
obtype='Z500',
localize_radius=2000,
assimilate_this=True,
error=10)
#Miami
ob2 = Observation(value=5900,
time=datetime(2017, 9, 6, 12),
lat=25.75,
lon=279.62,
obtype='Z500',
localize_radius=2000,
assimilate_this=True,
error=10)
#Tampa Bay
ob3 = Observation(value=5870,
time=datetime(2017, 9, 6, 12),
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# 10 sea level pressure observations, 3 hours after model is run. Accuracy is
# approx 0.01 in mercury, 33.8 Pa, rounded to 34
#key west
ob1 = Observation(value=101440,
time=datetime(2017, 9, 6, 3),
lat=24.555,
lon=278.248,
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=34)
#Miami
ob2 = Observation(value=101470,
time=datetime(2017, 9, 6, 3),
lat=25.791,
lon=279.684,
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=34)
#Tampa Bay
ob3 = Observation(value=101510,
time=datetime(2017, 9, 6, 3),
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# 10 sea level pressure observations, 3 hours after model is run. Accuracy is
# approx 0.01 in mercury, 33.8 Pa, rounded to 34
#key west
ob1 = Observation(value=101440,
time=datetime(2017, 9, 6, 3),
lat=24.555,
lon=278.248,
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=34)
#Miami
ob2 = Observation(value=101470,
time=datetime(2017, 9, 6, 3),
lat=25.791,
lon=279.684,
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=34)
#Tampa Bay
ob3 = Observation(value=101510,
time=datetime(2017, 9, 6, 3),
def get_gdas_obs(atime,
vrbls=['Z500'],
sampling_interval=5,
cutofflat=None,
loc='GC2000',
err=100.,
errvar_file=None):
"""
Gets a list of 'observations' from the CFSv2 f00 forecast (i.e., GDAS
analysis) grid.
Requires:
atime ------------> A datetime object indicating the time of the desired
observations.
vrbls ------------> A list of variables to get observations of.
Default: 500 mb heights.
sampling_interval-> An integer specifying the interval at which to sample
the gridded analysis for observations.
Default: Sample every 5 degrees lat/lon. (1 deg grid)
cutofflat --------> Grid points poleward of this latitude are ignored.
Returns:
observations -----> A list of Observation objects.
"""
# Set the localization radius
if loc[:2] == 'GC':
locrad = int(loc[2:])
else:
locrad = 2000.0 # kilometers
# Corresponding parameters in the analysis file
apars = {'Z500': 'HGT_500mb', 'CHI200': 'VPOT_200mb'}
# Set I/O variable(s)
anl_file = '/home/disk/vader2/njweber2/research/subseasonal/efa/' + \
'verification/analyses_01Dec2015-01Mar2016.nc'
# Load the data
vardata = {}
with Dataset(anl_file, 'r') as ncdata:
# Load dimensions
timenums = ncdata.variables['time']
times = num2date(timenums[:], timenums.units)
lats = ncdata.variables['latitude'][:]
lons = ncdata.variables['longitude'][:]
# Cut off polar grid points:
if cutofflat is not None:
yi = ut.nearest_ind(lats, -cutofflat)
yf = ut.nearest_ind(lats, cutofflat) + 1
else:
yi = 0
yf = len(lats)
lats = lats[yi:yf]
# Get a lat/lon value for every location
lonarr, latarr = np.meshgrid(lons, lats)
# Sample only a fraction of the points in the analysis
lonarr = lonarr[::sampling_interval, ::sampling_interval].flatten()
latarr = latarr[::sampling_interval, ::sampling_interval].flatten()
# Find the desired time
t_ind = np.where(times == atime)[0][0]
# Load the variable
for var in vrbls:
vardata[var] = ncdata.variables[apars[var]][t_ind, yi:yf, :]
# Sample only a fraction of the points in the analysis
vardata[var] = vardata[
var][::sampling_interval, ::sampling_interval].flatten()
if var == 'CHI200': vardata[var] /= 10**6 # to 10^6 m^2 s^-1
# Load the ob error variance file, if necessary
if errvar_file is not None:
with Dataset(errvar_file, 'r') as ncdata:
errvar_la = ncdata.variables['lat'][:]
errvar_lo = ncdata.variables['lon'][:]
errvar_xy = ncdata.variables['errvar'][:]
# Create Observation objects for each of the sampled locations
observations = []
for x in range(len(lonarr)):
# Assign the ob error variance
if errvar_file is None:
oberr = err
else:
la_i = ut.nearest_ind(errvar_la, latarr[x])
lo_i = ut.nearest_ind(errvar_lo, lonarr[x])
oberr = errvar_xy[la_i, lo_i]
# Create an Observation object for each variable
for var in vrbls:
obsject = Observation(value=vardata[var][x],
obtype=var,
time=atime,
error=oberr,
lat=latarr[x],
lon=lonarr[x],
localize_radius=locrad,
description='GDAS')
obsject.assimilate_this = True
observations.append(obsject)
return observations
ncdata.variables[var][:])
lonarr, latarr = np.meshgrid(lons, lats)
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(allvars,
{'validtime' : ftimes,
'lat' : (['y','x'], latarr),
'lon' : (['y','x'], lonarr),
'mem' : mems,
})
# Creating 2 dummy obs to test EFA- if exact coordinates in the lat/lon,
# interpolate will fail.
#Annette Island, AK
ob1 = Observation(value=5820, time=datetime(2017,9,6,0),lat=55.05,lon=228.41,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Prince George, BC
ob2 = Observation(value=5890, time=datetime(2017,9,6,0),lat=53.9,lon=237.21,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Edmonton Stony Plain, AB
ob3 = Observation(value=5870, time=datetime(2017,9,6,0),lat=53.53,lon=245.9,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Port Hardy, BC
ob4 = Observation(value=5870, time=datetime(2017,9,6,0),lat=50.68,lon=232.64,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Quillayute
ob5 = Observation(value=5910, time=datetime(2017,9,6,0),lat=47.95,lon=235.45,
#get the lat/lon of the station
ob_lat = float(ob_split[1])
ob_lon = float(ob_split[2])
#get longitude positive-definite- ie -130 lon is 230 E
if ob_lon < 0:
ob_lon = ob_lon + 360
ob_elev = float(ob_split[3])
ob_time = float(ob_split[4])
time2 = datetime.utcfromtimestamp(ob_time)
ob_value = float(ob_split[5])
#call function to check elevation to see whether or not to assimilate ob
TorF = efa.check_elevation(lats,lons,elevs,ob,ob_lat,ob_lon,ob_elev)
obser = Observation(value=ob_value, time=time2,
lat=ob_lat,lon=ob_lon, obtype=o_type, localize_radius=1000,
assimilate_this=TorF, error=1)
observations.append(obser)
# Put the state class object and observation objects into EnSRF object
assimilator = EnSRF(statecls, observations, loc='GC')
# Update the prior with EFA- post_state is an EnsembleState object
post_state, post_obs = assimilator.update()
state=post_state
outdir = '/home/disk/hot/stangen/Documents/atms544/posterior/'+ensemble_type+'/'+m_dict[month]+year+'/'
#Create output directories if they don't yet exit
if (os.path.isdir(outdir)):
pass
statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# 10 sea level pressure observations, 3 hours after model is run. Accuracy is
# approx 0.01 in mercury, 33.8 Pa, rounded to 34
#Ketchikan
ob1 = Observation(
value=290.95,
time=datetime(2017, 9, 6, 6),
lat=55.356,
lon=228.286, #101210
obtype='T2M',
localize_radius=2000,
assimilate_this=True,
error=0.3)
#Seatac
ob2 = Observation(
value=294.85,
time=datetime(2017, 9, 6, 6),
lat=47.445,
lon=237.686, #101280
obtype='T2M',
localize_radius=2000,
assimilate_this=True,
error=0.3)
#Spokane
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# 10 sea level pressure observations, 3 hours after model is run. Accuracy is
# approx 0.01 in mercury, 33.8 Pa, rounded to 34
#key west
ob1 = Observation(value=101440,
time=datetime(2017, 9, 6, 3),
lat=24.555,
lon=278.248,
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=100000)
#Miami
ob2 = Observation(value=101470,
time=datetime(2017, 9, 6, 3),
lat=25.791,
lon=279.684,
obtype='MSLP',
localize_radius=2000,
assimilate_this=True,
error=100000)
#Tampa Bay
ob3 = Observation(value=101510,
time=datetime(2017, 9, 6, 3),
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(
allvars, {
'validtime': ftimes,
'lat': (['y', 'x'], latarr),
'lon': (['y', 'x'], lonarr),
'mem': mems,
})
# Creating 2 dummy obs to test EFA- if exact coordinates in the lat/lon,
# interpolate will fail.
#key west
ob1 = Observation(value=5900,
time=datetime(2017, 9, 6, 0),
lat=24.55,
lon=278.21,
obtype='Z500',
localize_radius=2000,
assimilate_this=True,
error=10)
#Miami
ob2 = Observation(value=5890,
time=datetime(2017, 9, 6, 0),
lat=25.75,
lon=279.62,
obtype='Z500',
localize_radius=2000,
assimilate_this=True,
error=10)
#Tampa Bay
ob3 = Observation(value=5880,
time=datetime(2017, 9, 6, 0),
ncdata.variables[var][:])
lonarr, latarr = np.meshgrid(lons, lats)
# Package into an EnsembleState object knowing the state and metadata
statecls = EnsembleState.from_vardict(allvars,
{'validtime' : ftimes,
'lat' : (['y','x'], latarr),
'lon' : (['y','x'], lonarr),
'mem' : mems,
})
# Creating 2 dummy obs to test EFA- if exact coordinates in the lat/lon,
# interpolate will fail.
#key west
ob1 = Observation(value=5900, time=datetime(2017,9,6,0),lat=24.55,lon=278.21,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Miami
ob2 = Observation(value=5890, time=datetime(2017,9,6,0),lat=25.75,lon=279.62,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Tampa Bay
ob3 = Observation(value=5880, time=datetime(2017,9,6,0),lat=27.70,lon=277.6,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Tallahassee
ob4 = Observation(value=5870, time=datetime(2017,9,6,0),lat=30.45,lon=275.7,
obtype = 'Z500', localize_radius=2000, assimilate_this=True,
error=10)
#Jacksonville
ob5 = Observation(value=5870, time=datetime(2017,9,6,0),lat=30.50,lon=278.3,