def HRRS_mean_ztrop_to_csv(DR,hostname='taurus',debug=False):
"""
Given a certain daterange, retrieve available high res radiosonde data,
compute the average tropopause height per station, and store in a
csv file.
"""
from TIL import ztrop
# first read in station information as a dataframe
stationdata = HRRS_station_data(hostname)
# because the HRRS data are sorted by years, loop over the years in the daterange
y0 = DR[0].year
yf = DR[len(DR)-1].year
years = range(y0,yf+1,1)
for YYYY in years:
# load a list of the available stations for that year
Slist = HRRS_stations_available_per_year(YYYY)
# also compute the subset of the requested daterange that fits into this year.
year_daterange = dart.daterange(date_start=datetime.datetime(YYYY,1,1,0,0,0), periods=365*4, DT='6H')
DR2 = set(year_daterange).intersection(DR)
# also find the dir where the station data live
datadir = es.obs_data_paths('HRRS',hostname)
# initialize empty dictionary to hold average tropoopause heights per station
ztrop_dict=dict()
# now loop over available stations, and for each one, retrieve the data
# that fit into the requested daterange
for s in Slist:
ztrop_list=[] # empty list to hold tropopause heights for all available obs per station
# loop over dates, and retrieve data if available
for dd in DR2:
datestr = dd.strftime("%Y%m%d%H")
ff = datadir+'/'+str(YYYY)+'/'+str(s)+'/'+str(s)+'-'+datestr+'_mod.dat'
if os.path.exists(ff):
if debug:
print(ff)
# read in the station data
D = read_HRRS_data(ff)
# compute tropopause height
z=D['Alt']/1E3 # Altitude in km
T=D['Temp']+273.15 # Temp in Kelvin
ztropp=ztrop(z=z,T=T,debug=debug,hostname=hostname)
# add to list if not none
if ztropp is not None:
ztrop_list.append(ztropp)
# average the tropopause heights and add to dictionary
ztrop_dict[s]=np.mean(ztrop_list)
# turn dict into data frame
ZT=pd.Series(data=ztrop_dict, name='ztrop_mean')
if debug:
print(ZT)
# turn dataframe into csv file
hrrs_path = es.obs_data_paths('HRRS',hostname)
datestr = DR[0].strftime("%Y%m%d")+'-'+DR[len(DR)-1].strftime("%Y%m%d")+'.csv'
fname=hrrs_path+'/'+'mean_tropopause_height_per_station_'+datestr
print('storing file '+fname)
ZT.to_csv(fname, index=True, sep=',',header=True)
return(ZT)
def HRRS_mean_ztrop_to_csv(DR, hostname='taurus', debug=False):
"""
Given a certain daterange, retrieve available high res radiosonde data,
compute the average tropopause height per station, and store in a
csv file.
"""
from TIL import ztrop
# first read in station information as a dataframe
stationdata = HRRS_station_data(hostname)
# because the HRRS data are sorted by years, loop over the years in the daterange
y0 = DR[0].year
yf = DR[len(DR) - 1].year
years = range(y0, yf + 1, 1)
for YYYY in years:
# load a list of the available stations for that year
Slist = HRRS_stations_available_per_year(YYYY)
# also compute the subset of the requested daterange that fits into this year.
year_daterange = dart.daterange(date_start=datetime.datetime(
YYYY, 1, 1, 0, 0, 0),
periods=365 * 4,
DT='6H')
DR2 = set(year_daterange).intersection(DR)
# also find the dir where the station data live
datadir = es.obs_data_paths('HRRS', hostname)
# initialize empty dictionary to hold average tropoopause heights per station
ztrop_dict = dict()
# now loop over available stations, and for each one, retrieve the data
# that fit into the requested daterange
for s in Slist:
ztrop_list = [
] # empty list to hold tropopause heights for all available obs per station
# loop over dates, and retrieve data if available
for dd in DR2:
datestr = dd.strftime("%Y%m%d%H")
ff = datadir + '/' + str(YYYY) + '/' + str(s) + '/' + str(
s) + '-' + datestr + '_mod.dat'
if os.path.exists(ff):
if debug:
print(ff)
# read in the station data
D = read_HRRS_data(ff)
# compute tropopause height
z = D['Alt'] / 1E3 # Altitude in km
T = D['Temp'] + 273.15 # Temp in Kelvin
ztropp = ztrop(z=z, T=T, debug=debug, hostname=hostname)
# add to list if not none
if ztropp is not None:
ztrop_list.append(ztropp)
# average the tropopause heights and add to dictionary
ztrop_dict[s] = np.mean(ztrop_list)
# turn dict into data frame
ZT = pd.Series(data=ztrop_dict, name='ztrop_mean')
if debug:
print(ZT)
# turn dataframe into csv file
hrrs_path = es.obs_data_paths('HRRS', hostname)
datestr = DR[0].strftime("%Y%m%d") + '-' + DR[len(DR) - 1].strftime(
"%Y%m%d") + '.csv'
fname = hrrs_path + '/' + 'mean_tropopause_height_per_station_' + datestr
print('storing file ' + fname)
ZT.to_csv(fname, index=True, sep=',', header=True)
return (ZT)
def TP_based_HRRS_data(ff,vertical_res_km=50E-3,debug=False,hostname='taurus'):
"""
Given a single high-res radiosonde data sounding (identified by its
full file path, ff)
load the data from the sounding and compute the temperature data
as a function of distance from the thermal tropopause.
This is done by:
1. reading in the data as a pandas data frame
2. computing the height of the tropopause
3. computin the altitude of each data point relative to the tropopause
4. using a cubic spline to create evenly-spaced temperatures on a vertical
grid with 50m spacing.
This procedure is based on Birner et al. 2002 (http://doi.wiley.com/10.1029/2002GL015142)
Here the LR tropopause follows the WMO criterion. Quoting Birner et al. (2002):
The thermal TP is defined as the lowest level where the temperature lapse rate falls
below 2 K/km and its average between this level and all higher levels within 2 km remains below this value [WMO, 1957].
INPUTS:
ff: the full path to the HRRS profile that we will load
vertical_res_km: vertical resolution of the grid to which we interpolate, in km. The default is 50m, which is roughly the vertical resolution of the HRRS obs.
"""
# read in the data as a data frame
DF0 = read_HRRS_data(ff)
# drop unnecessary columns
useless_cols=['Time','Dewpt','RH','Ucmp','Vcmp','spd','dir','Lat','Lon',
'Wcmp', 'Ele', 'Azi', 'Qp', 'Qt', 'Qrh', 'Qu', 'Qv', 'QdZ']
DF0.drop(useless_cols,inplace=True,axis=1)
# get rid of NaNs
DF=DF0.dropna()
if debug:
print('Loading file '+ff)
# load interpolate function from scipy
from scipy.interpolate import interp1d
# compute the height of the lapse-tropopause from the altitude array
z=DF['Alt']/1E3 # Altitude in km
T=DF['Temp']+273.15 # Temp in Kelvin
P=DF['Press']
N2=DF['N2']
from TIL import ztrop
ztropp=ztrop(z=z,T=T,debug=debug,hostname=hostname)
if ztropp is not None:
# extract the station number from the file path
file_components=ff.split('/')
station=file_components[len(file_components)-2]
# retrieve the mean tropopause height for this station
# TODO: right noew this reads in a csv file of mean heights for Jan 2010.
# ...need to make this more dynamic and obviously not user and host specific
ZT=pd.read_csv('/data/c1/lneef/HRRS//mean_tropopause_height_per_station_20100101-20100131.csv',index_col=0)
ztrop_mean = ZT.loc[int(station)].ztrop_mean
# now compute the altitude relative to the tropopause, plus mean tropopause height
zTP = DF['Alt']*1E-3-ztropp+ztrop_mean
# interpolate temp, pressure to this new coordinate
fT = interp1d(zTP, T, kind='linear')
fP = interp1d(zTP, P, kind='linear')
# create a regularly spaced grid (in km)
zTPgrid=np.arange(0.0,26.0, vertical_res_km)
# select whatever part of the regular grid fits into the range sampled by this sounding
select = np.where(np.logical_and(zTPgrid>min(zTP), zTPgrid<max(zTP)))
zTPnew=zTPgrid[select]
# now compute the variables on this grid using the interpolate function
Tnew = fT(zTPnew)
Pnew = fP(zTPnew)
# N2 comes out quite noisy when computed from raw radiosonde observations.
# The spline (needed to get the obs on a common grid) is an opportunity for smoothing
# the temperature field a bit, which will yield a smoother N2 profile -- so just recompute
# N2 here
from TIL import Nsq
N2new = Nsq(Tnew,zTPnew,Pnew)
# now create a new dataframe with the TP-based heights
new_data={'Press':Pnew,'Temp':Tnew,'Alt':zTPnew,'N2':N2new,'ztropp':ztropp}
Dout = pd.DataFrame(data=new_data)
else:
if debug:
print('No clear lapse-rate tropopause found for the following sounding:')
print(ff)
print('Returning None')
Dout=None
return(Dout)
def TP_based_HRRS_data(ff,
vertical_res_km=50E-3,
debug=False,
hostname='taurus'):
"""
Given a single high-res radiosonde data sounding (identified by its
full file path, ff)
load the data from the sounding and compute the temperature data
as a function of distance from the thermal tropopause.
This is done by:
1. reading in the data as a pandas data frame
2. computing the height of the tropopause
3. computin the altitude of each data point relative to the tropopause
4. using a cubic spline to create evenly-spaced temperatures on a vertical
grid with 50m spacing.
This procedure is based on Birner et al. 2002 (http://doi.wiley.com/10.1029/2002GL015142)
Here the LR tropopause follows the WMO criterion. Quoting Birner et al. (2002):
The thermal TP is defined as the lowest level where the temperature lapse rate falls
below 2 K/km and its average between this level and all higher levels within 2 km remains below this value [WMO, 1957].
INPUTS:
ff: the full path to the HRRS profile that we will load
vertical_res_km: vertical resolution of the grid to which we interpolate, in km. The default is 50m, which is roughly the vertical resolution of the HRRS obs.
"""
# read in the data as a data frame
DF0 = read_HRRS_data(ff)
# drop unnecessary columns
useless_cols = [
'Time', 'Dewpt', 'RH', 'Ucmp', 'Vcmp', 'spd', 'dir', 'Lat', 'Lon',
'Wcmp', 'Ele', 'Azi', 'Qp', 'Qt', 'Qrh', 'Qu', 'Qv', 'QdZ'
]
DF0.drop(useless_cols, inplace=True, axis=1)
# get rid of NaNs
DF = DF0.dropna()
if debug:
print('Loading file ' + ff)
# load interpolate function from scipy
from scipy.interpolate import interp1d
# compute the height of the lapse-tropopause from the altitude array
z = DF['Alt'] / 1E3 # Altitude in km
T = DF['Temp'] + 273.15 # Temp in Kelvin
P = DF['Press']
N2 = DF['N2']
from TIL import ztrop
ztropp = ztrop(z=z, T=T, debug=debug, hostname=hostname)
if ztropp is not None:
# extract the station number from the file path
file_components = ff.split('/')
station = file_components[len(file_components) - 2]
# retrieve the mean tropopause height for this station
# TODO: right noew this reads in a csv file of mean heights for Jan 2010.
# ...need to make this more dynamic and obviously not user and host specific
ZT = pd.read_csv(
'/data/c1/lneef/HRRS//mean_tropopause_height_per_station_20100101-20100131.csv',
index_col=0)
ztrop_mean = ZT.loc[int(station)].ztrop_mean
# now compute the altitude relative to the tropopause, plus mean tropopause height
zTP = DF['Alt'] * 1E-3 - ztropp + ztrop_mean
# interpolate temp, pressure to this new coordinate
fT = interp1d(zTP, T, kind='linear')
fP = interp1d(zTP, P, kind='linear')
# create a regularly spaced grid (in km)
zTPgrid = np.arange(0.0, 26.0, vertical_res_km)
# select whatever part of the regular grid fits into the range sampled by this sounding
select = np.where(
np.logical_and(zTPgrid > min(zTP), zTPgrid < max(zTP)))
zTPnew = zTPgrid[select]
# now compute the variables on this grid using the interpolate function
Tnew = fT(zTPnew)
Pnew = fP(zTPnew)
# N2 comes out quite noisy when computed from raw radiosonde observations.
# The spline (needed to get the obs on a common grid) is an opportunity for smoothing
# the temperature field a bit, which will yield a smoother N2 profile -- so just recompute
# N2 here
from TIL import Nsq
N2new = Nsq(Tnew, zTPnew, Pnew)
# now create a new dataframe with the TP-based heights
new_data = {
'Press': Pnew,
'Temp': Tnew,
'Alt': zTPnew,
'N2': N2new,
'ztropp': ztropp
}
Dout = pd.DataFrame(data=new_data)
else:
if debug:
print(
'No clear lapse-rate tropopause found for the following sounding:'
)
print(ff)
print('Returning None')
Dout = None
return (Dout)