n_ttas
except NameError:
n_ttas = {
'nhours1':{'w_events':[],'w_hours':[],
'w_czd':[],'w_bby':[],
'n_events':[],'n_hours':[],
'n_czd':[],'n_bby':[]},
}
for p in params:
print p
n_events = list()
n_hours = list()
r_czd = list()
r_bby = list()
for y in years:
tta = tta_analysis(y)
tta.start_df_layer(**p)
s = pd.Series(tta.tta_dates)
if s.size == 0:
ev = 0 # events
ho = 0 # hours
cz = 0
bb = 0
else:
sdiff = s-s.shift()
ev = np.sum( sdiff>h ) + 1
ho = tta.tta_hours
cz = tta.tta_rainfall_czd
bb = tta.tta_rainfall_bby
av_time = ho/float(ev)
print template.format(y,ev,ho,av_time,cz,bb)
def process(year=[],wdsurf=None,
wdwpro=None,rainbb=None,
raincz=None, nhours=None):
binss={'wdir':np.arange(0,370,10),
'wspd':np.arange(0,36,1)}
target = ['wdir','wspd']
arrays = {}
for t in target:
first = True
for y in year:
print('Processing year {}'.format(y))
' tta analysis '
tta = tta_analysis(y)
tta.start_df(wdir_surf=wdsurf,
wdir_wprof=wdwpro,
rain_bby=rainbb,
rain_czd=raincz,
nhours=nhours)
' retrieve dates '
include_dates = tta.include_dates
tta_dates = tta.tta_dates
notta_dates = tta.notta_dates
' read wprof '
wprof_df = parse_data.windprof(y)
wprof = wprof_df.dframe[t]
' wprof partition '
wprof = wprof.loc[include_dates] # all included
wprof_tta = wprof.loc[tta_dates] # only tta
wprof_notta = wprof.loc[notta_dates]# only notta
s1 = np.squeeze(pandas2stack(wprof))
s2 = np.squeeze(pandas2stack(wprof_tta))
s3 = np.squeeze(pandas2stack(wprof_notta))
if first:
wp = s1
wp_tta = s2
wp_notta = s3
first = False
else:
wp = np.hstack((wp,s1))
wp_tta = np.hstack((wp_tta,s2))
wp_notta = np.hstack((wp_notta, s3))
_,wp_hours = wp.shape
_,tta_hours = wp_tta.shape
_,notta_hours = wp_notta.shape
arrays[t]=[wp,wp_tta,wp_notta]
' makes CFAD '
hist_array_spd = np.empty((40,len(binss['wspd'])-1,3))
hist_array_dir = np.empty((40,len(binss['wdir'])-1,3))
cfad_array_spd = np.empty((40,len(binss['wspd'])-1,3))
cfad_array_dir = np.empty((40,len(binss['wdir'])-1,3))
average_spd = np.empty((40,3))
average_dir = np.empty((40,3))
median_spd = np.empty((40,3))
median_dir = np.empty((40,3))
for k,v in arrays.iteritems():
hist_array = np.empty((40,len(binss[k])-1,3))
cfad_array = np.empty((40,len(binss[k])-1,3))
average = np.empty((40,3))
median = np.empty((40,3))
wp = v[0]
wp_tta = v[1]
wp_notta = v[2]
for hgt in range(wp.shape[0]):
row1 = wp[hgt,:]
row2 = wp_tta[hgt,:]
row3 = wp_notta[hgt,:]
for n,r in enumerate([row1,row2,row3]):
' following CFAD Yuter et al (1995) '
freq,bins=np.histogram(r[~np.isnan(r)],
bins=binss[k])
hist_array[hgt,:,n] = freq
cfad_array[hgt,:,n] = 100.*(freq/float(freq.sum()))
bin_middle = (bins[1:]+bins[:-1])/2.
average[hgt,n] = np.sum(freq*bin_middle)/freq.sum()
median[hgt,n] = np.percentile(r[~np.isnan(r)],50)
if k == 'wspd':
hist_array_spd = hist_array
cfad_array_spd = cfad_array
average_spd = average
median_spd = median
else:
hist_array_dir = hist_array
cfad_array_dir = cfad_array
average_dir = average
median_dir = median
return [hist_array_spd,
hist_array_dir,
cfad_array_spd,
cfad_array_dir,
binss['wspd'],
binss['wdir'],
wprof_df.hgt,
wp_hours,
tta_hours,
notta_hours,
average_spd,
average_dir,
median_spd,
median_dir]
def processv2(year=[],wdsurf=None,
wdwpro=None,rainbb=None,
raincz=None, nhours=None):
''' v2: target loop moved into year loop '''
binss={'wdir': np.arange(0,370,10),
'wspd': np.arange(0,36,1),
'u': np.arange(-15,21,1),
'v': np.arange(-14,21,1),
}
target = ['wdir','wspd']
arrays = {}
wsp = np.empty((40,1))
wsp_tta = np.empty((40,1))
wsp_notta = np.empty((40,1))
wdr = np.empty((40,1))
wdr_tta = np.empty((40,1))
wdr_notta = np.empty((40,1))
for y in year:
print('Processing year {}'.format(y))
' tta analysis '
tta = tta_analysis(y)
tta.start_df(wdir_surf = wdsurf,
wdir_wprof = wdwpro,
rain_bby = rainbb,
rain_czd = raincz,
nhours = nhours)
' retrieve dates '
include_dates = tta.include_dates
tta_dates = tta.tta_dates
notta_dates = tta.notta_dates
' read wprof '
wprof_df = parse_data.windprof(y)
for n,t in enumerate(target):
wprof = wprof_df.dframe[t]
' wprof partition '
wprof = wprof.loc[include_dates] # all included
wprof_tta = wprof.loc[tta_dates] # only tta
wprof_notta = wprof.loc[notta_dates]# only notta
s1 = np.squeeze(pandas2stack(wprof))
if wprof_tta.size > 0:
s2 = np.squeeze(pandas2stack(wprof_tta))
ttaok = True
else:
ttaok =False
s3 = np.squeeze(pandas2stack(wprof_notta))
if t == 'wdir':
wdr = np.hstack((wdr,s1))
if ttaok is True:
if s2.ndim == 1:
s2=np.expand_dims(s2,axis=1)
wdr_tta = np.hstack((wdr_tta,s2))
wdr_notta = np.hstack((wdr_notta, s3))
else:
wsp = np.hstack((wsp,s1))
if ttaok is True:
if s2.ndim == 1:
s2=np.expand_dims(s2,axis=1)
wsp_tta = np.hstack((wsp_tta,s2))
wsp_notta = np.hstack((wsp_notta, s3))
arrays['wdir']=[wdr,wdr_tta,wdr_notta]
arrays['wspd']=[wsp,wsp_tta,wsp_notta]
uw = -wsp*np.sin(np.radians(wdr))
uw_tta = -wsp_tta*np.sin(np.radians(wdr_tta))
uw_notta = -wsp_notta*np.sin(np.radians(wdr_notta))
vw = -wsp*np.cos(np.radians(wdr))
vw_tta = -wsp_tta*np.cos(np.radians(wdr_tta))
vw_notta = -wsp_notta*np.cos(np.radians(wdr_notta))
arrays['u']=[uw,uw_tta,uw_notta]
arrays['v']=[vw,vw_tta,vw_notta]
''' total hours, first rows are empty '''
_,wp_hours = wsp.shape
_,tta_hours = wsp_tta.shape
_,notta_hours = wsp_notta.shape
wp_hours -= 1
tta_hours-= 1
notta_hours -= 1
' initialize arrays '
hist_array_spd = np.empty((40,len(binss['wspd'])-1,3))
hist_array_dir = np.empty((40,len(binss['wdir'])-1,3))
cfad_array_spd = np.empty((40,len(binss['wspd'])-1,3))
cfad_array_dir = np.empty((40,len(binss['wdir'])-1,3))
average_spd = np.empty((40,3))
average_dir = np.empty((40,3))
median_spd = np.empty((40,3))
median_dir = np.empty((40,3))
' loop for variable (wdir,wspd) '
for k,v in arrays.iteritems():
hist_array = np.empty((40,len(binss[k])-1,3))
cfad_array = np.empty((40,len(binss[k])-1,3))
average = np.empty((40,3))
median = np.empty((40,3))
' extract value'
wp = v[0]
wp_tta = v[1]
wp_notta = v[2]
' makes CFAD '
for hgt in range(wp.shape[0]):
row1 = wp[hgt,:]
row2 = wp_tta[hgt,:]
row3 = wp_notta[hgt,:]
for n,r in enumerate([row1,row2,row3]):
' following CFAD Yuter et al (1995) '
freq,bins=np.histogram(r[~np.isnan(r)],
bins=binss[k])
hist_array[hgt,:,n] = freq
cfad_array[hgt,:,n] = 100.*(freq/float(freq.sum()))
bin_middle = (bins[1:]+bins[:-1])/2.
average[hgt,n] = np.sum(freq*bin_middle)/freq.sum()
median[hgt,n] = np.percentile(r[~np.isnan(r)],50)
if k == 'wspd':
hist_array_spd = hist_array
cfad_array_spd = cfad_array
average_spd = average
median_spd = median
elif k == 'wdir':
hist_array_dir = hist_array
cfad_array_dir = cfad_array
average_dir = average
median_dir = median
elif k == 'u':
hist_array_u = hist_array
cfad_array_u = cfad_array
average_u = average
median_u = median
elif k == 'v':
hist_array_v = hist_array
cfad_array_v = cfad_array
average_v = average
median_v = median
return [hist_array_spd,
hist_array_dir,
hist_array_u,
hist_array_v,
cfad_array_spd,
cfad_array_dir,
cfad_array_u,
cfad_array_v,
binss['wspd'],
binss['wdir'],
binss['u'],
binss['v'],
wprof_df.hgt,
wp_hours,
tta_hours,
notta_hours,
average_spd,
average_dir,
average_u,
average_v,
median_spd,
median_dir,
median_u,
median_v,
]
# 'wdir_layer': [0,500],
# 'rain_czd': 0.25,
# 'nhours': 2
# } for a in [140,160]]
#
# params = params_nh + params_wth
try:
results
except NameError:
# results = {nparam:list() for nparam in range(len(params))}
results = {nparam: pd.DataFrame() for nparam in range(len(params))}
first = True
for y in years:
print y
tta = tta_analysis(y)
if first:
for n, p in enumerate(params):
tta.start_df_layer(**p)
# results[n] = tta.print_stats(return_results=True,
# skip_print=True)
results[n] = results[n].append(tta.df[tta.df.tta])
first = False
else:
for n, p in enumerate(params):
tta.start_df_layer(**p)
r = tta.print_stats(return_results=True, skip_print=True)
results[n] = np.vstack((results[n], r))
first = True
def params(wdir_surf=None,wdir_wprof=None,
wdir_thres=None,wdir_layer=None,
rain_bby=None, rain_czd=None,
nhours=None):
years = [1998]+range(2001,2013)
if wdir_surf is not None:
params = dict(wdir_surf = wdir_surf,
wdir_wprof = wdir_wprof,
rain_bby = rain_bby,
rain_czd = rain_czd,
nhours = nhours)
elif wdir_thres is not None:
params = dict(wdir_thres = wdir_thres,
wdir_layer = wdir_layer,
rain_bby = rain_bby,
rain_czd = rain_czd,
nhours = nhours)
print(params)
for y in years:
tta=tta_analysis(y)
if wdir_surf is not None:
tta.start_df(**params)
elif wdir_thres is not None:
tta.start_df_layer(**params)
if y == 1998:
results = tta.print_stats(header=True,return_results=True)
else:
r = tta.print_stats(return_results=True)
results = np.vstack((results,r))
''' print totals '''
print('-'*((6*13)+12))
yer_col = '{:6}'
bby_col = '{:7.0f}{:7.0f}{:7.0f}'
czd_col = '{:7.0f}{:7.0f}{:7.0f}'
rto_col = '{:6.1f}{:6.1f}'
hrs_col = '{:6.0f}{:6.0f}'
prc_col = '{:6.0f}{:6.0f}'
bby_total = results[:,0].sum()
bby_tta = results[:,1].sum()
bby_notta = results[:,2].sum()
czd_total = results[:,3].sum()
czd_tta = results[:,4].sum()
czd_notta = results[:,5].sum()
tta_ratio = czd_tta/bby_tta
notta_ratio = czd_notta/bby_notta
tta_hours = results[:,8].sum()
notta_hours = results[:,9].sum()
rain_perc_bby = np.round(100.*(bby_tta/bby_total),0).astype(int)
rain_perc_czd = np.round(100.*(czd_tta/czd_total),0).astype(int)
col0 = yer_col.format('')
col1 = bby_col.format(bby_total, bby_tta, bby_notta)
col2 = czd_col.format(czd_total, czd_tta, czd_notta)
col3 = rto_col.format(tta_ratio, notta_ratio)
col4 = hrs_col.format(tta_hours, notta_hours)
col5 = prc_col.format(rain_perc_bby, rain_perc_czd)
print(col0+col1+col2+col3+col4+col5)
U = pd.Series(name='U-comp')
V = pd.Series(name='V-comp')
try:
results
except NameError:
results = collections.OrderedDict()
for tr in thres:
results[tr] = {'U': U, 'V': V}
for year in years:
for p in params:
tta = tta_analysis(year=year)
tta.start_df_layer(**p)
tta_dates = tta.tta_dates
" parse surface and profile obs "
bby = parse_data.surface('bby', year=year)
wpr = parse_data.windprof(year=year)
wpr_tta = wpr.dframe.loc[tta_dates]
wdr_tta = wpr_tta['wdir']
wsp_tta = wpr_tta['wspd']
bby_tta = bby.dframe.loc[tta_dates]
" append surface values to windprof "