def load_BCs(fields, season=None, timesel='0011-01-01,0011-12-31'):
simcbase='kemctl1'
simpbase='kem1pert2'
obsbase = 'kemnsidc'
flddat={}; fldcdat={}; fldpdat={}; sicnpdat={}
for fii,field in enumerate(fields):
fnamesc={}; fnamesp={}
flddiff={}; fldcavg={}; fldpavg={}
print field
# Loop through sim data:
# load all 5 sim BCs
for ii in range(0,6):
# do obs data at ii=5
if ii==5:
simname=obsbase+'ctl'
fnamec = con.getBCfilenames(field,sim=simname)
print fnamec
simname=obsbase+'pert'
fnamep = con.getBCfilenames(field,sim=simname)
else:
simname=simcbase+'r'+str(ii+1)
fnamec = con.getBCfilenames(field,sim=simname)
print fnamec
simname=simpbase+'r'+str(ii+1)
fnamep = con.getBCfilenames(field,sim=simname)
fldc = cnc.getNCvar(fnamec,field.upper(),timesel=timesel)
fldp = cnc.getNCvar(fnamep,field.upper(),timesel=timesel)
if field=='sicn':
# save pert data to mask gt
if season==None:
sicnpdat[ii] = fldp
else:
sicnpdat[ii] = cutl.seasonalize_monthlyts(fldp,season=sea,climo=1)
if season==None:
flddiff[ii] = fldp-fldc
fldcavg[ii] = fldc
fldpavg[ii] = fldp
else:
flddiff[ii] = cutl.seasonalize_monthlyts(fldp-fldc,season=sea,climo=1)
fldcavg[ii] = cutl.seasonalize_monthlyts(fldc,season=sea,climo=1)
fldpavg[ii] = cutl.seasonalize_monthlyts(fldp,season=sea,climo=1)
fnamesc[ii]=fnamec
fnamesp[ii]=fnamep
flddat[field] = flddiff
fldcdat[field] = fldcavg
fldpdat[field] = fldpavg
return flddat, fldcdat, fldpdat, sicnpdat
def load_pifield(fdict,seas,conv=1,subsampyrs=11,numsamp=50,
styear=None,anomyears=None,local=False,detrend=True,
verb=False,addcyc=False):
""" TAKEN FROM load_canesmfield() in canesm_LE_composite.py
loads subsampled CGCM data from specified simulation
(assumes a long control, e.g. piControl)
number of total chunks will be determined by length of sim data
and numyrs (ntime / numyrs). First the simulation is chunked into
numyrs segments. Then 2 segments at a time are randomly chosen
(at least a decade apart) to generate anomalies. This is done
numsamp times (e.g. 50)
returns subsamp (subsampled anomalies)
styears (start index of chunking of long run)
anomyears (indices of anomaly differences)
"""
casename='piControl'
pidat = lcd.load_data(fdict,casename,local=local,conv=conv,verb=verb)
piseadat = cutl.seasonalize_monthlyts(pidat,season=seas)
if detrend:
piseadat = cutl.detrend(piseadat,axis=0)
# the data must be seasonalized before using this func.
pisea,styear,anomyears = leg.subsamp_anom_pi(piseadat, numyrs=subsampyrs,numsamp=numsamp,
styear=styear,anomyears=anomyears)
if addcyc:
st=pisea[...,-1]
pisea=np.dstack((pisea,st[...,None]))
return pisea,styear,anomyears
def load_field(fdict,casename,timesel,seas,ftype='fullts',conv=1,local=False,verb=False):
""" TAKEN FROM load_field() in canesm_le_composite.py
returns [numens x space.flat] or [numens]
"""
ledat = le.load_LEdata(fdict,casename,timesel=timesel,
rettype='ndarray',conv=conv,ftype=ftype,local=local,verb=verb)
# time needs to be first dimension
try:
if ledat.ndim==2:
ledat = ledat.T
elif ledat.ndim==3:
ledat = np.transpose(ledat,(1,0,2))
else:
print 'Loaded data is not 2 or 3 dimensions. Do not understand.'
raise Exception
except:
raise
lesea = cutl.seasonalize_monthlyts(ledat,season=seas).mean(axis=0) # numens x space.flat
return lesea
def load_field(fdict,casename,timesel,seas,ftype='fullts',conv=1,local=False,verb=False):
"""
returns [numens x space.flat] or [numens]
"""
ledat = le.load_LEdata(fdict,casename,timesel=timesel,
rettype='ndarray',conv=conv,ftype=ftype,local=local,verb=verb)
print '@@@ ledat.shape ' + str(ledat.shape) # why does the 3d data have space flattened already...
# time needs to be first dimension
try:
if ledat.ndim==2:
ledat = ledat.T
elif ledat.ndim==3:
ledat = np.transpose(ledat,(1,0,2))
else:
print 'Loaded data is not 2 or 3 dimensions. Do not understand.'
raise Exception
except:
raise
lesea = cutl.seasonalize_monthlyts(ledat,season=seas).mean(axis=0) # numens x space.flat
return lesea
def plotvert_allseas(fld, lev, lat,title='',units='',cmap='blue2red_w20',ptype=None,
cmin='',cmax='',axis=None, suppcb=0,climo=0,
pvals = None,sigtype='hatch',latlim=None,levlim=None,
addcontlines=True,screen=False):
""" default seasons = 'SON','DJF','MAM','JJA'
"""
import cccmautils as cutl
seasons = 'SON','DJF','MAM','JJA'
midx=0
fig,axs = plt.subplots(1,4)
fig.set_size_inches(12,4)
fig.subplots_adjust(hspace=.15,wspace=.05)
for axii,ax in enumerate(axs.flat):
#print seasons[midx]
tmpfld=np.squeeze(cutl.seasonalize_monthlyts(fld,season=seasons[midx],climo=climo))
#print tmpfld.shape # already 2D... only when climo=1?
#if seasons[midx]=='DJF':
# print tmpfld
if climo:
plotfld = tmpfld
else:
plotfld = np.mean(tmpfld,axis=0)
#print plotfld.shape
if axii==0:
vpparams = dict(cmin=cmin,cmax=cmax,cmap=cmap,ptype=ptype,
axis=ax,suppcb=True,units=units,
latlim=latlim,levlim=levlim,
addcontlines=addcontlines,screen=screen)
else:
vpparams = dict(cmin=cmin,cmax=cmax,cmap=cmap,ptype=ptype,
axis=ax,suppcb=True,units=units,
latlim=latlim,levlim=levlim,
addcontlines=addcontlines,screen=screen,
suppylab=True)
cf = vert_plot(plotfld,lev,lat,**vpparams)
ax.set_title(seasons[midx])
if pvals != None:
#addtsig(cf,pvals,lat,lon,sigtype=sigtype) # @@@ untested? no lat... 5/10/15
addtsig(cf,pvals,lat,lev,sigtype=sigtype)
midx = midx+1
cbar_ax = fig.add_axes([.91,.25, .02,.5])
fig.colorbar(cf,cax=cbar_ax)
plt.suptitle(title)
return fig
def map_allseas(fld, lat, lon,title='',units='',cmap='blue2red_w20',ptype='sq',
cmin='',cmax='',axis=None, suppcb=0,lmask=0,climo=0,flipmask=0,
pvals = None,sigtype='hatch',conts=None,latlim=None):
""" default seasons = 'SON','DJF','MAM','JJA'
"""
import cccmautils as cutl
seasons = 'SON','DJF','MAM','JJA'
midx=0
fig,axs = plt.subplots(1,4)
fig.set_size_inches(12,3)
fig.subplots_adjust(hspace=.15,wspace=.05)
for ax in axs.flat:
#print seasons[midx]
tmpfld=np.squeeze(cutl.seasonalize_monthlyts(fld,season=seasons[midx],climo=climo))
#print tmpfld.shape # already 2D... only when climo=1?
#if seasons[midx]=='DJF':
# print tmpfld
if climo:
plotfld = tmpfld
else:
plotfld = np.mean(tmpfld,axis=0)
#print plotfld.shape
bm,pc = kemmap(plotfld,lat,lon,cmin=cmin,cmax=cmax,cmap=cmap,ptype=ptype,\
axis=ax,suppcb=1,lmask=lmask,flipmask=flipmask,units=units,latlim=latlim)
ax.set_title(seasons[midx])
if pvals != None:
addtsigm(bm,pvals,lat,lon,sigtype=sigtype)
if conts != None:
# add specified contour(s)
lons, lats = np.meshgrid(lon,lat)
bm.contour(lons,lats,plotfld,[conts],colors='k',linewidths=2,latlon=True)
midx = midx+1
cbar_ax = fig.add_axes([.91,.25, .02,.5])
fig.colorbar(pc,cax=cbar_ax)
plt.suptitle(title)
return fig
) # time start year to end
else:
fldc = ncfilec.variables[ncfield][...] * conv
fldp1 = ncfilep1.variables[ncfield][...] * conv
fldp2 = ncfilep2.variables[ncfield][...] * conv # time start year to end
fldp3 = ncfilep3.variables[ncfield][...] * conv # time start year to end
nt, nlev, nlat = fldc.shape # if nt == 12 then it's a climo
# ############# set which perturbation run! ######
casenamep = casenamep2
fldp = fldp2
# ###############################################
seasfldc = np.squeeze(cutl.seasonalize_monthlyts(fldc, timeavg, climo=climo))
seasfldp = np.squeeze(cutl.seasonalize_monthlyts(fldp, timeavg, climo=climo))
print seasfldc.shape
lats, levs = np.meshgrid(lat, lev)
tstat, pval = sp.stats.ttest_ind(seasfldp, seasfldc, axis=0)
print tstat.shape
print timeavg
pval = ma.masked_invalid(pval) # pcolormesh needs masked_array
# SEASONAL MEAN CLIMO (CONTROL)
plotfld = np.average(seasfldc, 0)
fldczmsea = np.zeros((len(seasons),len(lat)))
fldp1zmsea = np.zeros((len(seasons),len(lat)))
fldp2zmsea = np.zeros((len(seasons),len(lat)))
fldp3zmsea = np.zeros((len(seasons),len(lat)))
tstatp1zmsea = np.zeros((len(seasons),len(lat)))
pvalp1zmsea = np.zeros((len(seasons),len(lat)))
tstatp2zmsea = np.zeros((len(seasons),len(lat)))
pvalp2zmsea = np.zeros((len(seasons),len(lat)))
tstatp3zmsea = np.zeros((len(seasons),len(lat)))
pvalp3zmsea = np.zeros((len(seasons),len(lat)))
# seasonalize first, then zonal mean
for sidx in range(0,len(seasons)):
fldczm = np.mean(cutl.seasonalize_monthlyts(fldc,season=seasons[sidx]), axis=2) # zm with time dim
fldp1zm = np.mean(cutl.seasonalize_monthlyts(fldp1,season=seasons[sidx]), axis=2)
fldp2zm = np.mean(cutl.seasonalize_monthlyts(fldp2,season=seasons[sidx]), axis=2)
fldp3zm = np.mean(cutl.seasonalize_monthlyts(fldp3,season=seasons[sidx]), axis=2)
tstatp1zm,pvalp1zm = sp.stats.ttest_ind(fldp1zm,fldczm,axis=0)
tstatp2zm,pvalp2zm = sp.stats.ttest_ind(fldp2zm,fldczm,axis=0)
tstatp3zm,pvalp3zm = sp.stats.ttest_ind(fldp3zm,fldczm,axis=0)
# now get the seasonal average (season x lat)
fldczmsea[sidx,:] = np.mean(fldczm,axis=0)
fldp1zmsea[sidx,:] = np.mean(fldp1zm,axis=0)
fldp2zmsea[sidx,:] = np.mean(fldp2zm,axis=0)
fldp3zmsea[sidx,:] = np.mean(fldp3zm,axis=0)
pvalp1zmsea[sidx,:] = pvalp1zm
cutl.annualize_monthlyts(nfldc),color=dodgerblue,linewidth=2)
plt.xlim(xlims)
plt.title('ANN NH SIA anom from 1979-89')
plt.grid()
if printtofile:
plt.savefig('CanESMens_OBSanom' + ctimstr + '_' + cfield + '_ANN_timeseries' + pstr + '.pdf')
# MINIMUM (SEPT)
plt.figure()
cii=0.25
#mosel=12; mostr='Dec'
mosel=9; mostr='Sep'
for ii in xrange(0,5):
plotfld = cutl.seasonalize_monthlyts(cfldpall[ii,:],mo=mosel)
plt.plot(years,plotfld,color=str(cii))
if zoom:
slope, intercept, r_value, p_value, std_err = sp.stats.linregress(hyears,plotfld[129:])
plt.plot(years,slope*years+intercept,color=str(cii))
cii=cii+0.15
plotfld = cutl.seasonalize_monthlyts( np.mean(cfldpall,axis=0),mo=mosel )
plt.plot(years,plotfld,color='k',linewidth=2)
if zoom:
slope, intercept, r_value, p_value, std_err = sp.stats.linregress(hyears,plotfld[129:])
plt.plot(years,slope*years+intercept,color='k',linewidth=2)
if afield=='sicn':
plotfld=cutl.seasonalize_monthlyts(hfldp,mo=mosel)
plt.plot(hyears,plotfld,color='green',linewidth=2)
fname=fdict[fkey]
conv=condt[fkey]
fld=cnc.getNCvar(fname,ncfield,timesel=timesel)*conv
fld=clear_belowsfc(fld, sfcp, lev)
print fname + ', fld.shape ' + str(fld.shape) # @@@
fldint=np.zeros(len(lat))
fldintza=np.zeros(len(lat))
latidx=0
for ll in lat: # do all lats
#latrad = np.deg2rad(ll)
fldsub = np.squeeze(fld[:,:,latidx,:-1]) # get just one lat and remove extra lon
fldtm = np.mean(cutl.seasonalize_monthlyts(fldsub,**seasonalizedt),axis=0) # time mean
#print fldtm
#fldint[latidx] = calc_PHT(fldtm, latidx, nlon, dp, calc=calctype)
fldint[latidx] = calc_PHT3(fldtm, ll, nlon, dp, vcalc=calctype, zcalc='zonint') # zonal integration
fldintza[latidx] = calc_PHT3(fldtm, ll, nlon, dp, vcalc=calctype, zcalc='zonavg') # zonal average
latidx += 1
fldintdt[fkey] = fldint
fldintzadt[fkey] = fldintza
flddf = pd.DataFrame(fldintdt)
fldtot = flddf.sum(axis=1)
fldzadf = pd.DataFrame(fldintzadt)
lefield1='zg50000.00'; lencfield1='zg'; comp1='Amon'; leconv1=1 #region1='bksmori'
lefield2='tas'; lencfield2='tas'; comp2='Amon'; #region2='eurasiamori'
fdict1 = {'field': lefield1+region1, 'ncfield': lencfield1, 'comp': comp1}
fdict2 = {'field': lefield2+region2, 'ncfield': lencfield2, 'comp': comp2}
ftype='fullts' # 'fullclimo' or 'climo' or 'fullts'
lefdict1 = {'field': lefield1+region1, 'ncfield': lencfield1, 'comp': comp1}
lefdict2 = {'field': lefield2+region2, 'ncfield': lencfield2, 'comp': comp2}
# historical
lecdat1 = le.load_LEdata(fdict1,'historical',timesel=timeselc, rettype='ndarray',conv=leconv1,ftype=ftype)
(numens1,ntime1) = lecdat1.shape
lepdat1=le.load_LEdata(fdict1,'historical',timesel=timeselp, rettype='ndarray',conv=leconv1,ftype=ftype)
lecsea1 = cutl.seasonalize_monthlyts(lecdat1.T,season=sea1).T
lepsea1 = cutl.seasonalize_monthlyts(lepdat1.T,season=sea1).T
lefld1=lepsea1.mean(axis=1)-lecsea1.mean(axis=1)
lecdat2 = le.load_LEdata(fdict2,'historical',timesel=timeselc, rettype='ndarray',conv=conv2,ftype=ftype)
(numens2,ntime2) = lecdat1.shape
lepdat2=le.load_LEdata(fdict2,'historical',timesel=timeselp, rettype='ndarray',conv=conv2,ftype=ftype)
lecsea2 = cutl.seasonalize_monthlyts(lecdat2.T,season=sea2).T
lepsea2 = cutl.seasonalize_monthlyts(lepdat2.T,season=sea2).T
lefld2=lepsea2.mean(axis=1)-lecsea2.mean(axis=1)
obs=plt.scatter(erareg.mean(),gisreg.mean(),color='blue',marker='s',s=8**2)
ledat=plt.scatter(lefld1,lefld2,color=ccm.get_linecolor('darkolivegreen3'),marker='*',s=5**2,alpha=0.5)
lemm, lebb, lerval, lepval, lestd_err = sp.stats.linregress(lefld1,lefld2)
axylims = ax.get_ylim()
superslopes[monii,superii] = slope[eii]
thismonskey = thismonskey + (mod,)
#superkeys[monii,superii] = mod
superii=superii+1
modtrenddt[mod] = slope # trends for e/ ens member in the model group
mntrnd = np.min(slope)
mxtrnd = np.max(slope)
avgtrnd = np.mean(slope)
mndt[mod] = mntrnd
mxdt[mod] = mxtrnd
avgdt[mod] = avgtrnd
# do obs e/ month
homon = cutl.seasonalize_monthlyts(had,mo=monii+1)
xx = np.arange(0,homon.shape[0])
hslope, intercept, r_value, p_value, std_err = sp.stats.linregress(xx,homon)
hmontrenddt[mo] = hslope
nomon = cutl.seasonalize_monthlyts(nsidc,mo=monii+1)
xx = np.arange(0,nomon.shape[0])
nslope, intercept, r_value, p_value, std_err = sp.stats.linregress(xx,nomon)
nmontrenddt[mo] = nslope
superkeys = superkeys + (thismonskey,)
montrenddt[mo] = modtrenddt
monmndt[mo]= mndt
monmxdt[mo]= mxdt
monavgdt[mo] = avgdt
flddiff=flddat[field]
for ii in range(0,6): # 6 sims
if len(fields)==1:
ax=axs.flatten()[ii]
else:
if fliptohoriz:
ax=axs[fii,ii]
else:
ax=axs[ii,fii]
plotfld=flddiff[ii]*conv
plotfld = cutl.seasonalize_monthlyts(plotfld,season=sea, climo=1)
if field=='gt' and masksicn:
tmp = sicnpdat[ii]
tmp = cutl.seasonalize_monthlyts(tmp,season=sea, climo=1)
plotfld[tmp>=0.15] = 0 # mask where model sees grid cell as sea ice (not SST)
bm,pc = cplt.kemmap(plotfld, lat, lon, ptype='nheur',
latlim=latlim, axis=ax, suppcb=True,
round=False,lcol=pparams['lcol'], coastres= 'c',
coastwidth= 0.5, area_thresh=70000,
lmask=True,**metap)
if len(fields)==1:
ax.set_title(ylabs[ii],fontsize=fsz)
else:
