def plotTimeSerieVariabilityAtWl_mean(ax=None, axmap=None, plotXLabel=True, plotYLabel=True, warmingLev=2.0, bsmp=None):
yrs, _, _, r8srs_, r4srs_ = loadWlVsScenChange.loadMeanChangesAtWl_nYearsAroundWLYear(warmingLev=warmingLev, nmodels=nmodels, excludedModels=excludedModels)
r8srs = r8srs_[:, [3], :, :]
r4srs = r4srs_[:, [3], :, :]
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
srstot = np.concatenate([r8srs, r4srs], 0)
rallmeanWl, rallsigmaWl = getMeanAndSigma(r8srs)
cndPosi = rallmeanWl > 0
cndNega = rallmeanWl < 0
cnd = cndPosi if positiveChanges else cndNega
r8meanWl, r8sigmaWl = getMeanAndSigma(r8srs)
cndMtx = np.tile(cnd, [r8srs.shape[0], r8srs.shape[1], 1, 1])
r8srs[~cndMtx] = np.nan
r8mean_ = np.nanmean(r8srs, 0)
r8mean = r8mean_.reshape(r8mean_.shape[0], r8mean_.shape[1]*r8mean_.shape[2])
#r8mean = np.nanmedian(r8mean, 1)
r8mean = np.nanmean(r8mean, 1)
r8spmean = r8srs.reshape(r8srs.shape[0], r8srs.shape[1], r8srs.shape[2]*r8srs.shape[3])
r8spmean = np.nanmean(r8spmean, 2)
r8skew = stats.skew(r8spmean[:, 0])
r8sigma = np.std(r8spmean[:, 0])
r8mean = np.nanmean(r8spmean, 0)
r4meanWl, r4sigmaWl = getMeanAndSigma(r4srs)
cndMtx = np.tile(cnd, [r4srs.shape[0], r4srs.shape[1], 1, 1])
r4srs[~cndMtx] = np.nan
r4mean_ = np.nanmean(r4srs, 0)
r4mean = r4mean_.reshape(r4mean_.shape[0], r4mean_.shape[1]*r4mean_.shape[2])
#r4mean = np.nanmedian(r4mean, 1)
r4mean = np.nanmean(r4mean, 1)
r4spmean = r4srs.reshape(r4srs.shape[0], r4srs.shape[1], r4srs.shape[2]*r4srs.shape[3])
r4spmean = np.nanmean(r4spmean, 2)
r4skew = stats.skew(r4spmean[:, 0])
r4sigma = np.std(r4spmean[:, 0])
r4mean = np.nanmean(r4spmean, 0)
ensTot = np.concatenate([r8spmean, r4spmean], 0)
stdDevTot = np.std(ensTot, 0)*100.
meanTot = (r4mean + r8mean)/2.*100.
yr = [-15, -10, -5, 0, 5, 10, 15]
if ax is None:
fg = plt.figure(figsize=[8.46, 4.98])
ax = fg.gca()
doPlotChanges(ax, warmingLev, meanTot, stdDevTot, r8spmean, r4spmean, r8mean, r4mean, plotXLabel, plotYLabel)
if not axmap is None:
bsmp = doPlotAreaMap(axmap, cnd, bsmp)
plt.tight_layout()
if ax is None:
fg.savefig('./ensemblesVariabilityAround2deg_mean.png', dpi=400)
return bsmp
def plotAllModelsWarmingLevel(warmingLev, rootDir='/ClimateRun4/multi-hazard/eva/'):
rcp8, rcp4 = 'rcp85', 'rcp45'
retPer = 100
models = """
IPSL-INERIS-WRF331F_BC
SMHI-RCA4_BC_CNRM-CERFACS-CNRM-CM5
SMHI-RCA4_BC_ICHEC-EC-EARTH
SMHI-RCA4_BC_IPSL-IPSL-CM5A-MR
SMHI-RCA4_BC_MOHC-HadGEM2-ES
SMHI-RCA4_BC_MPI-M-MPI-ESM-LR
CLMcom-CCLM4-8-17_BC_CNRM-CERFACS-CNRM-CM5
CLMcom-CCLM4-8-17_BC_ICHEC-EC-EARTH
CLMcom-CCLM4-8-17_BC_MPI-M-MPI-ESM-LR
DMI-HIRHAM5-ICHEC-EC-EARTH_BC
KNMI-RACMO22E-ICHEC-EC-EARTH_BC
"""
models = models.split()
wlYearR8 = getWarmingLevels(rcp8, warmingLev)
wlYearR4 = getWarmingLevels(rcp4, warmingLev)
outputPngFile = 'scatter_100yrlChange_wl' + str(warmingLev) + '.png'
nrow = 4
ncol = 3
fig = plt.figure(figsize=(8, 9.5))
gs = gridspec.GridSpec(nrow, ncol)
lon, lat = [], []
relChng8lst, relChng4lst = [], []
mp = None
for mdl, imdl in zip(models, range(1, len(models) + 1)):
irow = imdl // ncol
icol = imdl % ncol
print('plotting model ' + mdl + ' at ' + str(irow) + ', ' + str(icol))
ax = plt.subplot(gs[irow, icol])
ncFlNm = '_'.join(['projection_dis', rcp8, mdl, 'wuChang', 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR8[mdl]
wrmYear = int(round(wrmYear/5.)*5.)
relChngR8 = getRelChngs(ncFlPth, retPer, wrmYear)
ncFlNm = '_'.join(['projection_dis', rcp4, mdl, 'wuChang', 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR4[mdl]
wrmYear = int(round(wrmYear/5.)*5.)
relChngR4 = getRelChngs(ncFlPth, retPer, wrmYear)
plotSingleModel(ax, relChngR8, relChngR4, mdl)
relChng8lst.append(relChngR8)
relChng4lst.append(relChngR4)
ax = plt.subplot(gs[0, 0])
relChng8 = np.nanmean(np.array(relChng8lst), 0)
relChng4 = np.nanmean(np.array(relChng4lst), 0)
plotSingleModel(ax, relChng8, relChng4, 'Ensemble rcp85 vs rcp45,\nwrm. lev. ' + str(warmingLev) + '$^\circ$', bold=True, plotRegression=True)
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def plotAllModelsWarmingLevel(warmingLev,
rootDir='/ClimateRun4/multi-hazard/eva/'):
rcp8, rcp4 = 'rcp85', 'rcp45'
retPer = 100
models = modelStr.split()
wlYearR8 = getWarmingLevels(rcp8, warmingLev)
wlYearR4 = getWarmingLevels(rcp4, warmingLev)
outputPngFile = 'scatter_100yrlChange_wl' + str(warmingLev) + '.png'
nrow = 4
ncol = 3
fig = plt.figure(figsize=(8, 9.5))
gs = gridspec.GridSpec(nrow, ncol)
lon, lat = [], []
relChng8lst, relChng4lst = [], []
mp = None
vlsMask = None
for mdl, imdl in zip(models, range(1, len(models) + 1)):
irow = imdl // ncol
icol = imdl % ncol
print('plotting model ' + mdl + ' at ' + str(irow) + ', ' + str(icol))
ax = plt.subplot(gs[irow, icol])
ncFlNm = '_'.join(
['projection_dis', rcp8, mdl, wuStr, 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR8[mdl]
wrmYear = int(round(wrmYear / 5.) * 5.)
relChngR8, vlsMask = getRelChngs(ncFlPth, retPer, wrmYear, vlsMask)
ncFlNm = '_'.join(
['projection_dis', rcp4, mdl, wuStr, 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR4[mdl]
wrmYear = int(round(wrmYear / 5.) * 5.)
relChngR4, vlsMask = getRelChngs(ncFlPth, retPer, wrmYear, vlsMask)
plotSingleModel(ax, relChngR8, relChngR4, mdl)
relChng8lst.append(relChngR8)
relChng4lst.append(relChngR4)
ax = plt.subplot(gs[0, 0])
relChng8 = np.nanmean(np.array(relChng8lst), 0)
relChng4 = np.nanmean(np.array(relChng4lst), 0)
plotSingleModel(ax,
relChng8,
relChng4,
str(retPer) +
'-year ret. lev.\nensemble rcp85 vs rcp45,\nwrm. lev. ' +
str(warmingLev) + '$^\circ$C',
bold=True,
plotRegression=True)
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def extract1Model(inputNcFlPath, outputNcFlPath):
outRetPer = [30.]
outWarmingLevs = [1.5, 2.0, 3.0]
m = re.match('(.*)/lonlat_projection_windYMax_(.*)_(.*).nc', inputNcFlPath)
scen = m.group(1)[0]
mdl = m.group(2)[0]
wls = [getWarmingLevels(scen, wl) for wl in outWarmingLevs]
print(' loading the data ...')
dsin = netCDF4.Dataset(inputNcFlPath)
x = dsin.variables['lon'][:]
y = dsin.variables['lat'][:]
inRetPer = dsin.variables['retper'][:]
tmnc = dsin.variables['time']
tm = netCDF4.num2date(tmnc[:], tmnc.units, tmnc.calendar)
ystart = tm[0].year
rlYrs = ystart + np.arange(len(tm))
dsin.close()
output = []
for wl in outWarmingLevs:
wlyrs = wls[wl]
if wlyrs is None:
continue
wlyr = wlyrs[mdl]
iyr = np.where(rlYrs == wlyr)
wlRetPer =
def doCreateAllFiles(diagnosticsDataDir, outdir):
scenarios = ['rcp85', 'rcp45']
wls = getWarmingLevels.getWarmingLevels('rcp85', 1.5)
models = wls.keys()
for scn in scenarios:
for mdl in models:
generateFrqFileForModel(diagnosticsDataDir, scn, mdl, outdir)
def doLoadDiagnosticData(diagnosticDataDir, warmingLev):
import getWarmingLevels
print(' loading data at ' + str(warmingLev) + '°')
fls = [
f for f in os.listdir(diagnosticsDataDir) if re.match('(.*).npy', f)
]
fls.sort()
mtchs = [
re.match('droughtFrqDiagnostics_(.*)_(.*)_(.*)\.npy', f) for f in fls
]
mdls_ = [m.groups(0)[0] for m in mtchs]
scns_ = [m.groups(0)[1] for m in mtchs]
yrs = [int(float(m.groups(0)[2])) for m in mtchs]
dtprot = np.load(os.path.join(diagnosticsDataDir, fls[0]))
nx, ny = dtprot.shape
mns = []
mdls = set(mdls_)
scns = set(scns_)
for scn in scns:
if warmingLev == 0:
wls = {}
else:
wls = getWarmingLevels.getWarmingLevels(scn, warmingLev)
if wls is None:
for mdl in mdls:
mns.append(np.zeros([nx, ny]) * np.nan)
continue
for mdl in mdls:
print(' loading model / scenario == ' + mdl + ' / ' + scn)
if warmingLev == 0:
ywl = 1995
else:
ywl = wls[mdl]
ymin = max(ywl - 15, 1981)
ymax = min(ywl + 14, 2100)
dt = np.zeros([30, nx, ny]) * np.nan
yrsMdl = range(ymin, ymax + 1)
for iyr, yr in zip(range(len(yrsMdl)), yrsMdl):
flPth = os.path.join(
diagnosticsDataDir, '_'.join(
['droughtFrqDiagnostics', mdl, scn,
str(yr)])) + '.0.npy'
if not os.path.isfile(flPth):
print(' ... file ' + flPth +
' not found, skipping ...')
continue
dt[iyr] = np.load(flPth)
mn = np.nanmedian(dt, 0)
#mn = np.nanmean(dt, 0)
mns.append(mn)
mns = np.array(mns)
frq = np.nanmedian(mns, 0)
retper = 1 / frq
return retper, 1 / mns
def plotAllModelsWarmingLevel(scenario, warmingLev, rootDir='/ClimateRun4/multi-hazard/eva/'):
models = """
IPSL-INERIS-WRF331F_BC
SMHI-RCA4_BC_CNRM-CERFACS-CNRM-CM5
SMHI-RCA4_BC_ICHEC-EC-EARTH
SMHI-RCA4_BC_IPSL-IPSL-CM5A-MR
SMHI-RCA4_BC_MOHC-HadGEM2-ES
SMHI-RCA4_BC_MPI-M-MPI-ESM-LR
CLMcom-CCLM4-8-17_BC_CNRM-CERFACS-CNRM-CM5
CLMcom-CCLM4-8-17_BC_ICHEC-EC-EARTH
CLMcom-CCLM4-8-17_BC_MPI-M-MPI-ESM-LR
DMI-HIRHAM5-ICHEC-EC-EARTH_BC
KNMI-RACMO22E-ICHEC-EC-EARTH_BC
"""
models = models.split()
wlYear = getWarmingLevels(scenario, warmingLev)
outputPngFile = '100yrlChange_' + scenario + '_wl' + str(warmingLev) + '.png'
nrow = 4
ncol = 3
fig = plt.figure(figsize=(8, 9))
gs = gridspec.GridSpec(nrow, ncol + 1, width_ratios=[1,1,1,.1])
lon, lat = [], []
bslnlist, projlist = [], []
mp = None
for mdl, imdl in zip(models, range(1, len(models) + 1)):
irow = imdl // ncol
icol = imdl % ncol
print('plotting model ' + mdl + ' at ' + str(irow) + ', ' + str(icol))
ax = plt.subplot(gs[irow, icol])
wrmYear = wlYear[mdl]
wrmYear = int(round(wrmYear/5.)*5.)
ncFlNm = '_'.join(['projection_dis', scenario, mdl, 'wuChang', 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
mdlname = mdl.replace('BC_', '').replace('_BC', '')
lon, lat, bsln, proj, pcl, mp = plotSingleModel(ax, ncFlPth, mdlname, mp=mp, projYear=wrmYear)
bslnlist.append(bsln)
projlist.append(proj)
ax = plt.subplot(gs[0, 0])
bsln = np.nanmean(np.array(bslnlist), 0)
proj = np.nanmean(np.array(projlist), 0)
projVar = (proj - bsln)/bsln*100
plotSingleModelVar(ax, 'Ensemble ' + scenario + ', warming lev. ' + str(warmingLev) + '$^\circ$', lon, lat, projVar, mp, bold=True)
ax = plt.subplot(gs[:, 3])
#ax.get_xaxis().set_visible(False)
#ax.get_yaxis().set_visible(False)
cb = plt.colorbar(pcl, cax=ax)
cb.set_label(label='100-year event change (%)', fontsize=12)
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def plotEnsemblesWarmingLevel(warmingLev=2.0, rootDir='/ClimateRun4/multi-hazard/eva/'):
scenarios = ['rcp85', 'rcp45']
models = """
IPSL-INERIS-WRF331F_BC
SMHI-RCA4_BC_CNRM-CERFACS-CNRM-CM5
SMHI-RCA4_BC_ICHEC-EC-EARTH
SMHI-RCA4_BC_IPSL-IPSL-CM5A-MR
SMHI-RCA4_BC_MOHC-HadGEM2-ES
SMHI-RCA4_BC_MPI-M-MPI-ESM-LR
CLMcom-CCLM4-8-17_BC_CNRM-CERFACS-CNRM-CM5
CLMcom-CCLM4-8-17_BC_ICHEC-EC-EARTH
CLMcom-CCLM4-8-17_BC_MPI-M-MPI-ESM-LR
DMI-HIRHAM5-ICHEC-EC-EARTH_BC
KNMI-RACMO22E-ICHEC-EC-EARTH_BC
"""
models = models.split()
outputPngFile = '100yrlChange_ensembles_wl.png'
fig = plt.figure(figsize=(10, 4))
gs = gridspec.GridSpec(1, 3, width_ratios=[1,1,.1])
lon, lat = [], []
mp = None
for scenario, iscen in zip(scenarios, range(len(scenarios))):
bslnlist, projlist = [], []
wlYear = getWarmingLevels(scenario, warmingLev)
for mdl, imdl in zip(models, range(1, len(models) + 1)):
print('getting model ' + mdl)
wrmYear = wlYear[mdl]
wrmYear = int(round(wrmYear/5.)*5.)
ncFlNm = '_'.join(['projection_dis', scenario, mdl, 'wuChang', 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
mdlname = mdl.replace('BC_', '').replace('_BC', '')
lon, lat, bsln, proj, pcl, mp = plotSingleModel(None, ncFlPth, mdlname, mp=mp, projYear=wrmYear)
bslnlist.append(bsln)
projlist.append(proj)
ax = plt.subplot(gs[iscen])
bsln = np.nanmean(np.array(bslnlist), 0)
proj = np.nanmean(np.array(projlist), 0)
projVar = (proj - bsln)/bsln*100
plotSingleModelVar(ax, 'Ensemble ' + scenario + '\n wrm. lev. ' + str(warmingLev), lon, lat, projVar, mp, bold=True, titleFontSize=10)
ax = plt.subplot(gs[2])
#ax.get_xaxis().set_visible(False)
#ax.get_yaxis().set_visible(False)
cb = plt.colorbar(pcl, cax=ax)
cb.set_label(label='100-year event change (%)', fontsize=12)
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def loadRetPerAllYears(ncDir='/ClimateRun4/multi-hazard/eva', rlVarName='rl', retPer=100,
flpattern='projection_dis_{scen}_{mdl}_wuConst_statistics.nc', nmodels=-1, excludedModels=[]):
wlyR8 = getWarmingLevels('rcp85', 2.0)
models = wlyR8.keys()
if nmodels > -1:
models = models[:nmodels]
for mdl in excludedModels:
if mdl in models:
models.remove(mdl)
dsuparea = netCDF4.Dataset('upArea.nc')
upArea = dsuparea.variables['upArea'][:].transpose()
dsuparea.close()
tamask, _, _ = getAfricaAndTurkeyMask()
tamask = tamask.transpose()
vlsR8 = []
vlsR4 = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
ds = netCDF4.Dataset(flr8pth)
yr = ds.variables['year'][:]
upAreaMtx = np.tile(upArea, [yr.shape[0], 1, 1])
tamaskMtx = np.tile(tamask, [yr.shape[0], 1, 1])
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_==retPer)[0][0]
vl = ds.variables[rlVarName][rpIndx, :, :, :]
vl[upAreaMtx < 1e9] = np.nan
vl[~tamaskMtx] = np.nan
vlsR8.append(vl)
ds.close()
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
ds = netCDF4.Dataset(flr4pth)
vl = ds.variables[rlVarName][rpIndx, :, :, :]
vl[upAreaMtx < 1e9] = np.nan
vl[~tamaskMtx] = np.nan
vlsR4.append(vl)
ds.close()
vlsR8 = np.array(vlsR8)
vlsR4 = np.array(vlsR4)
return yr, vlsR8, vlsR4, models
def loadFrqDataForModel(diagnosticsDataDir, scn, mdl, warmingLev):
if warmingLev == 0:
ywl = 1995 # baseline
else:
wls = getWarmingLevels.getWarmingLevels(scn, warmingLev)
ywl = wls[mdl]
ymin = max(ywl-15, 1981)
ymax = min(ywl+14, 2100)
yrsMdl = range(ymin, ymax+1)
dt = None
for iyr, yr in zip(range(len(yrsMdl)), yrsMdl):
flPth = os.path.join(diagnosticsDataDir, '_'.join(['droughtFrqDiagnostics', mdl, scn, str(yr)])) + '.0.npy'
if not os.path.isfile(flPth):
print(' ... file ' + flPth + ' not found, skipping ...')
continue
dti = np.load(flPth)
nx, ny = dti.shape
dt = np.zeros([30, nx, ny])*np.nan if dt is None else dt
dt[iyr] = dti
mn = np.nanmedian(dt, 0)
return 1/mn
def getRcpEnsembleAtYear(ryear,
ncDir='/ClimateRun4/multi-hazard/eva',
bslnYear=1995,
retPer=100,
scen='rcp85'):
flpattern = 'projection_dis_{scen}_{mdl}_wuConst_statistics.nc'
wlyR = getWarmingLevels(scen, 2.0)
#wlyR4 = getWarmingLevels('rcp45', warmingLev)
models = wlyR.keys()
rl = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr = flpattern.format(scen=scen, mdl=mdl)
flrpth = os.path.join(ncDir, flr)
ryearInf = int(np.floor(ryear / float(5)) * 5)
print(' loading file ' + flrpth)
ds = netCDF4.Dataset(flrpth)
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_ == retPer)[0][0]
year_ = ds.variables['year'][:]
yIndxBsln = np.where(year_ == bslnYear)[0][0]
rlBslnR = ds.variables['rl'][rpIndx, yIndxBsln, :, :]
yIndx = np.where(year_ == ryearInf)[0][0]
rlR_ = ds.variables['rl'][rpIndx, yIndx:yIndx + 2, :, :]
ds.close()
#rlR8 = interp1d(year_[yIndx:yIndx+2], rlR8_, axis=0)(r8year)
rlR = interp1d(year_[yIndx:yIndx + 2], rlR_, axis=0)(ryear)
rRelChng = (rlR - rlBslnR) / rlBslnR
rl.append(rRelChng)
rl = np.array(rl)
ensembleRelChng = np.nanmean(rl, 0)
return ensembleRelChng
def loadWlVsScenChange2(ncDir='/ClimateRun4/multi-hazard/eva',
bslnYear=1995,
warmingLev=2,
retPer=100,
threshold=0,
rlVarName='rl'):
# computes the relative change of the ensemble
flpattern = 'projection_dis_{scen}_{mdl}_wuConst_statistics.nc'
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
dsuparea = netCDF4.Dataset('upArea.nc')
upArea = dsuparea.variables['upArea'][:].transpose()
dsuparea.close()
models = wlyR8.keys()
# menta
#models = [models[0], models[1]]
tamask, _, _ = getAfricaAndTurkeyMask()
tamask = tamask.transpose()
rl_bsln = []
rl_r4 = []
rl_r8 = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
r8year = wlyR8[mdl]
r8yearInf = int(np.floor(r8year / float(5)) * 5)
print(' rcp85 w.l. year: ' + str(r8year))
r4year = wlyR4[mdl]
r4yearInf = int(np.floor(r4year / float(5)) * 5)
print(' rcp45 w.l. year: ' + str(r4year))
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_ == retPer)[0][0]
year_ = ds.variables['year'][:]
yIndxBsln = np.where(year_ == bslnYear)[0][0]
rlBslnR8 = ds.variables[rlVarName][rpIndx, yIndxBsln, :, :]
yIndx = np.where(year_ == r8yearInf)[0][0]
rlR8_ = ds.variables[rlVarName][rpIndx, yIndx:yIndx + 2, :, :]
ds.close()
rlR8 = interp1d(year_[yIndx:yIndx + 2], rlR8_, axis=0)(r8year)
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
yIndx = np.where(year_ == r4yearInf)[0][0]
rlR4_ = ds.variables[rlVarName][rpIndx, yIndx:yIndx + 2, :, :]
ds.close()
rlR4 = interp1d(year_[yIndx:yIndx + 2], rlR4_, axis=0)(r4year)
rlR8[upArea < 1e9] = np.nan
rlR4[upArea < 1e9] = np.nan
rlR8[~tamask] = np.nan
rlR4[~tamask] = np.nan
rl_r8.append(rlR8)
rl_r4.append(rlR4)
rl_bsln.append(rlBslnR8)
rl_r8 = np.array(rl_r8)
rl_r4 = np.array(rl_r4)
rl_bsln = np.array(rl_bsln)
mnbsln = np.nanmean(rl_bsln, 0)
mnbsln[mnbsln < threshold] = np.nan
rc_r8 = (np.nanmean(rl_r8, 0) - mnbsln) / mnbsln
rc_r4 = (np.nanmean(rl_r4, 0) - mnbsln) / mnbsln
relChngDiff = rc_r8 - rc_r4
relChngDiff[~tamask] = np.nan
nmdl = rl_bsln.shape[0]
std_r8 = np.std(rl_r8 - rl_bsln, 0) / np.sqrt(
np.sum(rl_bsln**2., 0) / nmdl)
std_r4 = np.std(rl_r4 - rl_bsln, 0) / np.sqrt(
np.sum(rl_bsln**2., 0) / nmdl)
std_diff = np.std(rl_r8 - rl_r4, 0) / np.sqrt(
np.sum(rl_bsln**2., 0) / nmdl)
_, pval_r8 = stats.ttest_ind(rl_r8, rl_bsln, 0)
_, pval_r4 = stats.ttest_ind(rl_r4, rl_bsln, 0)
_, pval_diff = stats.ttest_ind(rl_r8, rl_r4, 0)
return relChngDiff, rc_r8, rc_r4, std_r8, std_r4, std_diff, pval_r8, pval_r4, pval_diff
def getGrossEnsembleAtYear(ryear,
ncDir='/ClimateRun4/multi-hazard/eva',
bslnYear=1995,
retPer=100,
threshold=200):
flpattern = 'projection_dis_{scen}_{mdl}_wuConst_statistics.nc'
wlyR8 = getWarmingLevels('rcp85', 2.0)
#wlyR4 = getWarmingLevels('rcp45', warmingLev)
models = wlyR8.keys()
tamask, _, _ = getAfricaAndTurkeyMask()
tamask = tamask.transpose()
rl_all = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
#r8year = wlyR8[mdl] + wlOffset
#r8yearInf = int(np.floor(r8year/float(5))*5)
#print(' rcp85 w.l. year: ' + str(r8year))
#r4year = wlyR4[mdl] + wlOffset
#r4yearInf = int(np.floor(r4year/float(5))*5)
#print(' rcp45 w.l. year: ' + str(r4year))
ryearInf = int(np.floor(ryear / float(5)) * 5)
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_ == retPer)[0][0]
year_ = ds.variables['year'][:]
yIndxBsln = np.where(year_ == bslnYear)[0][0]
rlBslnR8 = ds.variables['rl'][rpIndx, yIndxBsln, :, :]
#yIndx = np.where(year_==r8yearInf)[0][0]
yIndx = np.where(year_ == ryearInf)[0][0]
rlR8_ = ds.variables['rl'][rpIndx, yIndx:yIndx + 2, :, :]
ds.close()
#rlR8 = interp1d(year_[yIndx:yIndx+2], rlR8_, axis=0)(r8year)
rlR8 = interp1d(year_[yIndx:yIndx + 2], rlR8_, axis=0)(ryear)
r8RelChng = (rlR8 - rlBslnR8) / rlBslnR8
if threshold > 0:
cnd = rlBslnR8 < threshold
r8RelChng[cnd] = np.nan
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
rlBslnR4 = ds.variables['rl'][rpIndx, yIndxBsln, :, :]
#yIndx = np.where(year_==r4yearInf)[0][0]
yIndx = np.where(year_ == ryearInf)[0][0]
rlR4_ = ds.variables['rl'][rpIndx, yIndx:yIndx + 2, :, :]
ds.close()
#rlR4 = interp1d(year_[yIndx:yIndx+2], rlR4_, axis=0)(r4year)
rlR4 = interp1d(year_[yIndx:yIndx + 2], rlR4_, axis=0)(ryear)
r4RelChng = (rlR4 - rlBslnR4) / rlBslnR4
if threshold > 0:
cnd = rlBslnR4 < threshold
r4RelChng[cnd] = np.nan
r8RelChng[~tamask] = np.nan
r4RelChng[~tamask] = np.nan
rl_all.append(r8RelChng)
rl_all.append(r4RelChng)
rl_all = np.array(rl_all)
ensembleRelChngAll = np.nanmean(rl_all, 0)
return ensembleRelChngAll
def extract1Model(scen, model, inputHistNcFlPath, inputRcNcFlPath, outputNcFlPath, varnames=['pr']):
wls = [1.5, 2.0, 3.0]
bslnYear = 1995
print(' loading the data ...')
dsin = netCDF4.Dataset(inputHistNcFlPath)
x = dsin.variables['x'][:]
y = dsin.variables['y'][:]
val = None
for varname in varnames:
if varname in dsin.variables:
val = dsin.variables[varname][:]
break
if val is None:
raise Exception('variables ' + str(varnames) + ' not found in file ' + inputHistNcFlPath)
valHist = val
tmnc = dsin.variables['time']
valYrsHist = [d.year for d in netCDF4.num2date(tmnc[:], tmnc.units, tmnc.calendar)]
dsin.close()
dsin = netCDF4.Dataset(inputRcNcFlPath)
val = None
for varname in varnames:
if varname in dsin.variables:
val = dsin.variables[varname][:]
break
if val is None:
raise Exception('variables ' + str(varnames) + ' not found in file' + inputRcNcFlPath)
valRc = val
tmnc = dsin.variables['time']
valYrsRc = [d.year for d in netCDF4.num2date(tmnc[:], tmnc.units, tmnc.calendar)]
dsin.close()
val = np.concatenate([valHist, valRc], 0)
valYrs = np.concatenate([valYrsHist, valYrsRc], 0)
def getLonLat():
lonLatFile = 'lonlat.nc'
ds = netCDF4.Dataset(lonLatFile)
lon = ds.variables['lon'][:].transpose()
lat = ds.variables['lat'][:].transpose()
return lon, lat
def getValAtYear(valAll, yearAll, year, halfTmWindowSize=14):
yindx = np.where(yearAll == year)[0][0]
val = np.nanmean(valAll[yindx-halfTmWindowSize:yindx+halfTmWindowSize, :], 0).squeeze()
return val
bslnVal = getValAtYear(val, valYrs, bslnYear)
print(' elaborating ...')
noutWls = len(wls)
shp_ = bslnVal.shape
shpOut = [noutWls, shp_[0], shp_[1]]
outYrVal = np.zeros(shpOut)*np.nan
for wl, iwl in zip(wls, range(noutWls)):
ywls = getWarmingLevels(scen, wl)
if ywls == None:
continue
outYr = ywls[model]
yrVal = getValAtYear(val, valYrs, outYr)
outYrVal[iwl, :, :] = yrVal
bslnVal3d = np.tile(bslnVal, [3, 1, 1])
dlt = (outYrVal/bslnVal3d - 1)*100.
absDlt = outYrVal - bslnVal3d
print(' saving the output ...')
if os.path.isfile(outputNcFlPath):
os.remove(outputNcFlPath)
dsout = netCDF4.Dataset(outputNcFlPath, 'w')
dsout.createDimension('x', len(x))
dsout.createDimension('y', len(y))
dsout.createDimension('warming_lev', noutWls)
xnc = dsout.createVariable('x', 'f8', ('x'))
xnc.description = 'x'
xnc[:] = x
ync = dsout.createVariable('y', 'f8', ('y'))
ync.description = 'y'
ync[:] = y
lon, lat = getLonLat()
lon, lat = lon.transpose(), lat.transpose()
lonnc = dsout.createVariable('lon', 'f8', ('y', 'x'))
lonnc.description = 'longitude mtx'
lonnc[:] = lon
latnc = dsout.createVariable('lat', 'f8', ('y', 'x'))
latnc.description = 'latitude mtx'
latnc[:] = lat
wlsnc = dsout.createVariable('warming_lev', 'f4', ('warming_lev'))
wlsnc.description = 'warming_lev'
wlsnc[:] = wls
bslnYrNc = dsout.createVariable('baseline_year', 'i4')
bslnYrNc.description = 'baseline year'
bslnYrNc[:] = bslnYear
bslnValNc = dsout.createVariable('baseline_val', 'f4', ('y', 'x'))
bslnValNc.description = 'value at baseline (deg or mm or m3 or ...)'
bslnValNc[:] = bslnVal
valNc = dsout.createVariable('value', 'f4', ('warming_lev', 'y', 'x'))
valNc.description = 'values at warming levels (deg or mm or m3 or ...)'
valNc[:] = outYrVal
valNc = dsout.createVariable('value_perc_chng', 'f4', ('warming_lev', 'y', 'x'))
valNc.description = 'percentage change at warming levels (%)'
valNc[:] = dlt
valNc = dsout.createVariable('value_abs_chng', 'f4', ('warming_lev', 'y', 'x'))
valNc.description = 'absolute change at warming levels (deg or mm or m3 or ...)'
valNc[:] = absDlt
dsout.close()
def computeRlChngPValueAtWarmingLev(ncDir='/ClimateRun4/multi-hazard/eva', bslnYear=1995, scen='rcp85', warmingLev=2, retPer=100, minThreshold=0):
flpattern = 'projection_dis_{scen}_{mdl}_wuChang_statistics.nc'
wly = getWarmingLevels(scen, warmingLev)
models = wly.keys()
rl_bs = []
serl_bs = []
rl_rc = []
serl_rc = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr = flpattern.format(scen=scen, mdl=mdl)
flrpth = os.path.join(ncDir, flr)
rcyear = wly[mdl]
rcyearInf = int(np.floor(rcyear/float(5))*5)
print(' ' + scen + ' w.l. year: ' + str(rcyear))
print(' loading file ' + flrpth)
ds = netCDF4.Dataset(flrpth)
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_==retPer)[0][0]
year_ = ds.variables['year'][:]
yIndx = np.where(year_==rcyearInf)[0][0]
rlR_ = ds.variables['rl'][rpIndx, yIndx:yIndx+2, :, :]
serlR_ = ds.variables['se_rl'][rpIndx, yIndx:yIndx+2, :, :]
yBslnIndx = np.where(year_==bslnYear)[0][0]
rlBsln = ds.variables['rl'][rpIndx, yBslnIndx, :, :]
serlBsln = ds.variables['se_rl'][rpIndx, yBslnIndx, :, :]
ds.close()
rlR = interp1d(year_[yIndx:yIndx+2], rlR_, axis=0)(rcyear)
serlR = interp1d(year_[yIndx:yIndx+2], serlR_, axis=0)(rcyear)
rl_rc.append(rlR)
serl_rc.append(serlR)
rl_bs.append(rlBsln)
serl_bs.append(serlBsln)
rl_rc = np.array(rl_rc)
rl_rc[rl_rc <= .1] = .1
serl_rc = np.array(serl_rc)
serl_rc[serl_rc <= .1] = .1
rl_bs = np.array(rl_bs)
rl_bs[rl_bs <= .1] = .1
serl_bs = np.array(rl_bs)
serl_bs[serl_bs <= .1] = .1
cnd = np.nanmean(rl_rc, 0) < minThreshold
cnd = np.tile(cnd, [len(models), 1, 1])
rl_rc[cnd] = np.nan
serl_rc[cnd] = np.nan
rl_bs[cnd] = np.nan
serl_bs[cnd] = np.nan
pvalue = computeRlChngPValueMontecarlo(rl_bs, serl_bs, rl_rc, serl_rc)
#pvalue = computeRlChngPValueMeansOnly(rl_bs, rl_rc)
agrMdlCnt, std = countAgreeingModelsAndGetStdDev(rl_bs, rl_rc)
return pvalue, agrMdlCnt, std
def loadWlVsScenChange(
ncDir='/ClimateRun4/multi-hazard/eva',
bslnYear=1995,
warmingLev=2,
retPer=100,
threshold=0,
rlVarName='rl',
flpattern='projection_dis_{scen}_{mdl}_wuConst_statistics.nc',
nmodels=-1,
shapeParamNcVarName='',
minNumModels=8):
# computes the mean relative change
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
dsuparea = netCDF4.Dataset('upArea.nc')
upArea = dsuparea.variables['upArea'][:].transpose()
dsuparea.close()
maskInvalidShapeParamPts = shapeParamNcVarName != ''
models = wlyR8.keys()
if nmodels > -1:
models = models[:nmodels]
# menta
#models = [models[0], models[1]]
tamask, _, _ = getAfricaAndTurkeyMask()
tamask = tamask.transpose()
if minNumModels > nmodels:
minNumModels = nmodels
rl_r4 = []
rl_r8 = []
mdlCountR4 = np.zeros(upArea.shape)
mdlCountR8 = np.zeros(upArea.shape)
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
r8year = wlyR8[mdl]
r8yearInf = int(np.floor(r8year / float(5)) * 5)
print(' rcp85 w.l. year: ' + str(r8year))
r4year = wlyR4[mdl]
r4yearInf = int(np.floor(r4year / float(5)) * 5)
print(' rcp45 w.l. year: ' + str(r4year))
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_ == retPer)[0][0]
year_ = ds.variables['year'][:]
yIndxBsln = np.where(year_ == bslnYear)[0][0]
rlBslnR8 = ds.variables[rlVarName][rpIndx, yIndxBsln, :, :]
yIndx = np.where(year_ == r8yearInf)[0][0]
rlR8_ = ds.variables[rlVarName][rpIndx, yIndx:yIndx + 2, :, :]
rlR8 = interp1d(year_[yIndx:yIndx + 2], rlR8_, axis=0)(r8year)
if maskInvalidShapeParamPts:
shapeParam = ds.variables[shapeParamNcVarName][:]
cnd = shapeParam <= -.4
rlBslnR8[cnd] = np.nan
rlR8[cnd] = np.nan
ds.close()
r8RelChng = (rlR8 - rlBslnR8) / rlBslnR8
if threshold > 0:
cnd = rlBslnR8 < threshold
r8RelChng[cnd] = np.nan
#r8RelChng[r8RelChng < -.15] = np.nan
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
rlBslnR4 = ds.variables[rlVarName][rpIndx, yIndxBsln, :, :]
#rlBslnR4 = rlBslnR8
yIndx = np.where(year_ == r4yearInf)[0][0]
rlR4_ = ds.variables[rlVarName][rpIndx, yIndx:yIndx + 2, :, :]
rlR4 = interp1d(year_[yIndx:yIndx + 2], rlR4_, axis=0)(r4year)
if maskInvalidShapeParamPts:
shapeParam = ds.variables[shapeParamNcVarName][:]
cnd = shapeParam <= -.4
rlBslnR4[cnd] = np.nan
rlR4[cnd] = np.nan
ds.close()
r4RelChng = (rlR4 - rlBslnR4) / rlBslnR4
if threshold > 0:
cnd = rlBslnR4 < threshold
r4RelChng[cnd] = np.nan
#r4RelChng[r4RelChng < -.15] = np.nan
r8RelChng[upArea < 1e9] = np.nan
r4RelChng[upArea < 1e9] = np.nan
r8RelChng[~tamask] = np.nan
r4RelChng[~tamask] = np.nan
rl_r8.append(r8RelChng)
rl_r4.append(r4RelChng)
mdlCountR8 += ~np.isnan(r8RelChng)
mdlCountR4 += ~np.isnan(r4RelChng)
rl_r8 = np.array(rl_r8)
rl_r4 = np.array(rl_r4)
mdlCountMskR8 = mdlCountR8 >= minNumModels
mdlCountMskR4 = mdlCountR4 >= minNumModels
mskR8 = np.tile(mdlCountMskR8, [rl_r8.shape[0], 1, 1])
mskR4 = np.tile(mdlCountMskR4, [rl_r8.shape[0], 1, 1])
rl_r8[mskR8 != 1] = np.nan
rl_r4[mskR4 != 1] = np.nan
relChngDiff = np.nanmean(rl_r8 - rl_r4, 0)
relChngDiff[~tamask] = np.nan
return relChngDiff, np.nanmean(rl_r8, 0), np.nanmean(rl_r4,
0), rl_r8, rl_r4
def plotEnsembles(rootDir='/DATA/JEODPP/eos/projects/CRITECH/ADAPTATION/ClimateRuns/LisfloodEuroCordex/', wuChangStr='wuChang'):
warmingLevs = [1.5, 2.0]
rcp8, rcp4 = 'rcp85', 'rcp45'
retPer = 100
models = """
IPSL-INERIS-WRF331F_BC
SMHI-RCA4_BC_CNRM-CERFACS-CNRM-CM5
SMHI-RCA4_BC_ICHEC-EC-EARTH
SMHI-RCA4_BC_IPSL-IPSL-CM5A-MR
SMHI-RCA4_BC_MOHC-HadGEM2-ES
SMHI-RCA4_BC_MPI-M-MPI-ESM-LR
CLMcom-CCLM4-8-17_BC_CNRM-CERFACS-CNRM-CM5
CLMcom-CCLM4-8-17_BC_ICHEC-EC-EARTH
CLMcom-CCLM4-8-17_BC_MPI-M-MPI-ESM-LR
DMI-HIRHAM5-ICHEC-EC-EARTH_BC
KNMI-RACMO22E-ICHEC-EC-EARTH_BC
"""
models = models.split()
outputPngFile = 'scatter_minDis_' + wuChangStr + '_wls_ensemble.png'
fig = plt.figure(figsize=(10, 5))
gs = gridspec.GridSpec(1, 3, width_ratios=[1,1,.1])
lon, lat = [], []
for warmingLev, iwl in zip(warmingLevs, range(len(warmingLevs))):
wlYearR8 = getWarmingLevels(rcp8, warmingLev)
wlYearR4 = getWarmingLevels(rcp4, warmingLev)
relChng8lst, relChng4lst = [], []
for mdl, imdl in zip(models, range(1, len(models) + 1)):
print('getting model ' + mdl)
fldirHist = os.path.join(rootDir, 'historical', mdl, 'wuConst')
fldirR8 = os.path.join(rootDir, rcp8, mdl, wuChangStr)
wrmYear = wlYearR8[mdl]
cacheIdData = '_'.join([mdl, rcp8, 'wuChang'])
cacheIdResult = cacheIdData + '_' + str(warmingLev)
relChngR8 = getRelChngs(cacheIdData, cacheIdResult, fldirHist, fldirR8, wrmYear)
fldirR4 = os.path.join(rootDir, rcp4, mdl, wuChangStr)
wrmYear = wlYearR4[mdl]
cacheIdData = '_'.join([mdl, rcp4, 'wuChang'])
cacheIdResult = cacheIdData + '_' + str(warmingLev)
relChngR4 = getRelChngs(cacheIdData, cacheIdResult, fldirHist, fldirR4, wrmYear)
relChng8lst.append(relChngR8)
relChng4lst.append(relChngR4)
ax = plt.subplot(gs[iwl])
relChng8 = np.nanmean(np.array(relChng8lst), 0)
relChng4 = np.nanmean(np.array(relChng4lst), 0)
plotSingleModel(ax, relChng8, relChng4,
'Ensemble rcp85 vs rcp45, min. dis.\nwrm. lev. ' + str(warmingLev) + '$^\circ$',
bold=True, plotRegression=False, tickFontSize=10, labelFontSize=12, titleFontSize=13,
xLabelPad=-25, yLabelPad=-10, xlim=[-150, 150], ylim=[-150, 150])
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def extract1Model(scen, model, inputNcFlPath, outputNcFlPath):
wls = [1.5, 2.0, 3.0, 4.0]
bslnYear = 1995
outRetPer = [1.5, 2., 5, 10., 15., 20., 50., 100., 200., 500.]
print(' loading the data ...')
dsin = netCDF4.Dataset(inputNcFlPath)
x = dsin.variables['x'][:]
y = dsin.variables['y'][:]
inRetPer = dsin.variables['return_period'][:]
retLev = dsin.variables['rl_min_ws'][:]
rlYrs = dsin.variables['year'][:]
dsin.close()
def getLonLat():
lonLatFile = 'lonlat.nc'
ds = netCDF4.Dataset(lonLatFile)
lon = ds.variables['lon'][:].transpose()
lat = ds.variables['lat'][:].transpose()
return lon, lat
def getRetLevAtYear(retLevAll, yearAll, year, interpFunc=None):
interpFunc = interp1d(
yearAll,
retLevAll,
axis=1,
bounds_error=False,
fill_value='extrapolate') if interpFunc is None else interpFunc
retLev = interpFunc(year)
return retLev, interpFunc
def getOutRetLev(inRetLev, inRetPer, outRetPer):
interpFunc = interp1d(inRetPer, inRetLev, axis=0)
outRetLev = interpFunc(outRetPer)
return outRetLev
def getOutRetPer(inRetPer, retLevCurrent, retLevBsln):
nx = retLevCurrent.shape[1]
ny = retLevCurrent.shape[2]
nOutRL = retLevBsln.shape[0]
outRetPer = np.zeros([nOutRL, nx, ny]) * np.nan
for ix in range(nx):
for iy in range(ny):
rlcri = retLevCurrent[:, ix, iy]
if np.sum(np.isnan(rlcri)) > 0:
continue
rlbsln = retLevBsln[:, ix, iy]
outRetPer[:, ix,
iy] = interp1d(rlcri,
inRetPer,
axis=0,
fill_value='extrapolate')(rlbsln)
return outRetPer
def getSignificanceTest(retLev_KSTest, bslnRetLev_KSTest):
nx = retLev_KSTest.shape[1]
ny = retLev_KSTest.shape[2]
outSigTest = np.zeros([nx, ny]) * np.nan
for ix in range(nx):
for iy in range(ny):
bsln = bslnRetLev_KSTest[:, ix, iy].squeeze()
if np.sum(np.isnan(bsln)) > 0:
continue
futvl = retLev_KSTest[:, ix, iy].squeeze()
st, pval = ks_2samp(bsln, futvl)
outSigTest[ix, iy] = pval <= .05
pass
return outSigTest
interpFunc = None
bslnRetLev_, interpFunc = getRetLevAtYear(retLev, rlYrs, bslnYear,
interpFunc)
bslnRetLev = getOutRetLev(bslnRetLev_, inRetPer, outRetPer)
retPerKSTest = 20
bslnYears_KSTest = np.arange(bslnYear - 14, bslnYear + 16)
bslnRetLev_KSTest_, _ = getRetLevAtYear(retLev, rlYrs, bslnYears_KSTest,
interpFunc)
bslnRetLev_KSTest = getOutRetLev(bslnRetLev_KSTest_, inRetPer,
retPerKSTest)
del bslnRetLev_KSTest_
gc.collect()
print(' elaborating ...')
noutWls = len(wls)
shp_ = bslnRetLev.shape
shpOut = [noutWls, shp_[0], shp_[1], shp_[2]]
outYrRetPer = np.zeros(shpOut) * np.nan
outYrRetLev = np.zeros(shpOut) * np.nan
signTest = np.zeros([noutWls, shp_[1], shp_[2]]) * np.nan
for wl, iwl in zip(wls, range(noutWls)):
ywls = getWarmingLevels(scen, wl)
if ywls == None:
continue
outYr = ywls[model]
yrRetLev, interpFunc = getRetLevAtYear(retLev, rlYrs, outYr,
interpFunc)
outYrRetLev[iwl, :, :, :] = getOutRetLev(yrRetLev, inRetPer, outRetPer)
outYrRetPer[iwl, :, :, :] = getOutRetPer(inRetPer, yrRetLev,
bslnRetLev)
years_KSTest = np.arange(outYr - 15, outYr + 15)
retLev_KSTest_, _ = getRetLevAtYear(retLev, rlYrs, years_KSTest)
retLev_KSTest = getOutRetLev(retLev_KSTest_, inRetPer, retPerKSTest)
del retLev_KSTest_
gc.collect()
signTest[iwl, :, :] = getSignificanceTest(retLev_KSTest,
bslnRetLev_KSTest)
outYrRetPer[outYrRetPer < 1] = 1
print(' saving the output ...')
if os.path.isfile(outputNcFlPath):
os.remove(outputNcFlPath)
dsout = netCDF4.Dataset(outputNcFlPath, 'w')
dsout.createDimension('x', len(x))
dsout.createDimension('y', len(y))
dsout.createDimension('warming_lev', noutWls)
dsout.createDimension('baseline_rp', len(outRetPer))
xnc = dsout.createVariable('x', 'f8', ('x'))
xnc.description = 'x'
xnc[:] = x
ync = dsout.createVariable('y', 'f8', ('y'))
ync.description = 'y'
ync[:] = y
lon, lat = getLonLat()
lonnc = dsout.createVariable('lon', 'f8', ('x', 'y'))
lonnc.description = 'longitude mtx'
lonnc[:] = lon
latnc = dsout.createVariable('lat', 'f8', ('x', 'y'))
latnc.description = 'latitude mtx'
latnc[:] = lat
wlsnc = dsout.createVariable('warming_lev', 'f4', ('warming_lev'))
wlsnc.description = 'warming_lev'
wlsnc[:] = wls
bslnRpNc = dsout.createVariable('baseline_rp', 'f4', ('baseline_rp'))
bslnRpNc.description = 'return period at the baseline (years)'
bslnRpNc[:] = outRetPer
bslnYrNc = dsout.createVariable('baseline_year', 'i4')
bslnYrNc.description = 'baseline year'
bslnYrNc[:] = bslnYear
bslnRetLevNc = dsout.createVariable('baseline_return_level', 'f4',
('baseline_rp', 'x', 'y'))
bslnRetLevNc.description = 'return levels at baseline'
bslnRetLevNc[:] = bslnRetLev
retPerNc = dsout.createVariable('baseline_rp_shift', 'f4',
('warming_lev', 'baseline_rp', 'x', 'y'))
retPerNc.description = 'shifted return periods of the baseline return values (years)'
retPerNc[:] = outYrRetPer
retLevNc = dsout.createVariable('return_level', 'f4',
('warming_lev', 'baseline_rp', 'x', 'y'))
retLevNc.description = 'return levels at warming levels (m**3)'
retLevNc[:] = outYrRetLev
sigNc = dsout.createVariable('significance', 'f4',
('warming_lev', 'x', 'y'))
sigNc.description = 'significance of the projected change (ks test on the 20-y return level)'
sigNc[:] = signTest
dsout.close()
def extract1Model(scen, model, inputNcFlPath, outputNcFlPath):
wls = [1.5, 2.0, 3.0, 4.0]
bslnYear = 1995
print(' loading the data ...')
dsin = netCDF4.Dataset(inputNcFlPath)
x = dsin.variables['x'][:]
y = dsin.variables['y'][:]
vals = dsin.variables['year_mean'][:]
yrs = dsin.variables['year_all'][:]
dsin.close()
def getLonLat():
lonLatFile = 'lonlat.nc'
ds = netCDF4.Dataset(lonLatFile)
lon = ds.variables['lon'][:].transpose()
lat = ds.variables['lat'][:].transpose()
return lon, lat
def getValAtYear(vlsAll, yearAll, year):
iyear = np.where(yearAll == year)[0][0]
minIndx = max(iyear - 14, 0)
vly = np.nanmean(vlsAll[minIndx:iyear + 15, :, :], 0).squeeze()
return vly
bslnVal = getValAtYear(vals, yrs, bslnYear)
print(' elaborating ...')
noutWls = len(wls)
shp_ = bslnVal.shape
shpOut = [noutWls, shp_[0], shp_[1]]
outYrVal = np.zeros(shpOut) * np.nan
for wl, iwl in zip(wls, range(noutWls)):
ywls = getWarmingLevels(scen, wl)
if ywls == None:
continue
outYr = ywls[model]
yrVal = getValAtYear(vals, yrs, outYr)
outYrVal[iwl, :, :] = yrVal
print(' saving the output ...')
if os.path.isfile(outputNcFlPath):
os.remove(outputNcFlPath)
dsout = netCDF4.Dataset(outputNcFlPath, 'w')
dsout.createDimension('x', len(x))
dsout.createDimension('y', len(y))
dsout.createDimension('warming_lev', noutWls)
xnc = dsout.createVariable('x', 'f8', ('x'))
xnc.description = 'x'
xnc[:] = x
ync = dsout.createVariable('y', 'f8', ('y'))
ync.description = 'y'
ync[:] = y
lon, lat = getLonLat()
lonnc = dsout.createVariable('lon', 'f8', ('x', 'y'))
lonnc.description = 'longitude mtx'
lonnc[:] = lon
latnc = dsout.createVariable('lat', 'f8', ('x', 'y'))
latnc.description = 'latitude mtx'
latnc[:] = lat
wlsnc = dsout.createVariable('warming_lev', 'f4', ('warming_lev'))
wlsnc.description = 'warming_lev'
wlsnc[:] = wls
bslnYrNc = dsout.createVariable('baseline_year', 'i4')
bslnYrNc.description = 'baseline year'
bslnYrNc[:] = bslnYear
bslnValNc = dsout.createVariable('baseline_mean_dis', 'f4', ('x', 'y'))
bslnValNc.description = 'mean discharge at baseline (m**3)'
bslnValNc[:] = bslnVal
valNc = dsout.createVariable('mean_dis', 'f4', ('warming_lev', 'x', 'y'))
valNc.description = 'mean discharge at warming levels (m**3)'
valNc[:] = outYrVal
prcChngNc = dsout.createVariable('perc_change', 'f4',
('warming_lev', 'x', 'y'))
prcChngNc.description = 'change of mean discharge at warming levels (%)'
prcChngNc[:] = (outYrVal - bslnVal) / bslnVal * 100
dsout.close()
def plotTimeSerie30YVariability_hiExt(ax=None,
plotXLabel=True,
warmingLev=2.0):
yrs, rpR8Srs, rpR4Srs, models = loadWlVsScenChange.loadRetPerAllYears(
nmodels=nmodels)
rpR8 = getRelChng(yrs, rpR8Srs)
rpR4 = getRelChng(yrs, rpR4Srs)
wlyR8 = getWarmingLevels('rcp85', warmingLev)
r8srs = getNYrsTimeSeriesAroundWl(yrs, rpR8, wlyR8, models)
r8medianWl, r8sigmaWl = getMeanAndSigma(r8srs)
cndSignR8 = np.abs(r8medianWl) > r8sigmaWl
cndPosi = r8medianWl > 0
#cnd = np.logical_and(cndSignR8, cndPosi)
cnd = cndPosi
cndMtx = np.tile(cnd, [r8srs.shape[0], r8srs.shape[1], 1, 1])
r8srs[~cndMtx] = np.nan
r8median_ = np.median(r8srs, 0)
r8median = r8median_.reshape(r8median_.shape[0],
r8median_.shape[1] * r8median_.shape[2])
r8median = np.nanmedian(r8median, 1)
#r8median = np.nanmean(r8median, 1)
r8spmean = r8srs.reshape(r8srs.shape[0], r8srs.shape[1],
r8srs.shape[2] * r8srs.shape[3])
r8spmean = np.nanmedian(r8spmean, 2)
#r8spmean = np.nanmean(r8spmean, 2)
r8skew = stats.skew(r8spmean[:, 3])
r8sigma = np.std(r8spmean[:, 3])
r8median = np.nanmedian(r8spmean, 0)
wlyR4 = getWarmingLevels('rcp45', 2.0)
r4srs = getNYrsTimeSeriesAroundWl(yrs, rpR4, wlyR4, models)
r4medianWl, r4sigmaWl = getMeanAndSigma(r4srs)
cndSignR4 = np.abs(r4medianWl) > r4sigmaWl
cndPosi = r4medianWl > 0
cnd = cndPosi
cndMtx = np.tile(cnd, [r4srs.shape[0], r4srs.shape[1], 1, 1])
r4srs[~cndMtx] = np.nan
r4median_ = np.median(r4srs, 0)
r4median = r4median_.reshape(r4median_.shape[0],
r4median_.shape[1] * r4median_.shape[2])
r4median = np.nanmedian(r4median, 1)
#r4median = np.nanmean(r4median, 1)
r4spmean = r4srs.reshape(r4srs.shape[0], r4srs.shape[1],
r4srs.shape[2] * r4srs.shape[3])
r4spmean = np.nanmedian(r4spmean, 2)
#r4spmean = np.nanmean(r4spmean, 2)
r4skew = stats.skew(r4spmean[:, 3])
r4sigma = np.std(r4spmean[:, 3])
r4median = np.nanmedian(r4spmean, 0)
yr = [-15, -10, -5, 0, 5, 10, 15]
if ax is None:
fg = plt.figure(figsize=[8.46, 4.98])
#pmdlr8 = plt.plot(yr, r8spmean.transpose()*100, 'peachpuff', label='RCP8.5 models ($\sigma={sgm:1.2f}\%, skw.={s:1.2f}$)'.format(sgm=r8sigma*100, s=r8skew));
#pmdlr4 = plt.plot(yr, r4spmean.transpose()*100, 'skyblue', label='RCP4.5 models ($\sigma={sgm:1.2f}\%, skw.={s:1.2f})$'.format(sgm=r4sigma*100, s=r4skew));
pmdlr8 = plt.plot(yr,
r8spmean.transpose() * 100,
'peachpuff',
label='RCP8.5 models ($\sigma={sgm:1.2f}\%$'.format(
sgm=r8sigma * 100, s=r8skew))
pmdlr4 = plt.plot(yr,
r4spmean.transpose() * 100,
'skyblue',
label='RCP4.5 models ($\sigma={sgm:1.2f}\%$'.format(
sgm=r4sigma * 100, s=r4skew))
pmedr8 = plt.plot(yr,
r8median * 100,
'firebrick',
linewidth=6,
label='RCP8.5 mean')
pmedr4 = plt.plot(yr,
r4median * 100,
'royalblue',
linewidth=6,
label='RCP4.5 mean')
plt.grid('on')
plt.legend(handles=[pmdlr8[0], pmedr8[0], pmdlr4[0], pmedr4[0]])
if plotXLabel:
plt.xlabel('years to $2^\circ C w.l.$', fontsize=14)
plt.ylabel('$Q_{H100}$ % change', fontsize=14)
plt.tight_layout()
if ax is None:
fg.savefig('./ensemblesVariabilityAround2deg.png', dpi=400)
def plotTimeSerie30YVariability_hiExt(ax=None, plotXLabel=True, plotYLabel=True, warmingLev=2.0):
yrs, rpR8Srs, rpR4Srs, models = loadWlVsScenChange.loadRetPerAllYears(nmodels=nmodels, excludedModels=excludedModels)
rpR8 = getRelChng(yrs, rpR8Srs)
rpR4 = getRelChng(yrs, rpR4Srs)
wlyR8 = getWarmingLevels('rcp85', warmingLev)
r8srs = getNYrsTimeSeriesAroundWl(yrs, rpR8, wlyR8, models)
r8meanWl, r8sigmaWl = getMeanAndSigma(r8srs)
cndPosi = r8meanWl > 0
cndNega = r8meanWl < 0
cnd = cndPosi if positiveChanges else cndNega
cndMtx = np.tile(cnd, [r8srs.shape[0], r8srs.shape[1], 1, 1])
r8srs[~cndMtx] = np.nan
r8mean_ = np.nanmean(r8srs, 0)
r8mean = r8mean_.reshape(r8mean_.shape[0], r8mean_.shape[1]*r8mean_.shape[2])
#r8mean = np.nanmedian(r8mean, 1)
r8mean = np.nanmean(r8mean, 1)
r8spmean = r8srs.reshape(r8srs.shape[0], r8srs.shape[1], r8srs.shape[2]*r8srs.shape[3])
r8spmean = np.nanmean(r8spmean, 2)
#r8spmean = np.nanmean(r8spmean, 2)
r8skew = stats.skew(r8spmean[:, 3])
r8sigma = np.std(r8spmean[:, 3])
r8mean = np.nanmean(r8spmean, 0)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
r4srs = getNYrsTimeSeriesAroundWl(yrs, rpR4, wlyR4, models)
r4meanWl, r4sigmaWl = getMeanAndSigma(r4srs)
cndPosi = r4meanWl > 0
cndNega = r8meanWl < 0
cnd = cndPosi if positiveChanges else cndNega
cndMtx = np.tile(cnd, [r4srs.shape[0], r4srs.shape[1], 1, 1])
r4srs[~cndMtx] = np.nan
r4mean_ = np.nanmean(r4srs, 0)
r4mean = r4mean_.reshape(r4mean_.shape[0], r4mean_.shape[1]*r4mean_.shape[2])
#r4mean = np.nanmedian(r4mean, 1)
r4mean = np.nanmean(r4mean, 1)
r4spmean = r4srs.reshape(r4srs.shape[0], r4srs.shape[1], r4srs.shape[2]*r4srs.shape[3])
r4spmean = np.nanmean(r4spmean, 2)
#r4spmean = np.nanmean(r4spmean, 2)
r4skew = stats.skew(r4spmean[:, 3])
r4sigma = np.std(r4spmean[:, 3])
r4mean = np.nanmean(r4spmean, 0)
ensTot = np.concatenate([r8spmean, r4spmean], 0)
stdDevTot = np.std(ensTot, 0)*100.
meanTot = (r4mean + r8mean)/2.*100.
yr = [-15, -10, -5, 0, 5, 10, 15]
if ax is None:
fg = plt.figure(figsize=[8.46, 4.98])
pstdDev = plt.fill_between(yr, meanTot+stdDevTot, meanTot-stdDevTot, color='gainsboro', label='$\sigma={sgm:1.0f}\%$'.format(sgm=stdDevTot[int(np.floor(len(stdDevTot)/2))]))
#pmdlr8 = plt.plot(yr, r8spmean.transpose()*100, 'sandybrown', label='RCP8.5 models ($\sigma={sgm:1.2f}\%, skw.={s:1.2f}$)'.format(sgm=r8sigma*100, s=r8skew));
#pmdlr4 = plt.plot(yr, r4spmean.transpose()*100, 'skyblue', label='RCP4.5 models ($\sigma={sgm:1.2f}\%, skw.={s:1.2f})$'.format(sgm=r4sigma*100, s=r4skew));
#pmdlr8 = plt.plot(yr, r8spmean.transpose()*100, 'sandybrown', label='RCP8.5 models ($\sigma_{{RCP8.5}}={sgm:1.0f}\%$)'.format(sgm=r8sigma*100, s=r8skew));
#pmdlr4 = plt.plot(yr, r4spmean.transpose()*100, 'skyblue', label='RCP4.5 models ($\sigma_{{RCP4.5}}={sgm:1.0f}\%$)'.format(sgm=r4sigma*100, s=r4skew));
pmdlr8 = plt.plot(yr, r8spmean.transpose()*100, 'sandybrown', label='RCP8.5 models');
pmdlr4 = plt.plot(yr, r4spmean.transpose()*100, 'skyblue', label='RCP4.5 models');
#pmedr8 = plt.plot(yr, r8mean*100, 'firebrick', linewidth=6, label='RCP8.5 mean');
#pmedr4 = plt.plot(yr, r4mean*100, 'royalblue', linewidth=6, label='RCP4.5 mean');
pmn = plt.plot(yr, (r4mean+r8mean)/2.*100, 'k', linewidth=6, label='ens. mean');
plt.grid('on')
lgndLoc = 2 if positiveChanges else 3
#plt.legend(handles=[pmdlr8[0], pmedr8[0], pmdlr4[0], pmedr4[0], pstdDev], fontsize=12, loc=lgndLoc)
plt.legend(handles=[pmdlr8[0], pmdlr4[0], pmn[0], pstdDev], fontsize=12, loc=lgndLoc)
if plotXLabel:
plt.xlabel('years to $' + str(warmingLev) + '^\circ$C w.l.', fontsize=18)
else:
ax.set_xticklabels([])
if plotYLabel:
plt.ylabel('$Q_{H100}$ % change', fontsize=17)
else:
ax.set_yticklabels([])
ax.tick_params(axis="x", labelsize=axFontSize)
ax.tick_params(axis="y", labelsize=axFontSize)
ylm = ax.get_ylim()
dlt = max(ylm) - min(ylm)
pzr = plt.plot([0, 0], [min(ylm)-dlt*4, max(ylm)+dlt*4], 'k', linewidth=2)
pzr[0].set_zorder(1)
ax.set_ylim(ylm)
plt.tight_layout()
if ax is None:
fg.savefig('./ensemblesVariabilityAround2deg.png', dpi=400)
def extract1Model(model, inputNcFlPath, outputNcFlPath):
bslnYear = 1995
wls = [1.5, 2.0, 3.0]
noutWls = len(wls)
outRetPer = [10., 20., 50., 100., 200., 500.]
print(' loading the data ...')
dsin = netCDF4.Dataset(inputNcFlPath)
lon = dsin.variables['lon'][:]
lat = dsin.variables['lat'][:]
inRetPer = dsin.variables['return_period'][:]
retLev = dsin.variables['rl'][:]
rlYrs = dsin.variables['year'][:]
dsin.close()
def getRetLevAtYear(retLevAll, yearAll, year, interpFunc=None):
interpFunc = interp1d(yearAll, retLevAll,
axis=1) if interpFunc is None else interpFunc
retLev = interpFunc(year)
return retLev, interpFunc
def getOutRetLev(inRetLev, inRetPer, outRetPer):
interpFunc = interp1d(inRetPer, inRetLev, axis=0)
outRetLev = interpFunc(outRetPer)
return outRetLev
def getOutRetPer(inRetPer, retLevCurrent, retLevBsln):
nx = retLevCurrent.shape[1]
ny = retLevCurrent.shape[2]
nOutRL = retLevBsln.shape[0]
outRetPer = np.zeros([nOutRL, nx, ny]) * np.nan
for ix in range(nx):
for iy in range(ny):
rlcri = retLevCurrent[:, ix, iy]
if np.sum(np.isnan(rlcri)) > 0:
continue
rlbsln = retLevBsln[:, ix, iy]
outRetPer[:, ix,
iy] = interp1d(rlcri,
inRetPer,
axis=0,
fill_value='extrapolate')(rlbsln)
return outRetPer
interpFunc = None
bslnRetLev_, interpFunc = getRetLevAtYear(retLev, rlYrs, bslnYear,
interpFunc)
bslnRetLev = getOutRetLev(bslnRetLev_, inRetPer, outRetPer)
print(' elaborating ...')
shp_ = bslnRetLev.shape
shpOut = [noutWls, shp_[0], shp_[1], shp_[2]]
outYrRetPer = np.zeros(shpOut) * np.nan
outYrRetLev = np.zeros(shpOut) * np.nan
for wl, iwl in zip(wls, range(len(wls))):
wly = getWarmingLevels('rcp85', wl)
outYr = wly[model]
yrRetLev, interpFunc = getRetLevAtYear(retLev, rlYrs, outYr,
interpFunc)
outYrRetLev[iwl, :, :, :] = getOutRetLev(yrRetLev, inRetPer, outRetPer)
outYrRetPer[iwl, :, :, :] = getOutRetPer(inRetPer, yrRetLev,
bslnRetLev)
outYrRetPer[outYrRetPer < 1] = 1
print(' saving the output ...')
if os.path.isfile(outputNcFlPath):
os.remove(outputNcFlPath)
dsout = netCDF4.Dataset(outputNcFlPath, 'w')
dsout.createDimension('lat', len(lat))
dsout.createDimension('lon', len(lon))
dsout.createDimension('warming_lev', noutWls)
dsout.createDimension('baseline_rp', len(outRetPer))
xnc = dsout.createVariable('lon', 'f8', ('lon'))
xnc.description = 'longitude'
xnc.long_name = 'longitude'
xnc.standard_name = 'longitude'
xnc.units = 'degrees_east'
xnc[:] = lon
ync = dsout.createVariable('lat', 'f8', ('lat'))
ync.description = 'latitude'
ync.description = 'latitude'
ync.long_name = 'latitude'
ync.standard_name = 'latitude'
ync.units = 'degrees_north'
ync[:] = lat
wlnc = dsout.createVariable('warming_lev', 'f4', ('warming_lev'))
wlnc.description = 'warming level'
wlnc[:] = wls
bslnRpNc = dsout.createVariable('baseline_rp', 'f4', ('baseline_rp'))
bslnRpNc.description = 'return period at the baseline (years)'
bslnRpNc[:] = outRetPer
bslnYrNc = dsout.createVariable('baseline_year', 'i4')
bslnYrNc.description = 'baseline year'
bslnYrNc[:] = bslnYear
bslnRetLevNc = dsout.createVariable('baseline_return_level', 'f4',
('baseline_rp', 'lon', 'lat'))
bslnRetLevNc.description = 'return levels at baseline'
bslnRetLevNc[:] = bslnRetLev
retPerNc = dsout.createVariable(
'baseline_rp_shift', 'f4',
('warming_lev', 'baseline_rp', 'lon', 'lat'))
retPerNc.description = 'shifted return periods of the baseline return values (years)'
retPerNc[:] = outYrRetPer
retPerNc.coordinates = 'warming_lev baseline_rp lon lat'
retLevNc = dsout.createVariable(
'return_level', 'f4', ('warming_lev', 'baseline_rp', 'lon', 'lat'))
retLevNc.description = 'return levels at warming levels (m**3)'
retLevNc[:] = outYrRetLev
retLevNc.coordinates = 'warming_lev baseline_rp lon lat'
dsout.close()
def loadWlVsScenChangeYMax(
ncDir='/ClimateRun4/multi-hazard/eva',
bslnYear=1995,
warmingLev=2,
retPer=100,
threshold=0,
rlVarName='year_max',
windowHalfSize=10,
flpattern='projection_dis_{scen}_{mdl}_wuConst_statistics.nc'):
# computes the mean relative change
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
dsuparea = netCDF4.Dataset('upArea.nc')
upArea = dsuparea.variables['upArea'][:].transpose()
dsuparea.close()
models = wlyR8.keys()
# menta
#models = [models[0], models[1]]
tamask, _, _ = getAfricaAndTurkeyMask()
tamask = tamask.transpose()
rl_r4 = []
rl_r8 = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
r8year = wlyR8[mdl]
print(' rcp85 w.l. year: ' + str(r8year))
r4year = wlyR4[mdl]
print(' rcp45 w.l. year: ' + str(r4year))
hsz = windowHalfSize
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
year_ = ds.variables['year_all'][:]
yIndxBsln = np.where(year_ == bslnYear)[0][0]
rlBslnR8 = np.nanmean(
ds.variables[rlVarName][yIndxBsln - hsz:yIndxBsln + hsz, :, :], 0)
yIndx = np.where(year_ == r8year)[0][0]
rlR8 = np.nanmean(
ds.variables[rlVarName][yIndx - hsz:yIndx + hsz, :, :], 0)
ds.close()
r8RelChng = (rlR8 - rlBslnR8) / rlBslnR8
if threshold > 0:
cnd = rlBslnR8 < threshold
r8RelChng[cnd] = np.nan
#r8RelChng[r8RelChng < -.15] = np.nan
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
#rlBslnR4 = ds.variables[rlVarName][rpIndx, yIndxBsln, :, :]
rlBslnR4 = rlBslnR8
yIndx = np.where(year_ == r4year)[0][0]
rlR4 = np.nanmean(
ds.variables[rlVarName][yIndx - hsz:yIndx + hsz, :, :], 0)
ds.close()
r4RelChng = (rlR4 - rlBslnR4) / rlBslnR4
if threshold > 0:
cnd = rlBslnR4 < threshold
r4RelChng[cnd] = np.nan
#r4RelChng[r4RelChng < -.15] = np.nan
r8RelChng[upArea < 1e9] = np.nan
r4RelChng[upArea < 1e9] = np.nan
r8RelChng[~tamask] = np.nan
r4RelChng[~tamask] = np.nan
rl_r8.append(r8RelChng)
rl_r4.append(r4RelChng)
rl_r8 = np.array(rl_r8)
rl_r4 = np.array(rl_r4)
relChngDiff = np.nanmean(rl_r8 - rl_r4, 0)
relChngDiff[~tamask] = np.nan
return relChngDiff, np.nanmean(rl_r8, 0), np.nanmean(rl_r4,
0), rl_r8, rl_r4
def plotEnsembles_means(rootDir='/ClimateRun4/multi-hazard/eva/',
gs=None,
showLegend=True):
warmingLevs = [1.5, 2.0]
rcp8, rcp4 = 'rcp85', 'rcp45'
models = modelStr.split()
#models = [models[0], models[1]]
if nTestModels != -1:
models = models[:nTestModels]
outputPngFile = 'scatter_100yrlChange_mean_wls_ensemble_retPer.png'
if gs is None:
fig = plt.figure(figsize=(11, 5))
gs = gridspec.GridSpec(1, 3, width_ratios=[1, 1, .04])
ownsFig = True
else:
ownsFig = False
lon, lat = [], []
axs = []
lgndShown = not showLegend
# lims = {1.5: [-20, 50], 2.0: [-39, 80]}
# lims = {1.5: [-24, 60], 2.0: [-39, 80]}
densityBins = {1.5: 130, 2.0: 130}
lims = {1.5: [-10, 35], 2.0: [-10, 35]}
for warmingLev, iwl in zip(warmingLevs, range(len(warmingLevs))):
wlYearR8 = getWarmingLevels(rcp8, warmingLev)
wlYearR4 = getWarmingLevels(rcp4, warmingLev)
relChng8lst, relChng4lst = [], []
vlsMask = None
for mdl, imdl in zip(models, range(1, len(models) + 1)):
print('getting model ' + mdl)
ncFlNm = '_'.join(
['projection_dis', rcp8, mdl, wuStr, 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR8[mdl]
wrmYear = int(round(wrmYear / 5.) * 5.)
relChngR8, vlsMask = getRelChngs(ncFlPth,
None,
wrmYear,
ncvar='year_mean',
vlsMask=vlsMask)
ncFlNm = '_'.join(
['projection_dis', rcp4, mdl, wuStr, 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR4[mdl]
wrmYear = int(round(wrmYear / 5.) * 5.)
relChngR4, vlsMask = getRelChngs(ncFlPth,
None,
wrmYear,
ncvar='year_mean',
vlsMask=vlsMask)
relChng8lst.append(relChngR8)
relChng4lst.append(relChngR4)
ax = plt.subplot(gs[iwl])
#relChng8 = np.nanmean(np.array(relChng8lst), 0)
#relChng4 = np.nanmean(np.array(relChng4lst), 0)
relChng8 = np.nanmedian(np.array(relChng8lst), 0)
relChng4 = np.nanmedian(np.array(relChng4lst), 0)
showLegend = not lgndShown
dp = plotSingleModel(ax,
relChng8,
relChng4,
'mean discharge\nwrm. lev. ' + str(warmingLev) +
'$^\circ$C',
bold=True,
plotRegression=False,
tickFontSize=13,
labelFontSize=15,
titleFontSize=14,
xLabelPad=-30,
yLabelPad=-15,
showLegend=showLegend,
lim=lims[warmingLev])
plt.xlim(lims[warmingLev])
plt.ylim(lims[warmingLev])
lgndShown = True
axs.append(ax)
cax = plt.subplot(gs[2])
cb = plt.colorbar(dp[-1], ax=ax, cax=cax)
cax.tick_params(labelsize=13, rotation=90)
cb.set_label('density (pixel count)', fontsize=13)
[ax.set_aspect('auto') for ax in axs]
cax.set_aspect('auto')
plt.tight_layout()
if ownsFig:
fig.savefig(outputPngFile, dpi=300)
return axs
def plotTimeSerieVariabilityAtWl_hiExt(ax=None, axmap=None, plotXLabel=True, plotYLabel=True, warmingLev=2.0, bsmp=None):
yrs, rpR8Srs, rpR4Srs, models = loadWlVsScenChange.loadRetPerAllYears(nmodels=nmodels, excludedModels=excludedModels)
rpR8 = getRelChng(yrs, rpR8Srs)
rpR4 = getRelChng(yrs, rpR4Srs)
wlyR8 = getWarmingLevels('rcp85', warmingLev)
r8srs = getNYrsTimeSeriesAroundWl(yrs, rpR8, wlyR8, models)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
r4srs = getNYrsTimeSeriesAroundWl(yrs, rpR4, wlyR4, models)
srstot = np.concatenate([r8srs, r4srs], 0)
rallmeanWl, rallsigmaWl = getMeanAndSigma(r8srs)
cndPosi = rallmeanWl > 0
cndNega = rallmeanWl < 0
cnd = cndPosi if positiveChanges else cndNega
r8meanWl, r8sigmaWl = getMeanAndSigma(r8srs)
cndMtx = np.tile(cnd, [r8srs.shape[0], r8srs.shape[1], 1, 1])
r8srs[~cndMtx] = np.nan
r8mean_ = np.nanmean(r8srs, 0)
r8mean = r8mean_.reshape(r8mean_.shape[0], r8mean_.shape[1]*r8mean_.shape[2])
#r8mean = np.nanmedian(r8mean, 1)
r8mean = np.nanmean(r8mean, 1)
r8spmean = r8srs.reshape(r8srs.shape[0], r8srs.shape[1], r8srs.shape[2]*r8srs.shape[3])
r8spmean = np.nanmean(r8spmean, 2)
#r8spmean = np.nanmean(r8spmean, 2)
r8skew = stats.skew(r8spmean[:, 0])
r8sigma = np.std(r8spmean[:, 0])
r8mean = np.nanmean(r8spmean, 0)
r4meanWl, r4sigmaWl = getMeanAndSigma(r4srs)
cndMtx = np.tile(cnd, [r4srs.shape[0], r4srs.shape[1], 1, 1])
r4srs[~cndMtx] = np.nan
r4mean_ = np.nanmean(r4srs, 0)
r4mean = r4mean_.reshape(r4mean_.shape[0], r4mean_.shape[1]*r4mean_.shape[2])
#r4mean = np.nanmedian(r4mean, 1)
r4mean = np.nanmean(r4mean, 1)
r4spmean = r4srs.reshape(r4srs.shape[0], r4srs.shape[1], r4srs.shape[2]*r4srs.shape[3])
r4spmean = np.nanmean(r4spmean, 2)
#r4spmean = np.nanmean(r4spmean, 2)
r4skew = stats.skew(r4spmean[:, 0])
r4sigma = np.std(r4spmean[:, 0])
r4mean = np.nanmean(r4spmean, 0)
ensTot = np.concatenate([r8spmean, r4spmean], 0)
stdDevTot = np.std(ensTot, 0)*100.
meanTot = (r4mean + r8mean)/2.*100.
if ax is None:
fg = plt.figure(figsize=[5, 5])
ax = fg.gca()
doPlotChanges(ax, warmingLev, meanTot, stdDevTot, r8spmean, r4spmean, r8mean, r4mean, plotXLabel, plotYLabel)
if not axmap is None:
rallmeanWlCnd = rallmeanWl.copy()
rallmeanWlCnd[~cnd] = np.nan
bsmp = doPlotAreaMap(axmap, rallmeanWlCnd, bsmp)
plt.tight_layout()
if ax is None:
fg.savefig('./ensemblesVariabilityAround2deg.png', dpi=400)
return bsmp
def loadMdlsAtWl(ncDir='/ClimateRun4/multi-hazard/eva',
bslnYear=1995,
warmingLev=2,
rlVarName='rl',
numberOfModels=-1):
flpattern = 'projection_dis_{scen}_{mdl}_wuConst_statistics.nc'
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
models = wlyR8.keys()
models.sort()
numberOfModels = len(models) if numberOfModels == -1 else numberOfModels
#tamask, _, _ = getAfricaAndTurkeyMask()
#tamask = tamask.transpose()
bsln = []
rl_r4 = []
rl_r8 = []
for mdl, imdl in zip(models[:numberOfModels], range(numberOfModels)):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
r8year = wlyR8[mdl]
r8yearInf = int(np.floor(r8year / float(5)) * 5)
print(' rcp85 w.l. year: ' + str(r8year))
r4year = wlyR4[mdl]
r4yearInf = int(np.floor(r4year / float(5)) * 5)
print(' rcp45 w.l. year: ' + str(r4year))
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
retper_ = ds.variables['return_period'][:]
year_ = ds.variables['year'][:]
yIndxBsln = np.where(year_ == bslnYear)[0][0]
rlBslnR8 = ds.variables[rlVarName][:, yIndxBsln, :, :]
yIndx = np.where(year_ == r8yearInf)[0][0]
rlR8_ = ds.variables[rlVarName][:, yIndx:yIndx + 2, :, :]
ds.close()
rlR8 = interp1d(year_[yIndx:yIndx + 2], rlR8_, axis=1)(r8year)
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
rlBslnR4 = ds.variables[rlVarName][:, yIndxBsln, :, :]
yIndx = np.where(year_ == r4yearInf)[0][0]
rlR4_ = ds.variables[rlVarName][:, yIndx:yIndx + 2, :, :]
ds.close()
rlR4 = interp1d(year_[yIndx:yIndx + 2], rlR4_, axis=1)(r4year)
rlBsln = rlBslnR8
#rlBsln[~tamask] = np.nan
#rlR4[~tamask] = np.nan
#rlR8[~tamask] = np.nan
bsln.append(rlBsln)
rl_r8.append(rlR8)
rl_r4.append(rlR4)
bsln = np.array(bsln)
rl_r8 = np.array(rl_r8)
rl_r4 = np.array(rl_r4)
return bsln, rl_r8, rl_r4, retper_
def plotEnsembles(rootDir='/ClimateRun4/multi-hazard/eva/', retPer=100):
warmingLevs = [1.5, 2.0]
rcp8, rcp4 = 'rcp85', 'rcp45'
models = """
IPSL-INERIS-WRF331F_BC
SMHI-RCA4_BC_CNRM-CERFACS-CNRM-CM5
SMHI-RCA4_BC_ICHEC-EC-EARTH
SMHI-RCA4_BC_IPSL-IPSL-CM5A-MR
SMHI-RCA4_BC_MOHC-HadGEM2-ES
SMHI-RCA4_BC_MPI-M-MPI-ESM-LR
CLMcom-CCLM4-8-17_BC_CNRM-CERFACS-CNRM-CM5
CLMcom-CCLM4-8-17_BC_ICHEC-EC-EARTH
CLMcom-CCLM4-8-17_BC_MPI-M-MPI-ESM-LR
DMI-HIRHAM5-ICHEC-EC-EARTH_BC
KNMI-RACMO22E-ICHEC-EC-EARTH_BC
"""
models = models.split()
outputPngFile = 'scatter_100yrlChange_wls_ensemble_retPer' + str(
retPer) + '.png'
fig = plt.figure(figsize=(11, 5))
gs = gridspec.GridSpec(1, 3, width_ratios=[1, 1, .04])
lon, lat = [], []
axs = []
lgndShown = False
for warmingLev, iwl in zip(warmingLevs, range(len(warmingLevs))):
wlYearR8 = getWarmingLevels(rcp8, warmingLev)
wlYearR4 = getWarmingLevels(rcp4, warmingLev)
relChng8lst, relChng4lst = [], []
for mdl, imdl in zip(models, range(1, len(models) + 1)):
print('getting model ' + mdl)
ncFlNm = '_'.join(
['projection_dis', rcp8, mdl, 'wuChang', 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR8[mdl]
wrmYear = int(round(wrmYear / 5.) * 5.)
relChngR8 = getRelChngs(ncFlPth, retPer, wrmYear)
ncFlNm = '_'.join(
['projection_dis', rcp4, mdl, 'wuChang', 'statistics.nc'])
ncFlPth = os.path.join(rootDir, ncFlNm)
wrmYear = wlYearR4[mdl]
wrmYear = int(round(wrmYear / 5.) * 5.)
relChngR4 = getRelChngs(ncFlPth, retPer, wrmYear)
relChng8lst.append(relChngR8)
relChng4lst.append(relChngR4)
ax = plt.subplot(gs[iwl])
relChng8 = np.nanmean(np.array(relChng8lst), 0)
relChng4 = np.nanmean(np.array(relChng4lst), 0)
showLegend = not lgndShown
dp = plotSingleModel(
ax,
relChng8,
relChng4,
str(retPer) +
'-year ret. lev.\nensemble rcp85 vs rcp45,\nwrm. lev. ' +
str(warmingLev) + '$^\circ$',
bold=True,
plotRegression=False,
tickFontSize=13,
labelFontSize=15,
titleFontSize=14,
xLabelPad=-30,
yLabelPad=-15,
showLegend=showLegend)
lgndShown = True
axs.append(ax)
cax = plt.subplot(gs[2])
cb = plt.colorbar(dp[-1], ax=ax, cax=cax)
cax.tick_params(labelsize=13, rotation=90)
cb.set_label('density (pixel count)', fontsize=13)
[ax.set_aspect('auto') for ax in axs]
cax.set_aspect('auto')
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def computeRlChngPValueAtWarmingLevBtwScen(ncDir='/ClimateRun4/multi-hazard/eva', warmingLev=2, retPer=100, minThreshold=0):
flpattern = 'projection_dis_{scen}_{mdl}_wuChang_statistics.nc'
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
models = wlyR8.keys()
rl_r4 = []
serl_r4 = []
rl_r8 = []
serl_r8 = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flr8 = flpattern.format(scen='rcp85', mdl=mdl)
flr8pth = os.path.join(ncDir, flr8)
flr4 = flpattern.format(scen='rcp45', mdl=mdl)
flr4pth = os.path.join(ncDir, flr4)
r8year = wlyR8[mdl]
r8yearInf = int(np.floor(r8year/float(5))*5)
print(' rcp85 w.l. year: ' + str(r8year))
r4year = wlyR4[mdl]
r4yearInf = int(np.floor(r4year/float(5))*5)
print(' rcp45 w.l. year: ' + str(r4year))
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
retper_ = ds.variables['return_period'][:]
rpIndx = np.where(retper_==retPer)[0][0]
year_ = ds.variables['year'][:]
yIndx = np.where(year_==r8yearInf)[0][0]
rlR8_ = ds.variables['rl'][rpIndx, yIndx:yIndx+2, :, :]
serlR8_ = ds.variables['se_rl'][rpIndx, yIndx:yIndx+2, :, :]
ds.close()
rlR8 = interp1d(year_[yIndx:yIndx+2], rlR8_, axis=0)(r8year)
serlR8 = interp1d(year_[yIndx:yIndx+2], serlR8_, axis=0)(r8year)
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
yIndx = np.where(year_==r4yearInf)[0][0]
rlR4_ = ds.variables['rl'][rpIndx, yIndx:yIndx+2, :, :]
serlR4_ = ds.variables['se_rl'][rpIndx, yIndx:yIndx+2, :, :]
ds.close()
rlR4 = interp1d(year_[yIndx:yIndx+2], rlR4_, axis=0)(r4year)
serlR4 = interp1d(year_[yIndx:yIndx+2], serlR4_, axis=0)(r4year)
rl_r8.append(rlR8)
serl_r8.append(serlR8)
rl_r4.append(rlR4)
serl_r4.append(serlR4)
rl_r8 = np.array(rl_r8)
rl_r8[rl_r8 <= .1] = .1
serl_r8 = np.array(serl_r8)
serl_r8[serl_r8 <= .1] = .1
rl_r4 = np.array(rl_r4)
rl_r4[rl_r4 <= .1] = .1
serl_r4 = np.array(rl_r4)
serl_r4[serl_r4 <= .1] = .1
cnd = np.nanmean(rl_r8, 0) < minThreshold
cnd = np.tile(cnd, [len(models), 1, 1])
rl_r8[cnd] = np.nan
serl_r8[cnd] = np.nan
rl_r4[cnd] = np.nan
serl_r4[cnd] = np.nan
pvalue = computeRlChngPValueMontecarlo(rl_r4, serl_r4, rl_r8, serl_r8)
#pvalue = computeRlChngPValueMeansOnly(rl_r4, rl_r8)
agrMdlCnt, _ = countAgreeingModelsAndGetStdDev(rl_r4, rl_r8)
return pvalue, agrMdlCnt
def plotEnsembleChangVsConstWaterUse(rootDir='/DATA/JEODPP/eos/projects/CRITECH/ADAPTATION/ClimateRuns/LisfloodEuroCordex/'):
warmingLevs = [1.5, 2.0]
rcp8, rcp4 = 'rcp85', 'rcp45'
scenarios = [rcp8, rcp4]
wuConst, wuChang = 'wuConst', 'wuChang'
models = """
IPSL-INERIS-WRF331F_BC
SMHI-RCA4_BC_CNRM-CERFACS-CNRM-CM5
SMHI-RCA4_BC_ICHEC-EC-EARTH
SMHI-RCA4_BC_IPSL-IPSL-CM5A-MR
SMHI-RCA4_BC_MOHC-HadGEM2-ES
SMHI-RCA4_BC_MPI-M-MPI-ESM-LR
CLMcom-CCLM4-8-17_BC_CNRM-CERFACS-CNRM-CM5
CLMcom-CCLM4-8-17_BC_ICHEC-EC-EARTH
CLMcom-CCLM4-8-17_BC_MPI-M-MPI-ESM-LR
DMI-HIRHAM5-ICHEC-EC-EARTH_BC
KNMI-RACMO22E-ICHEC-EC-EARTH_BC
"""
models = models.split()
outputPngFile = 'scatter_minDis_wuChngVSwuConst_wls_ensemble.png'
fig = plt.figure(figsize=(4, 4))
gs = gridspec.GridSpec(2, 2)
for warmingLev, iwl in zip(warmingLevs, range(len(warmingLevs))):
print('elaborating wl ' + str(warmingLev))
for scen, iscen in zip(scenarios, range(len(scenarios))):
print(' scenario ' + scen)
wlYear = getWarmingLevels(scen, warmingLev)
relChngWuChangLst, relChngWuConstLst = [], []
for mdl, imdl in zip(models, range(1, len(models) + 1)):
print(' getting model ' + mdl)
fldirHist = os.path.join(rootDir, 'historical', mdl, wuConst)
fldirScenWuChang = os.path.join(rootDir, scen, mdl, wuChang)
wrmYear = wlYear[mdl]
cacheIdData = '_'.join([mdl, scen, wuChang])
cacheIdResult = cacheIdData + '_' + str(warmingLev)
relChngWuChang = getRelChngs(cacheIdData, cacheIdResult, fldirHist, fldirScenWuChang, wrmYear)
fldirScenWuConst = os.path.join(rootDir, scen, mdl, wuConst)
cacheIdData = '_'.join([mdl, scen, wuConst])
cacheIdResult = cacheIdData + '_' + str(warmingLev)
relChngWuConst = getRelChngs(cacheIdData, cacheIdResult, fldirHist, fldirScenWuConst, wrmYear)
relChngWuChangLst.append(relChngWuChang)
relChngWuConstLst.append(relChngWuConst)
ax = plt.subplot(gs[iwl, iscen])
relChngWuChang = np.nanmean(np.array(relChngWuChangLst), 0)
relChngWuConst = np.nanmean(np.array(relChngWuConstLst), 0)
plotSingleModel(ax, relChngWuChang, relChngWuConst,
'Ensemble, constant vs changing w.u.,\n minimum discharge\n' + scen + ', wrm. lev. ' + str(warmingLev) + '$^\circ$',
bold=True, plotRegression=False, plotIdealFit=False, tickFontSize=5, labelFontSize=7, titleFontSize=5,
xLabelPad=-25, yLabelPad=-10, xlabel='const. w.u. ($\Delta$ %)', ylabel='chng. w.u. ($\Delta$ %)', xlim=[-80, 170], ylim=[-80, 170])
plt.tight_layout()
fig.savefig(outputPngFile, dpi=300)
def loadMeanPrecipitationChangesAtWl(
ncRootDir='/DATA/ClimateData/cordexEurope/yearlymeans/',
bslnYear=1995,
warmingLev=2,
threshold=0,
flname='pr.nc',
nmodels=-1,
timeWindowHalfSize=15):
# computes the mean relative change
wlyR8 = getWarmingLevels('rcp85', warmingLev)
wlyR4 = getWarmingLevels('rcp45', warmingLev)
models = wlyR8.keys()
if nmodels > -1:
models = models[:nmodels]
tamask, _, _ = getAfricaAndTurkeyMask()
tamask = tamask
rlVarName1 = 'pr'
rlVarname2 = 'prAdjust'
rl_r4 = []
rl_r8 = []
for mdl, imdl in zip(models, range(len(models))):
print('model ' + mdl)
flhistpth = os.path.join(ncRootDir, mdl, 'historical', flname)
flr8pth = os.path.join(ncRootDir, mdl, 'rcp85', flname)
flr4pth = os.path.join(ncRootDir, mdl, 'rcp45', flname)
r8year = wlyR8[mdl]
r4year = wlyR4[mdl]
print(' loading file ' + flhistpth)
ds = netCDF4.Dataset(flhistpth)
rlVarName = 'pr' if 'pr' in ds.variables else 'prAdjust'
tmnc = ds.variables['time']
tm = netCDF4.num2date(tmnc[:], tmnc.units, tmnc.calendar)
yearHist = np.array([t.year for t in tm])
yIndxBsln = np.where(yearHist == bslnYear)[0][0]
minIndx = np.max([yIndxBsln - timeWindowHalfSize, 0])
maxIndx = np.min([yIndxBsln + timeWindowHalfSize, len(yearHist)])
rlBsln = np.nanmean(ds.variables[rlVarName][minIndx:maxIndx, :, :], 0)
ds.close()
print(' loading file ' + flr8pth)
ds = netCDF4.Dataset(flr8pth)
rlVarName = 'pr' if 'pr' in ds.variables else 'prAdjust'
tmnc = ds.variables['time']
tm = netCDF4.num2date(tmnc[:], tmnc.units, tmnc.calendar)
yearR = np.array([t.year for t in tm])
yIndx = np.where(yearR == r8year)[0][0]
minIndx = np.max([yIndx - timeWindowHalfSize, 0])
maxIndx = np.min([yIndx + timeWindowHalfSize, len(yearR)])
rlR8 = np.nanmean(ds.variables[rlVarName][minIndx:maxIndx, :, :], 0)
r8RelChng = (rlR8 - rlBsln) / rlBsln
if threshold > 0:
cnd = rlBslnR8 < threshold
r8RelChng[cnd] = np.nan
#r8RelChng[r8RelChng < -.15] = np.nan
ds.close()
print(' loading file ' + flr4pth)
ds = netCDF4.Dataset(flr4pth)
rlVarName = 'pr' if 'pr' in ds.variables else 'prAdjust'
tmnc = ds.variables['time']
tm = netCDF4.num2date(tmnc[:], tmnc.units, tmnc.calendar)
yearR = np.array([t.year for t in tm])
yIndx = np.where(yearR == r4year)[0][0]
minIndx = np.max([yIndx - timeWindowHalfSize, 0])
maxIndx = np.min([yIndx + timeWindowHalfSize, len(yearR)])
rlR4 = np.nanmean(ds.variables[rlVarName][minIndx:maxIndx, :, :], 0)
r4RelChng = (rlR4 - rlBsln) / rlBsln
if threshold > 0:
cnd = rlBslnR4 < threshold
r4RelChng[cnd] = np.nan
#r4RelChng[r4RelChng < -.15] = np.nan
ds.close()
#r8RelChng[upArea < 1e9] = np.nan
#r4RelChng[upArea < 1e9] = np.nan
r8RelChng[~tamask] = np.nan
r4RelChng[~tamask] = np.nan
rl_r8.append(r8RelChng)
rl_r4.append(r4RelChng)
rl_r8 = np.array(rl_r8)
rl_r4 = np.array(rl_r4)
relChngDiff = np.nanmean(rl_r8 - rl_r4, 0)
relChngDiff[~tamask] = np.nan
return relChngDiff, np.nanmean(rl_r8, 0), np.nanmean(rl_r4,
0), rl_r8, rl_r4