def plotEnglandPoints():
outPng = 'englandPoints_ANOVA.png'
warmingLev = 2.0
relChngDiffMean, rc_r8_mean, rc_r4_mean, rc_r8all_mean, rc_r4all_mean = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
_, _, _, pValueMean, _, _ = anovaAnalysis(relChngDiffMean, rc_r8_mean,
rc_r4_mean, rc_r8all_mean,
rc_r4all_mean)
relChngDiffExt, rc_r8_ext, rc_r4_ext, rc_r8all_ext, rc_r4all_ext = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
nmodels=nmodels,
rlVarName='rl',
retPer=100)
_, _, _, pValueExt, _, _ = anovaAnalysis(relChngDiffMean, rc_r8_mean,
rc_r4_mean, rc_r8all_mean,
rc_r4all_mean)
dslonlat = netCDF4.Dataset('lonlat.nc')
lon = dslonlat.variables['lon'][:].transpose()
lat = dslonlat.variables['lat'][:].transpose()
dslonlat.close()
cndlonlat = np.logical_and(lon < 2, lat > 50)
cndlonlatMtx = np.tile(cndlonlat, [rc_r8all_mean.shape[0], 1, 1])
cndpval = pValueMean <= .05
cndpvalMtx = np.tile(pvalcnd, [rc_r8all_mean.shape[0], 1, 1])
cnd = np.logical_and(cndpvalMtx, cndlonlatMtx)
rc_r8all_mean[~cnd] = np.nan
rc_r4all_mean[~cnd] = np.nan
shp = rc_r4all_mean.shape
rc_r8_ = rc_r8all_mean.reshape([shp[0], shp[1] * shp[2]])
rc_r4_ = rc_r4all_mean.reshape([shp[0], shp[1] * shp[2]])
mn_r8_mean = np.nanmean(rc_r8_, 1) * 100
mn_r4_mean = np.nanmean(rc_r4_, 1) * 100
rc_r8all_ext[~cnd] = np.nan
rc_r4all_ext[~cnd] = np.nan
shp = rc_r4all_ext.shape
rc_r8_ = rc_r8all_ext.reshape([shp[0], shp[1] * shp[2]])
rc_r4_ = rc_r4all_ext.reshape([shp[0], shp[1] * shp[2]])
mn_r8_ext = np.nanmean(rc_r8_, 1) * 100
mn_r4_ext = np.nanmean(rc_r4_, 1) * 100
f = plt.figure(figsize=(7, 3))
gs = gridspec.GridSpec(1, 2)
vv = np.concatenate([mn_r4_mean, mn_r8_mean, mn_r4_ext, mn_r8_ext], axis=0)
ylm = [np.min(vv) - 1, np.max(vv) + 1]
ax0 = plt.subplot(gs[0])
ones = np.ones(mn_r8_mean.shape)
plt.plot(ones / 3., mn_r4_mean, 'o', color='green', label='RCP4.5')
plt.plot(ones * 2. / 3., mn_r8_mean, 'o', color='navy', label='RCP8.5')
plt.xlim([0, 1])
plt.ylim(ylm)
plt.ylabel('projected change at $2^\circ C$ (%)')
plt.xticks([])
plt.legend(loc='lower right')
plt.text(.05, 17, 'a: mean dis.', fontsize=11)
plt.grid('on')
ax1 = plt.subplot(gs[1])
ones = np.ones(mn_r8_ext.shape)
plt.plot(ones / 3., mn_r4_ext, 'o', color='green', label='RCP4.5')
plt.plot(ones * 2. / 3., mn_r8_ext, 'o', color='navy', label='RCP8.5')
plt.xlim([0, 1])
plt.ylim(ylm)
ax1.set_yticklabels('')
plt.xticks([])
plt.text(.05, 17, 'b: extreme high dis.', fontsize=11)
plt.grid('on')
plt.tight_layout()
f.savefig(outPng, dpi=300)
def plotFigureANOVA(varType='extHigh', ncDir='/ClimateRun4/multi-hazard/eva'):
#varType can be extHigh, extLow, mean
outPng = 'anova_' + varType + '.png'
f = plt.figure(figsize=(9, 12))
gs = gridspec.GridSpec(3, 3, width_ratios=[1, 1, .05])
mp = None
#### at 1.5 ####
warmingLev = 1.5
if varType == 'mean':
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
descrTxt = 'Means'
sigmaMax = 25
elif varType == 'extHigh':
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
nmodels=nmodels,
rlVarName='rl',
retPer=100,
shapeParamNcVarName='shape_fit')
descrTxt = 'High extremes'
sigmaMax = 30
elif varType == 'extLow':
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
nmodels=nmodels,
rlVarName='rl_min',
retPer=15,
threshold=.1,
shapeParamNcVarName='')
descrTxt = 'Low extremes'
sigmaMax = 50
sigmaTot, sigmaWithin, sigmaBetween, pValue, effectSize, pValueTtest = anovaAnalysis(
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all)
printAnovaStats(sigmaTot, sigmaWithin, sigmaBetween, pValue, pValueTtest)
ax00 = plt.subplot(gs[0, 0])
plt00, mp = plotSigma(ax00,
sigmaWithin * 100,
mp,
'a: $\sigma_{within}, 1.5^\circ C$',
txt2=descrTxt,
sigmamax=sigmaMax)
ax10 = plt.subplot(gs[1, 0])
plt10, mp = plotSigma(ax10,
sigmaBetween * 100,
mp,
'c: $\sigma_{between}, 1.5^\circ C$',
sigmamax=sigmaMax)
ax20 = plt.subplot(gs[2, 0])
_, _, mp = plotPvalue(ax20, pValue, mp, 'e: p-val of interpathway diff.')
#### at 2.0 ####
warmingLev = 2.0
if varType == 'mean':
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
elif varType == 'extHigh':
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
nmodels=nmodels,
rlVarName='rl',
retPer=100)
elif varType == 'extLow':
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
nmodels=nmodels,
rlVarName='rl_min',
retPer=15,
threshold=.1)
sigmaTot, sigmaWithin, sigmaBetween, pValue, effectSize, pValueTtest = anovaAnalysis(
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all)
printAnovaStats(sigmaTot, sigmaWithin, sigmaBetween, pValue, pValueTtest)
ax01 = plt.subplot(gs[0, 1])
plt01, mp = plotSigma(ax01,
sigmaWithin * 100,
mp,
'b: $\sigma_{within}, 2.0^\circ C$',
sigmamax=sigmaMax)
ax11 = plt.subplot(gs[1, 1])
plt11, mp = plotSigma(ax11,
sigmaBetween * 100,
mp,
'd: $\sigma_{between}, 2.0^\circ C$',
sigmamax=sigmaMax)
ax21 = plt.subplot(gs[2, 1])
_, _, mp = plotPvalue(ax21, pValue, mp, 'f: p-val of interpathway diff.')
cax = plt.subplot(gs[:2, 2])
cb = plt.colorbar(plt01, ax=ax01, cax=cax)
cb.set_label('$\sigma$ (%)')
ax00.set_aspect('auto')
ax10.set_aspect('auto')
ax20.set_aspect('auto')
ax01.set_aspect('auto')
ax11.set_aspect('auto')
ax21.set_aspect('auto')
cax.set_aspect('auto')
plt.tight_layout()
f.savefig(outPng, dpi=300)
def plotGrossEnsembles_mean(ncDir='/ClimateRun4/multi-hazard/eva'):
outPng = 'wlRelChngGrossEnsembles_mean.png'
f = plt.figure(figsize=(9, 8))
gs = gridspec.GridSpec(2, 3, width_ratios=[1, 1, .05])
mp = None
warmingLev = 1.5
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4) / 2.
ax0 = plt.subplot(gs[0, 0])
pcl, mp = plotRelChngDiff(ax0,
rc_mega,
mp,
'a: $\Delta Q_M$ at $' + str(warmingLev) +
'^\circ$',
vmax=25)
#sigma_im = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
#sigmaT = getTimeSigmaGrossEnsemble(warmingLev, ncDir=ncDir)
#sigma = np.sqrt(sigma_im**2. + sigmaT**2.)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
ax1 = plt.subplot(gs[1, 0])
pcl, mp = plotSigma(
ax1,
sigma * 100,
rc_mega * 100,
mp,
'c: $\sigma_M$ (%) at $' + str(warmingLev) + '^\circ$',
prcTxtTmpl='% of pixel where $\|\Delta Q_M\| > \sigma_M$: {p:2.0f}%',
sigmamax=25)
warmingLev = 2.0
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4) / 2.
ax2 = plt.subplot(gs[0, 1])
pclChng, mp = plotRelChngDiff(ax2,
rc_mega,
mp,
'b: $\Delta Q_M$ at $' + str(warmingLev) +
'^\circ$',
vmax=25)
#sigma_im = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
#sigmaT = getTimeSigmaGrossEnsemble(warmingLev, ncDir=ncDir)
#sigma = np.sqrt(sigma_im**2. + sigmaT**2.)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
ax3 = plt.subplot(gs[1, 1])
pclSigma, mp = plotSigma(
ax3,
sigma * 100,
rc_mega * 100,
mp,
'd: $\sigma_M$ (%) at $' + str(warmingLev) + '^\circ$',
prcTxtTmpl='% of pixel where $\|\Delta Q_M\| > \sigma_M$: {p:2.0f}%',
sigmamax=25)
cax1 = plt.subplot(gs[0, 2])
cb = plt.colorbar(pclChng, ax=ax2, cax=cax1)
cb.set_label('$\Delta Q_M$ (%)')
cax2 = plt.subplot(gs[1, 2])
cb = plt.colorbar(pclSigma, ax=ax3, cax=cax2)
cb.set_label('$\sigma_M$ (%)')
ax0.set_aspect('auto')
ax1.set_aspect('auto')
ax2.set_aspect('auto')
ax3.set_aspect('auto')
cax1.set_aspect('auto')
cax2.set_aspect('auto')
plt.tight_layout()
f.savefig(outPng, dpi=300)
def plotGrossEnsembleChange(ncDir='/ClimateRun4/multi-hazard/eva'):
outPng = 'wlRelChngGrossEnsembles.png'
f = plt.figure(figsize=(9, 12))
gs = gridspec.GridSpec(3, 3, width_ratios=[1, 1, .05])
mp = None
def writeSigmaRatioTxt(ax, sigma, relChng, varname):
sigma_ratio = sigma / np.abs(relChng)
percSign = float(np.nansum(sigma_ratio <= 1)) / np.nansum(
np.logical_not(np.isnan(sigma_ratio)))
prcTxtTmpl = '% of pixel where $\|{varname}\| > \sigma$: {p:2.0f}%'
prcTxt = prcTxtTmpl.format(varname=varname, p=percSign * 100)
#prcTxt = '% of pixel where $\|\Delta ' + varname + '\| > \sigma$: {p:2.0f}%'.format(p=percSign*100)
plt.axes(ax)
txtpos = mp(-24.3, 70.3)
bb = {'boxstyle': 'square,pad=0', 'ec': 'none', 'fc': 'w'}
plt.annotate(prcTxt,
xy=txtpos,
xycoords='data',
xytext=txtpos,
textcoords='data',
fontsize=10,
bbox=bb)
warmingLev = 1.5
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4) / 2.
ax00 = plt.subplot(gs[0, 0])
pcl, mp = plotRelChngDiff(ax00,
rc_mega,
mp,
'a: $\Delta Q_{H100}$ at $' + str(warmingLev) +
'^\circ$',
vmax=50)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax00, sigma, rc_mega, '\Delta Q_{H100}')
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4) / 2.
ax10 = plt.subplot(gs[1, 0])
pcl, mp = plotRelChngDiff(ax10,
rc_mega,
mp,
'b: $\Delta Q_M$ at $' + str(warmingLev) +
'^\circ$',
vmax=50)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax10, sigma, rc_mega, '\Delta Q_M')
retPer = 15
rlVarName = 'rl_min'
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
rlVarName=rlVarName,
retPer=retPer,
nmodels=nmodels,
threshold=.1)
rc_mega = (rc_r8 + rc_r4) / 2.
ax20 = plt.subplot(gs[2, 0])
pcl, mp = plotRelChngDiff(ax20,
rc_mega,
mp,
'c: $\Delta Q_{L15}$ at $' + str(warmingLev) +
'^\circ$',
vmax=50)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax20, sigma, rc_mega, '\Delta Q_{L15}')
warmingLev = 2.0
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4) / 2.
ax01 = plt.subplot(gs[0, 1])
pcl, mp = plotRelChngDiff(ax01,
rc_mega,
mp,
'd: $\Delta Q_{H100}$ at $' + str(warmingLev) +
'^\circ$',
vmax=50)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax01, sigma, rc_mega, '\Delta Q_{H100}')
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(
ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4) / 2.
ax11 = plt.subplot(gs[1, 1])
pcl, mp = plotRelChngDiff(ax11,
rc_mega,
mp,
'e: $\Delta Q_M$ at $' + str(warmingLev) +
'^\circ$',
vmax=50)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax11, sigma, rc_mega, '\Delta Q_{H100}')
retPer = 15
rlVarName = 'rl_min'
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadWlVsScenChange(
ncDir=ncDir,
warmingLev=warmingLev,
rlVarName=rlVarName,
retPer=retPer,
nmodels=nmodels,
threshold=.1)
rc_mega = (rc_r8 + rc_r4) / 2.
ax21 = plt.subplot(gs[2, 1])
pcl, mp = plotRelChngDiff(ax21,
rc_mega,
mp,
'f: $\Delta Q_{L15}$ at $' + str(warmingLev) +
'^\circ$',
vmax=50)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax21, sigma, rc_mega, '\Delta Q_M')
cax00 = plt.subplot(gs[:, 2])
cb = plt.colorbar(pcl, ax=ax21, cax=cax00)
cb.set_label('%', fontsize=13)
cb.ax.tick_params(labelsize=11)
sigma = np.nanstd(np.concatenate([rc_r8all, rc_r4all], 0), 0)
writeSigmaRatioTxt(ax21, sigma, rc_mega, '\Delta Q_{L15}')
ax00.set_aspect('auto')
ax10.set_aspect('auto')
ax20.set_aspect('auto')
ax01.set_aspect('auto')
ax11.set_aspect('auto')
ax21.set_aspect('auto')
cax00.set_aspect('auto')
plt.tight_layout()
f.savefig(outPng, dpi=300)
def plotGrossEnsembles_mean(ncDir='/ClimateRun4/multi-hazard/eva'):
outPng = 'wlRelChngGrossEnsembles_mean.png'
f = plt.figure(figsize=(9,12))
gs = gridspec.GridSpec(3, 3, width_ratios=[1,1,.05])
mp = None
def writeSigmaRatioTxt(ax, sigma, relChng, varname):
sigma_ratio = sigma/np.abs(relChng)
percSign = float(np.nansum(sigma_ratio <= 1))/np.nansum(np.logical_not(np.isnan(sigma_ratio)))
prcTxtTmpl = '% of pixel where $\|{varname}\| > \sigma$: {p:2.0f}%'
prcTxt = prcTxtTmpl.format(varname=varname, p=percSign*100)
plt.axes(ax)
txtpos = mp(-24.3, 70.2)
bb = {'boxstyle': 'square,pad=0', 'ec': 'none', 'fc': 'w'}
plt.annotate(prcTxt, xy=txtpos, xycoords='data', xytext=txtpos, textcoords='data', fontsize=10, bbox=bb)
warmingLev = 1.5
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4)/2.
sigmaTot, sigmaWithin, sigmaBetween, pValue, effectSize = anovaAnalysis(relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all)
ax0 = plt.subplot(gs[0,0])
pcl, mp = plotRelChngDiff(ax0, rc_mega, sigmaTot, mp, 'a: $\Delta Q_M$ at $' + str(warmingLev) +'^\circ$', vmax=25)
printPosNegChanges(warmingLev, rc_mega)
writeSigmaRatioTxt(ax0, sigmaTot, rc_mega, '\Delta Q_M')
ax1 = plt.subplot(gs[1,0])
pcl, mp = plotSigma(ax1, sigmaWithin*100, rc_mega*100, mp, 'c: $\sigma_{within}$ (%) at $' + str(warmingLev) +'^\circ$',
prcTxtTmpl = '', sigmamax=25)
ax2 = plt.subplot(gs[2,0])
pcl, mp = plotSigma(ax2, sigmaBetween*100, rc_mega*100, mp, 'e: $\sigma_{between}$ (%) at $' + str(warmingLev) +'^\circ$',
prcTxtTmpl = '', sigmamax=25)
warmingLev = 2.0
relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all = ldEnsmbl.loadMeanChangesAtWl(ncDir=ncDir, warmingLev=warmingLev, nmodels=nmodels)
rc_mega = (rc_r8 + rc_r4)/2.
sigmaTot, sigmaWithin, sigmaBetween, pValue, effectSize = anovaAnalysis(relChngDiff, rc_r8, rc_r4, rc_r8all, rc_r4all)
ax3 = plt.subplot(gs[0,1])
pclChng, mp = plotRelChngDiff(ax3, rc_mega, sigmaTot, mp, 'b: $\Delta Q_M$ at $' + str(warmingLev) +'^\circ$', vmax=30)
printPosNegChanges(warmingLev, rc_mega)
writeSigmaRatioTxt(ax3, sigmaTot, rc_mega, '\Delta Q_M')
ax4 = plt.subplot(gs[1,1])
pclSigma, mp = plotSigma(ax4, sigmaWithin*100, rc_mega*100, mp, 'd: $\sigma_{within}$ (%) at $' + str(warmingLev) +'^\circ$',
prcTxtTmpl = '', sigmamax=30)
ax5 = plt.subplot(gs[2,1])
pclSigma, mp = plotSigma(ax5, sigmaBetween*100, rc_mega*100, mp, 'f: $\sigma_{between}$ (%) at $' + str(warmingLev) +'^\circ$',
prcTxtTmpl = '', sigmamax=30)
cax1 = plt.subplot(gs[0,2])
cb = plt.colorbar(pclChng, ax=ax2, cax=cax1)
cb.set_label('$\Delta Q_M$ (%)')
cax2 = plt.subplot(gs[1:,2])
cb = plt.colorbar(pclSigma, ax=ax3, cax=cax2)
cb.set_label('$\sigma_M$ (%)')
ax0.set_aspect('auto')
ax1.set_aspect('auto')
ax2.set_aspect('auto')
ax3.set_aspect('auto')
ax4.set_aspect('auto')
ax5.set_aspect('auto')
cax1.set_aspect('auto')
cax2.set_aspect('auto')
plt.tight_layout()
f.savefig(outPng, dpi=300)