def main():
### Get Nino3.4 Index
yrs= [2015,2020] # Starting year and ending year
#Nino3.4 (5N-5S, 170W-120W) [-170,-120,-5,5]
nn34= vf.get_sst_areamean_from_HadISST([-170,-120,-5,5],yrs,remove_AC=True)
### And other region
tio= vf.get_sst_areamean_from_HadISST([240,280,-10,0],yrs,remove_AC=True)
spo= vf.get_sst_areamean_from_HadISST([-170,-120,-40,-30],yrs,remove_AC=True)
###---
### Plotting setup
###---
fig=plt.figure()
fig.set_size_inches(6.4,5) ## (xsize,ysize)
###--- Suptitle
suptit="Auto-correlation Example [HadISST, 2015-20]"
fig.suptitle(suptit,fontsize=15,y=0.94,va='bottom',stretch='semi-condensed')
ax1= fig.add_subplot(111)
sub_tit= ''
data=[nn34,tio,spo]
vnames=['Ni{}o3.4'.format('\u00F1'),'Trop. IO','S. Pacific']
autocorr_plot(ax1,data,vnames,sub_tit)
### Show or save
outdir= '../Pics/'
out_fig_nm= outdir+'V02.autocorr_example.SST_AM+Nino34.png'
#fig.savefig(outfnm,dpi=100) # dpi: pixels per inch
fig.savefig(out_fig_nm,dpi=150,bbox_inches='tight') # dpi: pixels per inch
print(out_fig_nm)
plt.show()
return
def main():
### Get Nino3.4 Index
yrs = [2015, 2020] # Starting year and ending year
#Nino3.4 (5N-5S, 170W-120W) [-170,-120,-5,5]
nn34 = vf.get_sst_areamean_from_HadISST([-170, -120, -5, 5],
yrs,
remove_AC=True)
### And other region
tio = vf.get_sst_areamean_from_HadISST([240, 280, -10, 0],
yrs,
remove_AC=True)
spo = vf.get_sst_areamean_from_HadISST([-170, -120, -40, -30],
yrs,
remove_AC=True)
###---
### Plotting setup
###---
fig = plt.figure()
fig.set_size_inches(6, 8.5) ## (xsize,ysize)
fig.subplots_adjust(hspace=0.3)
###--- Suptitle
suptit = "Lead-Lag Correlation Example [HadISST,2015-20]"
fig.suptitle(suptit,
fontsize=15,
y=0.95,
va='bottom',
stretch='semi-condensed')
maxlag = 7
ax1 = fig.add_subplot(211)
sub_tit = '(a) Ni{}o3.4 vs. TIO'.format('\u00F1')
data = [nn34, tio]
vnames = ['Ni{}o3.4'.format('\u00F1'), 'TIO']
corr_crt = estimate_significant_corr_coef(nn34, tio)[1] # Choose 95% level
llcorr_plot(ax1, data, vnames, sub_tit, maxlag=maxlag, corr_crt=corr_crt)
ax1 = fig.add_subplot(212)
sub_tit = '(b) Ni{}o3.4 vs. SPO'.format('\u00F1')
data = [nn34, spo]
vnames = ['Ni{}o3.4'.format('\u00F1'), 'SPO']
corr_crt = estimate_significant_corr_coef(nn34, spo)[1] # Choose 95% level
llcorr_plot(ax1, data, vnames, sub_tit, maxlag=maxlag, corr_crt=corr_crt)
### Show or save
outdir = '../Pics/'
out_fig_nm = outdir + 'V04.estimate_corr_sig_level_example.SST_AM+Nino34.png'
#fig.savefig(outfnm,dpi=100) # dpi: pixels per inch
fig.savefig(out_fig_nm, dpi=150,
bbox_inches='tight') # dpi: pixels per inch
print(out_fig_nm)
plt.show()
return
def main():
### Get RMM index
tgt_dates = (date(2010, 1, 1), date(2020, 12, 31)) # Recent 5 years
time_info, pcs, phs, strs, miss_idx = vf.read_rmm_text(tgt_dates)
### Check missing
if miss_idx.sum() > 0:
print("There are missings:", miss_idx.sum())
sys.exit()
else:
print("No missings")
### Data collected
suptit = "Power Spectral Density of MJO Index [2010-20]"
data = [pcs[:, 0], pcs[:, 1], strs]
var_names = ['RMM1', 'RMM2', 'Amp']
outdir = '../Pics/'
out_fig_nm = outdir + 'V10.power_spectrum_RMM.png'
plot_data = dict(data=data,
var_names=var_names,
out_fnm=out_fig_nm,
suptit=suptit)
plot_PS(plot_data)
return
def main():
### Years to read data
yrs = [2015, 2020] # Starting year and ending year
### Get SST anomaly
area_range = [-180, 180, -60, 60] # [lon_range, lat_range]
sstano, lat_info, lon_info = vf.get_sst_ano_from_HadISST(area_range,
yrs,
remove_AC=True)
### Re-arange SST by 5deg x 5deg cells
nlat, nlon = lat_info['nlat'], lon_info['nlon']
scaler = 5
nlat2, nlon2 = nlat // scaler, nlon // scaler
sstano = sstano.reshape([-1, nlat2, scaler, nlon2, scaler]).swapaxes(2, 3)
sstano = sstano.reshape([-1, nlat2, nlon2, scaler**2])
ms_idx = np.isnan(sstano[0, :, :, :]).sum(
axis=2) > scaler**2 / 2 # missing if more than half is missing
data2test = np.nanmean(sstano[:, ~ms_idx, :], axis=2) # Mean for 5x5 cells
### Calculate dependency_level
dof_coef1 = np.apply_along_axis(get_dof_coef_log_r1, 0, data2test)
dof_coef2 = np.apply_along_axis(get_dof_coef_e_folding, 0, data2test)
print(dof_coef1.min(), np.median(dof_coef1))
print(dof_coef2.min(), np.median(dof_coef2))
### Split data for t-test
data_prv, data_post = data2test[:36, :], data2test[
36:, :] # First 3 years and later 3 years
### Calculate significance level of t-test by different dof setting
p_vals1 = get_ttest_pval_2d(dof_coef1, data_prv, data_post)
p_vals2 = get_ttest_pval_2d(dof_coef2, data_prv, data_post)
p_vals3 = get_ttest_pval_2d(np.ones_like(dof_coef1), data_prv,
data_post) # Assume no dependency
### Construct new map of p-values
new_map = np.full(ms_idx.shape, np.nan, dtype=float)
data = []
for pp in [p_vals1, p_vals2, p_vals3]:
new_map[~ms_idx] = pp
data.append(np.copy(new_map))
### Map info for displaying
lon0, dlon = lon_info['lon0'], lon_info['dlon']
lat0, dlat = lat_info['lat0'], lat_info['dlat']
img_bound = [
lon0 - dlon / 2, lon0 + dlon * (nlon + 0.5), lat0 - dlat / 2,
lat0 + dlat * (nlat + 0.5)
] # Exact range of data, necessary for imshow()
##-- Above bound is based on previous resolution, but it's ok since no change on area_boundary
### Prepare for plotting
suptit = "t-test for Mean Diff., 2015-17 vs. 2018-20 [HadISST]"
var_names = ['DoF= -N*log(r1)', 'DoF= N/(2*Te)', 'DoF= N']
outdir = '../Pics/'
out_fig_nm = outdir + 'V08.t-test_example_5deg_box.png'
plot_data = dict(data=data,
var_names=var_names,
out_fnm=out_fig_nm,
img_bound=img_bound,
suptit=suptit)
plot_map(plot_data)
return
def plot_map(pdata):
'''
Draw global map where dateline is on the center
'''
abc = 'abcdefgh'
###--- Create a figure
fig = plt.figure()
fig.set_size_inches(7, 8.5) ## (xsize,ysize)
###--- Suptitle
suptit = pdata['suptit']
fig.suptitle(suptit,
fontsize=16,
y=0.97,
va='bottom',
stretch='semi-condensed')
###--- Map Projection
center = 180 # Want to draw a map where dateline is on the center
proj = ccrs.PlateCarree(central_longitude=center)
data_crs = ccrs.PlateCarree()
map_extent = [0., 359.9, -60, 60] # Range to be shown
img_range = pdata['img_bound']
###--- Set range of values to be shown
#val_min, val_max= 0,0.5 <-- it is unnecessary for BoundaryNorm
p_val_levels = [0, 0.01, 0.02, 0.05, 0.1, 0.2]
###--- Color map
colors = ['darkred', '#FA325A', 'darkorange', 'forestgreen', '#C9CD71']
cm = cls.LinearSegmentedColormap.from_list("cm5", colors)
cm.set_under('1.')
cm.set_over('1.')
cm.set_bad('0.8') # For the gridcell of NaN
norm = cls.BoundaryNorm(p_val_levels, ncolors=cm.N, clip=False)
left, right, top, bottom = 0.07, 0.97, 0.93, 0.12
npnx, gapx, npny, gapy = 1, 0.05, len(pdata['data']), 0.06
lx = (right - left - gapx * (npnx - 1)) / npnx
ly = (top - bottom - gapy * (npny - 1)) / npny
ix, iy = left, top
props_imshow = dict(norm=norm,
origin='lower',
cmap=cm,
extent=img_range,
transform=data_crs)
for i, (data, vnm) in enumerate(zip(pdata['data'], pdata['var_names'])):
ax1 = fig.add_axes([ix, iy - ly, lx, ly], projection=proj)
ax1.set_extent(map_extent, crs=data_crs)
map1 = ax1.imshow(data, **props_imshow)
subtit = '({}) {}'.format(abc[i], vnm)
vf.map_common(ax1, subtit, data_crs, xloc=60, yloc=20)
iy = iy - ly - gapy
cb = vf.draw_colorbar(fig,
ax1,
map1,
type='horizontal',
size='panel',
gap=0.06,
extend='max')
cb.set_label('Significance level', fontsize=11)
cb.ax.set_xticklabels(
['{:.0f}%'.format((1 - val) * 100) for val in p_val_levels])
cb.ax.tick_params(labelsize=10)
##-- Seeing or Saving Pic --##
outfnm = pdata['out_fnm']
print(outfnm)
#fig.savefig(outfnm,dpi=100) # dpi: pixels per inch
fig.savefig(outfnm, dpi=150, bbox_inches='tight') # dpi: pixels per inch
# Defalut: facecolor='w', edgecolor='w', transparent=False
plt.show()
return
def main():
### Years to read data
yrs = [2015, 2020] # Starting year and ending year
### Get SST anomaly
area_range = [-180, 180, -60, 60] # [lon_range, lat_range]
sstano, lat_info, lon_info = vf.get_sst_ano_from_HadISST(area_range,
yrs,
remove_AC=True)
lat0, dlat, nlat = lat_info.values()
lon0, dlon, nlon = lon_info.values()
### Weight by latitude
lat_deg = np.arange(nlat) * dlat + lat0
lat_wt = np.sqrt(np.cos(lat_deg / 180 * np.pi)) # Or use np.deg2rad()
sstano = sstano * lat_wt[None, :, None]
### This is only for reducing computing time!!!
### Degrading data resolution (for convenience)
scaler = 2
nlat2, nlon2 = nlat // scaler, nlon // scaler
sstano = sstano.reshape([-1, nlat2, scaler, nlon2,
scaler]).mean(axis=(2, 4))
### Find location of missing values
ms_idx = np.isnan(sstano.mean(axis=0)) # one missing becomes missing
print("Missing ratio= {:.2f}%".format(ms_idx.sum() /
ms_idx.reshape(-1).shape[0] * 100))
### Get PCs and Eigenvectors
maxnum = 10
pc, evec, eval = PC_analysis(sstano[:, ~ms_idx], maxnum=maxnum)
'''### Check if it works
n1= sstano[:,~ms_idx][:,3600]
n2= np.dot(pc,evec)[:,3600]
plt.plot(n1); plt.plot(n2)
plt.show() ; sys.exit()'''
### Restore to lat-lon format
ev_map = np.full([maxnum, *ms_idx.shape], np.nan)
ev_map[:, ~ms_idx] = evec
img_bound = [
lon0 - dlon / 2, lon0 + dlon * (nlon + 0.5), lat0 - dlat / 2,
lat0 + dlat * (nlat + 0.5)
] # Exact range of data, necessary for imshow()
##-- Above bound is based on previous resolution, but it's ok since no change on area_boundary
### Prepare for plotting
suptit = 'PC and EOF of Global SST [HadISST, 2015-20]'
tgt_nums = [1, 2, 3]
mon_list = vf.get_monthly_dates(date(yrs[0], 1, 1),
date(yrs[1], 12, 31),
day=15,
include_date2=True)
outdir = '../Pics/'
out_fig_nm = outdir + 'V09.EOF_example_HadISST.png'
plot_data = dict(ev_map=ev_map,
pc=pc,
suptit=suptit,
out_fnm=out_fig_nm,
img_bound=img_bound,
mon_list=mon_list,
tgt_nums=tgt_nums)
plot_map(plot_data)
return
def plot_map(pdata):
'''
Draw PC time series on the top, and
draw global map where dateline is on the center
'''
###--- Create a figure
fig = plt.figure()
fig.set_size_inches(6, 10) ## (xsize,ysize)
###--- Suptitle
fig.suptitle(pdata['suptit'],
fontsize=16,
y=0.97,
va='bottom',
stretch='semi-condensed')
###--- Axes setting
nk = len(pdata['tgt_nums']) # Number of data to show
left, right, top, bottom = 0.07, 0.93, 0.925, 0.1
npnx, gapx, npny, gapy = 1, 0.05, nk + 1, 0.064
lx = (right - left - gapx * (npnx - 1)) / npnx
ly = (top - bottom - gapy * (npny - 1)) / npny
ix, iy = left, top
###--- Top panel: PC time series
ax1 = fig.add_axes([ix, iy - ly, lx, ly])
colors = plt.cm.tab10(np.linspace(0.05, 0.95, 10))
for i, k in enumerate(pdata['tgt_nums']):
ax1.plot_date(pdata['mon_list'],
pdata['pc'][:, k],
fmt='-',
c=colors[i],
lw=2.5,
alpha=0.85,
label='PC{}'.format(k))
iy = iy - ly - gapy
subtit = '(a) '
ax1.set_title(subtit, fontsize=12, ha='left', x=0.0)
ax1.legend(bbox_to_anchor=(0.08, 1.02, .92, .10),
loc='lower left',
ncol=nk,
mode="expand",
borderaxespad=0.,
fontsize=10)
ax1.axhline(y=0., c='k', lw=0.8, ls='--')
ax1.grid(ls=':')
ax1.xaxis.set_major_formatter(DateFormatter('%b%Y'))
ax1.yaxis.set_minor_locator(AutoMinorLocator(2))
ax1.yaxis.set_ticks_position('both')
ax1.tick_params(axis='both', labelsize=10)
###--- Next, draw global maps
###--- Map Projection
center = 180 # Want to draw a map where dateline is on the center
proj = ccrs.PlateCarree(central_longitude=center)
data_crs = ccrs.PlateCarree()
map_extent = [0., 359.9, -61, 61] # Range to be shown
img_range = pdata['img_bound']
val_max = max(np.nanmin(pdata['ev_map']) * -1, np.nanmax(pdata['ev_map']))
val_min, val_max = val_max * -0.9, val_max * 0.9
abc = 'abcdefgh'
###--- Color map
cm = plt.cm.get_cmap('RdBu_r').copy()
cm.set_bad('0.9') # For the gridcell of NaN
props = dict(vmin=val_min,
vmax=val_max,
origin='lower',
extent=img_range,
cmap=cm,
transform=data_crs)
for i, (data, k) in enumerate(zip(pdata['ev_map'], pdata['tgt_nums'])):
ax2 = fig.add_axes([ix, iy - ly, lx, ly], projection=proj)
ax2.set_extent(map_extent, crs=data_crs)
map1 = ax2.imshow(data, **props)
subtit = '({}) EOF{}'.format(abc[i + 1], k)
vf.map_common(ax2,
subtit,
data_crs,
xloc=60,
yloc=20,
gl_lab_locator=[True, True, False, True])
iy = iy - ly - gapy
vf.draw_colorbar(fig,
ax2,
map1,
type='horizontal',
size='panel',
gap=0.06,
extend='both',
width=0.02)
##-- Seeing or Saving Pic --##
outfnm = pdata['out_fnm']
print(outfnm)
#fig.savefig(outfnm,dpi=100) # dpi: pixels per inch
fig.savefig(outfnm, dpi=150, bbox_inches='tight') # dpi: pixels per inch
# Defalut: facecolor='w', edgecolor='w', transparent=False
plt.show()
return