def file_loop(passit):
ds = xr.open_dataset(passit)
dataset = passit[6:9]
ds['tir'].values = ds['tir']
bloblist = []
tirlist = []
lat = ds.lat
lon = ds.lon
for ids, day in enumerate(ds['tir']):
print('id', ids)
date = day.time
day.values = day / 100
if np.sum(day.values) == 0:
continue
img, nogood, t_thresh_size, t_thresh_cut, pix_nb = powerBlob_utils.filter_img(day.values, 5)
power = util.waveletT(img, dataset='METEOSAT5K_vera')
power_out = powerBlob_utils.find_scales_dominant(power, nogood, dataset=dataset)
if power_out is None:
continue
new_savet = (day.values*100).astype(np.int16)
bloblist.append(xr.DataArray(power_out.astype(np.int16), coords={'time': date, 'lat': lat, 'lon': lon},
dims=['lat', 'lon'])) # [np.newaxis, :])
tirlist.append(xr.DataArray(new_savet, coords={'time': date, 'lat': lat, 'lon': lon}, dims=['lat', 'lon']))
ds_mfg = xr.Dataset()
ds_mfg['blobs'] = xr.concat(bloblist, 'time')
ds_mfg['tir'] = xr.concat(tirlist, 'time')
ds_mfg.sel(lat=slice(5, 12), lon=slice(-13, 13))
savefile = passit.replace('cores_', 'coresPower_')
try:
os.remove(savefile)
except OSError:
pass
comp = dict(zlib=True, complevel=5)
enc = {var: comp for var in ds_mfg.data_vars}
ds_mfg.to_netcdf(path=savefile, mode='w', encoding=enc, format='NETCDF4')
print('Saved ' + savefile)
def _timeLoop(timeslice, inds_inter, weights_inter, shape_inter, tag):
try:
outt = u_int.interpolate_data(timeslice.values, inds_inter, weights_inter, shape_inter)
#print('Interpolated', id)
except ValueError:
print('Interpolation value error!!')
return
outt_out = outt.copy()
if np.sum(outt) == 0: # if temperature is all empty
print('Temperature empty, continue')
return
outt, nogood, t_thresh_size, t_thresh_cut, pix_nb = powerBlob_utils.filter_img(outt, 5)
wav = util.waveletT(outt, dataset='METEOSAT5K_vera')
outt[nogood] = np.nan
power_msg = powerBlob_utils.find_scales_dominant(wav, nogood, dataset=tag)
if power_msg is None: # if power calculation failed
print('Power calc fail, continue')
return
hour = timeslice['time.hour']
minute = timeslice['time.minute']
day = timeslice['time.day']
month = timeslice['time.month']
year = timeslice['time.year']
date = dt.datetime(year, month, day, hour, minute)
isnan = np.isnan(outt)
outt[isnan] = 0
new_savet = (outt_out * 100).astype(np.int16)
return date, power_msg.astype(np.int16), new_savet, outt, nogood, t_thresh_size, t_thresh_cut, pix_nb, wav
def file_loop(fi):
ret = []
print('Doing file: ' + fi)
dic = xr.open_dataset(fi)
id = fi.split('/')[-1]
outt = dic['tc_lag0'].values
outp = dic['p'].values
outpc = dic['pconv'].values
lon = dic['lon'].values
lat = dic['lat'].values
outt[np.isnan(outt)] = 150
outt[outt >= -40] = 150
grad = np.gradient(outt)
outt[outt == 150] = np.nan
outp[np.isnan(outt)] = np.nan
outpc[np.isnan(outt)] = np.nan
area = np.nansum(outt <= -40)
try:
bulk_pmax = np.max(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
except ValueError:
return ret
try:
bulk_pmin = np.min(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
except ValueError:
return ret
if (area * 25 < 15000) or (area * 25 > 800000) or (bulk_pmax > 200) or (bulk_pmin < 0):
print(area * 25)
print('throw out')
return
perc = np.percentile(outt[np.isfinite(outt)], 60) # 60
#perc = -60
clat = np.min(dic.lat.values) + ((np.max(dic.lat.values) - np.min(dic.lat.values)) * 0.5)
clon = np.min(dic.lon.values) + ((np.max(dic.lon.values) - np.min(dic.lon.values)) * 0.5)
if (clon > 28) or (clon < -17.2) or (clat < 4.1):
return
lat_min = np.min(dic.lat.values)
lat_max = np.max(dic.lat.values)
lon_min = np.min(dic.lon.values)
lon_max = np.max(dic.lon.values)
bulk_tmean = np.nanmean(outt)
bulk_tmin_p = np.min(outt[(np.isfinite(outp)) & (np.isfinite(outt))])
bulk_tmean_p = np.mean(outt[(np.isfinite(outp)) & (np.isfinite(outt))])
bulk_pmean = np.max(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
bulk_g30 = np.sum(outp[(np.isfinite(outp)) & (np.isfinite(outt))] >= 30)
o2 = outt.copy()
o2[np.isnan(o2)] = perc
nok = np.where(abs(grad[0]) > 80)
d = 2
i = nok[0]
j = nok[1]
for ii, jj in zip(i, j):
kern = o2[ii - d:ii + d + 1, jj - d:jj + d + 1]
o2[ii - d:ii + d + 1, jj - d:jj + d + 1] = ndimage.gaussian_filter(kern, 3, mode='nearest')
wav = util.waveletT(o2, 5)
arr = np.array(wav['scales'], dtype=str)
arrf = np.array(wav['scales'], dtype=float)
scale_ind = range(arr.size)
figure = np.zeros_like(outt)
wll = wav['t']
#maxs = np.zeros_like(wll)
yyy = []
xxx = []
scal = []
for nb in scale_ind[::-1]:
orig = float(arr[nb])
print(np.round(orig))
wl = wll[nb, :, :]
maxout = (
wl == ndimage.maximum_filter(wl, (5, 5), mode='constant', cval=np.amax(wl) + 1)) # (np.round(orig / 5))
try:
yy, xx = np.where((maxout == 1) & (outt <= -40) & (wl >= np.percentile(wl[wl >= 0.5], 90)) & (wl > orig ** .5))
except IndexError:
continue
for y, x in zip(yy, xx):
ss = orig
iscale = (np.ceil(ss / 2. / 5.)).astype(int)
# if ss <= 20:
# iscale = iscale + 1
ycirc, xcirc = ua.draw_cut_circle(x, y, iscale, outt)
figure[ycirc, xcirc] = np.round(orig)
xxx.append(x)
yyy.append(y)
scal.append(orig)
figure[np.isnan(outt)] = 0
figure[np.isnan(outp)] = 0
circle_val_all = np.sum(figure > 0)
xx = []
yy = []
for y, x, sc in zip(yyy[::-1], xxx[::-1], scal[::-1]):
if figure[y, x] == 0:
continue
# if sc < 150:
# continue
xx.append(x)
yy.append(y)
int_sc = np.round(sc)
radius = sc
iscale = (np.ceil(radius / 2. / 5.)).astype(int)
# if int_sc <= 20:
# iscale = iscale + 1
ycircf, xcircf = ua.draw_cut_circle(x, y, iscale, outt)
pos = np.where((figure[ycircf, xcircf] == int_sc))
if len(pos[0]) <= 3:
continue
circle_Tcenter = outt[y, x]
t_para = np.nanmean(outt[ycircf[pos], xcircf[pos]])
# if sc < 90:
# km, coords = u_gis.call_parallax_era(int(dic['time.month']), t_para, lon[y,x], lat[y,x], 0, 0)
# lx, ly = km
# lx = int(np.round(lx/5.))
# ly = int(np.round(ly/5.)) # km into pixels
# else:
lx = 0
ly = 0
if (int_sc >= 90):
# outp[ycircf[pos] - ly, xcircf[pos] - lx] = 1000
# outt[ycircf[pos], xcircf[pos]] = 1000
ppos = np.where(outp >= 30)
outt[np.isnan(outt)] = -40
f = plt.figure()
# plt.imshow(outp, cmap='jet', origin='lower')
# f = plt.figure()
# plt.imshow(outt, cmap='jet', origin='lower')
# f = plt.figure()
plt.imshow(outt, cmap='jet', origin='lower')
plt.contour(outp, cmap='viridis', vmin=20)
figure[figure<15] = np.nan
plt.contourf(figure, cmap='Reds', vmin=9)
plt.plot(ppos[1], ppos[0], 'ro')
r = 20
kernel = tm_utils.cut_kernel(outp, x-lx, y-ly, r)
kernelt = tm_utils.cut_kernel(outt, x, y, r)
if kernel.shape != (r * 2 + 1, r * 2 + 1):
kernel = np.zeros((41, 41)) + np.nan
if kernelt.shape != (r * 2 + 1, r * 2 + 1):
kernelt = np.zeros((41, 41)) + np.nan
circle_p = outp[ycircf[pos]-ly, xcircf[pos]-lx]
#outp[ycircf[pos] - ly, xcircf[pos] - lx] = np.nan
circle_pc = outpc[ycircf[pos]-ly, xcircf[pos]-lx]
#outpc[ycircf[pos] - ly, xcircf[pos] - lx] = np.nan
circle_t = outt[ycircf[pos], xcircf[pos]]
circle_valid = np.sum(np.isfinite(circle_p))
#
#
# if (int_sc >= 150) | (int_sc == 15):
# # outp[ycircf[pos] - ly, xcircf[pos] - lx] = 1000
# # outt[ycircf[pos], xcircf[pos]] = 1000
# f = plt.figure()
# plt.imshow(outp, cmap='jet')
# f = plt.figure()
# plt.imshow(outt, cmap='jet')
# f = plt.figure()
# plt.imshow(figure, cmap='jet')
if ((circle_valid) < 3):
continue
circle_sum = np.nansum(circle_p)
circle_nz = np.nansum(circle_p > 0.1)
circle_g30 = np.nansum(circle_p >= 30)
try:
circle_max = np.nanmax(circle_p)
except ValueError:
circle_max = np.nan
try:
circle_p99 = np.percentile(circle_p[circle_p >= 0.1], 99)
except IndexError:
circle_p99 = np.nan
try:
circle_p95 = np.percentile(circle_p[circle_p >= 0.1], 95)
except IndexError:
circle_p95 = np.nan
try:
circle_p90 = np.percentile(circle_p[circle_p >= 0.1], 90)
except IndexError:
circle_p90 = np.nan
#maxs[posi, y, x] = 1
ret.append((kernel, kernelt, int_sc, id, dic['time.hour'].values.tolist(),
clat, clon, lat_min, lat_max, lon_min, lon_max, area,
bulk_pmax, bulk_pmean, bulk_tmean, bulk_tmean_p, bulk_tmin_p, bulk_g30,
len(ycircf), circle_Tcenter, circle_p, circle_t, circle_valid, circle_sum,
circle_nz, circle_g30, circle_max, circle_p99, circle_p95, circle_p90, circle_val_all, circle_pc))
return ret
def run(fi):
ret = []
print('Doing file: ' + fi)
dic = xr.open_dataset(fi)
outt = dic['tc_lag0'].values
outp = dic['p'].values
outpc = dic['pconv'].values
outplot = outp.copy()
outt[np.isnan(outt)] = 150
outt[outt >= -40] = 150
grad = np.gradient(outt)
outt[outt == 150] = np.nan
outp[np.isnan(outt)] = np.nan
outpc[np.isnan(outt)] = np.nan
area = np.sum(outt <= -40)
try:
bulk_pmax = np.max(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
except ValueError:
return ret
try:
bulk_pmin = np.min(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
except ValueError:
return ret
if (area * 25 < 15000) or (area * 25 > 800000) or (bulk_pmax > 200) or (bulk_pmin < 0):
print(area*25)
print('throw out')
return
perc = np.percentile(outt[np.isfinite(outt)], 60) # 60
print('perc:',perc)
perc=-60
clat = np.min(dic.lat.values) + ((np.max(dic.lat.values) - np.min(dic.lat.values)) * 0.5)
clon = np.min(dic.lon.values) + ((np.max(dic.lon.values) - np.min(dic.lon.values)) * 0.5)
if (clon > 28) or (clon < -17.2) or (clat < 4.1):
return
figure = np.zeros_like(outt)
o2 = outt.copy()
o2[np.isnan(o2)] = perc
nok = np.where(abs(grad[0]) > 80)
d = 2
i = nok[0]
j = nok[1]
for ii, jj in zip(i, j):
kern = o2[ii - d:ii + d + 1, jj - d:jj + d + 1]
o2[ii - d:ii + d + 1, jj - d:jj + d + 1] = ndimage.gaussian_filter(kern, 3, mode='nearest')
wav = util.waveletT(o2, 5)
arr = np.array(wav['scales'], dtype=str)
arrf = np.array(wav['scales'], dtype=float)
scale_ind = range(arr.size)
yp, xp = np.where(outp > 30)
figure = np.zeros_like(outt)
wll = wav['t']
maxs = np.zeros_like(wll)
yyy=[]
xxx=[]
scal=[]
for nb in scale_ind[::-1]:
orig = float(arr[nb])
print(np.round(orig))
wl = wll[nb, :, :]
maxout = (
wl == ndimage.maximum_filter(wl, (5,5), mode='constant', cval=np.amax(wl) + 1)) # (np.round(orig / 5))
try:
yy, xx = np.where((maxout == 1) & (outt <= -40) & (wl >= np.percentile(wl[wl >= 0.5], 90)) & (wl > orig**.5))# & (wl > orig**.5) )) #(wl >= np.percentile(wl[wl >= 0.5], 90)))# & (wl > orig**.5))#& (wl >= np.percentile(wl[wl >= 0.5], 90))) #)& (wl > orig**.5) (wl >= np.percentile(wl[wl >= 0.1], 90)) )#(wl > orig**.5))# & (wlperc > orig**.5))# & (wlperc > np.percentile(wlperc[wlperc>=0.1], 80)))# & (wlperc > np.percentile(wlperc[wlperc>=0.1], 80) )) # & (wl100 > 5)
except IndexError:
continue
for y, x in zip(yy, xx):
ss = orig
iscale = (np.ceil(ss / 2. / 5.)).astype(int)
if ss <= 20:
iscale = iscale+1
ycirc, xcirc = ua.draw_cut_circle(x, y, iscale, outp)
figure[ycirc, xcirc] = np.round(orig)
xxx.append(x)
yyy.append(y)
scal.append(orig)
figure[np.isnan(outt)]=0
##file 130!!! nR
spos = np.where(np.array(scal, dtype=int) == 15)
figure[figure == 0] = np.nan
f = plt.figure(figsize = (7,6), dpi=300)
ax2 = f.add_subplot(111)
# ax2.autoscale = False
ttest = outplot.copy()
ttest = ttest + 1
ttest[np.isnan(ttest)] = 0
ax2.contourf(np.arange(wll.shape[2]) , np.arange(wll.shape[1]) , outplot, cmap='Blues', zorder=1)
ax2.imshow(outt, cmap='Greys', vmax=-40, zorder=2)
mt = ax2.imshow(figure, cmap='OrRd_r', vmax=180, zorder=3)
ax2.contour(np.arange(wll.shape[2]), np.arange(wll.shape[1]), ttest, cmap='Blues', levels=[-0.5, 0.5], zorder=4)
plt.plot(np.array(xxx)[spos], np.array(yyy)[spos], 'wo', markersize=3, label='Wavelet power maximum', zorder=5)
ax2.invert_yaxis()
ax2.set_xlim(20, 140)
ax2.set_ylim(20, 140)
ax2.set_xticklabels(np.arange(100, 800, 100)-100)
ax2.set_yticklabels(np.arange(100, 800, 100)-100)
ax2.plot(xp , yp , 'o', markersize=3, label='Rain > 30mm h$^{-1}$', zorder=10)
ax2.set_xlabel('Location (km)')
ax2.set_ylabel('Location (km)')
plt.colorbar(mt, label = 'Scale (km)')
plt.tight_layout()
plt.show()
spath = '/users/global/cornkle/C_paper/wavelet/figs/paper/'
plt.savefig(spath + '/method_circles2.png', dpi=300)
f = plt.figure(figsize = (7,6), dpi=300)
ax2 = f.add_subplot(111)
# ax2.autoscale = False
ttest = outplot.copy()
ttest = ttest + 1
ttest[np.isnan(ttest)] = 0
ax2.imshow(outt, cmap='viridis', vmax=-40, zorder=2)
ax2.invert_yaxis()
ax2.set_xlim(20, 140)
ax2.set_ylim(20, 140)
ax2.set_xticklabels(np.arange(100, 800, 100)-100)
ax2.set_yticklabels(np.arange(100, 800, 100)-100)
ax2.plot(xp , yp , 'o', markersize=3, label='Rain > 30mm h$^{-1}$', zorder=10)
ax2.set_xlabel('Location (km)')
ax2.set_ylabel('Location (km)')
plt.tight_layout()
plt.show()
spath = '/users/global/cornkle/C_paper/wavelet/figs/paper/'
plt.savefig(spath + '/method_temp.png', dpi=300)
#bla = wcno.file_loop(files[130])
f = plt.figure(figsize = (6.5,11), dpi=300)
gridspec.GridSpec(4,1)
posi = 116
ax1 = plt.subplot2grid((4,1),(0,0),rowspan=2)
ax2 = plt.subplot2grid((4,1),(2,0))
ax3 = plt.subplot2grid((4,1),(3,0))
lev = np.arange(-90,-39,4)
ax1.contourf(np.arange(wll.shape[2]) * 5, np.arange(wll.shape[1]) * 5, outplot, cmap='Blues')
mt = ax1.contourf(np.arange(wll.shape[2])*5,np.arange(wll.shape[1])*5,outt, cmap='Greys', vmax=-40, levels = lev)
ax1.plot(np.arange(wll.shape[2])*5, [posi*5]*len(np.arange(wll.shape[2])*5), linestyle = '--', linewidth=2, color = 'black')
ttest = outplot.copy()
ttest = ttest+1
ttest[np.isnan(ttest)] = 0
ax1.contour(np.arange(wll.shape[2]) * 5, np.arange(wll.shape[1]) * 5, ttest, cmap='Blues' , levels=[-0.5,0.5])
ax1.invert_yaxis()
ax1.set_xlim(100,700)
ax1.set_ylim(100, 700)
ax1.set_xticklabels(np.arange(100, 800, 100) - 100)
ax1.set_yticklabels(np.arange(100, 800, 100) - 100)
ax1.plot(xp*5, yp*5, 'o', markersize=3, label='Rain > 30mm h$^{-1}$')
ax1.legend(loc=4)
ax1.set_ylabel('Location (km)')
ax1.set_title(str(dic['time.year'].values)+'-'+str(dic['time.month'].values)+'-'+str(dic['time.day'].values)+' '+str(dic['time.hour'].values)+':'+str(dic['time.minute'].values)+'UTC')
colors = cm.viridis(np.linspace(0, 1, len([0,1, 2,5,10,20,40,60,80])))
ax2.plot(np.arange(wll.shape[2])*5, outt[posi,:], color='r') #118
ax2.set_xlim(100, 700)
ax2.set_xticklabels(np.arange(100, 800, 100) - 100)
ax2.set_ylabel('Cloud-top temperature ($^{\circ}$C)')
ax22 = ax2.twinx()
ax22.set_xlim(100, 700)
ax22.plot(np.arange(wll.shape[2])*5, outp[posi,:])
ax22.set_ylabel('Rain (mm h$^{-1}$)')
print(np.nanmax(wll[:,posi,:]))
mp = ax3.contourf(np.arange(wll.shape[2])*5, arr,wll[:,posi,:], levels=[0,1, 2,5,10,20,40,80,100], colors=colors)
maxs = np.mean(maxs[:,posi-1:posi+2, :], 1) # -1, +2
ppos = np.where(maxs)
for p1, p2 in zip(ppos[1], ppos[0]):
ax3.errorbar((np.arange(wll.shape[2])*5)[p1], arrf[p2], xerr=arrf[p2]/2, fmt='o', ecolor='white', color='white', capthick=3, ms=3, elinewidth=0.7)
ax3.set_xlim(100,700)
ax3.set_xticklabels(np.arange(100, 800, 100) - 100)
ax3.set_ylim(15, 180)
ax3.set_xlabel('Location (km)')
ax3.set_ylabel('Length scale (km)')
plt.tight_layout()
f.subplots_adjust(right=0.86)
cax = f.add_axes([0.87, 0.545, 0.025, 0.415])
cb = plt.colorbar(mt, cax=cax, label='Cloud-top temperature ($^{\circ}$C)')
cb.ax.tick_params(labelsize=12)
cax = f.add_axes([0.87, 0.065, 0.025, 0.175])
cb = plt.colorbar(mp, cax=cax, label='Wavelet power')
cb.ax.tick_params(labelsize=12)
fsiz = 14
x = 0.02
plt.annotate('a)', xy=(x, 0.96), xytext=(0, 4), size=fsiz, xycoords=('figure fraction', 'figure fraction'),
textcoords='offset points')
plt.annotate('b)', xy=(x, 0.51), xytext=(0, 4), size=fsiz, xycoords=('figure fraction', 'figure fraction'),
textcoords='offset points')
plt.annotate('c)', xy=(x, 0.245), xytext=(0, 4), size=fsiz, xycoords=('figure fraction', 'figure fraction'),
textcoords='offset points')
plt.show()
spath = '/users/global/cornkle/C_paper/wavelet/figs/paper/'
plt.savefig(spath+'/method2.png', dpi=300)
dic.close()
plt.close('all')
def file_loop(passit):
gridd = passit[0]
inds_inter = gridd[0]
weights_inter = gridd[1]
shape_inter = gridd[2]
grid_inter = gridd[3]
m = passit[1]
file = passit[2]
tag = passit[3]
# if tag == 'mfg':
# strr = file.split(os.sep)[-2]
# if (np.int(strr[4:6]) != 9):
# print('Skip month')
# return
# else:
# strr = file.split(os.sep)[-1]
# if (np.int(strr[4:6]) != 9):
# print('Skip month')
# return
strr = file.split(os.sep)[-1]
if ((strr[-2::]) != '00') & ((strr[-2::]) != '30'):
print('Skip minute')
return
# if not ((np.int(strr[8:10]) >= 20)): #& (np.int(strr[8:10]) <= 19) ): #((np.int(strr[8:10]) > 3)): #not ((np.int(strr[8:10]) >= 16) & (np.int(strr[8:10]) <= 19) ): #& (np.int(strr[8:10]) < 18): #(np.int(strr[4:6]) != 6) & #(np.int(strr[8:10]) != 3) , (np.int(strr[8:10]) > 3)
# print('Skip hour')
# return
lon, lat = grid_inter.ll_coordinates
file = file + '.gra'
print('Doing file: ' + file)
try:
mdic = m.read_data(file, llbox=[-25, 20, 2, 25])
except FileNotFoundError:
print('File not found')
return
if not mdic:
print('File missing')
return
timeslice = mdic['t']
#test = timeslice.copy(deep=True)
try:
outt = u_int.interpolate_data(timeslice.values, inds_inter, weights_inter, shape_inter)
print('Interpolated')
except ValueError:
print('Interpolation value error!!')
ipdb.set_trace()
savet = outt.copy()
t_thresh_size = -40
t_thresh_cut = -50
core_min = -50
outt[outt>=t_thresh_size] = 0
outt[np.isnan(outt)] = 0
labels, numL = label(outt)
u, inv = np.unique(labels, return_inverse=True)
n = np.bincount(inv)
pix_nb = 28
badinds = u[(n < pix_nb)] # all blobs with more than 1000 pixels = 25,000km2 (meteosat regridded 5km), 200pix = 5000km2, 8pix = 200km2
# scale 30km, radius 15km ca. 700km2 circular area equals 28 pix
for bi in badinds:
inds = np.where(labels == bi)
outt[inds] = 0
outt[outt >=t_thresh_cut] = 150
grad = np.gradient(outt)
outt[outt == 150] = np.nan
nogood = np.isnan(outt) # filters edge maxima later, no maxima in -40 edge area by definition!
tdiff = np.nanmax(outt)-np.nanmin(outt) # define background temperature for image
if tdiff > 28: # temp difference of 28 degrees
xmin = 15
else:
xmin = 10
outt[nogood] = t_thresh_cut-xmin
nok = np.where(abs(grad[0]) > 80)
d = 2
i = nok[0]
j = nok[1]
# edge smoothing for wavelet application
for ii, jj in zip(i, j):
kern = outt[ii - d:ii + d + 1, jj - d:jj + d + 1]
outt[ii - d:ii + d + 1, jj - d:jj + d + 1] = ndimage.gaussian_filter(kern, 3, mode='nearest')
wav = util.waveletT(outt, dataset='METEOSAT10K')
outt[nogood] = np.nan
arr = np.array(wav['scales'], dtype=str)
scale_ind = range(arr.size)
figure = np.zeros_like(outt)
wll = wav['t']
maxoutt = (
wll == ndimage.maximum_filter(wll, (5,4,4), mode='reflect', # 5,4,4
cval=np.amax(wll) + 1)) # (np.round(orig / 5))
yyy = []
xxx = []
scal = []
for nb in scale_ind[::-1]:
orig = float(arr[nb])
scale = int(np.round(orig))
print(np.round(orig))
wl = wll[nb, :, :]
maxout = maxoutt[nb, :, :]
try:
yy, xx = np.where((maxout == 1) & (outt <= core_min) & (wl > orig**.5)) #& (wl > orig**.5)) # # &
except IndexError:
continue
print(outt[yy,xx])
for y, x in zip(yy, xx):
ss = orig
iscale = (np.ceil(ss / 2. / 5.)).astype(int)
ycirc, xcirc = ua.draw_cut_circle(x, y, iscale, outt)
figure[ycirc, xcirc] = scale #outt
figure[y,x] = scale * -1
xxx.append(x)
yyy.append(y)
scal.append(orig)
figure[np.isnan(outt)] = 0
hour = timeslice['time.hour']
minute = timeslice['time.minute']
day = timeslice['time.day']
month = timeslice['time.month']
year = timeslice['time.year']
date = dt.datetime(year, month, day, hour, minute)
isnan = np.isnan(savet)
savet[isnan]=0
new_savet = (savet*100).astype(np.int16)
bloblist = xr.DataArray(figure.astype(np.int16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon']) #[np.newaxis, :])
tirlist = xr.DataArray(new_savet, coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
power1 = xr.DataArray((np.mean(wav['t'][0:5,:,:], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
power2 = xr.DataArray((np.mean(wav['t'][13:17, :, :], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
power3 = xr.DataArray((np.mean(wav['t'][29:32, :, :], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
power4 = xr.DataArray((np.mean(wav['t'][-5:-3, :, :], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
ds = xr.Dataset()
ds['blobs'] = bloblist
ds['tir'] = tirlist
# ds['tir_w'] = xr.DataArray(outt, coords={'time': date, 'lat': lat[:, 0], 'lon': lon[0, :]}, dims=['lat', 'lon'])
ds['power15-19km'] = power1
ds['power32-38km'] = power2
ds['power80-90km'] = power3
ds['power160-170km'] = power4
ds.attrs['radii']=(np.ceil(wav['scales'] / 2. / 5.)).astype(np.uint8)
ds.attrs['scales_rounded'] = np.round(wav['scales']).astype(np.uint8)
ds.attrs['scales_original'] = wav['scales']
ds.attrs['cutout_T'] = t_thresh_size
ds.attrs['core_minT'] = core_min
ds.attrs['cutout_minPixelNb'] = pix_nb
ds = ds.sel(lat=slice(2,17), lon=slice(-18,13)) #[-14, 2.5, 4, 11.5] cutout to remove dodgy boundaries
print('Did ', file)
return (ds)
def file_loop(passit):
gridd = passit[0]
inds = gridd[0]
weights = gridd[1]
shape = gridd[2]
grid = gridd[3]
m = passit[1]
file = passit[2]
#min = ['00', '30']
strr = file.split(os.sep)[-1]
if ((np.int(strr[4:6]) != 9) & (np.int(strr[4:6]) != 6)):
print('Skip month')
return
if ((strr[-6:-5]) != '00') & ((strr[-6:-5]) != '30'):
print('Skip minute')
return
# if not ((np.int(strr[8:10]) >= 20)): #& (np.int(strr[8:10]) <= 19) ): #((np.int(strr[8:10]) > 3)): #not ((np.int(strr[8:10]) >= 16) & (np.int(strr[8:10]) <= 19) ): #& (np.int(strr[8:10]) < 18): #(np.int(strr[4:6]) != 6) & #(np.int(strr[8:10]) != 3) , (np.int(strr[8:10]) > 3)
# print('Skip hour')
# return
lon, lat = grid.ll_coordinates
#file = files+min+'.gra'
print('Doing file: ' + file)
try:
mdic = m.read_data(file, llbox=[-15, 1.5, 5, 11.5])
except FileNotFoundError:
print('File not found')
return
if not mdic:
print('File missing')
return
outt = u_int.interpolate_data(mdic['t'].values, inds, weights, shape)
savet = outt.copy()
t_thresh_size = -40
t_thresh_cut = -50
core_min = -50
outt[outt>=t_thresh_size] = 0
outt[np.isnan(outt)] = 0
labels, numL = label(outt)
u, inv = np.unique(labels, return_inverse=True)
n = np.bincount(inv)
pix_nb = 28
badinds = u[(n < pix_nb)] # all blobs with more than 1000 pixels = 25,000km2 (meteosat regridded 5km), 200pix = 5000km2, 8pix = 200km2
# scale 30km, radius 15km ca. 700km2 circular area equals 28 pix
for bi in badinds:
inds = np.where(labels == bi)
outt[inds] = 0
outt[outt >=t_thresh_cut] = 150
grad = np.gradient(outt)
outt[outt == 150] = np.nan
nogood = np.isnan(outt) # filters edge maxima later, no maxima in -40 edge area by definition!
tdiff = np.nanmax(outt)-np.nanmin(outt) # define background temperature for image
if tdiff > 28: # temp difference of 28 degrees
xmin = 15
else:
xmin = 10
outt[nogood] = t_thresh_cut-xmin
nok = np.where(abs(grad[0]) > 80)
d = 2
i = nok[0]
j = nok[1]
# edge smoothing for wavelet application
for ii, jj in zip(i, j):
kern = outt[ii - d:ii + d + 1, jj - d:jj + d + 1]
outt[ii - d:ii + d + 1, jj - d:jj + d + 1] = ndimage.gaussian_filter(kern, 3, mode='nearest')
wav = util.waveletT(outt, dataset='METEOSAT5K')
outt[nogood] = np.nan
arr = np.array(wav['scales'], dtype=str)
scale_ind = range(arr.size)
figure = np.zeros_like(outt)
wll = wav['t']
maxoutt = (
wll == ndimage.maximum_filter(wll, (5,4,4), mode='reflect', # 5,4,4
cval=np.amax(wll) + 1)) # (np.round(orig / 5))
yyy = []
xxx = []
scal = []
for nb in scale_ind[::-1]:
orig = float(arr[nb])
scale = int(np.round(orig))
print(np.round(orig))
wl = wll[nb, :, :]
maxout = maxoutt[nb, :, :]
try:
yy, xx = np.where((maxout == 1) & (outt <= core_min) & (wl > orig**.5)) #& (wl > orig**.5)) # # &
except IndexError:
continue
print(outt[yy,xx])
for y, x in zip(yy, xx):
ss = orig
iscale = (np.ceil(ss / 2. / 5.)).astype(int)
ycirc, xcirc = ua.draw_cut_circle(x, y, iscale, outt)
figure[ycirc, xcirc] = scale #outt
figure[y,x] = scale * -1
xxx.append(x)
yyy.append(y)
scal.append(orig)
figure[np.isnan(outt)] = 0
# f = plt.figure()
# plt.contourf(outt)
# plt.contour(figure)
# figure[figure == 0] = np.nan
# f = plt.figure()
# f.add_subplot(111)
# plt.imshow(outt, cmap='inferno')
# plt.imshow(figure, cmap='viridis')
# plt.colorbar()
# plt.plot(xxx, yyy, 'yo', markersize=3)
# ax = f.add_subplot(132, projection=ccrs.PlateCarree())
# plt.contourf(lon, lat, figure, cmap='viridis', transform=ccrs.PlateCarree())
# ax.coastlines()
# ax.add_feature(cartopy.feature.BORDERS, linestyle='--');
#
# plt.colorbar()
# f.add_subplot(131)
# plt.imshow(outt, cmap='inferno')
#
#
#plt.show()
hour = mdic['time.hour']
minute = mdic['time.minute' ]
day = mdic['time.day']
month = mdic['time.month']
year = mdic['time.year']
date = dt.datetime(year, month, day, hour, minute)
ds = xr.Dataset()
isnan = np.isnan(savet)
savet[isnan]=0
new_savet = (savet*100).astype(np.int16)
ds['blobs'] = xr.DataArray(figure.astype(np.int16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon']) #[np.newaxis, :]
ds['tir'] = xr.DataArray(new_savet, coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
# ds['tir_w'] = xr.DataArray(outt, coords={'time': date, 'lat': lat[:, 0], 'lon': lon[0, :]}, dims=['lat', 'lon'])
ds['power15-19km'] = xr.DataArray((np.mean(wav['t'][0:5,:,:], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
ds['power32-38km'] = xr.DataArray((np.mean(wav['t'][13:17, :, :], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
ds['power80-90km'] = xr.DataArray((np.mean(wav['t'][29:32, :, :], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
ds['power160-170km'] = xr.DataArray((np.mean(wav['t'][-5:-3, :, :], axis=0)*100).astype(np.uint16), coords={'time': date, 'lat': lat[:,0], 'lon':lon[0,:]}, dims=['lat', 'lon'])
ds.attrs['radii']=(np.ceil(wav['scales'] / 2. / 5.)).astype(np.uint8)
ds.attrs['scales_rounded'] = np.round(wav['scales']).astype(np.uint8)
ds.attrs['scales_original'] = wav['scales']
ds.attrs['cutout_T'] = t_thresh_size
ds.attrs['core_minT'] = core_min
ds.attrs['cutout_minPixelNb'] = pix_nb
print('Did ', file)
return (ds)
def file_loop(fi):
ret = []
print('Doing file: ' + fi)
dic = xr.open_dataset(fi)
id = fi.split('/')[-1]
outt = dic['tc_lag0'].values
outp = dic['p'].values
outt[np.isnan(outt)] = 150
outt[outt >= -40] = 150
grad = np.gradient(outt)
outt[outt == 150] = np.nan
outp[np.isnan(outt)] = np.nan
area = np.sum(outt <= -40)
try:
bulk_pmax = np.max(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
except ValueError:
return ret
try:
bulk_pmin = np.min(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
except ValueError:
return ret
# 2004-07-17_04:09:00_80.nc (127) is a huge circular storm. good example for organisation
#area gt 3000km2 cause that's 30km radius if circular
if (area * 25 < 15000) or (area * 25 > 800000) or (bulk_pmax > 200) or (bulk_pmin < 0): #or (np.sum(np.isfinite(outp)) < (np.sum(np.isfinite(outt))*0.1)):
print(area*25)
print('throw out')
return
print('Area', area*25)
perc = np.percentile(outt[np.isfinite(outt)], 60) # 60
clat = np.min(dic.lat.values) + ((np.max(dic.lat.values) - np.min(dic.lat.values)) * 0.5)
clon = np.min(dic.lon.values) + ((np.max(dic.lon.values) - np.min(dic.lon.values)) * 0.5)
lat_min = np.min(dic.lat.values)
lat_max = np.max(dic.lat.values)
lon_min = np.min(dic.lon.values)
lon_max = np.max(dic.lon.values)
bulk_tmean = np.nanmean(outt)
lat = dic['lat'].values
bulk_tmin_p = np.min(outt[(np.isfinite(outp)) & (np.isfinite(outt))])
bulk_tmean_p = np.mean(outt[(np.isfinite(outp)) & (np.isfinite(outt))])
bulk_pmean = np.max(outp[(np.isfinite(outp)) & (np.isfinite(outt))])
bulk_g30 = np.sum(outp[(np.isfinite(outp)) & (np.isfinite(outt))] >= 30)
o2 = outt.copy()
o2[np.isnan(o2)] = perc
nok = np.where(abs(grad[0]) > 80)
d = 2
i = nok[0]
j = nok[1]
for ii, jj in zip(i, j):
kern = o2[ii - d:ii + d + 1, jj - d:jj + d + 1]
o2[ii - d:ii + d + 1, jj - d:jj + d + 1] = ndimage.gaussian_filter(kern, 3, mode='nearest')
wav = util.waveletT(o2, 5)
wl = wav['t'] # [nb, :, :]
# maxout = (
# wl == ndimage.maximum_filter(wl, (6, 6, 5), mode='constant',
# cval=np.amax(wl) + 1)) # (np.round(orig / 5))
#(3, 3 ,len(wav['scales'])-1)
arr = np.array(wav['scales'], dtype=str)
scale_ind = range(arr.size)
wlperc = wav['t'].copy()
figure = np.zeros_like(outt)
#
# for s in range(wlperc.shape[0]):
# wlperc[s,:,:][wlperc[s,:,:] < np.percentile(wlperc[s,:,:][wlperc[s,:,:]>=0.05], 70)] = 0
#
# labels, numL = label(wlperc)
# for s in [0,1]:
#
# f = plt.figure()
# plt.imshow(labels[s,:,:])
#
# return
yp, xp = np.where(outp > 30)
size = [15,20,30,40,50,60,70,80,90,100,120,140,160,180,200,220]
for sc in wav['scales'][::-1]:
print(sc)
mimin = np.argmin(np.abs(size-sc))
idd = wav['scales'].tolist().index(sc)
wll = wlperc[idd,:,:].copy()
# wll[wll < np.percentile(wll[wll >= 0.01], 90)] = 0
# wll[wll < np.percentile(np.arange(wll.min(), wll.max()+0.1,0.1), 5)] = 0
# wll[wll < np.percentile(wll[wll >= 0.05], 75)] = 0
wll[(wll < sc**.5)] = 0
labels, numL = label(wll)
for ll in np.unique(labels):
if ll==0:
continue
pos = np.where(labels==ll)
if np.max(wll[pos]) < np.percentile(wll[wll >= 1], 50):
continue
figure[pos] = size[mimin]
#figure[np.nonzero(labels)] = size[mimin]
figure[np.isnan(outt)]=0
# pdb.set_trace()
figure[figure==0]=np.nan
f = plt.figure()
ax1=f.add_subplot(131)
plt.imshow(outt)
plt.imshow(figure, cmap='viridis')
ax2 = f.add_subplot(132)
plt.imshow(figure, cmap='viridis')
plt.plot(xp, yp, 'yo', markersize=3)
ax3 = f.add_subplot(133)
plt.imshow(outt)
ax1.invert_yaxis()
ax2.invert_yaxis()
ax3.invert_yaxis()
plt.show()
return
for sc in np.unique(figure):
if sc == 0:
continue
perblob = figure.copy()
perblob[perblob!=sc] = 0
labels_blob, numL = label(perblob)
for blob in np.unique(labels_blob):
if blob == 0:
continue
pos = np.where(labels_blob == blob) #(figure == sc)
# if len(pos[0]) < 5:
# continue
circle_p = outp[pos]
circle_t = outt[pos]
circle_valid = np.sum(np.isfinite(circle_p))
# if no valid temperature pixels (noise from wavelet outside of storm) or very little valid TRMM pixels
# (very small overlap of TRMM and MSG), continue!
if (np.sum(np.isfinite(circle_t)) <= 0 ):
continue
if ((circle_valid) < 3 ):
continue
circle_sum = np.nansum(circle_p)
## some rain at least
# if circle_sum < 0.1:
# continue
circle_Tcenter = np.nanmin(outt[pos])
y, x = np.where((outt == circle_Tcenter) & (labels_blob == blob) & (figure == sc))
# figure[figure==0]=np.nan
# f = plt.figure()
# f.add_subplot(131)
# plt.imshow(outt)
# plt.imshow(figure, cmap='viridis')
# f.add_subplot(132)
# plt.imshow(figure, cmap='viridis')
# plt.plot(xp, yp, 'yo', markersize=3)
# plt.plot(pos[1], pos[0], 'ro', markersize=3)
# f.add_subplot(133)
# plt.imshow(outt)
# plt.show()
r = 20
kernel = tm_utils.cut_kernel(outp, x[0], y[0], r)
kernelt = tm_utils.cut_kernel(outt, x[0], y[0], r)
if kernel.shape != (r * 2 + 1, r * 2 + 1):
kernel = np.zeros((41, 41)) + np.nan
if kernelt.shape != (r * 2 + 1, r * 2 + 1):
kernelt = np.zeros((41, 41)) + np.nan
#
# if np.nansum(kernel) < 1:
# continue
circle_nz = np.nansum(circle_p>0.1)
circle_g30 = np.nansum(circle_p >= 30)
try:
circle_max = np.nanmax(circle_p)
except ValueError:
circle_max = np.nan
try:
circle_p99 = np.percentile(circle_p[circle_p>=0.1], 99)
except IndexError:
circle_p99 = np.nan
try:
circle_p95 = np.percentile(circle_p[circle_p>=0.1], 95)
except IndexError:
circle_p95 = np.nan
try:
circle_p90 = np.percentile(circle_p[circle_p>=0.1], 90)
except IndexError:
circle_p90 = np.nan
#### HOW TO GIVE BACK THE MAX SCALE PER SYSTEM??
ret.append((kernel, kernelt, sc, id, dic['time.hour'].values.tolist(),
clat, clon, lat_min, lat_max, lon_min, lon_max, area,
bulk_pmax, bulk_pmean, bulk_tmean, bulk_tmean_p, bulk_tmin_p, bulk_g30,
circle_Tcenter, circle_p, circle_t, circle_valid, circle_sum,
circle_nz, circle_g30, circle_max, circle_p99, circle_p95, circle_p90))
figure[figure==0]=np.nan
f = plt.figure()
f.add_subplot(133)
plt.imshow(outt, cmap='inferno')
plt.imshow(figure, cmap='viridis')
f.add_subplot(132)
plt.imshow(figure, cmap='viridis')
plt.colorbar()
plt.plot(xp, yp, 'yo', markersize=3)
f.add_subplot(131)
plt.imshow(outt, cmap='inferno')
plt.plot(xp, yp, 'yo', markersize=3)
plt.show()
# #
#
#f = plt.figure()
# fcnt = 0
# vv = 4
# for s in scale_ind:
#
# pos = np.where((maxout[s, :, :] == 1) & (outt <= -40))
#
# if len(pos[0]) == 0:
# continue
# fcnt+=1
# ax = f.add_subplot(vv,vv,fcnt)
# ax.imshow(wl[s,:,:])
#
# # plt.plot(xp, yp, 'yo', markersize=3)
#
# ax.set_title(str(wav['scales'][s]))
#
# ax = f.add_subplot(vv, vv, fcnt+1)
# plt.imshow(figure, cmap='viridis')
# plt.plot(xp, yp, 'yo', markersize=3)
# ax = f.add_subplot(vv, vv, fcnt + 2)
# plt.imshow(outt)
# ax = f.add_subplot(vv, vv, fcnt + 3)
# plt.imshow(outp)
# plt.plot(xp, yp, 'yo', markersize=3)
# ax = f.add_subplot(vv, vv, fcnt + 4)
# plt.imshow(figure, cmap='viridis')
#
#
# plt.show()
#
# dic.close()
return ret
def circle():
matplotlib.rc('xtick', labelsize=10)
matplotlib.rc('ytick', labelsize=10)
ellipse = np.zeros((100,100))+5
short =25
xcirc, ycirc = ua.draw_circle(50,50,short)
ellipse[ycirc,xcirc] = -80
nb = 21
wav = util.waveletT(ellipse, dx=1, dist=0.08, start=15, nb=nb)#dx=5, dist=0.08,start=15,nb=15 )
wll = wav['t']
arr = np.round(wav['scales'])
print('AVAIL WAVELET SCALES: ', arr)
maxs = np.zeros_like(wll)
yl = []
xl = []
for nb in range(wav['t'].shape[0]):
orig = float(wav['scales'][nb])
wl = wav['t'][nb, :, :]
maxout = (
wl == ndimage.maximum_filter(wl, (5, 5), mode='constant', cval=np.amax(wl) + 1)) # (np.round(orig / 5))
try:
yy, xx = np.where((maxout == 1) & (wl > orig ** .5))
except IndexError:
continue
for y, x in zip(yy, xx):
#print(arr[nb],y,x)
maxs[nb,y,x] = 1
#print('Power value',wll[nb,y,x])
yl.append(y)
xl.append(x)
print('finish loop')
maxout2 = (
wll == ndimage.maximum_filter(wll, (5, 5,5), mode='reflect', cval=np.amax(wl) + 1)) # (np.round(orig / 5))
zl, yl, xl = np.where((maxout2==1) & (wll > arr.repeat(100*100,axis=0).reshape((nb+1,100,100)) ** .5))
wlmax = np.max(wll[zl,yl,xl])
pl = np.where(wll == wlmax)
zll, yll, xll = np.where((maxs == 1))
wllmax = np.max(wll[zll,yll,xll])
pll = np.where(wll == wllmax)
print('Max point scales: ', arr[zl])
amax = np.unravel_index(np.argmax(wll), wll.shape)
print('Totalmax whole domain', arr[amax[0]])
print('Totalmax 3d', arr[pl[0]])
print('Totalmax 2d', arr[pll[0]])
print('Scale of perfect circle', (2*short+1)*wav['res'])
print('Pixel-adjusted perfect circle', (2 *short + 1 -2) * wav['res'])
print('Max scale in pixelacross', arr[amax[0]]/wav['res'])
print('Power max', np.max(wll))
f = plt.figure(figsize = (6.5,11), dpi=300)
#
gridspec.GridSpec(3,1)
posi = 56#116 ## 118
ax1 = plt.subplot2grid((3,1),(0,0),rowspan=2)
ax3 = plt.subplot2grid((3,1),(2,0))
#
lev = np.arange(-90,-39,4)
mt = ax1.contourf(np.arange(wll.shape[2])*5,np.arange(wll.shape[1])*5,ellipse, cmap='Greys', vmax=-40, levels = lev)
ax1.plot(np.arange(wll.shape[2])*5, [posi*5]*len(np.arange(wll.shape[2])*5), linestyle = '--', linewidth=2, color = 'black')
ax1.invert_yaxis()
ax1.legend(loc=4)
ax1.set_ylabel('Spatial extent (km)')
#colors = cm.viridis(np.linspace(0, 1, len([0,1, 2,5,10,20,40,60,80,100])))
#
mp = ax3.contourf(np.arange(wll.shape[2])*5, arr,wll[:,posi,:], levels=[0,1, 2,5,10,20,40,80,100, 130, 150, 180, 200, 300,400], cmap='viridis')
maxs = np.mean(maxs[:,posi-1:posi+2, :], 1) # -1, +2
#ax3.contour(np.arange(wll.shape[2])*5, arr,maxs, cmap='Greys_r')
ppos = np.where(maxs)
#for p1, p2 in zip(ppos[1], ppos[0]):
# ax3.errorbar((np.arange(wll.shape[2])*5)[p1], arr[p2], xerr=arr[p2]/2, fmt='o', ecolor='white', color='white', capthick=3, ms=3, elinewidth=0.7)
#ax3.set_xlim(100,700)
ax3.set_ylim(15, 180)
ax3.set_xlabel('Spatial extent (km)')
ax3.set_ylabel('Length scale (km)')
#plt.tight_layout()
f.subplots_adjust(right=0.86)
cax = f.add_axes([0.87, 0.545, 0.025, 0.415])
cb = plt.colorbar(mt, cax=cax, label='Cloud-top temperature ($^{\circ}$C)')
cb.ax.tick_params(labelsize=12)
cax = f.add_axes([0.87, 0.065, 0.025, 0.175])
cb = plt.colorbar(mp, cax=cax, label='Wavelet power')
cb.ax.tick_params(labelsize=12)
plt.show()
f = plt.figure()
plt.imshow(ellipse)
print(xl,yl)
plt.plot(xl,yl, 'ro')
plt.show()
f = plt.figure()
pos = np.argmin(np.abs((2*short+1)*wav['res']-arr))
plt.imshow(wll[pos, :,:])
print(xl,yl)
plt.plot(xl,yl, 'ro')
plt.show()
f = plt.figure()
plt.imshow(wll[amax[0], :,:])
print(xl,yl)
plt.plot(xl,yl, 'ro')
plt.show()
plt.figure()
plt.plot(wll[amax[0],50,:])
plt.plot(ellipse[50, :])
plt.plot(wav['coeffs'][amax[0],50,:])
print(np.sum(wll[amax[0],50,:]>0))
print(amax[0])
def circle():
matplotlib.rc('xtick', labelsize=10)
matplotlib.rc('ytick', labelsize=10)
ellipse = np.zeros((100, 100)) + 5
short = 25
xcirc, ycirc = ua.draw_circle(50, 50, short)
ellipse[ycirc, xcirc] = -80
nb = 21
wav = util.waveletT(ellipse, dx=1, dist=0.08, start=15,
nb=nb) #dx=5, dist=0.08,start=15,nb=15 )
wll = wav['t']
arr = np.round(wav['scales'])
print('AVAIL WAVELET SCALES: ', arr)
maxs = np.zeros_like(wll)
yl = []
xl = []
for nb in range(wav['t'].shape[0]):
orig = float(wav['scales'][nb])
wl = wav['t'][nb, :, :]
maxout = (wl == ndimage.maximum_filter(wl, (5, 5),
mode='constant',
cval=np.amax(wl) + 1)
) # (np.round(orig / 5))
try:
yy, xx = np.where((maxout == 1) & (wl > orig**.5))
except IndexError:
continue
for y, x in zip(yy, xx):
#print(arr[nb],y,x)
maxs[nb, y, x] = 1
#print('Power value',wll[nb,y,x])
yl.append(y)
xl.append(x)
print('finish loop')
maxout2 = (wll == ndimage.maximum_filter(wll, (5, 5, 5),
mode='reflect',
cval=np.amax(wl) + 1)
) # (np.round(orig / 5))
zl, yl, xl = np.where((maxout2 == 1) & (
wll > arr.repeat(100 * 100, axis=0).reshape((nb + 1, 100, 100))**.5))
wlmax = np.max(wll[zl, yl, xl])
pl = np.where(wll == wlmax)
zll, yll, xll = np.where((maxs == 1))
wllmax = np.max(wll[zll, yll, xll])
pll = np.where(wll == wllmax)
print('Max point scales: ', arr[zl])
amax = np.unravel_index(np.argmax(wll), wll.shape)
print('Totalmax whole domain', arr[amax[0]])
print('Totalmax 3d', arr[pl[0]])
print('Totalmax 2d', arr[pll[0]])
print('Scale of perfect circle', (2 * short + 1) * wav['res'])
print('Pixel-adjusted perfect circle', (2 * short + 1 - 2) * wav['res'])
print('Max scale in pixelacross', arr[amax[0]] / wav['res'])
print('Power max', np.max(wll))
f = plt.figure(figsize=(6.5, 11), dpi=300)
#
gridspec.GridSpec(3, 1)
posi = 56 #116 ## 118
ax1 = plt.subplot2grid((3, 1), (0, 0), rowspan=2)
ax3 = plt.subplot2grid((3, 1), (2, 0))
#
lev = np.arange(-90, -39, 4)
mt = ax1.contourf(np.arange(wll.shape[2]) * 5,
np.arange(wll.shape[1]) * 5,
ellipse,
cmap='Greys',
vmax=-40,
levels=lev)
ax1.plot(np.arange(wll.shape[2]) * 5,
[posi * 5] * len(np.arange(wll.shape[2]) * 5),
linestyle='--',
linewidth=2,
color='black')
ax1.invert_yaxis()
ax1.legend(loc=4)
ax1.set_ylabel('Spatial extent (km)')
#colors = cm.viridis(np.linspace(0, 1, len([0,1, 2,5,10,20,40,60,80,100])))
#
mp = ax3.contourf(
np.arange(wll.shape[2]) * 5,
arr,
wll[:, posi, :],
levels=[0, 1, 2, 5, 10, 20, 40, 80, 100, 130, 150, 180, 200, 300, 400],
cmap='viridis')
maxs = np.mean(maxs[:, posi - 1:posi + 2, :], 1) # -1, +2
#ax3.contour(np.arange(wll.shape[2])*5, arr,maxs, cmap='Greys_r')
ppos = np.where(maxs)
#for p1, p2 in zip(ppos[1], ppos[0]):
# ax3.errorbar((np.arange(wll.shape[2])*5)[p1], arr[p2], xerr=arr[p2]/2, fmt='o', ecolor='white', color='white', capthick=3, ms=3, elinewidth=0.7)
#ax3.set_xlim(100,700)
ax3.set_ylim(15, 180)
ax3.set_xlabel('Spatial extent (km)')
ax3.set_ylabel('Length scale (km)')
#plt.tight_layout()
f.subplots_adjust(right=0.86)
cax = f.add_axes([0.87, 0.545, 0.025, 0.415])
cb = plt.colorbar(mt, cax=cax, label='Cloud-top temperature ($^{\circ}$C)')
cb.ax.tick_params(labelsize=12)
cax = f.add_axes([0.87, 0.065, 0.025, 0.175])
cb = plt.colorbar(mp, cax=cax, label='Wavelet power')
cb.ax.tick_params(labelsize=12)
plt.show()
f = plt.figure()
plt.imshow(ellipse)
print(xl, yl)
plt.plot(xl, yl, 'ro')
plt.show()
f = plt.figure()
pos = np.argmin(np.abs((2 * short + 1) * wav['res'] - arr))
plt.imshow(wll[pos, :, :])
print(xl, yl)
plt.plot(xl, yl, 'ro')
plt.show()
f = plt.figure()
plt.imshow(wll[amax[0], :, :])
print(xl, yl)
plt.plot(xl, yl, 'ro')
plt.show()
plt.figure()
plt.plot(wll[amax[0], 50, :])
plt.plot(ellipse[50, :])
plt.plot(wav['coeffs'][amax[0], 50, :])
print(np.sum(wll[amax[0], 50, :] > 0))
print(amax[0])