def find_seabreezes(region, start_year, final_year, lcstimelen):
years = range(start_year, final_year)
# ---- File reading and formatting ---- #
departures = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_departuretimelen' +
f'_{lcstimelen}_v2.nc',
year_range=years,
reverseLat=True).isel(time=slice(None, None))
ftle_array = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_lcstimelen' + f'_{lcstimelen}_v2.nc',
year_range=years).isel(time=slice(None, None))
landSea_mask = xr.open_dataarray(
'/gws/nopw/j04/primavera1/observations/ERA5/landSea_mask.nc')
landSea_mask = landSea_mask.squeeze('time').drop('time')
landSea_mask.coords['longitude'] = (
landSea_mask.coords['longitude'].values + 180) % 360 - 180
landSea_mask = landSea_mask.sortby('latitude').sortby('longitude')
landSea_mask = landSea_mask.interp_like(ftle_array, method='nearest')
landSea_mask.values = landSea_mask.values.round(0)
departures.coords['time'] = \
pd.date_range(start=f"{start_year}-01-01T00:00:00",
periods=departures.time.shape[0], freq='6H')
departures = departures.sortby('latitude')
ftle_array = xr.apply_ufunc(lambda x: np.log(x), ftle_array**0.5)
threshold = ftle_array.quantile(0.6)
dep_x, dep_y = departures.x_departure, departures.y_departure
dep_lat, dep_lon = dep_y.where(ftle_array > threshold,
np.nan), dep_x.where(
ftle_array > threshold, np.nan)
# dep_lat = dep_y.copy(data=dep_lat)
# dep_lon = dep_x.copy(data=dep_lon)
dep_mask = xr.full_like(dep_x, 1)
# Selecting only continental points (sort of)
dep_lat = dep_lat.where(landSea_mask == 1, drop=True)
dep_lon = dep_lon.where(landSea_mask == 1, drop=True)
times = dep_lat.time.values
array_list = apply_binary_mask(times, dep_lat, dep_lon, landSea_mask)
seabreeze = dep_lat.copy(data=np.stack(array_list))
breezetime = seabreeze.where(
seabreeze == 1, drop=True).groupby('time.hour').sum('time').stack(
points=['latitude', 'longitude']).groupby('points').apply(
np.argmax).unstack()
# seabreeze = dep_lat.groupby('time').apply(dummy)
seabreeze.to_netcdf(
f'/group_workspaces/jasmin4/upscale/gmpp/convzones/seabreeze_{start_year}_{final_year}_{region}_lcstimelen_{lcstimelen}.nc'
)
return seabreeze
def filter(region, years):
ftle_array = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_lcstimelen_1.nc',
year_range=years)
ftle_array = ftle_array.expand_dims('dummy')
encoder = xru.autoencoder(n_components=7,
dims_to_reduce=['dummy'],
alongwith=['time'],
mode='emd',
parallel=True).fit(ftle_array)
transformed = encoder.transform(ftle_array)
transformed = transformed.transpose('time', 'dim_to_groupby',
'encoded_dims')
new_values = transformed.values.reshape(transformed.shape[0],
ftle_array.shape[2],
ftle_array.shape[3],
transformed.shape[2])
new_array = xr.DataArray(new_values,
dims=['time', 'latitude', 'longitude', 'modes'],
coords=[
transformed.time.values,
ftle_array.latitude.values,
ftle_array.longitude.values,
transformed.encoded_dims.values
])
import os
if __name__ == '__main__':
region = 'SACZ_big'
years = range(1995, 2005)
lcstimelen = 16
basin = 'Uruguai'
season = 'DJF'
basin_origin = 'amazon'
MAG = xr.open_dataset('~/phdlib/convlib/data/xarray_mair_grid_basins.nc')
mask = MAG[basin]
departures = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_departuretimelen' +
f'_{lcstimelen}_v2.nc',
year_range=years,
season=season,
set_date=True,
lcstimelen=lcstimelen,
binary_mask=mask)
ftle_array = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_lcstimelen' + f'_{lcstimelen}_v2.nc',
year_range=years,
season=season,
binary_mask=mask,
lcstimelen=lcstimelen,
set_date=True)
pr = read_nc_files(region=region,
basepath='/gws/nopw/j04/primavera1/observations/ERA5/',
def _identify_features(self, array):
array = (array - array.min()) / array.max()
for time in array.time.values:
array2D = array.sel(time=time).values
array2D.shape()
blobs_doh = blob_doh(array2D, max_sigma=30, threshold=.01)
print(';a')
pass
def track(self, array):
identified_array = self._identify_features(array)
if __name__ == '__main__':
array = read_nc_files(year_range=range(2000, 2001))
array = array.sel(latitude=slice(-30, None),
longitude=slice(-60, -30),
time=slice(None, 10))
threshold = array.quantile(0.7)
array = array.where(array > threshold, 0)
array = array.where(array < threshold, 1)
#labeled_array = xr.apply_ufunc(lambda x: measure.label(x), array)
labeled_array = xr.apply_ufunc(lambda x: measure.label(x),
array.groupby('time'))
array = array.expand_dims(groups=np.unique(labeled_array.values))
array_of_sizes = np.zeros(
[len(array.time.values),
len(array.groups.values)])
region = 'SACZ_big'
years = range(2020, 2021)
lcstimelen = 6
basin = 'Tiete'
season = 'DJF'
MAG = xr.open_dataset('data/xarray_mair_grid_basins.nc')
MAG = MAG[basin]
MAG = MAG.rename({'lat': 'latitude', 'lon': 'longitude'})
datapath_local = '/home/gab/phd/data/'
ftle_array = read_nc_files(region=region,
basepath=datapath_local + 'FTLE_ERA5/',
filename='SL_attracting_{year}_lcstimelen' +
f'_{lcstimelen}_v2.nc',
year_range=years,
season=None,
lcstimelen=lcstimelen,
set_date=True)
ftle_array = xr.apply_ufunc(lambda x: np.log(x), ftle_array**
0.5) * 6 / lcstimelen
#ftle_array.roll(time=-2, roll_coords=False)
u = read_nc_files(region={
'latitude': slice(10, -70),
'longitude': slice(-100, -1)
},
basepath=datapath_local + '/ERA5/ERA5',
filename='viwve_ERA5_6hr_{year}010100-{year}123118.nc',
year_range=years,
transformLon=True,
reverseLat=False,
import xarray as xr
from convlib.xr_tools import read_nc_files, createDomains
import numpy as np
area = {'latitude': slice(-15, -10), 'longitude': slice(-50, -38)}
region = "SACZ_big"
years = range(1995, 2005)
lcstimelen = 8 # 8
season = "DJF"
CZ = 1
ftle_array = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_lcstimelen' + f'_{lcstimelen}_v2.nc',
year_range=years,
season=season)
ftle_array = xr.apply_ufunc(lambda x: np.log(x), ftle_array**0.5)
ftle_array = ftle_array.sel(area).groupby('time').mean()
ftle_array.to_netcdf('/home/users/gmpp/sacz_index.nc')
axarr[2].coastlines()
# axarr.add_feature(states_provinces, edgecolor='gray')
plt.savefig(f'./tempfigs/SL{time}.png')
plt.close()
domains = dict(AITCZ=dict(latitude=slice(-5, 15), longitude=slice(-50, -13)),
SACZ=dict(latitude=slice(-40, -5), longitude=slice(-62, -20)))
if __name__ == '__main__':
region = "SACZ_small"
lcs_time_len = 1
arr1 = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}' + f'_lcstimelen_1_v2.nc',
year_range=range(1990, 2000))
arr4 = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}' + f'_lcstimelen_4_v2.nc',
year_range=range(1990, 2000))
# arr = xr.open_dataarray('/home/users/gmpp/out/SL_repelling_1980_1998.nc')
sel_hour = 0
hours = [pd.to_datetime(x).hour for x in arr.time.values]
#arr = xr.apply_ufunc(lambda x: np.log(x), arr ** 0.5)
array_mean = arr.where(
np.array(hours).reshape(arr.shape[0], 1, 1) == sel_hour).groupby(
def seabreeze():
region = 'NEBR'
lcstimelen = 1
vmin = 0
vmax = 0.6
start_year, final_year = 1995, 1996
years = range(start_year, final_year)
# seabreeze = xr.open_dataarray(f'/group_workspaces/jasmin4/upscale/gmpp/convzones/seabreeze_{start_year}_{final_year}_{region}.nc')
seabreeze = find_seabreezes(region, start_year, final_year, lcstimelen)
pr_array = read_nc_files(
region=region,
basepath='/gws/nopw/j04/primavera1/observations/ERA5/',
filename='pr_ERA5_6hr_{year}010100-{year}123118.nc',
year_range=years,
transformLon=True,
reverseLat=True)
pr_array = pr_array.sortby('latitude') * 6 * 3600 # total mm in 6h
pr_array = pr_array.sel(latitude=seabreeze.latitude,
longitude=seabreeze.longitude,
time=seabreeze.time,
method='nearest')
masked_precip = pr_array.where(seabreeze == 1, 0)
#
# for time in seabreeze.time:
# fig = plt.figure(figsize=[10, 10])
# gs = gridspec.GridSpec(1, 1, width_ratios=[1])
#
# axs = {'mask': fig.add_subplot(gs[0, 0], projection=proj)}
#
# seabreeze.sel(time=time).plot(transform=ccrs.PlateCarree(), add_colorbar=True,
# ax=axs['mask'], vmin=0, vmax=1)
# axs['mask'].add_feature(cfeature.NaturalEarthFeature(
# 'cultural', 'admin_1_states_provinces_lines', scale='50m',
# edgecolor='gray', facecolor='none'))
# axs['mask'].add_feature(cfeature.RIVERS)
# axs['mask'].add_feature(cfeature.COASTLINE)
# plt.savefig('tempfigs/seabreeze/{}.png'.format(time.values))
# plt.close()
prec_frac = masked_precip.groupby('time.hour').sum(
'time') / pr_array.groupby('time.hour').sum('time')
total = 4 * seabreeze.groupby('time.hour').sum(
'time') / seabreeze.time.values.shape[0]
prec_frac = prec_frac / total
# total = total - total.mean('hour')
hours = total.hour.values
proj = ccrs.PlateCarree()
fig = plt.figure(figsize=[20, 20])
gs = gridspec.GridSpec(2, 3, width_ratios=[1, 1, 0.05])
axs = {}
axs[hours[0]] = fig.add_subplot(gs[0, 0], projection=proj)
axs[hours[1]] = fig.add_subplot(gs[0, 1], projection=proj)
axs[hours[2]] = fig.add_subplot(gs[1, 0], projection=proj)
axs[hours[3]] = fig.add_subplot(gs[1, 1], projection=proj)
# Disable axis ticks
for ax in axs.values():
ax.tick_params(bottom=False,
labelbottom=False,
left=False,
labelleft=False)
# Add titles
for name, ax in axs.items():
ax.set_title(name)
for hour in hours:
plot = prec_frac.sel(hour=hour).plot.contourf(
ax=axs[hour],
transform=ccrs.PlateCarree(),
add_colorbar=False,
add_labels=False,
levels=5,
vmax=4,
vmin=0,
cmap='Blues')
CS = total.sel(hour=hour).plot.contour(ax=axs[hour],
levels=5,
vmin=vmin,
vmax=vmax,
cmap="Reds")
axs[hour].clabel(CS, inline=True, fontsize=15)
axs[hour].coastlines(color='black')
axs[hour].add_feature(
cfeature.NaturalEarthFeature('cultural',
'admin_1_states_provinces_lines',
scale='50m',
edgecolor='gray',
facecolor='none'))
cbar_gs = gridspec.GridSpecFromSubplotSpec(1,
1,
subplot_spec=gs[:, 2],
wspace=2.5)
cax = fig.add_subplot(cbar_gs[0])
plt.colorbar(plot, cax)
plt.savefig(
f'tempfigs/seabreeze/seabreeze_{region}_lcstimelen_{lcstimelen}.png')
plt.close()
total.stack(points=['latitude', 'longitude']).mean('points').plot()
plt.savefig('tempfigs/seabreeze/ciclo_diurno.png')
def precip_fraction(region, years):
if region == 'AITCZ':
proj = ccrs.PlateCarree()
figsize1 = (20, 11)
figsize2 = (30, 20)
elif 'SACZ' in region:
figsize1 = (20, 11)
figsize2 = (20, 20)
proj = ccrs.Orthographic(-40, -20)
print('*---- Reading files ----*')
ftle_array = read_nc_files(
region=region,
basepath='/group_workspaces/jasmin4/upscale/gmpp/convzones/',
filename='SL_repelling_{year}_lcstimelen_1.nc',
year_range=years)
pr_array = read_nc_files(
region=region,
basepath='/gws/nopw/j04/primavera1/observations/ERA5/',
filename='pr_ERA5_6hr_{year}010100-{year}123118.nc',
year_range=years,
transformLon=True,
reverseLat=True)
pr_array = pr_array.sortby('latitude') * 6 * 3600 # total mm in 6h
ftle_array = xr.apply_ufunc(lambda x: np.log(x), ftle_array**0.5)
threshold = ftle_array.quantile(0.6)
pr_array = pr_array.sel(latitude=ftle_array.latitude,
longitude=ftle_array.longitude,
method='nearest')
masked_precip = pr_array.where(ftle_array > threshold, 0)
# ------ Total precip ------ #
fig = plt.figure(figsize=figsize1)
gs = gridspec.GridSpec(1, 3, width_ratios=[1, 1, 0.02])
axs = {}
axs['Total'] = fig.add_subplot(gs[0, 0], projection=proj)
axs['Conv. zone events'] = fig.add_subplot(gs[0, 1], projection=proj)
# Disable axis ticks
for ax in axs.values():
ax.tick_params(bottom=False,
labelbottom=False,
left=False,
labelleft=False)
# Add titles
for name, ax in axs.items():
ax.set_title(name)
vmin = 0
vmax = pr_array.sum('time').quantile(0.95)
masked_precip.sum('time').plot(ax=axs['Conv. zone events'],
transform=ccrs.PlateCarree(),
add_colorbar=False,
add_labels=False,
vmin=vmin,
vmax=vmax,
cmap='nipy_spectral')
total = pr_array.sum('time').plot(ax=axs['Total'],
transform=ccrs.PlateCarree(),
add_colorbar=False,
add_labels=False,
vmin=vmin,
vmax=vmax,
cmap='nipy_spectral')
axs['Total'].coastlines(color='black')
axs['Conv. zone events'].coastlines(color='black')
cbar_gs = gridspec.GridSpecFromSubplotSpec(1,
1,
subplot_spec=gs[:, 2],
wspace=2.5)
cax = fig.add_subplot(cbar_gs[0])
plt.colorbar(total, cax)
plt.savefig(f'tempfigs/diagnostics/total_{region}.png')
plt.close()
# ---- Seasonal plots ---- #
fraction = masked_precip / pr_array
fraction = fraction.groupby('time.season').mean('time')
seasons = fraction.season.values
fig = plt.figure(figsize=figsize2)
gs = gridspec.GridSpec(2, 3, width_ratios=[1, 1, 0.05])
axs = {}
axs[seasons[0]] = fig.add_subplot(gs[0, 0], projection=proj)
axs[seasons[1]] = fig.add_subplot(gs[0, 1], projection=proj)
axs[seasons[2]] = fig.add_subplot(gs[1, 0], projection=proj)
axs[seasons[3]] = fig.add_subplot(gs[1, 1], projection=proj)
# Disable axis ticks
for ax in axs.values():
ax.tick_params(bottom=False,
labelbottom=False,
left=False,
labelleft=False)
# Add titles
for name, ax in axs.items():
ax.set_title(name)
vmin = 0
vmax = fraction.quantile(0.99)
for season in seasons:
plot = fraction.sel(season=season).plot(ax=axs[season],
transform=ccrs.PlateCarree(),
add_colorbar=False,
add_labels=False,
vmin=vmin,
vmax=vmax,
cmap='nipy_spectral')
axs[season].coastlines(color='black')
cbar_gs = gridspec.GridSpecFromSubplotSpec(1,
1,
subplot_spec=gs[:, 2],
wspace=2.5)
cax = fig.add_subplot(cbar_gs[0])
plt.colorbar(plot, cax)
plt.savefig(f'tempfigs/diagnostics/seasons_{region}.png')
plt.close()