def myplot(vari, depi, varol, varoc, depo, figfile):
def getdv(var, dep):
v = var[1, :, 1, :]
if dep[:].ndim==var.ndim:
dep = dep[1, :, 1, :]
elif dep[:].ndim==3:
dep = dep[:, 1, :]
xb, yb = meshcells(v.getAxis(-1), dep)
return xb, yb, v.asma()
xbi, ybi, vi = getdv(vari, depi)
xbo, ybo, vol = getdv(varol, depo)
_, _, voc = getdv(varoc, depo)
vmin, vmax = minmax(vi, vol, voc)
kw = dict(vmin=vmin, vmax=vmax)
P.figure(figsize=(10, 4))
ax = P.subplot(131)
P.pcolormesh(xbi, ybi, vi, **kw)
P.title('Original')
P.subplot(132, sharex=ax, sharey=ax)
P.pcolormesh(xbo, ybo, vol, **kw)
P.title('Linear')
P.subplot(133, sharex=ax, sharey=ax)
P.pcolormesh(xbo, ybo, voc, **kw)
P.ylim(ymin=min(ybi.min(), ybo.min()), ymax=max(ybi.max(), ybo.max()))
P.title('Cellave')
P.tight_layout()
P.savefig(figfile)
def plot(xx, yy, target, label, figfiles, figfile, lon=None, lat=None, show=False):
xs, ys, mask = coord2slice(target, lon=lon, lat=lat)
P.figure(figsize=(6, 3.5))
P.title('Target=%(label)s / select: lon=%(lon)s, lat=%(lat)s'%locals())
add_grid((xx, yy))
xx = xx.asma()
yy = yy.asma()
if isinstance(lon, tuple):
P.axvline(lon[0], color='m', ls='--', lw=2)
P.axvline(lon[1], color='m', ls='--', lw=2)
elif isinstance(lon, slice):
i, j, k = lon.indices(xx.shape[1])
P.plot(xx[:, i], yy[:, i], 'c--', lw=2)
P.plot(xx[:, j-1], yy[:, j-1], 'c--', lw=2)
if isinstance(lat, tuple):
P.axhline(lat[0], color='m', ls='--', lw=2)
P.axhline(lat[1], color='m', ls='--', lw=2)
elif isinstance(lat, slice):
i, j, k = lat.indices(yy.shape[0])
P.plot(xx[i], yy[i], 'c--', lw=2)
P.plot(xx[j-1], yy[j-1], 'c--', lw=2)
P.xticks(N.arange(xx.min()-1, xx.max()+1))
P.yticks(N.arange(yy.min()-1, yy.max()+1))
xxi, yyi = xx, yy
xx = xx[ys, xs]
yy = yy[ys, xs]
# mask = mask[ys, xs]
xxb, yyb = meshbounds(xx, yy)
P.pcolor(xxb, yyb, mask, shading='faceted')
P.scatter(xx.ravel(), yy.ravel(), c=(0, 1, 0))
P.grid('on')
P.axis('image')
P.tight_layout()
i = len(figfiles)
savefig = figfile%i
if os.path.exists(savefig): os.remove(savefig)
P.savefig(savefig)
figfiles.append(savefig)
if show: P.show()
else: P.close()
continue
if rank==0:
print >>f, tool.upper(), method, ':', '%5.1f'%(time()-t0), 'seconds', psinfo()
frac = diag['dstAreaFractions']
if frac is not None:
mask = frac<=1.e-3
frac[mask] = 1.
frac = N.resize(frac, varo.shape)
mask = N.resize(mask, varo.shape)
varo[:] /= frac
varo[:] = MV2.masked_where(mask, varo, copy=0)
# del r
gc.collect()
if rank==0:
print >>f, ' plot'
P.figure(figsize=(12, 6))
P.subplots_adjust(right=0.9)
P.subplot(121)
P.pcolormesh(xi, yi, vari[0,0].asma(),vmin=vmin,vmax=vmax)
P.axis([xmin, xmax, ymin, ymax])
P.colorbar()
P.title('Original')
P.subplot(122)
P.pcolormesh(xo, yo, varo[0,0].asma(),vmin=vmin,vmax=vmax)
P.axis([xmin, xmax, ymin, ymax])
P.title(tool.upper()+' / '+method.upper())
P.colorbar(extend='min')#cax=P.axes([0.92, 0.3, 0.02, 0.6]))
figfile = basefile+'_%(tool)s_%(method)s.png'%vars()
if os.path.exists(figfile): os.remove(figfile)
P.savefig(figfile)
P.close()
cmap_name = "CMRmap" cmap_old = P.get_cmap(cmap_name) # Max cmap_10 = desaturate(cmap_old, 1) # Medium cmap_05 = desaturate(cmap_old, 0.5) # Min cmap_00 = desaturate(cmap_old, 0) # Plot fig = P.figure(figsize=(6, 2)) kw = dict(fig=fig, figsize=None, aspect=.15, show=False, close=False) P.subplot(221) plot_cmap(cmap_old, title='Original', **kw) P.subplot(222) plot_cmap(cmap_00, title='f = 0.0', **kw) P.subplot(223) plot_cmap(cmap_05, title='f = 0.5', **kw) P.subplot(224) plot_cmap(cmap_10, title='f = 1.0', **kw) P.tight_layout() P.figtext(.5, 1, 'Desaturate '+cmap_name, va='top', ha='center', size=12) P.savefig(code_file_name(ext='png')) P.close()
vari[int(nyi*0.4):int(nyi*0.4)+3, int(nxi*0.4):int(nxi*0.4)+2] = N.ma.masked
xxib, yyib = meshcells(xi, yi)
nxo = 40
nyo = 25
xo = N.linspace(int(nxi*0.2),int(nxi*1.2),nxo)
yo = N.linspace(int(-nyi*0.2),int(nyi*0.8),nyo)
xxob, yyob = meshcells(xo, yo)
vari.shape = (1, )+vari.shape
varo = N.ma.masked_values(dstwgt(vari.filled(), xi, yi, xo, yo, mv, 0), mv)
kw = dict(vmin=vari.min(), vmax=vari.max())
axlims = [min(xi.min(), xo.min()), max(xi.max(), xo.max()),
min(yi.min(), yo.min()), max(yi.max(), yo.max())]
P.figure(figsize=(8, 4))
P.subplot(211)
P.pcolor(xxib, yyib, vari[0], **kw)
P.axis(axlims)
P.title('Original')
P.subplot(212)
P.pcolor(xxob, yyob, varo[0], **kw)
P.axis(axlims)
P.title('Distance weight')
P.tight_layout()
figfile = code_file_name(ext='png')
if os.path.exists(figfile): os.remove(figfile)
P.savefig(figfile)
P.show()
P.close()
"""Test function :func:`~vacumm.misc.grid.kriging.cloud_split`"""
nps = [100, 60, 40]
npmax = 90
dxy = 5
from vcmq import P, simple_colors, savefigs, code_file_name, N
from vacumm.misc.grid.kriging import cloud_split
# Input clouds
xx = []
yy = []
P.figure(figsize=(7, 3.5))
P.subplot(121, aspect=1)
colors = simple_colors[1:]
for i, np in enumerate(nps):
xc = P.rand(1)*dxy*2.5
yc = P.rand(1)*dxy*2.5
x = P.rand(np)*dxy+xc
y = P.rand(np)*dxy+yc
xx.extend(x)
yy.extend(y)
P.scatter(x, y, c=colors[i:i+1], s=40)
P.title('Original (mixed clouds)')
# Spliting
indices, centroids = cloud_split(xx, yy, getdist=False, getcent=True, npmax=npmax)
# Plot
P.subplot(122, aspect=1)
xx = N.array(xx)
create_time, CurvedInterpolator, rotate_grid)
# Curved grid
nxy = 10
nt = 5
lon = N.arange(nxy * 1.)
lat = N.arange(nxy * 1.)
time = create_time((nt, ), 'years since 2000')
gridi = rotate_grid((lon, lat), 30)
xxi = gridi.getLongitude()[:].filled()
yyi = gridi.getLatitude()[:].filled()
vari = MV2.resize(yyi, (nt, nxy, nxy))
vari.setAxis(0, time)
set_grid(vari, gridi)
kw = dict(vmin=vari.min(), vmax=vari.max())
P.figure(figsize=(10, 3.5))
P.subplot(131, aspect=1)
P.contourf(xxi, yyi, vari[0].asma(), **kw)
add_grid(gridi, edges=False, centers=-1)
xylims = (xxi.min(), xxi.max(), yyi.min(), yyi.max())
P.axis(xylims)
P.title('Curved grid')
# Interpolate to grid
xg, yg = N.meshgrid(N.arange(-3.5, 14.5), N.arange(-3.5, 14.5))
nxyg = xg.shape
cig = CurvedInterpolator(gridi, (xg, yg), g2g=True)
varog = cig(vari)
P.subplot(132, aspect=1)
P.scatter(xg, yg, c=varog[0].asma(), s=120, linewidth=0, **kw)
add_grid(gridi, edges=False, centers=-1)
nb = 10
xxbi, yybi = meshbounds(xxi, yyi)
# Output grid
grido = rotate_grid((N.linspace(0, 6, 50) - 1, N.linspace(0, 4, 35) + 1.), -20)
xxo = grido.getLongitude()[:].filled()
yyo = grido.getLatitude()[:].filled()
xxbo, yybo = meshbounds(xxo, yyo)
# Nearest
varo = nearest2d(vari, xxi, yyi, xxo, yyo, nb)
# Plot
vmin = varo.min()
vmax = varo.max()
P.figure(figsize=(8, 4))
P.subplot(121, aspect=1)
P.pcolor(xxbi, yybi, vari[0], vmin=vmin, vmax=vmax)
add_grid(grido)
P.title('original')
P.subplot(122, aspect=1)
P.pcolor(xxbo, yybo, varo[0], vmin=vmin, vmax=vmax)
add_grid(gridi)
P.title('nearest2d')
P.axis('image')
figfile = code_file_name(ext='png')
if os.path.exists(figfile): os.remove(figfile)
P.savefig(figfile)
P.show()
P.close()
"""Test function :func:`~vacumm.misc.grid.kriging.cloud_split`"""
nps = [100, 60, 40]
npmax = 90
dxy = 5
from vcmq import P, simple_colors, savefigs, code_file_name, N
from vacumm.misc.grid.kriging import cloud_split
# Input clouds
xx = []
yy = []
P.figure(figsize=(7, 3.5))
P.subplot(121, aspect=1)
colors = simple_colors[1:]
for i, np in enumerate(nps):
xc = P.rand(1) * dxy * 2.5
yc = P.rand(1) * dxy * 2.5
x = P.rand(np) * dxy + xc
y = P.rand(np) * dxy + yc
xx.extend(x)
yy.extend(y)
P.scatter(x, y, c=colors[i:i + 1], s=40)
P.title('Original (mixed clouds)')
# Spliting
indices, centroids = cloud_split(xx,
yy,
getdist=False,
getcent=True,
npmax=npmax)
diag = {'dstAreaFractions':None}
varo = vari.regrid(grido, tool='esmf', method='conservative',
diag=diag, coordSys='cart')
# Norm
frac = diag['dstAreaFractions']
mask = frac==0.
frac[mask]=1.
varo[:] /= frac
varo[:] = MV2.masked_where(mask, varo, copy=0)
# Plot
rc('font', size=9)
kw = dict(vmin=vari.min(), vmax=vari.max())
axis = [xxbi.min(), xxbi.max(), yybo.min(), yybo.max()]
P.figure(figsize=(7, 3))
P.subplot(121, aspect=1)
P.pcolormesh(xxbi, yybi, vari, **kw)
P.colorbar(shrink=0.7)
add_grid(gridi, color='0.5')
add_grid(grido)
P.axis(axis)
P.title('Original: max=%g min=%g'%(vari.max(), vari.min()))
P.subplot(122, aspect=1)
P.pcolormesh(xxbo, yybo, varo, **kw)
P.colorbar(shrink=0.7)
add_grid(gridi)
add_grid(grido, color='0.5')
P.axis(axis)
P.title('Regridded: max=%g min=%g'%(varo.max(), varo.min()))
P.tight_layout()
from vacumm.misc.grid.misc import rotate_grid
from vacumm.misc.grid.regridding import CurvedInterpolator
from vcmq import P, N, set_grid, plot2d, MV2, add_grid, code_file_name, os
# Curved grid
nxy = 10
lon = N.arange(nxy*1.)
lat = N.arange(nxy*1.)
gridi = rotate_grid((lon, lat), 30)
xxi = gridi.getLongitude()[:].filled()
yyi = gridi.getLatitude()[:].filled()
vari = MV2.array(yyi)
set_grid(vari, gridi)
kw = dict(vmin=vari.min(), vmax=vari.max())
P.figure(figsize=(10, 3.5))
P.subplot(131, aspect=1)
P.contourf(xxi, yyi, vari.asma(), **kw)
add_grid(gridi, edges=False, centers=-1)
xylims = (xxi.min(), xxi.max(), yyi.min(), yyi.max())
P.axis(xylims)
P.title('Curved grid')
# Interpolate to grid
xg, yg = N.meshgrid(N.arange(-3.5, 14.5), N.arange(-3.5, 14.5))
nxyg = xg.shape
cig = CurvedInterpolator(gridi, (xg, yg), g2g=True)
varog = cig(vari)
P.subplot(132, aspect=1)
P.scatter(xg, yg, c=varog.asma(), s=120, linewidth=0, **kw)
add_grid(gridi, edges=False, centers=-1)
# Generate random field
xg, yg, zzg = gridded_gauss3()
xi, yi, zi = random_gauss3(np=npi)
# Interpolate to random points
xo, yo = random_points(np=npo)
zo = krig(xi, yi, zi, xo, yo)
# Plot
# - source data
axis = [xg.min(), xg.max(), yg.min(), yg.max()]
kw = dict(vmin=zzg.min(), vmax=zzg.max())
kwim = dict(extent=axis, interpolation='bilinear', origin='lower', alpha=.2, **kw)
kwsc = dict(lw=0.2, **kw)
P.figure(figsize=(6, 3.5))
P.subplot(121)
P.title('Source field')
P.imshow(zzg, **kwim)
P.scatter(xi, yi, c=zi, **kwsc)
P.axis(axis)
# - interpolated data
P.subplot(122)
P.title('Interpolated points')
P.imshow(zzg, **kwim)
P.scatter(xo, yo, c=zo, marker='s', s=30, **kwsc)
P.axis(axis)
P.tight_layout()
savefigs(code_file_name(), verbose=False)
P.close()
xg, yg, zzg = gridded_gauss3()
xi, yi, zi = random_gauss3(np=npi)
zi = N.vstack((zi, zi)) # simulate several time steps for multifit
# Init kriger
kriger = OrdinaryCloudKriger(xi, yi, zi, npmax=npmax, nproc=nproc)
# Output points
xo, yo = random_points(np=npo)
# Interpolate
t0 = time()
zo = kriger(xo, yo)
# Plot
P.figure(figsize=(6, 8))
axis = [xg.min(), xg.max(), yg.min(), yg.max()]
kw = dict(vmin=zi.min(), vmax=zi.max())
kwim = dict(extent=axis, interpolation='bilinear', origin='lower', alpha=.2, **kw)
kwsc = dict(lw=0.2, **kw)
P.subplot(211)
P.imshow(zzg, **kwim)
P.scatter(xi, yi, c=zi if zi.ndim==1 else zi[0], s=20, **kwsc)
P.title('Input points')
P.subplot(212)
P.imshow(zzg, **kwim)
P.scatter(xo, yo, c=zo if zo.ndim==1 else zo[0], s=40, **kwsc)
P.title('Interpolated points')
P.tight_layout()
savefigs(code_file_name(), verbose=False)
P.close()
dyo = (yo[1] - yo[0]) * 0.49
yyi = N.resize(yi, vari.shape) + N.random.uniform(-dyi, dyi, vari.shape)
yyo = N.resize(yo, (nx, len(yo))) + N.random.uniform(-dyo, dyo, (nx, len(yo)))
yyib, xxib = meshcells(yyi, x)
yyob, xxob = meshcells(yyo, x)
varon = N.ma.masked_values(
interp1dxx(vari.filled(), yyi, yyo, mv, 0, extrap=0), mv)
varol = N.ma.masked_values(
interp1dxx(vari.filled(), yyi, yyo, mv, 1, extrap=0), mv)
varoh = N.ma.masked_values(
interp1dxx(vari.filled(), yyi, yyo, mv, 3, extrap=0), mv)
kw = dict(vmin=vari.min(), vmax=vari.max())
axlims = [x[0], x[-1], yo[0], yo[-1]]
P.figure(figsize=(8, 8))
P.subplot(221)
P.pcolor(xxib, yyib, vari)
P.axis(axlims)
P.title('Original')
P.subplot(222)
P.pcolor(xxob, yyob, varon, **kw)
P.axis(axlims)
P.title('Nearest1dxx')
P.subplot(223)
P.pcolor(xxob, yyob, varol, **kw)
P.axis(axlims)
P.title('Linear1dxx')
P.subplot(224)
P.pcolor(xxob, yyob, varoh, **kw)
P.axis(axlims)
yi = N.arange(nxy * 1.)
gridi = rotate_grid((xi, yi), 30)
xxi = gridi.getLongitude().getValue()
yyi = gridi.getLatitude().getValue()
zzi = N.ma.array(yyi)
zzi[int(nxy * 0.3):int(nxy * 0.8), int(nxy * 0.3):int(nxy * 0.8)] = N.ma.masked
zzi.shape = 1, nxy, nxy
# Output positions
no = 1000
xo = N.random.uniform(-nxy / 4., nxy + nxy / 4., no)
yo = N.random.uniform(-nxy / 4., nxy + nxy / 4., no)
# Interpolate
mv = zzi.get_fill_value()
zo = nearest2dto1dc(xxi, yyi, zzi.filled(mv), xo, yo, mv)
zo = N.ma.masked_values(zo, mv)
# Plot
kw = dict(vmin=zzi.min(), vmax=zzi.max())
P.figure(figsize=(6, 6))
P.subplot(111, aspect=1)
P.contourf(xxi, yyi, zzi[0], **kw)
add_grid((xxi, yyi), edges=False, centers=True, marker='o')
P.scatter(xo, yo, c=zo[0], s=50, **kw)
P.title('nearest2dto1dc')
figfile = code_file_name(ext='png')
if os.path.exists(figfile): os.remove(figfile)
P.savefig(figfile)
P.close()
"""Test the :func:`~vacumm.misc.grid.masking.create_polygon` function""" from vcmq import N, P, create_polygon, plot_polygon, code_file_name from _geoslib import Polygon # Data xx = N.array([0., 5., 4., 1.]) yy = N.array([0., 0., 2., 2.]) # From Polygon P.figure(figsize=(5, 5)) p0 = create_polygon(Polygon(N.array([xx, yy]).T)) plot_polygon(p0, color='b', label='From poly') # From data p1 = create_polygon(N.array([xx, yy]).T-.25) plot_polygon(p1, color='r', label='From data') # From transposed data p2 = create_polygon(N.array([xx, yy])-0.5) plot_polygon(p2, color='g', label='From transposed data') # From min/max p3 = create_polygon([xx.min(), yy.min(), xx.max(), yy.max()]) plot_polygon(p3, color='m', label='From min/max') # With projection proj = lambda x, y: (x*1.5, y*1.5) p4 = create_polygon(p0, proj=proj) plot_polygon(p4, color='cyan', label='With projection') # Save
# Imports from vcmq import P, darken, code_file_name, plot_cmap cmap_name = "CMRmap" cmap_old = P.get_cmap(cmap_name) # Max cmap_10 = darken(cmap_old, 1) # Medium cmap_05 = darken(cmap_old, 0.5) # Min cmap_00 = darken(cmap_old, 0) # Plot fig = P.figure(figsize=(6, 2)) kw = dict(fig=fig, figsize=None, aspect=.15, show=False, close=False) P.subplot(221) plot_cmap(cmap_old, title='Original', **kw) P.subplot(222) plot_cmap(cmap_00, title='f = 0.0', **kw) P.subplot(223) plot_cmap(cmap_05, title='f = 0.5', **kw) P.subplot(224) plot_cmap(cmap_10, title='f = 1.0', **kw) P.figtext(.5, 1, 'Darken ' + cmap_name, va='top', ha='center', size=12) P.tight_layout() P.savefig(code_file_name(ext='png')) P.close()
method='conservative',
diag=diag,
coordSys='cart')
# Norm
frac = diag['dstAreaFractions']
mask = frac == 0.
frac[mask] = 1.
varo[:] /= frac
varo[:] = MV2.masked_where(mask, varo, copy=0)
# Plot
rc('font', size=9)
kw = dict(vmin=vari.min(), vmax=vari.max())
axis = [xxbi.min(), xxbi.max(), yybo.min(), yybo.max()]
P.figure(figsize=(7, 3))
P.subplot(121, aspect=1)
P.pcolormesh(xxbi, yybi, vari, **kw)
P.colorbar(shrink=0.7)
add_grid(gridi, color='0.5')
add_grid(grido)
P.axis(axis)
P.title('Original: max=%g min=%g' % (vari.max(), vari.min()))
P.subplot(122, aspect=1)
P.pcolormesh(xxbo, yybo, varo, **kw)
P.colorbar(shrink=0.7)
add_grid(gridi)
add_grid(grido, color='0.5')
P.axis(axis)
P.title('Regridded: max=%g min=%g' % (varo.max(), varo.min()))
P.tight_layout()
fill_value=1e20)
N.random.seed(0)
xo = N.random.uniform(lon0, lon1, np)
yo = N.random.uniform(lat0, lat1, np)
zo = N.random.uniform(dep0, dep1, np)
to = comptime(N.random.uniform(reltime(time0, time.units).value,
reltime(time1, time.units).value, np),
time.units)
# Rectangular xyzt with 1d z
vo = grid2xy(vi, xo=xo, yo=yo, zo=zo, to=to, method='linear')
von = grid2xy(vi, xo=xo, yo=yo, zo=zo, to=to, method='nearest')
assert vo.shape==(ne, np)
N.testing.assert_allclose(vo[0], yo)
kwp = dict(vmin=vi.min(), vmax=vi.max())
P.figure(figsize=(6, 3))
P.subplot(121)
P.scatter(xo, yo, c=vo[0], cmap='jet', **kwp)
add_grid(vi.getGrid())
P.title('linear4d')
P.subplot(122)
P.scatter(xo, yo, c=von[0], cmap='jet', **kwp)
add_grid(vi.getGrid())
P.title('nearest4d')
P.figtext(.5, .98, 'grid2xy in 4D', va='top', ha='center', weight='bold')
P.tight_layout()
P.savefig(code_file_name(ext='.png'))
P.close()
# Reversed z and y
vi_revz = vi[:, :, ::-1, ::-1, :]
zzov = griddata(xi, yi, zi, (xr, yr), method='nat', ext=True, sub=10)
# -> Tester la methode "carg"
# -> Testez parametre sub=...
# -> Essayer avec GridData
zov2 = xy2xy(xi, yi, zi, xo, yo)
# Krigeage
from vacumm.misc.grid.kriging import krig
zzok = krig(xi, yi, zi, xxr.ravel(), yyr.ravel(), nproc=1).reshape(zzr.shape)
# -> Tester nproc et npmax
# Plots
from vcmq import meshbounds, P
xxrb, yyrb = meshbounds(xr, yr)
P.figure(figsize=(10, 8))
axis = [xxrb.min(), xxrb.max(), yyrb.min(), yyrb.max()]
#
P.subplot(332)
P.pcolormesh(xxrb, yyrb, zzr, **vminmax)
P.scatter(xi, yi, c='k')
P.title('Original')
P.axis(axis)
#
P.subplot(334)
P.pcolormesh(xxrb, yyrb, zzoc, **vminmax)
P.title('CDAT/Natgrid')
P.axis(axis)
#
P.subplot(335)
P.pcolormesh(xxrb, yyrb, zzork, **vminmax)
# Interpolate to random points
xo, yo = random_points(np=npo)
zo = krig(xi, yi, zi, xo, yo)
# Plot
# - source data
axis = [xg.min(), xg.max(), yg.min(), yg.max()]
kw = dict(vmin=zzg.min(), vmax=zzg.max())
kwim = dict(extent=axis,
interpolation='bilinear',
origin='lower',
alpha=.2,
**kw)
kwsc = dict(lw=0.2, **kw)
P.figure(figsize=(6, 3.5))
P.subplot(121)
P.title('Source field')
P.imshow(zzg, **kwim)
P.scatter(xi, yi, c=zi, **kwsc)
P.axis(axis)
# - interpolated data
P.subplot(122)
P.title('Interpolated points')
P.imshow(zzg, **kwim)
P.scatter(xo, yo, c=zo, marker='s', s=30, **kwsc)
P.axis(axis)
P.tight_layout()
savefigs(code_file_name(), verbose=False)
P.close()
taxi.toRelativeTime("hours since 2000")
ctimesi = taxi.asComponentTime()
ct0 = round_date(ctimesi[0], "hour")
ct1 = round_date(ctimesi[-1], "hour")
taxo = create_time(lindates(ct0, ct1, 1, "hour"), taxi.units)
# Lag error
# - estimation
els = []
lags = N.arange(1, 6)
for lag in lags:
els.append(N.sqrt(((sp[lag:] - sp[:-lag]) ** 2).mean()))
els = N.array(els)
a, b, _, _, _ = linregress(lags, els)
# - plot
P.figure(figsize=(6, 6))
P.subplot(211)
P.plot(lags, els, "o")
P.plot([0, lags[-1]], [b, a * lags[-1] + b], "g")
P.axhline(b, color="0.8", ls="--")
P.ylim(ymin=0)
P.xlabel("Lag [hour]")
P.ylabel("Error [m s-1]")
add_key(1)
P.title("Linear lag error model")
# Interpolation
sph, speh = regrid1d(sp, taxo, method="cellerr", erri=spe, errl=-a, geterr=True)
# Time zoom for plot clarity
tzoom = (ct1.sub(7, cdtime.Hour), ctimesi[-1])
zo = sck(xo, yo)
zoe = scke(xo, yo)
zzg = zo.reshape(ny, nx)
zzge = zoe.reshape(ny, nx)
# Plot
vmin = min(zi.min(), zo.min(), zoe.min())
vmax = max(zi.max(), zo.max(), zoe.max())
vmax = max(abs(vmin), abs(vmax))
vmin = -vmax
cmap = 'cmocean_tempo'
cmap = 'cmocean_delta'
kw = dict(vmin=vmin, vmax=vmax)
P.figure(figsize=(6, 5.5))
ax = P.subplot(221)
P.pcolormesh(xxg, yyg, zzg, cmap=cmap, **kw)
P.colorbar()
P.scatter(xi, yi, c=zi, s=100, cmap=cmap, **kw)
P.title('Without obs error')
P.subplot(222, sharex=ax, sharey=ax)
P.pcolormesh(xxg, yyg, zzge, cmap=cmap, **kw)
P.colorbar()
P.scatter(xi, yi, c=zi, s=100, cmap=cmap, **kw)
#P.axis('image')
P.title('With obs error')
P.subplot(223, sharex=ax, sharey=ax)
nx = ny = 50
np = 500
mtype = 'gauss'
dmax = 2.5
from vcmq import N, P, code_file_name, savefigs
from vacumm.misc.grid.kriging import gridded_gauss3, random_gauss3, variogram, variogram_fit
# Generate random field
xxg, yyg, zzg = gridded_gauss3(nx=nx, ny=ny)
x, y, z = random_gauss3(np=np)
# Variogram from data
d, v = variogram(x, y, z, dmax=dmax)
# Variogram fit with null nugget
vm = variogram_fit(x, y, z, mtype, n=0, dmax=dmax)
D = N.linspace(0, d.max())
V = vm(D)
# Compare
P.figure(figsize=(6, 4))
P.title('Variogram')
P.plot(d, v, 'ob', label='From data')
P.plot(D, V, '-r', label='Fitted model (%s)'%mtype)
P.legend(loc='best')
P.ylim(ymin=0)
savefigs(code_file_name(ext=False), pdf=True, verbose=False)
P.close()
# Output
grido = create_grid(xi[:]+2*2.5, yi[:]+3*1.5)
xo = grido.getLongitude()
yo = grido.getLatitude()
xo.setBounds(bounds1d(xo))
yo.setBounds(bounds1d(yo))
xxob, yyob = meshbounds(xo, yo)
# Regridding
varo, wo = vari.regrid(grido, tool='regrid2', returnTuple=1)
# Plot
kw = dict(fill='pcolor', contour=False, xhide=True, yhide=True,
xticks=[], yticks=[], cmap='jet',
colorbar=False, show=False)
P.figure(figsize=(6, 3))
p = plot2d(vari, subplot=131, title='Original', **kw)
add_grid(gridi)
add_grid(grido)
P.axis('image')
p = plot2d(varo, subplot=132, title='Regridded', **kw)
add_grid(gridi)
add_grid(grido)
P.axis('image')
P.subplot(133)
P.pcolor(xxob, yyob, wo)
add_grid(grido)
P.title("Output weights")
P.tight_layout()
savefigs(code_file_name(),pdf=True, verbose=False)
P.close()
vari[nx/3:2*nx/3, nyi/3:2*nyi/3] = N.ma.masked
x = N.arange(nx)
dyi = (yi[1]-yi[0])*0.49
dyo = (yo[1]-yo[0])*0.49
yyi = N.resize(yi, vari.shape)+N.random.uniform(-dyi, dyi, vari.shape)
yyo = N.resize(yo, (nx, len(yo)))+N.random.uniform(-dyo, dyo, (nx, len(yo)))
yyib, xxib = meshcells(yyi, x)
yyob, xxob = meshcells(yyo, x)
varon = N.ma.masked_values(interp1dxx(vari.filled(), yyi, yyo, mv, 0, extrap=0), mv)
varol = N.ma.masked_values(interp1dxx(vari.filled(), yyi, yyo, mv, 1, extrap=0), mv)
varoh = N.ma.masked_values(interp1dxx(vari.filled(), yyi, yyo, mv, 3, extrap=0), mv)
kw = dict(vmin=vari.min(), vmax=vari.max())
axlims = [x[0], x[-1], yo[0], yo[-1]]
P.figure(figsize=(8, 8))
P.subplot(221)
P.pcolor(xxib, yyib, vari)
P.axis(axlims)
P.title('Original')
P.subplot(222)
P.pcolor(xxob, yyob, varon, **kw)
P.axis(axlims)
P.title('Nearest1dxx')
P.subplot(223)
P.pcolor(xxob, yyob, varol, **kw)
P.axis(axlims)
P.title('Linear1dxx')
P.subplot(224)
P.pcolor(xxob, yyob, varoh, **kw)
P.axis(axlims)
xg, yg, zzg = gridded_gauss3()
xi, yi, zi = random_gauss3(np=npi)
zi = N.vstack((zi, zi)) # simulate several time steps for multifit
# Init kriger
kriger = OrdinaryCloudKriger(xi, yi, zi, npmax=npmax, nproc=nproc)
# Output points
xo, yo = random_points(np=npo)
# Interpolate
t0 = time()
zo = kriger(xo, yo)
# Plot
P.figure(figsize=(6, 8))
axis = [xg.min(), xg.max(), yg.min(), yg.max()]
kw = dict(vmin=zi.min(), vmax=zi.max())
kwim = dict(extent=axis,
interpolation='bilinear',
origin='lower',
alpha=.2,
**kw)
kwsc = dict(lw=0.2, **kw)
P.subplot(211)
P.imshow(zzg, **kwim)
P.scatter(xi, yi, c=zi if zi.ndim == 1 else zi[0], s=20, **kwsc)
P.title('Input points')
P.subplot(212)
P.imshow(zzg, **kwim)
P.scatter(xo, yo, c=zo if zo.ndim == 1 else zo[0], s=40, **kwsc)
