def plot_potential_skillgain():
res = pd.read_csv('/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv', index_col=0)
modes = ['abs', 'anom_lst', 'anom_lt', 'anom_st']
f = plt.figure(figsize=(15,8))
for i, mode in enumerate(modes):
R = res[f'ubrmse_grid_{mode}_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P = res[f'ubrmse_grid_{mode}_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R2 = res[f'r2_grid_{mode}_m_CLSM_tc_ASCAT_SMAP_CLSM']
K = P / (R + P)
NSR = (1 - R2) / R2
R2upd = 1 / (1 + (1 - K) * NSR)
res['gain_pot'] = R2upd - R2
plt.subplot(2, 2, i+1)
plot_ease_img(res, 'gain_pot', fontsize=12, cbrange=[-0.2,0.2], cmap=cc.cm.bjy, log_scale=False, title=mode, plot_cb=True)
fout = f'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/plots/skillgain_pot_Pcorr_simple.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_kalman_gains():
res = pd.read_csv('/Users/u0116961/Documents/work/validation_good_practice/CI80/ASCAT_SMAP_CLSM/result.csv', index_col=0)
# k_repr_abs = [(1,0), (2,0), (3,0), (1,0.01), (1,0.02)]
# k_repr_lt = [(1,0), (2,0), (3,0), (1,0.01), (1,0.02)]
# k_repr_lst = [(1,0), (2,0), (3,0), (1,0.01), (1,0.02)]
# k_repr_st = [(1,0), (1.5,0), (2.5,0), (1,0.005), (1,0.01)]
k_repr_abs = [(0.8,0), (1,0), (1.5,0)]
k_repr_lt = [(0.8,0), (1,0), (1.5,0)]
k_repr_lst = [(0.8,0), (1,0), (1.5,0)]
k_repr_st = [(0.8,0), (1,0), (1.5,0)]
modes = ['abs', 'anom_lt', 'anom_lst', 'anom_st']
f = plt.figure(figsize=(23,10))
for j, (mode, k_repr) in enumerate(zip(modes,[k_repr_abs,k_repr_lt,k_repr_lst,k_repr_st])):
for i, k in enumerate(k_repr):
k_m, k_a = k
tag_r = f'ubrmse_grid_{mode}_m_SMAP_tc_ASCAT_SMAP_CLSM'
tag_p = f'ubrmse_grid_{mode}_m_CLSM_tc_ASCAT_SMAP_CLSM'
res['tmp_tag'] = 1 / (1 + (res[tag_r] * k_m + k_a)**2 / res[tag_p] ** 2)
res.loc[res['tmp_tag']<0.5,'tmp_tag'] **= 1.75
plt.subplot(3, 4, i*4 + 1 + j)
plot_ease_img(res, 'tmp_tag', fontsize=12, cbrange=[0,1], cmap='seismic_r', log_scale=False, title=f'{mode}: k_m = {k_m} / k_a = {k_a}', plot_cb=True)
fout = f'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/plots/obs_weight.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_TCA_reliability():
res = pd.read_csv('/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv', index_col=0)
modes = ['abs', 'anom_lst', 'anom_lt', 'anom_st']
f = plt.figure(figsize=(15,8))
cmap = mpl.cm.get_cmap("magma", 8)
for i, mode in enumerate(modes):
p_a_s = res[f'p_grid_{mode}_p_ASCAT_SMAP']
p_a_c = res[f'p_grid_{mode}_p_ASCAT_CLSM']
p_s_c = res[f'p_grid_{mode}_p_SMAP_CLSM']
r_a_s = res[f'r_grid_{mode}_p_ASCAT_SMAP']
r_a_c = res[f'r_grid_{mode}_p_ASCAT_CLSM']
r_s_c = res[f'r_grid_{mode}_p_SMAP_CLSM']
res[f'p_mask_{mode}'] = ((p_a_s <= 0.05)&(r_a_s > 0.2)) * 1 + \
((p_a_c <= 0.05)&(r_a_c > 0.2)) * 2 + \
((p_s_c <= 0.05)&(r_s_c > 0.2)) * 4
plt.subplot(2, 2, i+1)
plot_ease_img(res, f'p_mask_{mode}', fontsize=12, cbrange=[0,8], cmap=cmap, log_scale=False, title=mode, plot_cb=True)
plt.tight_layout()
plt.show()
def plot_orthogonality_check(dir_out):
sensors = ['ASCAT', 'SMAP', 'CLSM']
res = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
figsize = (11, 3)
fontsize = 10
# cb = [-0.02, 0.02]
cb = [-0.02, 0.02]
f = plt.figure(figsize=figsize)
valid = pd.Series(True, index=res.index)
for mode in ['anom_lt', 'anom_st', 'anom_lst']:
r_asc_smap = res[f'r_grid_{mode}_p_ASCAT_SMAP']
r_asc_clsm = res[f'r_grid_{mode}_p_ASCAT_CLSM']
r_smap_clsm = res[f'r_grid_{mode}_p_SMAP_CLSM']
thres = 0.2
valid &= ((r_asc_smap > thres) & (r_asc_smap > thres) &
(r_asc_smap > thres))
ind_valid = res[valid].index
for i, s in enumerate(sensors[1::]):
plt.subplot(1, 2, i + 1)
tag_lt = 'ubrmse_grid_anom_lt_m_' + s + '_tc_ASCAT_SMAP_CLSM'
tag_st = 'ubrmse_grid_anom_st_m_' + s + '_tc_ASCAT_SMAP_CLSM'
tag_lst = 'ubrmse_grid_anom_lst_m_' + s + '_tc_ASCAT_SMAP_CLSM'
res['diff'] = (res[tag_lst] - np.sqrt(res[tag_lt]**2 + res[tag_st]**2))
im_r = plot_ease_img(res.reindex(ind_valid),
'diff',
fontsize=fontsize + 2,
cbrange=cb,
cmap=cc.cm.bjy,
print_meanstd=True)
plt.title(s, fontsize=fontsize)
plot_centered_cbar(f,
im_r,
2,
fontsize=fontsize - 2,
bottom=0.00,
hspace=0.030,
pad=0.02,
wdth=0.04)
fout = dir_out / 'orthogonality_verification.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_tca_uncertainties(dir_out):
sensors = ['ASCAT', 'SMAP', 'CLSM']
res = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
figsize = (15, 5)
fontsize = 12
cb = [0, 0.04]
modes = ['anom_lst', 'anom_lt', 'anom_st']
titles = ['Anomalies', 'LF signal', 'HF signal']
labels = [
'$\widehat{std}(\epsilon_{\Theta,smap})$',
'$\widehat{std}(\epsilon_{\Theta,clsm})$'
]
f = plt.figure(figsize=figsize)
n = 0
pos = []
for s, l in zip(sensors[1::], labels):
for mode, title in zip(modes, titles):
n += 1
plt.subplot(2, 3, n)
tag = 'ubrmse_grid_' + mode + '_m_' + s + '_tc_ASCAT_SMAP_CLSM'
r_asc_smap = res[f'r_grid_{mode}_p_ASCAT_SMAP']
r_asc_clsm = res[f'r_grid_{mode}_p_ASCAT_CLSM']
r_smap_clsm = res[f'r_grid_{mode}_p_SMAP_CLSM']
thres = 0.2
ind_valid = res[(r_asc_smap > thres) & (r_asc_smap > thres) &
(r_asc_smap > thres)].index
im_r = plot_ease_img(res.reindex(ind_valid),
tag,
fontsize=fontsize,
cbrange=cb,
cmap='viridis',
print_mean=True)
# if (n == 6) | (n == 7):
# pos += [im_r.axes.get_position()]
if s == 'SMAP':
plt.title(title, fontsize=fontsize)
if mode == 'anom_lst':
plt.ylabel(l, fontsize=fontsize)
plot_centered_cbar(f, im_r, 3, fontsize=fontsize - 2, bottom=0.07)
fout = dir_out / 'tca_uncertainties.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_eval():
analysis_only = True
relative = True
snr = False
pc = 'noPcorr'
ref_exp = f'OL_{pc}'
# ref_exp = f'DA_{pc}_4K'
ana = 'ana_' if analysis_only else ''
rel = 'rel_' if relative else ''
res = pd.read_csv(Path(
f'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/validation/ascat_eval.csv'
),
index_col=0)
fbase = Path(
f'~/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/plots/ascat_eval/ref_{ref_exp}'
).expanduser()
if not fbase.exists():
Path.mkdir(fbase, parents=True)
# topo = pd.read_csv('/Users/u0116961/data_sets/ASCAT/static_layer/topographic_complexity_conus.csv', index_col=0, squeeze=True, header=None)
cbr_r = [0.6, 1.4] if relative else [0.2, 0.7]
cbr_r = [-3, 3] if snr else [-0.2, 0.2] if relative else [0.2, 0.7]
cbr_len = [2e2, 2e3]
runs = [f'OL_{pc}'] + [
f'DA_{pc}_{err}'
for err in ['4K', 'abs', 'anom_lt', 'anom_lst', 'anom_st']
]
modes = ['abs', 'anom_st', 'anom_lt', 'anom_lst']
# params = ['len',]
# cbranges = [cbr_len]
params = ['r_corr']
cbranges = [cbr_r, cbr_r]
# r_tca_
# 'len', 'r', 'r_corr', 'ana_len', 'ana_r', 'ana_r_corr'
fontsize = 12
for p, cbr in zip(params, cbranges):
# cmap = 'seismic_r' if ('r' in p) & relative else 'YlGn'
cmap = cc.cm.bjy
for i, m in enumerate(modes):
f = plt.figure(figsize=(25, 9))
for cnt, r in enumerate(runs):
ncols = 3
plt.subplot(2, ncols, cnt + 1)
col = f'{ana}{p}_{r}_{m}'
ref_col = f'{ana}{p}_{ref_exp}_{m}'
if relative and (p != 'len'):
if snr:
SNR1 = 10**np.log10(res[col]**2 / (1 - res[col]**2))
SNR2 = 10**np.log10(res[ref_col]**2 /
(1.01 - res[ref_col]**2))
res[f'{col}_diff'] = SNR1 - SNR2
else:
res[f'{col}_diff'] = res[col]**2 - res[ref_col]**2
# res[f'{col}_diff'] = (res[col]**2 - res[ref_col]**2) / res[ref_col]**2
# res.loc[topo.reindex(res.index) > 15, f'{col}_diff'] = np.nan
ext = '_diff'
else:
ext = ''
log_scale = True if 'len' in p else False
im = plot_ease_img(res,
f'{col}{ext}',
title=f'{r}',
cmap=cmap,
cbrange=cbr,
fontsize=fontsize,
print_median=True,
log_scale=log_scale)
# im = plot_ease_img(res, f'{col}{ext}' , title=f'{e} / {m}', cmap=cmap, cbrange=cbr, fontsize=fontsize, print_median=True, log_scale=log_scale)
# if (i == 2) & (j == 1):
# ax = im.axes
plot_centered_cbar(f, im, ncols, fontsize=fontsize)
fname = f'{ana}rel_snr_{m}' if snr else f'{ana}{rel}{p}_{m}.png'
f.savefig(fbase / fname, dpi=300, bbox_inches='tight')
plt.close()
def plot_smos_eval_diffs():
difference = True
sub = 'skillgain' if difference else ''
res = pd.read_csv(Path(
f'~/Documents/work/MadKF/CLSM/SMOS40/validation/multiperiod/ascat/ascat_eval_smos.csv'
).expanduser(),
index_col=0)
fbase = Path(f'~/Documents/work/MadKF/CLSM/SMOS40/plots/multiperiod/ascat'
).expanduser() / sub
n_subits = 4
exps = ['SMOS40_it621', 'SMOS40_it622', 'SMOS40_it623', 'SMOS40_it624']
modes = ['absolute', 'shortterm', 'longterm']
per = [
'2010-2020', '2010-2015', '2015-2020', '2010-2012', '2012-2015',
'2015-2017', '2017-2020'
]
p = 'r'
cbr = [-0.2, 0.2]
fontsize = 10
cmap = 'seismic_r' if ('r' in p) else 'YlGn'
combs = list(combinations(np.arange(7), 2))
combs = [combs[i] for i in [0, 1, 6, 7, 8, 13, 14, 15, 18, 20]]
for j, e in enumerate(exps):
froot = fbase / e
if not froot.exists():
Path.mkdir(froot, parents=True)
for i, m in enumerate(modes):
f = plt.figure(figsize=(26, 6))
for k, comb in enumerate(combs):
plt.subplot(2, 5, k + 1)
if difference:
res[f'diff_{k}'] = (res[f'p{comb[0]}_{p}_{e}_{m}'] - res[f'p{comb[0]}_{p}_open_loop_{m}']) \
- (res[f'p{comb[1]}_{p}_{e}_{m}'] - res[f'p{comb[1]}_{p}_open_loop_{m}'])
else:
res[f'diff_{k}'] = res[f'p{comb[0]}_{p}_{e}_{m}'] - res[
f'p{comb[1]}_{p}_{e}_{m}']
title = f'{per[comb[0]]} minus {per[comb[1]]}'
im = plot_ease_img(res,
f'diff_{k}',
title=title,
cmap=cmap,
cbrange=cbr,
fontsize=fontsize,
print_median=True)
plot_centered_cbar(f, im, 5, fontsize=fontsize, col_offs=0)
f.savefig(froot / f'{m}.png', dpi=300, bbox_inches='tight')
plt.close()
def plot_smos_eval():
res = pd.read_csv(Path(
f'~/Documents/work/MadKF/CLSM/SMOS40/validation/multiperiod/ascat_eval_smos.csv'
).expanduser(),
index_col=0)
fbase = Path(
f'~/Documents/work/MadKF/CLSM/SMOS40/plots/multiperiod/per_period'
).expanduser()
if not fbase.exists():
Path.mkdir(fbase, parents=True)
cbr_r = [0, 0.8]
# cbr_r = [-0.2,0.2]
# cbr_len = [2e2, 2e3]
cbr_len = [0, 250]
n_subits = 4
exps = ['SMOS40_it621', 'SMOS40_it622', 'SMOS40_it623', 'SMOS40_it624']
modes = ['absolute', 'shortterm', 'longterm']
# ref_exp = 'SMOS40_it622'
params = ['r', 'len']
cbranges = [cbr_r, cbr_len]
# params = ['r',]
# cbranges = [cbr_r,]
fontsize = 12
for p, cbr in zip(params, cbranges):
cmap = 'seismic_r' if ('r' in p) else 'YlGn'
for per in np.arange(7):
f = plt.figure(figsize=(25, 10))
for i, m in enumerate(modes):
for j, e in enumerate(exps):
plt.subplot(len(modes), len(exps), len(exps) * i + j + 1)
col = f'p{per}_{p}_{e}_{m}'
im = plot_ease_img(res,
f'{col}',
title=f'{e} / {m}',
cmap=cmap,
cbrange=cbr,
fontsize=fontsize,
print_median=True,
log_scale=False)
plot_centered_cbar(f, im, len(exps), fontsize=fontsize)
f.savefig(fbase / f'{p}_per{per}.png',
dpi=300,
bbox_inches='tight')
plt.close()
def plot_ascat_smap_eval(iteration):
analysis_only = True
relative = True
ref_exp = 'open_loop'
sub = 'ana_' if analysis_only else ''
fext = '_rel' if relative else '_abs'
res_ascat = pd.read_csv(
f'/Users/u0116961/Documents/work/MadKF/CLSM/SMAP/validation/iter_{iteration}/ascat_eval.csv',
index_col=0)
res_smap = pd.read_csv(
f'/Users/u0116961/Documents/work/MadKF/CLSM/SMAP/validation/iter_{iteration}/smap_eval.csv',
index_col=0)
fbase = Path(
f'/Users/u0116961/Documents/work/MadKF/CLSM/SMAP/plots/iter_{iteration}'
)
if not fbase.exists():
Path.mkdir(fbase, parents=True)
cbr_r = [0.6, 1.4] if relative else [0.2, 0.7]
cbr_r = [-0.4, 0.4] if relative else [0.2, 0.7]
cbr_len = [500, 1500]
exps = ['SMOSSMAP_short', 'SMAP_it11']
modes = ['absolute', 'longterm', 'shortterm']
# params = ['len', 'r']
# cbranges = [cbr_len, cbr_r]
params = ['r']
cbranges = [cbr_r]
fontsize = 12
for p, cbr in zip(params, cbranges):
cmap = 'seismic_r' if (p == 'r') & relative else 'YlGn'
f = plt.figure(figsize=(22, 8))
for j, e in enumerate(exps):
for i, m in enumerate(modes):
plt.subplot(len(exps), len(modes), len(modes) * j + i + 1)
col = f'{sub}{p}_{e}_{m}'
if relative and (p != 'len'):
ref_col = f'{sub}{p}_{ref_exp}_{m}'
res_ascat[f'{col}_diff'] = (res_ascat[col] -
res_ascat[ref_col]) / (
1 - res_ascat[ref_col])
res_smap[f'{col}_diff'] = (res_smap[col] -
res_smap[ref_col]) / (
1 - res_smap[ref_col])
ext = '_diff'
else:
ext = ''
if j == 0:
im = plot_ease_img(res_ascat,
f'{col}{ext}',
title=f'Skill gain {m} (ref: ASCAT)',
cmap=cmap,
cbrange=cbr,
fontsize=fontsize,
print_median=True)
else:
im = plot_ease_img(res_smap,
f'{col}{ext}',
title=f'Skill gain {m} (ref: SMAP)',
cmap=cmap,
cbrange=cbr,
fontsize=fontsize,
print_median=True)
f.subplots_adjust(wspace=0.04, hspace=0.025, bottom=0.06)
# pos1 = f.axes[-3].get_position()
# pos2 = f.axes[-4].get_position()
# x1 = (pos1.x0 + pos1.x1) / 2
# x2 = (pos2.x0 + pos2.x1) / 2
pos = f.axes[-2].get_position()
x1 = pos.x0
x2 = pos.x1
cbar_ax = f.add_axes([x1, 0.03, x2 - x1, 0.03])
cbar = f.colorbar(im, orientation='horizontal', cax=cbar_ax)
for t in cbar.ax.get_xticklabels():
t.set_fontsize(fontsize)
f.savefig(fbase / f'ascat_smap_eval.png', dpi=300, bbox_inches='tight')
plt.close()
def plot_ascat_eval():
analysis_only = True
relative = True
ref_exp = 'open_loop'
sub = 'ana_' if analysis_only else ''
fext = '_rel' if relative else '_abs'
res = pd.read_csv(
'/Users/u0116961/Documents/work/LDAS/2020-03_scaling/validation/ascat_eval.csv',
index_col=0)
cbr_r = [0.6, 1.4] if relative else [0.2, 0.7]
cbr_r = [-0.3, 0.3] if relative else [0.2, 0.7]
cbr_len = [500, 1500]
exps = ['SMOSSMAP_short', 'MadKF_SMOS40']
# exps = ['SMOS_long', 'SMOSSMAP_long', 'SMOS_short', 'SMOSSMAP_short', 'SMOSSMAP_short_PCA', 'SMAP_short']
modes = ['absolute', 'longterm', 'shortterm']
params = ['len', 'r']
cbranges = [cbr_len, cbr_r]
# params = ['r']
# cbranges = [cbr_r]
fontsize = 12
for p, cbr in zip(params, cbranges):
cmap = 'seismic_r' if (p == 'r') & relative else 'YlGn'
f = plt.figure(figsize=(10, 8))
for i, m in enumerate(modes):
for j, e in enumerate(exps):
plt.subplot(len(modes), len(exps), len(exps) * i + j + 1)
col = f'{sub}{p}_{e}_{m}' if p != 'len' else f'{sub}{p}_{e}{m}'
if relative and (p != 'len'):
ref_col = f'{sub}{p}_{ref_exp}_{m}'
res[f'{col}_diff'] = (res[col] -
res[ref_col]) / (1 - res[ref_col])
ext = '_diff'
else:
ext = ''
im = plot_ease_img(res,
f'{col}{ext}',
title=f'{e} / {m}',
cmap=cmap,
cbrange=cbr,
fontsize=fontsize,
print_median=True)
# if (i == 2) & (j == 1):
# ax = im.axes
f.subplots_adjust(wspace=0.04, hspace=0.025, bottom=0.06)
pos1 = f.axes[-2].get_position()
pos2 = f.axes[-1].get_position()
x1 = (pos1.x0 + pos1.x1) / 2
x2 = (pos2.x0 + pos2.x1) / 2
cbar_ax = f.add_axes([x1, 0.03, x2 - x1, 0.03])
cbar = f.colorbar(im, orientation='horizontal', cax=cbar_ax)
for t in cbar.ax.get_xticklabels():
t.set_fontsize(fontsize)
fout = f'/Users/u0116961/Documents/work/LDAS/2020-03_scaling/plots/ascat_eval_{p}{sub}{fext}.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_potential_skillgain_decomposed():
res = pd.read_csv('/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv', index_col=0)
R_abs = res[f'ubrmse_grid_abs_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_abs = res[f'ubrmse_grid_abs_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_anom_lt = res[f'ubrmse_grid_anom_lt_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_anom_lt = res[f'ubrmse_grid_anom_lt_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_anom_lst = res[f'ubrmse_grid_anom_lst_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_anom_lst = res[f'ubrmse_grid_anom_lst_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_anom_st = res[f'ubrmse_grid_anom_st_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_anom_st = res[f'ubrmse_grid_anom_st_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_clim = (R_abs - R_anom_lt - R_anom_st).abs()
P_clim = (P_abs - P_anom_lt - P_anom_st).abs()
# Baseline estimates
R2 = res[f'r2_grid_abs_m_CLSM_tc_ASCAT_SMAP_CLSM']
SNR = R2 / (1 - R2)
SIG = SNR * P_abs
modes = ['abs', 'anom_lt', 'anom_lst', 'anom_st']
result = pd.DataFrame(index=res.index, columns=modes)
result['row'] = res.row
result['col'] = res.col
for cnt, i in enumerate(res.index):
print(f'{cnt} / {len(res)}')
R11 = R_clim.loc[i]
R22 = R_anom_lt.loc[i]
R33 = R_anom_st.loc[i]
P11 = P_clim.loc[i]
P22 = P_anom_lt.loc[i]
P33 = P_anom_st.loc[i]
r_c_l = 0.2
r_c_s = 0.8
r_l_s = 0.2
R12 = r_c_l * np.sqrt(R11 * R22)
R13 = r_c_s * np.sqrt(R11 * R33)
R23 = r_l_s * np.sqrt(R22 * R33)
P12 = r_c_l * np.sqrt(P11 * P22)
P13 = r_c_s * np.sqrt(P11 * P33)
P23 = r_l_s * np.sqrt(P22 * P33)
S = np.matrix([[R11, R12, R13, 0, 0, 0 ],
[R12, R22, R23, 0, 0, 0 ],
[R13, R23, R33, 0, 0, 0 ],
[0, 0, 0, P11, P12, P13],
[0, 0, 0, P12, P22, P23],
[0, 0, 0, P13, P23, P33]])
for mode in modes:
P = res.loc[i, f'ubrmse_grid_{mode}_m_CLSM_tc_ASCAT_SMAP_CLSM'] ** 2
R = res.loc[i, f'ubrmse_grid_{mode}_m_SMAP_tc_ASCAT_SMAP_CLSM'] ** 2
K = P / (R + P)
A = np.matrix([K, K, K, 1-K, 1-K, 1-K])
P_upd = (A * S * A.T)[0,0]
NSR_upd = P_upd / SIG.loc[i]
R2upd = 1 / (1 + NSR_upd)
result.loc[i, mode] = R2upd - R2.loc[i]
f = plt.figure(figsize=(15,8))
for i, mode in enumerate(modes):
plt.subplot(2, 2, i+1)
plot_ease_img(result, mode, fontsize=12, cbrange=[-0.2,0.2], cmap=cc.cm.bjy, log_scale=False, title=mode, plot_cb=True)
fout = f'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/plots/skillgain_pot_Pcorr_corr_asym.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_uncertainty_ratios(dir_out):
sensors = ['ASCAT', 'SMAP', 'CLSM']
res = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
res_tc = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
tg = GEOSldas_io().grid.tilegrids
res_cols = res.col.values - tg.loc['domain', 'i_offg']
res_rows = res.row.values - tg.loc['domain', 'j_offg']
figsize = (16, 4)
fontsize = 10
cb = [-10, 10]
cmap = cc.cm.bjy
modes = ['4K', 'anom_lst', 'anom_lt', 'anom_st']
titles = ['4K benchmark', 'Anomalies', 'LF signal', 'HF signal']
ios = [
GEOSldas_io('ObsFcstAna',
exp=f'NLv4_M36_US_DA_SMAP_Pcorr_{mode}').timeseries
for mode in modes
]
io_ol = GEOSldas_io('ObsFcstAna', exp=f'NLv4_M36_US_OL_Pcorr').timeseries
grid = GEOSldas_io().grid
f = plt.figure(figsize=figsize)
for n, (mode, title, io_da) in enumerate(zip(modes, titles, ios)):
if n > 0:
plt.subplot(2, 4, n + 1)
tagP = 'ubrmse_grid_' + mode + '_m_CLSM_tc_ASCAT_SMAP_CLSM'
tagR = 'ubrmse_grid_' + mode + '_m_SMAP_tc_ASCAT_SMAP_CLSM'
res['tmp'] = 10 * np.log10(res[tagP]**2 / res[tagR]**2)
r_asc_smap = res_tc[f'r_grid_{mode}_p_ASCAT_SMAP']
r_asc_clsm = res_tc[f'r_grid_{mode}_p_ASCAT_CLSM']
r_smap_clsm = res_tc[f'r_grid_{mode}_p_SMAP_CLSM']
thres = 0.2
ind_valid = res_tc[(r_asc_smap > thres) & (r_asc_smap > thres) &
(r_asc_smap > thres)].index
img = plot_ease_img(res.reindex(ind_valid),
'tmp',
fontsize=fontsize,
cbrange=cb,
cmap=cmap,
plot_cb=False)
plt.title(title, fontsize=fontsize)
if n == 1:
plt.ylabel('TCA unc. ratio', fontsize=fontsize)
for n, (mode, title, io_da) in enumerate(zip(modes, titles, ios)):
if mode != '4K':
r_asc_smap = res_tc[f'r_grid_{mode}_p_ASCAT_SMAP']
r_asc_clsm = res_tc[f'r_grid_{mode}_p_ASCAT_CLSM']
r_smap_clsm = res_tc[f'r_grid_{mode}_p_SMAP_CLSM']
thres = 0.2
ind_valid = res_tc[(r_asc_smap > thres) & (r_asc_smap > thres) &
(r_asc_smap > thres)].index
else:
ind_valid = res.index
avg = np.full(io_da['obs_obsvar'].shape[1::], np.nan)
for spc in range(4):
tmp1 = io_da['obs_obsvar'][:, spc, :, :].values
tmp2 = io_ol['obs_fcstvar'][:, spc, :, :].values
avg[spc, :, :] = np.nanmean(tmp2 / tmp1, axis=0)
avg = np.nanmean(avg, axis=0)
res['avg'] = 10 * np.log10(avg[res_rows, res_cols])
plt.subplot(2, 4, n + 5)
img = plot_ease_img(res.reindex(ind_valid),
'avg',
fontsize=fontsize,
cbrange=cb,
cmap=cmap,
plot_cb=False)
if n == 0:
plt.title(title, fontsize=fontsize)
plt.ylabel('Ens. var. ratio', fontsize=fontsize)
plot_centered_cbar(f, img, 4, fontsize=fontsize, bottom=0.07)
fout = dir_out / 'uncertainty_ratio.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()
def plot_predicted_skillgain(dir_out):
res = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
R_anom_lst = res[f'ubrmse_grid_anom_lst_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_anom_lst = res[f'ubrmse_grid_anom_lst_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_anom_lt = res[f'ubrmse_grid_anom_lt_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_anom_lt = res[f'ubrmse_grid_anom_lt_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_anom_st = res[f'ubrmse_grid_anom_st_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
P_anom_st = res[f'ubrmse_grid_anom_st_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R_lt_st_arr = (R_anom_lst - (R_anom_lt + R_anom_st)) / 2
P_lt_st_arr = (P_anom_lst - (P_anom_lt + P_anom_st)) / 2
P_lt_st_arr_rho = P_lt_st_arr / np.sqrt(P_anom_lt * P_anom_st)
# Baseline estimates
# R2 = res[f'r2_grid_abs_m_CLSM_tc_ASCAT_SMAP_CLSM']
R2 = res[f'r2_grid_anom_lst_m_CLSM_tc_ASCAT_SMAP_CLSM']
SNR = R2 / (1 - R2)
SIG = SNR * P_anom_lst
modes = ['anom_lt', 'anom_st', 'anom_lst', 'anom_lst']
titles = [
'LF signal', 'HF signal', 'Anomalies (lumped)', 'Anomalies (joint)'
]
result = pd.DataFrame(index=res.index, columns=modes)
result['row'] = res.row
result['col'] = res.col
for i, mode in enumerate(modes):
P = res[f'ubrmse_grid_{mode}_m_CLSM_tc_ASCAT_SMAP_CLSM']**2
R = res[f'ubrmse_grid_{mode}_m_SMAP_tc_ASCAT_SMAP_CLSM']**2
K = P / (R + P)
if i < 3:
K_4K = P / (4**2 + P)
P_upd = K * R + (1 - K) * P
NSR_upd = P_upd / SIG
R2upd = 1 / (1 + NSR_upd)
result[f'{i}_4K'] = np.sqrt(R2upd) - np.sqrt(R2)
if i < 2:
# Single signal assimilation
# R2 = res[f'r2_grid_{mode}_m_CLSM_tc_ASCAT_SMAP_CLSM']
# NSR = (1 - R2) / R2
# R2upd = 1 / (1 + (1 - K) * NSR)
P_upd = (1 - K) * P
NSR_upd = P_upd / SIG
R2upd = 1 / (1 + NSR_upd)
result[i] = np.sqrt(R2upd) - np.sqrt(R2)
result[f'P_upd_{i}'] = P_upd
elif i == 2:
# Joint signal assimilation
result[i] = np.nan
for cnt, idx in enumerate(res.index):
print(f'{cnt} / {len(res)}')
R11 = R_anom_lt.loc[idx]
R22 = R_anom_st.loc[idx]
R12 = R_lt_st_arr.loc[idx]
P11 = P_anom_lt.loc[idx]
P22 = P_anom_st.loc[idx]
P12 = P_lt_st_arr.loc[idx]
S = np.matrix([[R11, R12, 0, 0], [R12, R22, 0, 0],
[0, 0, P11, P12], [0, 0, P12, P22]])
A = np.matrix(
[K.loc[idx], K.loc[idx], 1 - K.loc[idx], 1 - K.loc[idx]])
P_upd = (A * S * A.T)[0, 0]
NSR_upd = P_upd / SIG.loc[idx]
R2upd = 1 / (1 + NSR_upd)
result.loc[idx, i] = np.sqrt(R2upd) - np.sqrt(R2.loc[idx])
else:
P_upd = result[f'P_upd_0'] + result[
f'P_upd_1'] + 2 * P_lt_st_arr_rho * np.sqrt(
result[f'P_upd_0'] * result[f'P_upd_1'])
NSR_upd = P_upd / SIG
R2upd = 1 / (1 + NSR_upd)
result[i] = np.sqrt(R2upd) - np.sqrt(R2)
f = plt.figure(figsize=(23, 7))
for i, title in enumerate(titles):
plt.subplot(2, 4, i + 1)
im = plot_ease_img(result,
i,
fontsize=12,
cbrange=[-0.2, 0.2],
cmap=cc.cm.bjy,
log_scale=False,
title=title,
plot_cb=False)
if i < 3:
plt.subplot(2, 4, i + 5)
im = plot_ease_img(result,
f'{i}_4K',
fontsize=12,
cbrange=[-0.2, 0.2],
cmap=cc.cm.bjy,
log_scale=False,
title=title + ' (4K)',
plot_cb=False)
plot_centered_cbar(f, im, 3, fontsize=10)
plt.tight_layout()
plt.show()
def plot_ascat_eval_relative(res_path, dir_out):
refs = ['Pcorr_OL', 'Pcorr_4K']
runs = [['Pcorr_anom_lst', 'Pcorr_anom_lt_ScDY', 'Pcorr_anom_st_ScYH'],
'Pcorr_LTST']
titles = [
'R$_{TC}$ - R$_{OL}$ (Individual)', 'R$_{TC}$ - R$_{OL}$ (Joint)',
'R$_{TC}$ - R$_{4K}$ (Individual)', 'R$_{TC}$ - R$_{4K}$ (Joint)'
]
res = pd.read_csv(res_path / 'ascat_eval.csv', index_col=0)
res_tc = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
modes = ['anom_lst', 'anom_lt', 'anom_st']
labels = [f'Anomaly skill', f'LF skill', f'HF skill']
cb_r = [-0.2, 0.2]
fontsize = 14
f = plt.figure(figsize=(22, 8))
for i, (m, label) in enumerate(zip(modes, labels)):
for j, ref in enumerate(refs):
ref_col = f'ana_r_corr_{ref}_{m}'
res[ref_col][res[ref_col] < 0] = 0
res[ref_col][res[ref_col] > 1] = 1
for k, run in enumerate(runs):
r = run if k == 1 else run[i]
ax = plt.subplot(3, 4, i * 4 + j * 2 + k + 1)
col = f'ana_r_corr_{r}_{m}'
res[col][res[col] < 0] = 0
res[col][res[col] > 1] = 1
res['diff'] = res[col] - res[ref_col]
if i == 0:
title = titles[j * 2 + k]
else:
title = ''
if (j == 0) & (k == 0):
ylabel = label
else:
ylabel = ''
r_asc_smap = res_tc[f'r_grid_{m}_p_ASCAT_SMAP']
r_asc_clsm = res_tc[f'r_grid_{m}_p_ASCAT_CLSM']
r_smap_clsm = res_tc[f'r_grid_{m}_p_SMAP_CLSM']
thres = 0.2
ind_valid = res_tc[(r_asc_smap > thres) & (r_asc_smap > thres)
& (r_asc_smap > thres)].index
im = plot_ease_img(res.reindex(ind_valid),
'diff',
title=title,
cmap=cc.cm.bjy,
cbrange=cb_r,
fontsize=fontsize,
print_mean=True,
plot_cb=False)
ax.set_ylabel(ylabel)
plot_centered_cbar(f, im, 4, fontsize=fontsize - 2, pad=0.02)
f.savefig(dir_out / f'ascat_eval_rel.png', dpi=300, bbox_inches='tight')
plt.close()
def plot_ascat_eval_absolute(res_path, dir_out):
runs = [
'Pcorr_OL', 'Pcorr_4K',
['Pcorr_anom_lst', 'Pcorr_anom_lt_ScDY', 'Pcorr_anom_st_ScYH'],
'Pcorr_LTST'
]
titles = [
'Open-loop', '4K Benchmark', 'Individual assimilation',
'Joint assimilation'
]
res = pd.read_csv(res_path / 'ascat_eval.csv', index_col=0)
res_tc = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
modes = ['anom_lst', 'anom_lt', 'anom_st']
cb_r = [0.6, 1]
fontsize = 14
f = plt.figure(figsize=(22, 8))
for i, m in enumerate(modes):
for cnt, (run, tit) in enumerate(zip(runs, titles)):
r = run if cnt != 2 else run[i]
ax = plt.subplot(3, 4, i * 4 + cnt + 1)
col = f'ana_r_corr_{r}_{m}'
res[col][res[col] < 0] = 0
res[col][res[col] > 1] = 1
if i == 0:
title = tit
else:
title = ''
if cnt == 0:
if i == 0:
ylabel = 'Anomaly skill'
elif i == 1:
ylabel = 'LF skill'
else:
ylabel = 'HF skill'
else:
ylabel = ''
r_asc_smap = res_tc[f'r_grid_{m}_p_ASCAT_SMAP']
r_asc_clsm = res_tc[f'r_grid_{m}_p_ASCAT_CLSM']
r_smap_clsm = res_tc[f'r_grid_{m}_p_SMAP_CLSM']
thres = 0.2
ind_valid = res_tc[(r_asc_smap > thres) & (r_asc_smap > thres) &
(r_asc_smap > thres)].index
im = plot_ease_img(res.reindex(ind_valid),
col,
title=title,
cmap='viridis',
cbrange=cb_r,
fontsize=fontsize,
print_mean=True,
plot_cb=False)
ax.set_ylabel(ylabel)
plot_centered_cbar(f, im, 4, fontsize=fontsize - 2, pad=0.02)
f.savefig(dir_out / f'ascat_eval_abs.png', dpi=300, bbox_inches='tight')
plt.close()
def plot_filter_diagnostics(res_path, dir_out):
fname = res_path / 'filter_diagnostics.nc'
res = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
res_tc = pd.read_csv(
'/Users/u0116961/Documents/work/MadKF/CLSM/SM_err_ratio/GEOSldas/sm_validation/Pcorr/result.csv',
index_col=0)
tg = GEOSldas_io().grid.tilegrids
res_cols = res.col.values - tg.loc['domain', 'i_offg']
res_rows = res.row.values - tg.loc['domain', 'j_offg']
r_asc_smap = res_tc[f'r_grid_anom_lst_p_ASCAT_SMAP']
r_asc_clsm = res_tc[f'r_grid_anom_lst_p_ASCAT_CLSM']
r_smap_clsm = res_tc[f'r_grid_anom_lst_p_SMAP_CLSM']
thres = 0.2
ind_valid = res_tc[(r_asc_smap > thres) & (r_asc_smap > thres) &
(r_asc_smap > thres)].index
fontsize = 14
root = Path('/Users/u0116961/data_sets/GEOSldas_runs')
runs = [run.name for run in root.glob('*_DA_SMAP_*')]
runs += ['NLv4_M36_US_OL_Pcorr', 'NLv4_M36_US_OL_noPcorr']
tags = ['OL_Pcorr', 'Pcorr_4K', f'Pcorr_anom_lst', 'Pcorr_LTST']
iters = [np.where([tag in run for run in runs])[0][0] for tag in tags]
titles = ['Open-loop', '4K constant', 'Anomalies']
labels = ['H pol. / Asc.', 'V pol. / Asc.']
with Dataset(fname) as ds:
lons = ds.variables['lon'][:]
lats = ds.variables['lat'][:]
lons, lats = np.meshgrid(lons, lats)
var = 'innov_autocorr'
cbrange = [0, 0.7]
step = 0.2
cmap = 'viridis'
f = plt.figure(figsize=(19, 6))
for j, (spc, label) in enumerate(zip([0, 2], labels)):
for i, (it_tit, it) in enumerate(zip(titles, iters)):
title = it_tit if j == 0 else ''
plt.subplot(2, 3, j * 3 + i + 1)
data = ds.variables[var][:, :, it, spc]
res['tmp'] = data[res_rows, res_cols]
im = plot_ease_img(res.reindex(ind_valid),
'tmp',
fontsize=fontsize,
cbrange=cbrange,
cmap=cmap,
title=title,
plot_cb=False,
print_meanstd=True)
if i == 0:
plt.ylabel(label, fontsize=fontsize)
plot_centered_cbar(f, im, 3, fontsize=fontsize - 2, bottom=0.07)
fout = dir_out / f'{var}.png'
f.savefig(fout, dpi=300, bbox_inches='tight')
plt.close()