def apply_mask(resolution, surface, mask_filename=None, path=None):
if path is None:
path = './fortran/data/src'
if resolution == 0.25:
mask = read_surface('mask_glbl_qrtd.dat', path)
surface = surface + mask
return surface
elif resolution == 0.5:
mask = read_surface('mask_glbl_hlfd.dat', path)
surface = surface + mask
return surface
else:
print("Mask with correct resolution not found.")
return surface
def easy_plot(dat_file, dat_path, product, bds, figs_dir, figname, log=True):
surface = read_surface('MPI-ESM1-2-HR_cs.dat', cs)
fig = plot(surface, bds=bds, product=product)
fig.savefig(figs_dir+figname+'_cs', dpi=300)
plt.close()
surface = read_surface('MPI-ESM1-2-HR_cs.dat', cs)
fig = plot(surface, bds=bds, product=product, extent='gs')
fig.savefig(figs_dir+figname+'_gs', dpi=300)
plt.close()
surface = read_surface('MPI-ESM1-2-HR_cs.dat', cs)
fig = plot(surface, bds=bds, product=product, extent='ag')
fig.savefig(figs_dir+figname+'_ag', dpi=300)
plt.close()
if log:
surface = read_surface('MPI-ESM1-2-HR_cs.dat', cs)
fig = plot(surface, bds=bds, product=product, log=True)
fig.savefig(figs_dir+figname+'_cs_log', dpi=300)
plt.close()
surface = read_surface('MPI-ESM1-2-HR_cs.dat', cs)
fig = plot(surface, bds=bds, product=product, extent='gs', log=True)
fig.savefig(figs_dir+figname+'_gs_log', dpi=300)
plt.close()
surface = read_surface('MPI-ESM1-2-HR_cs.dat', cs)
fig = plot(surface, bds=bds, product=product, extent='ag', log=True)
fig.savefig(figs_dir+figname+'_ag_log', dpi=300)
plt.close()
return surface
def main():
mdts = '../a_mdt_data/computations/mdts/'
mss = '../a_mdt_data/computations/mss/'
geoids = '../a_mdt_data/computations/geoids/'
cs = '../a_mdt_data/computations/currents/'
masks = '../a_mdt_data/computations/masks/'
figs_dir = './figs/'
mask = read_surface('mask_rr0004.dat', '../a_mdt_data/computations/masks/')
prior_dir = '../a_mdt_data/computations/currents/prior_geodetic_gauss'
regions = '../a_mdt_data/cdae_testing_data_128/cs/geodetic/'
geoid = read_surface(f'GO_CONS_GCF_2_TIM_R6_do0280_rr0004.dat', geoids)
print(np.mean(geoid), np.max(geoid), np.min(geoid), np.std(geoid), np.shape(geoid))
geoid = (geoid - geoid.mean())/geoid.std()
print(np.mean(geoid), np.max(geoid), np.min(geoid), np.std(geoid), np.shape(geoid))
nemo = read_surface(f'orca0083_12th_cs_band20.dat', cs)
# plot(nemo, product='cs', coastlines=True, bds=1.4)#, extent='ku')
cmip = read_surface(f'MPI-ESM1-2-HR_cs.dat', cs)
# plot(cls, product='cs', coastlines=True, bds=2)#, extent='gs')
dtu18_eigen = read_surface(f'dtu18_eigen-6c4_do0280_rr0004_cs_band20.dat', cs)
# plot(dtu18_eigen, product='cs', coastlines=True, bds=2)#, extent='gs')
dtu18_gtim6 = read_surface(f'dtu18_GTIM_R6_do0280_rr0004_cs.dat', cs)
# plot(dtu18_gtim6, product='cs', coastlines=True, bds=2)#, extent='gs')
dtu18_gtim5 = read_surface(f'dtu18_GTIM_R5_do0280_rr0004_cs.dat', cs)
# plot(dtu18_gtim5, product='cs', coastlines=True, bds=2)#, extent='gs')
dtu18_gtim5[dtu18_gtim5<1.4] = 0
# plot(dtu18_gtim5, product='cs', coastlines=True, bds=2)
test = dtu18_gtim6 - dtu18_eigen
test[test<1.4] = 0
filenames = glob.glob((os.path.join(directory, '*.dat')))
# filenames2 = glob.glob((os.path.join(directory2, 'cls18_cs_band10.dat')))
# filenames = glob.glob((os.path.join(directory, '*.dat')))
# filenames = [os.path.split(fname)[-1] for fname in filenames]
# filenames = [fname[:len(fname)-4] for fname in filenames]
# cls18 = read_surface(filenames2[0])
# cls18[cls18==0] = np.nan
# cls18=apply_gaussian(cls18, sigma=6) # <----- take out
# cls18[np.isnan(cls18)] = 0
# cls18 = bound_arr(cls18, -20, 20)
# cls18 = resize(cls18, (720, 1440), order=2)
for fname in filenames:
arr = read_surface(fname)
arr[arr == 0] = np.nan
# arr=apply_gaussian(arr, sigma=6) # <----- take out
arr[np.isnan(arr)] = 0
arr = bound_arr(arr, -20, 20)
arr = resize(arr, (720, 1440), order=2)
fname = os.path.split(fname)[-1]
fname = fname[:len(fname) - 4]
print('running')
for x, y in zip(x_coords, y_coords):
# print(x,y)
region = arr[y:y + size, x:x + size]
if mdt:
nans = np.isnan(arr)
arr[nans] = 0
# if (np.count_nonzero(region==0)/(region.size) < 0.25):
def main():
mdts = '../a_mdt_data/computations/mdts/'
gauss_filtered = '../a_mdt_data/computations/mdts/gauss_filtered/'
cs_path = '../a_mdt_data/computations/currents/'
gauss_cs_path = '../a_mdt_data/computations/currents/gauss_filtered/'
masks = '../a_mdt_data/computations/masks/'
mask = read_surface('mask_rr0008.dat', masks)
mask_4 = read_surface('mask_rr0004.dat', masks)
fig_dir = 'figs/mdt_plots/dtu18_gtim5_filtered/'
dtu18_gtim5_4 = read_surface('sh_mdt_DTU18_GTIM5_L280.dat', mdts)
dtu18_gtim5 = read_surface('dtu18_gtim5_do0280_rr0008.dat', mdts)
# dtu18_gtim5 = resize(dtu18_gtim5, (1440, 2880), order=3)
dtu18_gtim5[np.isnan(dtu18_gtim5)] = 0
# dtu18_gtim5 = np.clip(dtu18_gtim5, -1.4, 1.4)
# dtu18_gtim5 = norm(dtu18_gtim5)
# filt8 = apply_filter(dtu18_gtim5_4, 'boxcar', 8)
# filt12 = apply_filter(dtu18_gtim5_4, 'boxcar', 12)
# filt16 = apply_filter(dtu18_gtim5_4, 'boxcar', 16)
# filt20 = apply_filter(dtu18_gtim5_4, 'boxcar', 20)
# multi_plot([filt8, filt12, filt16, filt20], extent='na', coastlines=True)
cls_mdt = read_surface('cls18_mdt.dat', mdts)
cls_mdt[np.isnan(cls_mdt)] = 0
cls_downscaled = rescale(cls_mdt, 0.5)
# cls_downscaled = np.clip(cls_downscaled, -1.4, 1.4)
# cls_downscaled = norm(cls_downscaled)
nemo_mdt = read_surface('orca0083_mdt.dat', mdts)
nemo_mdt[np.isnan(nemo_mdt)] = 0
nemo_downscaled = resize(nemo_mdt, (720, 1440))
# nemo_downscaled = np.clip(nemo_downscaled, -1.4, 1.4)
# nemo_downscaled = norm(nemo_downscaled)
# # Gaussian filter
filter_widths = []
gauss_mdts = []
gauss_mdt_plots = []
for i in range(32):
k = (i + 1) / 4
mdt = gaussian_filter(dtu18_gtim5, sigma=k)
mdt[np.isnan(mdt)] = 0
gauss_mdts.append(mdt)
r = int((int((4 * k) + 0.5)) / 4 * 111)
filter_widths.append(r)
if r in (138, 249, 360, 471):
# resized_mdt = resize(mdt, (1440, 2880), order=3)
gauss_mdt_plots.append(mdt)
# multi_plot(gauss_mdt_plots, extent='na', coastlines=True)
res1 = dtu18_gtim5 - gauss_mdt_plots[0]
res2 = dtu18_gtim5 - gauss_mdt_plots[1]
res3 = dtu18_gtim5 - gauss_mdt_plots[2]
res4 = dtu18_gtim5 - gauss_mdt_plots[3]
multi_plot([res1, res2, res3, res4],
product='resid',
extent='na',
coastlines=True)
plt.show()
# Plot Gaussian filter widths: for r = 5, 9, 13, 17
# gauss_plots = []
# for i in range(4):
# k = ((i+1)*4)
# image = resize(gauss_mdts[k], (1440, 2880), order=3) + mask
# gauss_plots.append(image)
# plot(image, product='mdt', extent='na')
# plt.show()
# plt.close()
# # Load gaussian filtered currents
# gauss_cs = []
# gauss_cs_plots = []
# for i in range(32):
# k = (i+1)/4
# r = int((int((4*k)+0.5))/4 * 111)
# cs = read_surface(f'dtu18_gtim5_gauss_{r}km_cs.dat', gauss_cs_path)
# cs[np.isnan(cs)] = 0
# gauss_cs.append(cs)
# if r in (138, 249, 360, 471):
# resized_cs = resize(cs, (1440, 2880), order=3) + mask
# gauss_cs_plots.append(resized_cs)
# multi_plot(gauss_cs_plots, product='cs', extent='na', coastlines=True)
# multi_plot(gauss_cs_plots, product='cs', extent='na', coastlines=True)
# pmf_0 = read_surface('dtu18_gtim5_land0_pmf_300i_16k_02g.dat', mdts)
# pmf_1 = read_surface('dtu18_gtim5_land0_pmf_500i_16k_02g.dat', mdts)
# pmf_2 = read_surface('dtu18_gtim5_land0_pmf_700i_16k_02g.dat', mdts)
# pmf_3 = read_surface('dtu18_gtim5_land0_pmf_900i_16k_02g.dat', mdts)
# pmf_0[np.isnan(pmf_0)] = 0
# pmf_1[np.isnan(pmf_1)] = 0
# pmf_2[np.isnan(pmf_2)] = 0
# pmf_3[np.isnan(pmf_3)] = 0
# pmf_0 = resize(pmf_0, (1440, 2880), order=3)
# pmf_1 = resize(pmf_1, (1440, 2880), order=3)
# pmf_2 = resize(pmf_2, (1440, 2880), order=3)
# pmf_3 = resize(pmf_2, (1440, 2880), order=3)
# multi_plot([pmf_0, pmf_1, pmf_2, pmf_3], product='mdt', extent='na', coastlines=True)
# plt.show()
# PMF Filter
pmf_mdts, iterations = pmf(dtu18_gtim5, iterations=900)
extent = 'na'
# write_surface('dtu18_gtim5_land0_pmf_30i_16k_05g.dat', pmf_mdts[29], mdts)
# write_surface('dtu18_gtim5_land0_pmf_50i_16k_05g.dat', pmf_mdts[49], mdts)# overwrite=True)
# write_surface('dtu18_gtim5_land0_pmf_70i_16k_05g.dat', pmf_mdts[69], mdts)# overwrite=True)
# write_surface('dtu18_gtim5_land0_pmf_90i_16k_05g.dat', pmf_mdts[89], mdts)# overwrite=True)
res1 = dtu18_gtim5 - pmf_mdts[299]
res2 = dtu18_gtim5 - pmf_mdts[499]
res3 = dtu18_gtim5 - pmf_mdts[699]
res4 = dtu18_gtim5 - pmf_mdts[899]
multi_plot([res1, res2, res3, res4],
product='resid',
extent='na',
coastlines=True)
plt.show()
# get pmf from interior na
pmf_regions = [
extract_region(pmf_mdt, (-60, -40), (10, 30)) for pmf_mdt in pmf_mdts
]
# rmse between pmf and gaussian[25] same extent
g_region = extract_region(gauss_mdts[25], (-60, -40), (10, 30))
rmses = []
for pmf_region in pmf_regions:
rmses.append(np.sqrt(np.nanmean(((g_region - pmf_region)**2))))
plt.plot(iterations, rmses)
plt.grid(b=True, which='major', color='#888888', linestyle='-')
plt.grid(b=True, which='minor', color='#999999', linestyle='-', alpha=0.2)
# plt.xlim([100, 600])
plt.minorticks_on()
# plt.ylim([-0.05, 0.5])
# plt.title('')
plt.xlabel('Iterations')
plt.ylabel('RMSE (m)')
# plt.yscale("log")
plt.show()
# Original MDT
plot(resize(dtu18_gtim5, (1440, 2880), order=3) + mask,
product='mdt',
extent=extent,
coastlines=True)
plt.show()
# # Gaussian mdt filtered 200km radius
# image = resize(gauss_mdts[6], (1440, 2880), order=3) + mask
# plot(image, product='mdt', extent=extent)
# plt.show()
# Gaussian: Residual from original mdt
gauss_residual = dtu18_gtim5 - gauss_mdts[12]
image = resize(gauss_residual, (1440, 2880), order=3) + mask
plot(image,
product='mdt',
extent=extent,
low_bd=-0.15,
up_bd=0.15,
coastlines=True)
plt.show()
# PMF filtered mdt 350 iterations
image = resize(pmf_mdts[349], (1440, 2880), order=3) + mask
plot(image, product='mdt', extent=extent, coastlines=True)
plt.show()
# PMF: Residual from original mdt
pmf_residual = dtu18_gtim5 - pmf_mdts[349]
image = resize(pmf_residual, (1440, 2880), order=3) + mask
plot(image,
product='mdt',
extent=extent,
low_bd=-0.15,
up_bd=0.15,
coastlines=True)
plt.show()
# PMF - gaussian MDT residual
pmf_gauss = pmf_mdts[349] - gauss_mdts[12]
image = resize(pmf_gauss, (1440, 2880), order=3) + mask
plot(image,
product='mdt',
extent=extent,
low_bd=-0.1,
up_bd=0.1,
coastlines=True)
plt.show()
lon_range, lat_range = EXTENTS['na']
gauss_mdts = [
extract_region(mdt, lon_range, lat_range) for mdt in gauss_mdts
]
cls_downscaled = extract_region(cls_downscaled, lon_range, lat_range)
nemo_downscaled = extract_region(nemo_downscaled, lon_range, lat_range)
# RMSE
rmse_cls = []
ss_cls = []
for mdt in gauss_mdts:
# print(mdt.shape, cls_downscaled.shape)
rmse = np.sqrt(np.nanmean(((mdt - cls_downscaled)**2)))
rmse_cls.append(rmse)
ss = ssim(mdt, cls_downscaled)
ss_cls.append(ss)
rmse_cls = np.array(rmse_cls)
rmse_cls = norm(rmse_cls) #+ 0.01
ss_cls = np.array(ss_cls)
rmse_nemo = []
ss_nemo = []
for mdt in gauss_mdts:
rmse = np.sqrt(np.nanmean(((mdt - nemo_downscaled)**2)))
rmse_nemo.append(rmse)
ss = ssim(mdt, nemo_downscaled)
ss_nemo.append(ss)
rmse_nemo = np.array(rmse_nemo)
rmse_nemo = norm(rmse_nemo) #+ 0.01
ss_nemo = np.array(ss_nemo)
rmse_summed = rmse_cls + rmse_nemo
rmse_summed = norm(rmse_summed) + 0.01
fig, axs = plt.subplots(1, 2)
axs[0].plot(filter_widths, rmse_cls)
axs[0].plot(filter_widths, rmse_nemo)
# axs[0].plot(filter_widths, rmse_summed)
axs[0].grid(b=True, which='major', color='#888888', linestyle='-')
# Show the minor grid lines with very faint and almost transparent grey lines
axs[0].grid(b=True,
which='minor',
color='#999999',
linestyle='-',
alpha=0.2)
# axs[0].set_xlim([0, 900])
axs[0].minorticks_on()
# axs[0].set_ylim([-0.05, 1])
# axs[0].title('')
# axs[0].set_xlabel('Gaussian kernel half-width radius (km)')
# axs[0].set_ylabel('RMSE')
# axs[0].yscale("log")
axs[1].plot(filter_widths, ss_cls)
axs[1].plot(filter_widths, ss_nemo)
# axs[1].plot(filter_widths, ss_summed)
axs[1].grid(b=True, which='major', color='#888888', linestyle='-')
# Show the minor grid lines with very faint and almost transparent grey lines
axs[1].grid(b=True,
which='minor',
color='#999999',
linestyle='-',
alpha=0.2)
# axs[1].set_xlim([0, 900])
axs[1].minorticks_on()
# axs[1].ylim([-0.05, 1])
# axs[1].title('')
axs[1].set_xlabel('Gaussian kernel half-width radius (km)')
# axs[1].set_ylabel('Structural Similarity Index Measurement')
# axs[1].yscale("log")
plt.show()
plt.close()
plt.plot(filter_widths, rmse_cls)
plt.plot(filter_widths, rmse_nemo)
# plt.plot(filter_widths, rmse_summed)
plt.grid(b=True, which='major', color='#888888', linestyle='-')
# Show the minor grid lines with very faint and almost transparent grey lines
plt.grid(b=True, which='minor', color='#999999', linestyle='-', alpha=0.2)
# plt.xlim([0, 900])
plt.minorticks_on()
# plt.ylim([-0.05, 0.6])
# plt.title('')
plt.xlabel('Gaussian filter half-width radius (km)')
plt.ylabel('RMSE')
# plt.yscale("log")
plt.show()
plt.plot(filter_widths, ss_cls)
plt.plot(filter_widths, ss_nemo)
# plt.plot(filter_widths, ss_summed)
plt.grid(b=True, which='major', color='#888888', linestyle='-')
# Show the minor grid lines with very faint and almost transparent grey lines
plt.grid(b=True, which='minor', color='#999999', linestyle='-', alpha=0.2)
# plt.xlim([100, 600])
plt.minorticks_on()
plt.ylim([-0.05, 0.5])
# plt.title('')
plt.xlabel('Gaussian kernel half-width radius (km)')
plt.ylabel('Structural Similarity Index Measurement')
# plt.yscale("log")
plt.show()
rmse_pmf = []
for mdt in pmf_mdts:
rmse = np.sqrt(np.nanmean(((mdt - cls_mdt)**2)))
rmse_pmf.append(rmse)
rmse_pmf = np.array(rmse_pmf)
rmse_pmf = norm(rmse_pmf)
plt.plot(iterations, rmse_pmf)
plt.show()
region = extract_region(arr, (tile_pt[0], tile_pt[0] + tile_size),
(tile_pt[1], tile_pt[1] + tile_size))
if valid_region(region):
region[np.isnan(region)] = 0
regions.append(region)
# plt.imshow(region)
# plt.show()
return regions, tile_pts
cmip6_file = 'cmip6_historical_mdts_yr5_meta.txt'
cmip6_path = '../a_mdt_data/datasets/cmip6/'
mdt_path = "../a_mdt_data/computations/mdts/"
cs_path = "../a_mdt_data/computations/currents/"
masks = '../a_mdt_data/computations/masks/'
mask = read_surface('mask_rr0004.dat', masks)
training_fnames = [
'dtu18_GO_CONS_GCF_2_DIR_R5_do0280_rr0004.dat', #300-
'dtu18_GO_CONS_GCF_2_TIM_R6_do0280_rr0004.dat', #300-
'dtu18_GO_CONS_GCF_2_SPW_R5_do0280_rr0004.dat', #330-
'dtu18_GO_CONS_GCF_2_SPW_R4_do0280_rr0004.dat', #280-
'dtu18_goco06s_do0280_rr0004.dat', #300-
# gdir6 moves to testing
'dtu18_GO_CONS_GCF_2_TIM_R5_do0280_rr0004.dat', #280-
'dtu18_GO_CONS_GCF_2_SPW_R2_do0240_rr0004.dat', #240-
'dtu18_GTIM5_R6e_do0280_rr0004.dat', #300 - not available on home pc
# 'dtu18_eigen-6c4_do0280_rr0004.dat', #2190-
# 'dtu18_egm2008_do0280_rr0004.dat', #2190-
# 'dtu18_geco_do0280_rr0004.dat', #2190-
# 'dtu18_IGGT_R1_do0240_rr0004.dat', #240