plt.ylim(0, 3e10)
plt.xticks([fs / (24 * 7), fs / (24), fs / (12), fs / (8)],
labels=['weekly', 'daily', '12-hours', '8-hours'])
plt.xlabel(r'Periodicity ($\frac{1}{f}$)')
plt.ylabel(r'PSD of PM$_{10}$ at Station 1')
plt.tight_layout()
plt.axes([.65, .6, .2, .2])
plt.plot(f, psd_pm10, 'k')
plt.xlim(0, 1 / (3600 * 24 * 7 * 25))
plt.xticks([fs / (24 * 365)], labels=['annually'])
plt.savefig("psd_pm10.pdf", format='pdf')
# =============================================================================
# Covariance matrix
# =============================================================================
Sigma_no2 = functions.covmat(acf_no2, time[0:24 * 7])
plt.figure(figsize=(5, 4))
plt.imshow(Sigma_no2, cmap='gray')
plt.colorbar()
plt.xticks(np.arange(0, 24 * 7, 24))
plt.yticks(np.arange(0, 24 * 7, 24))
plt.xlabel(r'Time (hours)')
plt.ylabel(r'Time (hours)')
Sigma_pm10 = functions.covmat(acf_pm10, time[0:24 * 7])
plt.figure(figsize=(5, 4))
plt.imshow(Sigma_pm10, cmap='gray')
plt.colorbar()
plt.xticks(np.arange(0, 24 * 7, 24))
mu_test_prior = mean_busday
else:
mu_test_prior = mean_holiday
meteo_test = meteo[:, index_test[0] - 24:index_test[0]]
feature_test = np.empty((6, 24))
feature_test[0:4, :] = meteo_test[[0, 1, 2, 4], :]
feature_test[4, :] = np.sin(np.deg2rad(meteo_test[3, :]))
feature_test[5, :] = np.cos(np.deg2rad(meteo_test[3, :]))
# mu_test_prior = NN_model.predict(np.concatenate(feature_test).reshape(1, 144))
# mu_test_prior = mu_test_prior.reshape(5, 24)
cov_test_prior = np.zeros((5, 24, 24))
for i in range(5):
cov_test_prior[i, :, :] = functions.covmat(acf[i, :], time_test)
# fig, axs = plt.subplots(1, 5, figsize=(20, 4), sharey=True)
# for i in range(5):
# axs[i].plot(np.exp(mu_test_prior[i, :]), 'k')
# axs[i].plot(no2[i, index_test], 'k:')
# axs[i].set_xlim([0, 23])
# axs[i].set_xlabel(r'Time (hour)')
# axs[i].set_ylim([0, .8*np.max(no2)])
# axs[0].set_ylabel(r'NO$_2$ ($\mu g/m^3$)')
#
# fig, axes = plt.subplots(nrows=1, ncols=5, figsize=(24,4), sharey=True)
# i = 0
# for ax in axes.flat:
# print(i)
# im = ax.imshow(cov_test_prior[i, :, :], cmap = 'binary', vmin = 0, vmax = np.max(cov_test_prior))