obs_swe = obs_swe[~np.isnan(obs_swe)]
for i, run_id in enumerate(dict_in.keys()):
if hy == '2017-18':
sim_swe = dict_in[run_id]['states_output'].swe.copy().truncate(
after='2017-12-31 23:00:00')
if hy == '2019-20':
sim_swe = dict_in[run_id]['states_output'].swe.copy().truncate(
before='2019-05-01 00:00:00', after='2019-11-12 00:00:00')
# assert np.all(sim_swe.index == obs_swe.index) #TODO fix index so both timezone aware
# cut to just period with observed swe
sim_swe = sim_swe[obs_ind]
if model == 'fsm2':
sim_swe = sim_swe.values
stats_store['ns'][i] = nash_sut(sim_swe, obs_swe)
stats_store['bias'][i] = mean_bias(sim_swe, obs_swe)
stats_store['rmsd'][i] = rmsd(sim_swe, obs_swe)
stats_store['mae'][i] = mean_absolute_error(sim_swe, obs_swe)
stats_store['run_id'][i] = run_id
if model == 'fsm2':
sim_nml = dict_in[run_id]['namelist']
for key in sim_nml['params'].keys():
param_store[key][i] = sim_nml['params'][key]
# param_store['exe'][i] = dict_in[run_id]['exe'][1] #
param_store['density'][i] = dict_in[run_id]['exe'][0]
param_store['exchng'][i] = dict_in[run_id]['exe'][1]
param_store['hydrol'][i] = dict_in[run_id]['exe'][2]
elif model == 'clark' or model == 'eti':
for key in dict_in[run_id]['config'].keys():
# plt.figure()
# plt.imshow(np.mean(modis_fsca_rs, axis=0),origin=0)
# plt.figure()
# plt.imshow(np.mean(model_fsca_rs, axis=0),origin=0)
ns_array = np.zeros((ny_out, nx_out))
mbd_array = np.zeros((ny_out, nx_out))
rmsd_array = np.zeros((ny_out, nx_out))
mae_array = np.zeros((ny_out, nx_out))
for i in range(ny_out):
for j in range(nx_out):
obs = np.convolve(modis_fsca_rs[:, i, j], np.ones((smooth_period,)) / smooth_period, mode='same')
mod = np.convolve(model_fsca_rs[:, i, j], np.ones((smooth_period,)) / smooth_period, mode='same')
ns_array[i, j] = nash_sut(mod, obs)
mbd_array[i, j] = mean_bias(mod, obs)
rmsd_array[i, j] = rmsd(mod, obs)
mae_array[i, j] = mean_absolute_error(mod, obs)
modis_mean = np.mean(modis_fsca_rs, axis=0)
model_mean = np.mean(model_fsca_rs, axis=0)
s_ns.append(ns_array)
s_bias.append(mbd_array)
s_rmse.append(rmsd_array)
s_mae.append(mae_array)
s_obs.append(modis_mean)
s_mod.append(model_mean)
for i,j,name in zip([161,147,127,107,186,125],[83,102,59,88,21,34],['Pisa','Dunstan','Hector','Old Man','Earnslaw','Lochy']):
inp_doy,
inp_hourdec,
dtstep=1800,
init_swe=init_swe,
init_d_snow=init_d_snow,
inp_sw=inp_sw,
which_melt='dsc_snow',
**config)
daily_swe3 = []
for k in range(47, len(st_swe3[:, 0]), 48):
daily_swe3.append(st_swe3[k, 0])
mb_sim = -1 * np.diff(np.asarray(daily_swe3))
ns_array[i, j] = nash_sut(mb_sim, daily_sfc_melt)
mbd_array[i, j] = mean_bias(mb_sim, daily_sfc_melt)
rmsd_array[i, j] = rmsd(mb_sim, daily_sfc_melt)
h_ns_array[i, j] = nash_sut(-1 * np.diff(st_swe3[:, 0]),
hourly_seb_melt)
h_mbd_array[i, j] = mean_bias(-1 * np.diff(st_swe3[:, 0]),
hourly_seb_melt)
h_rmsd_array[i, j] = rmsd(-1 * np.diff(st_swe3[:, 0]), hourly_seb_melt)
np.savetxt(
outfolder + '/daily_sfc_ns_2010_TF_TT_ROS{}_ta_m_tt{}.txt'.format(
config['ros'], config['ta_m_tt']), ns_array)
np.savetxt(
outfolder + '/daily_sfc_mbd_2010_TF_TT_ROS{}_ta_m_tt{}.txt'.format(
config['ros'], config['ta_m_tt']), mbd_array)