def plotcost_ens_inc(vect=1):
"""Using test_cost plots convergance of cost fn gradient for decreasing
value of alpha.
"""
sns.set_context('poster', font_scale=1.5, rc={'lines.linewidth':1, 'lines.markersize':6})
fig, ax = plt.subplots(nrows=1, ncols=1,)#figsize=(10,10))
sns.set_style('whitegrid')
power=np.arange(1,10,1)
xlist = [10**(-x) for x in power]
jj = fdj.FourDEnVar(assim=True)
#jj.output_dir="output/ens_run500"
tstlist = [abs(test_cost_ens_inc(jj, x, vect)-1) for x in xlist]
ax.loglog(xlist, tstlist, 'k', marker='x', mew=1, ms=8)
#font = {'size' : 24}
#matplotlib.rc('font', **font)
plt.xlabel(r'$\eta$')
plt.ylabel(r'$|f(\eta) - 1|$')
#plt.title('test of the gradient of the cost function')
print tstlist
#plt.show()
return ax, fig
def save_plots(xa_ens_pickle, out_dir):
"""
Function saving plots from data assimilatoin experiment output
:param xa_ens_pickle: location of pickled posterior ensemble arr (str)
:param out_dir: directory to save plots in (str)
:return: string confirming plots have been saved (str)
"""
dat_xt = nc.Dataset('output/background/xb0.daily.nc', 'r')
jda = fourdenvar.FourDEnVar()
obs = es.obs_fn()
date = nc.num2date(dat_xt.variables['time'][:],
dat_xt.variables['time'].units)
xb_dir = es.output_directory + '/ensemble' + str(es.seed_value)
xa_dir = es.output_directory + '/ensemble_xa_' + str(es.seed_value)
#plot SM
fig, ax = plot_twin_spread(date[:],
'smcl',
xb_dir,
xa_dir,
ob_times=obs['sm_times'],
obs=obs['sm_obs'],
err=obs['sm_err'],
ylab=r'Soil Moisture (m$^{3}$ m$^{-3}$)')
fig.savefig(out_dir + '/sm.png', bbox_inches='tight')
xb_dir2 = es.output_directory + '/background_tst'
xa_dir2 = es.output_directory + '/analysis_med'
fig, ax = calc_cosmos_stats(date[:], xb_dir2, xa_dir2)
fig.savefig(out_dir + '/cosmos_median.png', bbox_inches='tight')
#plot distribution
#true_params = {'alpha_io': 5.5e-02, 'neff_io': 5.7e-04, 'fd_io': 9.6e-03, 'mu_io': 2.0e-02, 'nu_io': 4.0e+00,
# 'gamma_io': 1.76e+01, 'delta_io':-3.3e-01}
xa_ens = pickle.load(open(xa_ens_pickle, 'rb'))
p_keys = [
'oneovernminusone', 'oneoveralpha', 'satcon', 'vsat', 'vcrit', 'vwilt',
'hcap', 'hcon'
]
fig, ax = plot_mult_dist(xa_ens, jda.xbs, p_keys)
fig.savefig(out_dir + '/distributions.png', bbox_inches='tight')
return 'plots saved!'
:return: string confirming if the ensemble has been run (str)
"""
print 'Running ensemble'
mp.freeze_support()
with poolcontext(processes=es.num_processes) as pool:
res = pool.map(
partial(ens_member_run, seed_val=seed_val, params=params, xa=xa),
enumerate(x_ens))
pool.close()
pool.join()
return 'Ensemble has been run'
if __name__ == "__main__":
# instantiate JULES data assimilation class
jda = fourdenvar.FourDEnVar(seed_val=int(sys.argv[1]))
seed_val = int(sys.argv[1])
params = jda.p_keys
# if 'run_xb' is in system arguments then run JULES with prior parameters
if 'run_xb' in sys.argv:
nml_dir = 'output_seed' + str(seed_val) + '_xb/'
if not os.path.exists(nml_dir):
os.makedirs(nml_dir)
for file in glob.glob(es.nml_directory + '/*.nml'):
sh.copy(file, nml_dir)
rj = rjda.RunJulesDa(params=params, values=jda.xb, nml_dir=nml_dir)
rj.run_jules_dic(output_name='xb' + str(jda.seed_val),
out_dir=es.output_directory + '/background/')
sh.rmtree(nml_dir)
# remove any old output in folders
old_outs = glob.glob(es.output_directory + '/ensemble' + str(seed_val) +
def save_plots(xa_ens_pickle, out_dir):
"""
Function saving plots from data assimilatoin experiment output
:param xa_ens_pickle: location of pickled posterior ensemble arr (str)
:param out_dir: directory to save plots in (str)
:return: string confirming plots have been saved (str)
"""
dat_xt = nc.Dataset('output/model_truth/mod_truth.daily.nc', 'r')
jda = fourdenvar.FourDEnVar()
obs = es.obs_fn()
date = nc.num2date(dat_xt.variables['time'][:],
dat_xt.variables['time'].units)
xb_dir = es.output_directory + '/ensemble' + str(es.seed_value)
xa_dir = es.output_directory + '/ensemble_xa_' + str(es.seed_value)
#plot GPP
fig, ax = plot_twin_spread(
date[:],
'gpp',
xb_dir,
xa_dir,
xt_var=1000 * 60 * 60 * 24 * dat_xt.variables['gpp'][:, 7, 0, 0],
ob_pos=obs['gpp_pos'],
obs=obs['gpp_obs'],
err=obs['gpp_err'],
ylab=r'Gross Primary Productivity (g C m$^{-2}$ day$^{-1}$)')
fig.savefig(out_dir + '/gpp.png', bbox_inches='tight')
#plot LAI
fig, ax = plot_twin_spread(date[:],
'lai',
xb_dir,
xa_dir,
xt_var=dat_xt.variables['lai'][:, 7, 0, 0],
ob_pos=obs['lai_pos'],
obs=obs['lai_obs'],
err=obs['lai_err'],
ylab=r'Leaf Area Index (m$^{2}$ m$^{-2}$)')
fig.savefig(out_dir + '/lai.png', bbox_inches='tight')
#plot Canht
fig, ax = plot_twin_spread(date[:],
'canht',
xb_dir,
xa_dir,
xt_var=dat_xt.variables['canht'][:, 7, 0, 0],
ob_pos=obs['canht_pos'],
obs=obs['canht_obs'],
err=obs['canht_err'],
ylab=r'Canopy Height (m)')
fig.savefig(out_dir + '/canht.png', bbox_inches='tight')
#plot harvestable material
fig, ax = plot_twin_spread(date[:],
'cropharvc',
xb_dir,
xa_dir,
xt_var=dat_xt.variables['cropharvc'][:, 2, 0,
0],
ylab=r'Harvestable Material (kg C m$^{-2}$)',
lvl_idx=2)
fig.savefig(out_dir + '/harvc.png', bbox_inches='tight')
#plot distribution
true_params = {
'alpha_io': 5.5e-02,
'neff_io': 5.7e-04,
'fd_io': 9.6e-03,
'mu_io': 2.0e-02,
'nu_io': 4.0e+00,
'gamma_io': 1.76e+01,
'delta_io': -3.3e-01
}
xa_ens = pickle.load(open(xa_ens_pickle, 'rb'))
fig, ax = plot_mult_dist(xa_ens, jda.xbs, jda.p_keys, true_params)
fig.savefig(out_dir + '/distributions.png', bbox_inches='tight')
return 'plots saved!'