"""Reproduce results from `bib.sakov2008deterministic`."""
import numpy as np
import dapper.mods as modelling
from dapper.mods.QG import LP_setup, model_config, sample_filename, shape
from dapper.tools.localization import nd_Id_localization
############################
# Time series, model, initial condition
############################
model = model_config("sakov2008", {})
Dyn = {
'M': np.prod(shape),
'model': model.step,
'noise': 0,
}
# Considering that I have 8GB mem on the Mac, and the estimate:
# ≈ (8 bytes/float)*(129² float/stat)*(7 stat/k) * K,
# it should be possible to run experiments of length (K) < 8000.
t = modelling.Chronology(dt=model.prms['dtout'], dko=1, T=1500, BurnIn=250)
# In my opinion the burn in should be 400.
# Sakov also used 10 repetitions.
X0 = modelling.RV(M=Dyn['M'], file=sample_filename)
############################
# Observation settings
"""Reproduce experiments from
'Application of a hybrid EnKF-OI to ocean forecasting'
by F. Counillon, P. Sakov, and L. Bertino (2009)."""
import dapper as dpr
from dapper.mods.QG import model_config
from dapper.mods.QG.sakov2008 import HMM
dt = 1.25 * 10 # 10 steps between obs (also requires dkObs=1)
HMM.t = dpr.Chronology(dt=dt, dkObs=1, T=1000 * dt, BurnIn=10 * dt)
HMM.Dyn.model = model_config("counillon2009_ens", {
"dtout": dt,
'RKH2': 2.0e-11
}).step
truth_model = model_config("counillon2009_truth", {"dtout": dt}).step
####################
# Suggested tuning
####################
# Reproduce Table 1 results.
# - Note that Counillon et al:
# - Report forecast rmse's (but they are pretty close to analysis rmse anyways).
# - Use enkf-matlab which has a bug which cause them to report the
# wrong localization radius (see mods/QG/sakov2008.py).
# Eg. enkf-matlab radius 15 (resp 25) corresponds to
# DAPPER radius 10.6 (resp 17.7).
# R = 17.7 # equiv. to R=25 in enkf-matlab
# from dapper.mods.QG.counillon2009 import HMM, truth_model # rmse.f:
# xps += LETKF(mp=True, N=25,infl=1.15,taper='Gauss',loc_rad=R) # 1.11
x0 = KS.x0
dt = KS.dt
N = KS.Nx
Nx = len(x0)
T = 1e3
eps = 0.0002
# n0 ≈ 140
if mod == "QG":
from dapper.mods.QG import model_config, sample_filename, shape
# NB: There may arise an ipython/multiprocessing bug/issue.
# Ref https://stackoverflow.com/a/45720872 . If so, set mp=False,
# or run outside of ipython. However, I did not encounter it lately.
model = model_config("sakov2008", {}, mp=True)
step = model.step
Nx = np.prod(shape)
ii = np.random.choice(np.arange(Nx), 100, False)
T = 1000.0
dt = model.prms['dtout']
x0 = np.load(sample_filename)['sample'][-1]
eps = 0.01 # ensemble rescaling
N = 300
########################
# Reference trajectory
########################
# NB: Arbitrary, coz models are autonom. But dont use nan coz QG doesn't like it.
t0 = 0.0
# Main
###########
# Load or generate time-series data of a simulated state and obs:
fname = dpr.rc.dirs.data / "QG-ts-en.npz"
np.random.seed(123)
# ensemble size needs to be at least Ne=2 for plotting to be true
plotting = True
Ne = 2
try:
with np.load(fname) as data:
E1 = np.squeeze(data['ens'][:, 0, :])
E2 = np.squeeze(data['ens'][:, 1, :])
except FileNotFoundError:
sample = gen_ensemble_sample(model_config("sample_generation", {}), 400,
Ne, 10, 10)
E1 = np.squeeze(sample[:, 0, :])
E2 = np.squeeze(sample[:, 1, :])
np.savez(fname, ens=sample)
if plotting == True:
# Create figure
fig, (ax1, ax2) = plt.subplots(ncols=2,
sharex=True,
sharey=True,
figsize=(12, 6))
for ax in (ax1, ax2):
ax.set_aspect('equal', 'box')
ax1.set_title(r'Ensemble member 1')
ax2.set_title(r'Ensemble member 2')