# As in Lorenz 1996 "Predictability..."
from common import *
from mods.LorenzUV.core import model_instance
LUV = model_instance(nU=36, J=10, F=10)
nU = LUV.nU
################
# Full
################
#
t = Chronology(dt=0.005, dtObs=0.05, T=4**3, BurnIn=6)
Dyn = {
'M': LUV.M,
'model': with_rk4(LUV.dxdt, autonom=True),
'noise': 0,
'jacob': LUV.dfdx,
}
X0 = GaussRV(C=0.01 * eye(LUV.M))
R = 1.0
jj = arange(LUV.nU)
Obs = partial_direct_Obs(LUV.M, jj)
Obs['noise'] = R
other = {'name': rel_path(__file__, 'mods/') + '_full'}
# Uses nU, J, F as in core.py, which is taken from Wilks2005.
# Obs settings taken from different places (=> quasi-linear regime).
from common import *
from mods.LorenzUV.core import model_instance
LUV = model_instance()
nU = LUV.nU
# Wilks2005 uses dt=1e-4 with RK4 for the full model,
# and dt=5e-3 with RK2 for the forecast/truncated model.
# As berry2014linear notes, this is possible coz
# "numerical stiffness disappears when fast processes are removed".
################
# Full
################
#t = Chronology(dt=0.001,dtObs=0.05,T=4**3,BurnIn=6) # allows using rk2
t = Chronology(dt=0.005, dtObs=0.05, T=4**3, BurnIn=6) # requires rk4
f = {
'm': LUV.m,
'model': with_rk4(LUV.dxdt, autonom=True),
'noise': 0,
'jacob': LUV.dfdx,
'plot': LUV.plot_state
}
X0 = GaussRV(C=0.01 * eye(LUV.m))