def L96_table():
import dapper.mods.Lorenz96 as model
from dapper.mods.Lorenz96.sakov2008 import HMM as _HMM
model.Force = 8.0 # undo pinheiro2019
HMM = _HMM.copy()
HMM.t.BurnIn = 0
HMM.t.KObs = 10
dpr.set_seed(3000)
# xps
xps = dpr.xpList()
xps += da.Climatology()
xps += da.OptInterp()
xps += da.Var3D(xB=0.02)
xps += da.ExtKF(infl=6)
xps += da.EnKF("PertObs", N=40, infl=1.06)
xps += da.EnKF("Sqrt", N=28, infl=1.02, rot=True)
xps += da.EnKF_N(N=24, rot=True)
xps += da.EnKF_N(N=24, rot=True, xN=2)
xps += da.iEnKS("Sqrt", N=40, infl=1.01, rot=True)
xps += da.LETKF(N=7, rot=True, infl=1.04, loc_rad=4)
xps += da.SL_EAKF(N=7, rot=True, infl=1.07, loc_rad=6)
xps.launch(HMM, store_u=True)
table = xps.tabulate_avrgs(statkeys, decimals=4)
return table.splitlines(True)
def L63_gen():
from dapper.mods.Lorenz63.sakov2012 import HMM as _HMM
HMM = _HMM.copy()
HMM.tseq.BurnIn = 0
HMM.tseq.Ko = 10
# xps
xps = dpr.xpList()
xps += da.Climatology()
xps += da.OptInterp()
xps += da.Persistence()
xps += da.PreProg(lambda k, xx, yy: xx[k])
xps += da.Var3D(xB=0.1)
xps += da.ExtKF(infl=90)
xps += da.EnKF("Sqrt", N=3, infl=1.30)
xps += da.EnKF("Sqrt", N=10, infl=1.02, rot=True)
xps += da.EnKF("PertObs", N=500, infl=0.95, rot=False)
xps += da.EnKF_N(N=10, rot=True)
xps += da.iEnKS("Sqrt", N=10, infl=1.02, rot=True)
xps += da.PartFilt(N=100, reg=2.4, NER=0.3)
xps += da.PartFilt(N=800, reg=0.9, NER=0.2)
xps += da.PartFilt(N=4000, reg=0.7, NER=0.05)
xps += da.PFxN(xN=1000, N=30, Qs=2, NER=0.2)
xps += da.OptPF(N=100, Qs=2, reg=0.7, NER=0.3)
xps += da.EnKS("Serial", N=30, Lag=1)
xps += da.EnRTS("Serial", N=30, DeCorr=0.99)
for xp in xps:
xp.seed = 3000
# Run
xps.launch(HMM, False, store_u=True)
return xps
def L63_gen():
from dapper.mods.Lorenz63.sakov2012 import HMM as _HMM
HMM = _HMM.copy()
HMM.t.BurnIn = 0
HMM.t.KObs = 10
dpr.set_seed(3000)
# xps
xps = dpr.xpList()
xps += da.Climatology()
xps += da.OptInterp()
xps += da.Var3D(xB=0.1)
xps += da.ExtKF(infl=90)
xps += da.EnKF("Sqrt", N=3, infl=1.30)
xps += da.EnKF("Sqrt", N=10, infl=1.02, rot=True)
xps += da.EnKF("PertObs", N=500, infl=0.95, rot=False)
xps += da.EnKF_N(N=10, rot=True)
xps += da.iEnKS("Sqrt", N=10, infl=1.02, rot=True)
xps += da.PartFilt(N=100, reg=2.4, NER=0.3)
xps += da.PartFilt(N=800, reg=0.9, NER=0.2)
xps += da.PartFilt(N=4000, reg=0.7, NER=0.05)
xps += da.PFxN(xN=1000, N=30, Qs=2, NER=0.2)
# Run
xps.launch(HMM, False, store_u=True)
return xps
def test_L63():
from dapper.mods.Lorenz63.sakov2012 import HMM as _HMM
xps = dpr.xpList()
xps += da.EnKF("Sqrt", N=10, infl=1.02, rot=True)
xps += da.PartFilt(N=20, reg=2.4, NER=0.3)
xps += da.OptInterp()
# xps += da.iEnKS('Sqrt', N=10, infl=1.02,rot=True)
HMM = _HMM.copy()
HMM.tseq.BurnIn = HMM.tseq.dto
HMM.tseq.Ko = 1
xps.launch(
HMM,
free=False,
statkeys=True,
liveplots=None,
store_u=False,
fail_gently=False,
save_as=False,
)
print(xps.tabulate_avrgs(["rmse.a"]))
# Disabled coz magic doesn't work any more on python 3.7
# spell_out(HMM)
# spell_out(xps[-1])
# spell_out(xps[-1].stats)
# spell_out(xps[-1].avrgs)
assert True # An assertion for pytest to count
return HMM, xps # Return useful stuff
def L96_gen():
import dapper.mods.Lorenz96 as model
from dapper.mods.Lorenz96.sakov2008 import HMM as _HMM
model.Force = 8.0 # undo pinheiro2019
HMM = _HMM.copy()
HMM.tseq.BurnIn = 0
HMM.tseq.Ko = 10
# xps
xps = dpr.xpList()
xps += da.Climatology()
xps += da.OptInterp()
xps += da.Persistence()
xps += da.PreProg(lambda k, xx, yy: xx[k])
xps += da.Var3D(xB=0.02)
xps += da.ExtKF(infl=6)
xps += da.EnKF("PertObs", N=40, infl=1.06)
xps += da.EnKF("Sqrt", N=28, infl=1.02, rot=True)
xps += da.EnKF_N(N=24, rot=True)
xps += da.EnKF_N(N=24, rot=True, xN=2)
xps += da.iEnKS("Sqrt", N=40, infl=1.01, rot=True)
xps += da.LETKF(N=7, rot=True, infl=1.04, loc_rad=4)
xps += da.SL_EAKF(N=7, rot=True, infl=1.07, loc_rad=6)
for xp in xps:
xp.seed = 3000
xps.launch(HMM, store_u=True)
return xps
def test_L63():
from dapper.mods.Lorenz63.sakov2012 import HMM as _HMM
xps = dpr.xpList()
xps += da.EnKF('Sqrt', N=10, infl=1.02, rot=True)
xps += da.PartFilt(N=20, reg=2.4, NER=0.3)
xps += da.OptInterp()
# xps += da.iEnKS('Sqrt', N=10, infl=1.02,rot=True)
HMM = _HMM.copy()
HMM.t.BurnIn = HMM.t.dtObs
HMM.t.KObs = 1
xps.launch(HMM,
free=False,
liveplots="all",
store_u=False,
fail_gently=False)
for xp in xps:
xp.stats.replay("all")
xp.stats.replay(t2=1)
xp.stats.replay(t2=0.0)
xp.stats.replay(t2=0.3)
xp.stats.replay(t2=0.8)
xp.stats.replay(t2=0.8, t1=0.2)
xp.stats.replay(t2=np.inf)
xp.stats.replay(t2=np.inf, speed=1)
xp.stats.replay(t2=np.inf, pause_a=0, pause_f=0)
assert True # An assertion for pytest to count
return HMM, xps # Return useful stuff
def test_L96():
xps = dpr.xpList()
from dapper.mods.Lorenz96.sakov2008 import HMM as _HMM
xps += da.EnKF('PertObs', N=40, infl=1.06)
xps += da.EnKF('Serial', N=28, infl=1.02, rot=True)
xps += da.OptInterp()
xps += da.Var3D(xB=0.02)
xps += da.ExtKF(infl=10)
xps += da.LETKF(N=6, rot=True, infl=1.05, loc_rad=4, taper='Step')
# from dapper.mods.Lorenz96.bocquet2015loc import HMM
# xps += da.EnKF_N( N=24, rot=True ,infl=1.01)
# xps += da.PartFilt(N=3000,NER=0.20,reg=1.2)
# xps += da.PFxN( N=1000,xN=100, NER=0.9,Qs=0.6)
HMM = _HMM.copy()
# HMM.t.BurnIn = 10*HMM.t.dtObs
# HMM.t.KObs = 30
HMM.t.BurnIn = HMM.t.dtObs
HMM.t.KObs = 2
xps.launch(HMM,
free=False,
liveplots="all",
store_u=False,
fail_gently=False,
save_as=False)
for xp in xps:
xp.stats.replay("all")
xp.stats.replay(t2=1)
xp.stats.replay(t2=0.0)
xp.stats.replay(t2=0.3)
xp.stats.replay(t2=0.8)
xp.stats.replay(t2=0.8, t1=0.2)
xp.stats.replay(t2=np.inf)
xp.stats.replay(t2=np.inf, speed=1)
xp.stats.replay(t2=np.inf, pause_a=0, pause_f=0)
assert True # An assertion for pytest to count
return HMM, xps # Return useful stuff
def test_L96():
xps = dpr.xpList()
from dapper.mods.Lorenz96.sakov2008 import HMM as _HMM
xps += da.EnKF("PertObs", N=40, infl=1.06)
xps += da.EnKF("Serial", N=28, infl=1.02, rot=True)
xps += da.OptInterp()
xps += da.Var3D(xB=0.02)
xps += da.ExtKF(infl=10)
xps += da.LETKF(N=6, rot=True, infl=1.05, loc_rad=4, taper="Step")
# from dapper.mods.Lorenz96.bocquet2015loc import HMM
# xps += da.EnKF_N( N=24, rot=True ,infl=1.01)
# xps += da.PartFilt(N=3000,NER=0.20,reg=1.2)
# xps += da.PFxN( N=1000,xN=100, NER=0.9,Qs=0.6)
HMM = _HMM.copy()
# HMM.tseq.BurnIn = 10*HMM.tseq.dto
# HMM.tseq.Ko = 30
HMM.tseq.BurnIn = HMM.tseq.dto
HMM.tseq.Ko = 2
xps.launch(
HMM,
free=False,
statkeys=True,
liveplots=None,
store_u=False,
fail_gently=False,
save_as=False,
)
print(xps.tabulate_avrgs(["rmse.a"]))
# spell_out(HMM)
# spell_out(xps[-1])
# spell_out(xps[-1].stats)
# spell_out(xps[-1].avrgs)
assert True # An assertion for pytest to count
return HMM, xps # Return useful stuff
"""Settings from `bib.wiljes2016second`.""" import numpy as np import dapper.mods as modelling from dapper.mods.Lorenz63.sakov2012 import HMM as _HMM from dapper.mods.Lorenz63.sakov2012 import Nx HMM = _HMM.copy() HMM.tseq = modelling.Chronology(0.01, dko=12, T=4**5, BurnIn=4) jj = np.array([0]) Obs = modelling.partial_Id_Obs(Nx, jj) Obs['noise'] = 8 HMM.Obs = modelling.Operator(**Obs) #################### # Suggested tuning #################### # Reproduce benchmarks for NETF and ESRF (here EnKF-N) from left pane of Fig 1. # from dapper.mods.Lorenz63.wiljes2017 import HMM # rmse.a reported by DAPPER / PAPER: # ------------------------------------------------------------------------------ # HMM.tseq.Ko = 10**2 # xps += OptInterp() # 5.4 / N/A # xps += Var3D(xB=0.3) # 3.0 / N/A # xps += EnKF_N(N=5) # 2.68 / N/A # xps += EnKF_N(N=30,rot=True) # 2.52 / 2.5 # xps += LNETF(N=40,rot=True,infl=1.02,Rs=1.0,loc_rad='NA') # 2.61 / ~2.2 # xps += PartFilt(N=35 ,reg=1.4,NER=0.3) # 2.05 / 1.4 * # *: tuning settings not given #
import dapper as dpr
import dapper.da_methods as da
# #### Load experiment setup: the hidden Markov model (HMM)
from dapper.mods.Lorenz63.sakov2012 import HMM # isort:skip
# #### Generate the same random numbers each time this script is run
seed = dpr.set_seed(3000)
# #### Simulate synthetic truth (xx) and noisy obs (yy)
HMM.t.T = 30 # shorten experiment
xx, yy = HMM.simulate()
# #### Specify a DA method configuration ("xp" for "experiment")
xp = da.EnKF('Sqrt', N=10, infl=1.02, rot=True)
# xp = da.Var3D()
# xp = da.PartFilt(N=100, reg=2.4, NER=0.3)
# #### Assimilate yy, knowing the HMM; xx is used to assess the performance
xp.assimilate(HMM, xx, yy, liveplots=not nb)
# #### Average the time series of various statistics
xp.stats.average_in_time()