Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
Development of Water Isotope Ra2o Data Assimila2on System
with Ensemble Kalman Filter
Kei Yoshimura AORI, Univ Tokyo
Yoshimura, Miyoshi, Kanamitsu, 2013 Yoshimura, Miyoshi, Kanamitsu, 2014
H 18O H
H D 16O
Stable Water Isotopes and Hydrologic Cycle
• SWI have integrated records of phase changes during its transport.
Iso Var
Iso. Var.
River Flow
H
H
18O
16O H
H
<
H D 16O
0.20% 0.016%
H2O
H218O HDO
(‘Heavy’ Water) Easy to condense
(‘Light’ Water) Easy to evaporate
“Fractionation” causes large heterogeneity in time and space.
Jasechko et al., 2013
• TranspiraIon represents 80~90 % of terrestrial ET.
⎥⎦
⎤⎢⎣
⎡
−
−=
ET
EET
ETT
δδδδ ET:ET T:Transp. E:Evap.
δET, δE, δT:Respective isotope ratios
Transpiration δT
≈δS
Evaporation δE
≈δS-εET δET
Soil moisture δS
Isotopes in GCM/RCM
• Incorporate water isotopes as passive tracers in GCMs/RCMs. Whenever water phase change takes place, isotopic water (HDO, H2
18O) behave differently to ordinary water (H2O).
Risi et al. 2008 Courtesy of JMA
H18OH
HD16O
Typical convecIve precipitaIon process
Valida2on: Comparison in δ18O in Precip
SWING-2 o Kick-off in 17-19 November 2008 in IAEA
HQ; chaired by C. Sturm, K. Yoshimura & D. Noone.
o More isotopic AGCMs (at least 9) and 2 isotopic RCMs.
o Add nudging experiments to focus on only isotopic parameterizations and on more realistic reconstruction of isotopic variations.
o More focused on hydrologic cycle than climatology
o Endorsed by GHP/CEOP in 2008-2010
Courtesy of L. StoZ
Forward Proxy Modeling of δ18O in cellulose
Sea water δ18O derived from coral and model (temperature effect removed by Sr/Ca)
δ18Osw records are well reproduced both seasonally and inter-‐annually in various sites.
-‐0.5
0
0.5
-‐0.3
0.7
Nauru (1S 166E) r=0.52
Papua NewGuinea (5S 144E) r=0.73
Vanuatu (16S 167E) r = 0.42 0
0.5
1 Seychells (5S 54E) r = 0.41
-‐0.3 -‐0.1 0.1 0.3 0.5
1975 1980 1985 1990 1995 2000
Clipperton (10N 108W) r= 0.46
:Model :Coral record
δ18 Osw
(‰)
-‐0.5
0
0.5
1
1975 1980 1985 1990 1995 2000
-‐2
0
2 Pohnpei (6N 158E) r=0.52
Courtesy of K. Kojima
Way forward: Isotope Reanalysis
H18OH
δ18O, δD T, P, etc
Regression Analysis, etc
ConvenIonal paleo-‐climatology Empirical relaIonship between δ-‐T or -‐P.
Ongoing approach: Forward proxy modeling Proxy-‐Climate relaIonship may vary in space and Ime
Forward Proxy Modeling with GCMs
ObjecIve analysis with DA
Toward “Real” Isotope Reanalysis: Data Assimila2on of Isotope
R.S. HDO
H18OH
HD16O
IsoGSM
LETKF
[ ]
afofa
Tf
Tf
m
H δXXyKXX
RXHδUUDδXKXXδXxxX
+−+=
=
−==−−
))((
)(
,11
1 !
Targets: ü First global 4D analyses for vapor isotopes.
ü Accurate Precip. isotopes in fine resolution.
ü Possibility of improvement on other dynamical fields.
SCIAMACHY/Envisat: surface vapor HDO (Frankenberg et al., 2009, Science)
TES/Aura: mid troposphere vapor HDO (Worden et al., 2007, Nature)
12
Local Ensemble Transformed Kalman Filter (Miyoshi and Yamane, 2008)
• Not only the assimilated variables, but also other variables will be corrected to be a consistent field.
Idealized Experiments (OSSE)
• Assume one realizaIon of AMIP runs as truth.
Grid with denser observa2ons, δ18O
NoObs
DeltaOnly
Grid with denser observa2ons, T
NoObs
DeltaOnly
(a) Bias, NoObs
(b) RMSD, NoObs
(c) Bias, DeltaOnly
(d) RMSD, DeltaOnly
T
What about more realis2c situa2on? Experiments with conven2onal measurement system
σ=0.995 U [m/s]
V [m/s]
T [K]
q [g/kg]
Ps [hPa]
δ18O [‰]
δD [‰]
UVTq 1.33 1.30 0.40 0.42 1.04 0.98 7.23
UVTq+δD 1.27 1.25 0.40 0.41 0.99 0.93 6.94
σ=0.8835 U [m/s]
V [m/s]
T [K]
q [g/kg]
δ18O [‰]
δD [‰]
UVTq 1.49 1.39 0.55 0.69 1.41 10.77
UVTq+δD 1.42 1.34 0.53 0.68 1.35 10.35
Global RMSD
Ul2mate goal: Climate Reanalysis
• Much longer records than man-‐made observaIon • Oceanic sediment δ18O (millions yBP) • Icesheet cores δ18O・δD (~800 kyBP) • Icecap cores δ18O・δD (~20 kyBP) • Speleothem δ18O (~2000 yBP) • Treering δ18O (~1000 yBP) • Coral δ18O (~400 yBP)
• Bridging data and physics, consistently!
19
H18OH
Summary
• Isotopic Data as input observaIon had posiIve impact on not only isotopic fields but also dynamical fields. • (Selfish) suggesIon for new observaIons: • Accuracy < Number of data • Temporal resoluIon < Longer data • Dense coverage < Sparse but equally distributed
• There is potenIal for dynamical constraint by isotopic proxy data for the past, but lots of technical obstacles exist.
Paired isotopic proxy data since 8,000yBP
Liu et al., 2014, Nature Comm.
LH minus MH (‰) PNA+ minus PNA-‐(‰) IsoGSM simulated precipitaIon δ18O difference
IsoGSM simula2ons
• Time-‐slice runs for MH and LH for 30 years, respecIvely. • SST anomaly simulated by IPSL-‐CGCM was forced.
Result (Global RMSE for δ18O, Wind, Temp, and Surface Pressure)
3.3
3.4
3.5
3.6
3.7
NoObs SCIA TES GNIP ALL
δ18O (‰) at 2m
4.15
4.2
4.25
4.3
4.35
4.4
NoObs SCIA TES GNIP ALL
Zonal Wind (m/s) at boZom
2 2.1 2.2 2.3 2.4 2.5 2.6
NoObs SCIA TES GNIP ALL
Air Temperature (K) at boZom
540
550
560
570
580
590
NoObs SCIA TES GNIP ALL
Surface Pressure (Pa)