OSE Experiments Using the JMA-MRI ENSO Forecasting System

1
OSE Experiments Using the JMA-MRI ENSO
Forecasting System
2nd GODAE OSE Workshop, Jun. 5th, Toulouse, France
Y. Fujii1, S. Matsumoto1, T. Yasuda1, M.
Kamachi1, K. Ando2 (1MRI/JMA, 2JAMSTEC)

• Outline
• Introduction
• Experimental Design
• Impacts of TAO/TRITON and Argo on the
  assimilation
• Impacts of TAO/TRITON and Argo on the forecast
• Impacts of satellite altimetry
• Summary

2
1. Introduction
3
Increase of observed profiles with ARGO floats
Jan. 2000
TAO/TRITON
ARGO
Others (XBT, CTD)
The number of the observed points by Argo floats
around the equator gets grater than that of
TAO-TRITON Buoy.
Jan. 2006
4
Background of this study
JAMSTEC would like to reduce the budget for
TRITON. (They want to assign more budgets to
Ocean bottom drilling, climate modeling research
with Earth Simulator, and so on.)
Questions
Is TRITON really valuable for JMAs operational
ENSO Forecasting System? Is it really necessary
to sustain both TRITON and Argo Networks? (It may
be oversampling.)
Joint Research of JAMSTEC and MRI
Check the impacts of TAO/TRITON and Argo Networks
(and satellite altimetry) on the JMAs ENSO
forecast system.
5
2. Experimental design
6
Outline of ENSO Forecasting in JMA
Observation
JRA25/JCDAS
Coupled Atmosphere-Ocean GCM
Atmospheric Data Assimilation System
JMA/MRI-CGCM
Initial State
Prediction
Ocean Data Assimilation System
MOVE/MRI.COM-G
Resolution Atmosphere TL95L40 Ocean
0.3-1ºx1º, L50 Coupling Interval 1 hour Heat
and momentum flux correction
Observation
7
Outline of the experiments
Assimilation(MOVE/MRI.COM-G) ? Jan. 2000-Dec.
2007 ALL ? Using all available data
(TAO/TRITON?10-day mean) NTT ? excluding
the data of TAO/TRITON NAF ? excluding the
data of ARGO floats NSH ? excluding
satellite altimetry data
Forecast (JMA/MRI-CGCM) ? 2004-2007 (16 cases)
Initial date Jan. 1st, Apr. 26th, Jul. 30th,
Oct. 28th (4 times a year) 13-month
forecasts using 11 ensemble members are
performed for ALL, NTT, NAF, and NSH
individually. Ensembles ? Assimilation runs
with perturbed SST Obs. The same flux
correction as in the JMA operation is used.
Forecasts biases are calculated for each lead
time, each forecasted month, and each
experiments (ALL, NTT, NAF, NSH) individually,
and removed from the forecasted fields.
8
3. Impact of TAO/TRITON and Argo Floats on
Assimilated fields
9
Variation of the impact on Z20 in the EQ PAC
m
m
10
Averaged T difference in the eq. Pac. (2004-07)
ALL-NAF
ALL-NTT
?C
?C
Contours represents temperature fields in ALL.
TAO/TRITON has an impact different from Argo(!?)
11
4. Impact of TAO/TRITON and Argo Floats on ENSO
Forecasting
12
Impact on 0-6 month forecast
RMSE of NTT or NAF RMSE of ALL
RMSE Improvement ratio
RMSE of ALL
13
Impact on 7-12 month forecast
The impact is large in the Indian Ocean for 7-12
month forecast Impact of Argo can be seen on the
equatorial and western tropical Pacific.
14
Impact on 0-6 month forecast score
Normalized RMSE by that of the persistency
forecast
15
Comparison with ECMWF
ECMWF
AME Absolute Mean Error
Balmaseda, M. A., and D. Anderson (2008b) Impact
on initialization strategies and observations on
seasonal forecast skill. Geophys. Res. Lett.,
submitted.
16
Changes of Spread from ALL
Spreads tends to increase without TAO/TRITON
data. Spreads tends to decrease without Argo
floats.
?
17
Example of forecasts
Initial 20051028
Initial 20070730
18
6. Summary
19
Summary
TAO/TRITON Array Remarkable positive impact on
NINO3 and NINO4 areas for 0-6 month SST
forecast. The Impact is not clear on western
tropical Pacific and for 7-12 month forecast.
Argo Floats Positive impact on NINO3, NINO4,
western tropical Pacific, and Western Indian
Ocean for 0-6 month forecast. The positive
impact remains for 7-12 month forecasts.
20
Things to Note
Results of OSE depends on the model, and data
assimilation scheme.
OSE should be performed with multi systems
(A better system in the future may use data more
effectively for improving forecast. It is
dangerous to judge the importance only with state
of art systems.)
We use 10-day mean observation data. The effect
of the high time-resolution of TAO/TRITON is not
evaluated.
Atmospheric reanalysis used in the data
assimilation run already uses information from
TAO/TRITON.
A longer time period is required for evaluating
the impacts of TAO/TRITON and Argo floats
correctly.
21
Thank You