• Arakawa, O., , and A. Kitoh, 2004: Comparison of local precipitation–SST relationship between the observation and a reanalysis dataset. Geophys. Res. Lett., 31 , L12206. doi:10.1029/2004GL020283.

    • Search Google Scholar
    • Export Citation
  • Balmaseda, M. A., , D. Dee, , A. Vialard, , and D. L. T. Anderson, 2007: A multivariate treatment of bias for sequential data assimilation: Application to the tropical oceans. Quart. J. Roy. Meteor. Soc., 133 , 167179.

    • Search Google Scholar
    • Export Citation
  • Bloom, S. C., , L. L. Takacs, , A. M. D. Silva, , and D. Ledvina, 1996: Data assimilation using incremental analysis updates. Mon. Wea. Rev., 124 , 12561271.

    • Search Google Scholar
    • Export Citation
  • CLS, 2004: SSALTO/DUACS user handbook: (M)SLA and (M)ADT near-real time and delayed time products. CLS-DOS-NT-04 103, Collecte Localisation Satellites (CLS), Toulouse, France, 42 pp.

    • Search Google Scholar
    • Export Citation
  • Conkright, M. E., and Coauthors, 2002: Introduction. Vol. 1, World Ocean Database 2001, NOAA Atlas NESDIS 42, 167 pp.

  • Fujii, Y., , and M. Kamachi, 2003: Three-dimensional analysis of temperature and salinity in the equatorial Pacific using a variational method with vertical coupled temperature-salinity empirical orthogonal function modes. J. Geophys. Res., 108 , 3297. doi:10.1029/2002JC001745.

    • Search Google Scholar
    • Export Citation
  • Fujii, Y., , S. Ishizaki, , and M. Kamachi, 2005: Application of nonlinear constraints in a three dimensional variational ocean analysis. J. Oceanogr., 61 , 655662.

    • Search Google Scholar
    • Export Citation
  • Hamilton, D., 1994: GTSPP builds an ocean temperature-salinity database. Earth Syst. Monit., 4 (4) 45.

  • Hayes, S. P., , L. J. Mangum, , J. Picaut, , A. Sumi, , and K. Takeuchi, 1991: TOGA-TAO: A moored array for real-time measurements in tropical Pacific Ocean. Bull. Amer. Meteor. Soc., 72 , 339347.

    • Search Google Scholar
    • Export Citation
  • Ishii, M., , A. Shouji, , S. Sugimoto, , and T. Matsumoto, 2005: Objective analyses of sea-surface temperature and marine meteorological variables for the 20th century using ICOADS and the Kobe collection. Int. J. Climatol., 25 , 865879.

    • Search Google Scholar
    • Export Citation
  • Ishikawa, I., , H. Tsujino, , M. Hirabara, , H. Nakano, , T. Yasuda, , and H. Ishizaki, 2005: Meteorological Research Institute Community Ocean Model (MRI.COM) manual. Tech. Rep. of the Meteorological Research Institute 47, Meteorological Research Institute, Tsukuba, Japan, 189 pp.

    • Search Google Scholar
    • Export Citation
  • Ishizaki, S., , S. Matsumoto, , Y. Fujii, , T. Yasuda, , and M. Kamachi, 2006: Correction of the zonal wind stress in the tropical Pacific using the ocean data assimilation system in Meteorological Research Institute. Proc. The Oceanographic Society of Japan 2006 Spring Meeting, Yokohama, Japan, The Oceanographic Society of Japan, 23.

    • Search Google Scholar
    • Export Citation
  • JMA, 2007: Outline of the operational numerical weather prediction at the Japan Meteorological Agency. Appendix to WMO Technical Progress Report on the Global Data-Processing and Forecasting System and Numerical Weather Prediction, Japan Meteorological Agency, Tokyo, Japan, 196 pp.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kanamitsu, M., and Coauthors, 2002: NCEP dynamical seasonal forecast system 2000. Bull. Amer. Meteor. Soc., 83 , 10191037.

  • Kuroda, Y., 2002: TRITON, present status and future plan. Report for the international workshop for review of the tropical moored buoy network, JAMSTEC, Yokosuka, Japan, 77 pp.

    • Search Google Scholar
    • Export Citation
  • Lin, J., 2007: The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere feedback analysis. J. Climate, 20 , 44974525.

  • Locarnini, R. A., , A. V. Mishonov, , J. I. Antonov, , T. P. Boyer, , and H. E. Garcia, 2006: Temperature. Vol. 1, World Ocean Atlas 2005, NOAA Atlas NESDIS 61, 182 pp.

    • Search Google Scholar
    • Export Citation
  • Luo, J-J., , S. Masson, , S. K. Behera, , and T. Yamagata, 2008: Extended ENSO predictions using a fully coupled ocean–atmospheric model. J. Climate, 21 , 8493.

    • Search Google Scholar
    • Export Citation
  • Madden, R. A., , and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40–50 day period. J. Atmos. Sci., 29 , 11091123.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and Coauthors, 1998: The Tropical Ocean-Global Atmosphere observing system: A decade of progress. J. Geophys. Res., 103 , (C7). 1416914240.

    • Search Google Scholar
    • Export Citation
  • Mizuta, R., and Coauthors, 2006: 20-km-mesh global climate simulations using JMA-GSM model—Mean climate states. J. Meteor. Soc. Japan, 84 , 165185.

    • Search Google Scholar
    • Export Citation
  • Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85 , 369432.

  • Stammer, D., 2007: Draft report of the first CLIVAR/GODAE meeting on ocean synthesis evaluation. International CLIVAR Project Office, 80 pp. [Available online at http://www.clivar.org/organization/gsop/synthesis/mit/synthesis_2.php].

    • Search Google Scholar
    • Export Citation
  • Sugiura, N., , T. Awaji, , S. Masuda, , T. Mochizuki, , T. Toyoda, , T. Miyama, , H. Igarashi, , and Y. Ishikawa, 2008: Development of a four-dimensional variational coupled data assimilation system for enhanced analysis and prediction of seasonal to interannual climate variations. J. Geophys. Res., 113 , C10017. doi:10.1029/2008JC004741.

    • Search Google Scholar
    • Export Citation
  • Tsujino, H., , and T. Yasuda, 2004: Formation and circulation of mode waters of the North Pacific in a high-resolution GCM. J. Phys. Oceanogr., 34 , 399415.

    • Search Google Scholar
    • Export Citation
  • Usui, N., , S. Ishizaki, , Y. Fujii, , H. Tsujino, , T. Yasuda, , and M. Kamachi, 2006: Meteorological Research Institute Multivariate Ocean Variational Estimation (MOVE) system: Some early results. Adv. Space Res., 37 , 806822.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , and Z. Fan, 1999: Choice of South Asian summer monsoon indices. Bull. Amer. Meteor. Soc., 80 , 629638.

  • Wang, B., , R. Wu, , and X. Fu, 2000: Pacific–East Asian teleconnection: How does ENSO affect East Asian climate? J. Climate, 13 , 15171536.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , R. Wu, , and T. Li, 2003: Atmosphere–warm ocean interaction and its impacts on Asian–Australian monsoon variation. J. Climate, 16 , 11951211.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , I-S. Kang, , and J-Y. Lee, 2004: Ensemble simulations of Asian–Australian monsoon variability by 11 AGCMs. J. Climate, 17 , 803818.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , Q. Ding, , X. Fu, , I-S. Kang, , K. Jin, , J. Shukla, , and F. Doblas-Reyes, 2005: Fundamental challenge in simulation and prediction of summer monsoon rainfall. Geophys. Res. Lett., 32 , L15711. doi:10.1029/2005GL022734.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., , and S. Yang, 1992: Monsoon and ENSO: Selectively interactive systems. Quart. J. Roy. Meteor. Soc., 118 , 877926.

  • Xie, P., , and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • Xie, S. P., , K. Hu, , J. Hafner, , H. Tokinaga, , Y. Du, , G. Huang, , and T. Sampe, 2009: Indian Ocean capacitor effect on Indo-Western Pacific climate during the summer following El Niño. J. Climate, 22 , 730747.

    • Search Google Scholar
    • Export Citation
  • Yamanaka, G., 2008: Discrepancies between observed and ocean general circulation model–simulated anomalies in recent SSTs of the tropical Indian Ocean caused by apparent trends in atmospheric reanalysis data. Geophys. Res. Lett., 35 , L18603. doi:10.1029/2008GL034737.

    • Search Google Scholar
    • Export Citation
  • Yasuda, T., , Y. Takaya, , C. Kobayashi, , M. Kamachi, , H. Kamahori, , and T. Ose, 2007: Asian monsoon predictability in JMA/MRI Seasonal Forecast System. CLIVAR Exchanges, No. 43, International CLIVAR Project Office, Southampton, United Kingdom, 18–24.

    • Search Google Scholar
    • Export Citation
  • Zhang, S., , M. J. Harrison, , A. T. Wittenberg, , and A. Rosati, 2005: Initialization of an ENSO forecast system using a parallelized ensemble filter. Mon. Wea. Rev., 133 , 31763201.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    Monthly climatology of SST (°C) in (left) January and (right) July for (top) MOVE-C RA, (middle) MOVE-G RA07, and (bottom) the CGCM run. Contours indicate the climatology, and color shading, the deviation from the climatology of COBE-SST.

  • View in gallery

    ACC maps of SST in MOVE-C RA with COBE-SST.

  • View in gallery

    Monthly climatology of precipitation (mm day−1) in (left) January and (right) July for (a) CMAP, (b) MOVE-C RA, (c) the AMIP run, and (d) the CGCM run.

  • View in gallery

    ACC maps of precipitation with CMAP in boreal winter, spring, summer, and fall in the tropical Pacific and Indian Ocean for (a) MOVE-C RA and (b) the AMIP run.

  • View in gallery

    Time series of precipitation (mm day−1) in the area east of the Philippines (5°–20°N, 125°–150°E) for MOVE-C RA (black solid line), AMIP run (gray solid line), and CMAP (black dashed line).

  • View in gallery

    Daily mean SLP (contour, hPa) and daily precipitation (color shading, mm day−1) every four days in July 1997 for (left) MOVE-C RA and (right) the AMIP run.

  • View in gallery

    SLP field (contour, hPa) and wind field at 850 hPa (arrow) averaged in the summer of 1997 for (a) JRA-25, (b) MOVE-C RA, and (c) the AMIP run.

  • View in gallery

    Climatological fields of SLP (contour, hPa) and zonal wind shear between 850 and 200 hPa (shading, m s−1) in summer for (a) JRA-25, (b) MOVE-C RA, and (c) the AMIP run.

  • View in gallery

    Yearly variations of (a) the DU2 index, (b) the W–Y index, and (c) the EPP index for MOVE-C RA (black solid line with open circles), the AMIP run (gray solid line with closed circle), and JRA-25 in (a) and (b) and CMAP in (c) (black dashed line with open square). “E” denotes an El Niño period defined by JMA.

  • View in gallery

    Maps of time lags (month) of precipitation change behind SST for (a) CMAP and COBE-SST, (b) MOVE-C RA, and (c) the AMIP run. Positions where the significance level is less than 99% are blank.

  • View in gallery

    Velocity potential fields (m2 s−1) at 200 hPa averaged in summer 1997 for MOVE-C RA (black contour) and the AMIP run (gray contour). Color shading represents the difference between MOVE-C RA and AMIP run.

  • View in gallery

    Maps of the correlation between SST and precipitation in summer (June–August) for (a) CMAP and COBE-SST, (b) MOVE-C RA, and (c) the AMIP run.

  • View in gallery

    Fig. A1. As in Fig. 1, but for OHC (°C): color shading represents the deviation of OHC calculated from WOA05.

  • View in gallery

    Fig. A2. Map of ACC between MOVE-C RA and MOVE-G RA07 for OHC.

  • View in gallery

    Fig. A3. Time–longitude section of monthly OHC anomaly (°C) at the equator in the Pacific between 2000 and 2005 for (left) MOVE-C RA and (right) MOVE-G RA07. Solid (dashed) lines indicate contours of positive (negative) values.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 50 50 10
PDF Downloads 14 14 2

Coupled Climate Simulation by Constraining Ocean Fields in a Coupled Model with Ocean Data

View More View Less
  • 1 Meteorological Research Institute, Tsukuba, Japan
© Get Permissions
Full access

Abstract

The authors developed a system for simulating climate variation by constraining the ocean component of a coupled atmosphere–ocean general circulation model (CGCM) through ocean data assimilation and conducted a climate simulation [Multivariate Ocean Variational Estimation System–Coupled Version Reanalysis (MOVE-C RA)]. The monthly variation of sea surface temperature (SST) is reasonably recovered in MOVE-C RA. Furthermore, MOVE-C RA has improved precipitation fields over the Atmospheric Model Intercomparison Project (AMIP) run (a simulation of the atmosphere model forced by observed daily SST) and the CGCM free simulation run. In particular, precipitation in the Philippine Sea in summer is improved over the AMIP run. This improvement is assumed to stem from the reproduction of the interaction between SST and precipitation, indicated by the lag of the precipitation change behind SST. Enhanced (suppressed) convection tends to induce an SST drop (rise) because of cloud cover and ocean mixing in the real world. A lack of this interaction in the AMIP run leads to overestimating the precipitation in the Bay of Bengal in summer. Because it is recovered in MOVE-C RA, the overestimate is suppressed. This intensifies the zonal Walker circulation and the monsoon trough, resulting in enhanced convection in the Philippine Sea. The spurious positive correlation between SST and precipitation around the Philippines in the AMIP run in summer is also removed in MOVE-C RA. These improvements demonstrate the effectiveness of simulating ocean interior processes with the ocean model and data assimilation for reproducing the climate variability.

Corresponding author address: Yosuke Fujii, Meteorological Research Institute, 1-1 Nagamine, Tsukuba, 305-0052, Japan. Email: yfujii@mri-jma.go.jp

Abstract

The authors developed a system for simulating climate variation by constraining the ocean component of a coupled atmosphere–ocean general circulation model (CGCM) through ocean data assimilation and conducted a climate simulation [Multivariate Ocean Variational Estimation System–Coupled Version Reanalysis (MOVE-C RA)]. The monthly variation of sea surface temperature (SST) is reasonably recovered in MOVE-C RA. Furthermore, MOVE-C RA has improved precipitation fields over the Atmospheric Model Intercomparison Project (AMIP) run (a simulation of the atmosphere model forced by observed daily SST) and the CGCM free simulation run. In particular, precipitation in the Philippine Sea in summer is improved over the AMIP run. This improvement is assumed to stem from the reproduction of the interaction between SST and precipitation, indicated by the lag of the precipitation change behind SST. Enhanced (suppressed) convection tends to induce an SST drop (rise) because of cloud cover and ocean mixing in the real world. A lack of this interaction in the AMIP run leads to overestimating the precipitation in the Bay of Bengal in summer. Because it is recovered in MOVE-C RA, the overestimate is suppressed. This intensifies the zonal Walker circulation and the monsoon trough, resulting in enhanced convection in the Philippine Sea. The spurious positive correlation between SST and precipitation around the Philippines in the AMIP run in summer is also removed in MOVE-C RA. These improvements demonstrate the effectiveness of simulating ocean interior processes with the ocean model and data assimilation for reproducing the climate variability.

Corresponding author address: Yosuke Fujii, Meteorological Research Institute, 1-1 Nagamine, Tsukuba, 305-0052, Japan. Email: yfujii@mri-jma.go.jp

1. Introduction

A coupled atmosphere–ocean general circulation model (CGCM) is a powerful tool for analyzing the variability in the climate system because it reflects the interaction between the atmosphere and the ocean. It also provides us a physical way to predict climate variation. However, it is impossible for a CGCM to reproduce real climate fluctuations with actual timing without inputting observed information. We can therefore only rely on data assimilation techniques to access the actual climate.

A conventional method to insert observed information into a CGCM is to nudge the sea surface temperature (SST) field toward observed SST. For example, Luo et al. (2008) reproduced the variability of the El Niño–Southern Oscillation (ENSO) by nudging the SST fields toward observations and successfully performed an ENSO forecasting experiment from the reproduced fields. The method is, however, inadequate because SST controls only a part of the climate variability. Some climate phenomena cannot be detected or separated from other phenomena with SST data alone. It should also be noted that the SST nudging technique does not allow representation of the coupled feedbacks. In addition, state-of-the-art CGCMs still suffer from model biases. Various sorts of observed data are required so as to mitigate this model deficiency. In fact, Luo et al. pointed out the necessity to assimilate both atmosphere and ocean interior observations.

It should also be noted that combined datasets made of uncoupled atmosphere and ocean data assimilation systems are inadequate for climate analysis because they do not correctly represent the interaction between the atmosphere and ocean (e.g., Arakawa and Kitoh 2004). Although this combination is also used for initial conditions of the atmosphere and ocean fields of CGCMs in many current operational systems of seasonal and ENSO forecasting (e.g., Kanamitsu et al. 2002; Yasuda et al. 2007), it is often pointed out that the inconsistency between the atmosphere and ocean fields can degrade the forecast accuracy (e.g., Luo et al. 2008; Sugiura et al. 2008). Therefore, some groups are now developing a coupled atmosphere–ocean data assimilation system in which both atmosphere and ocean observations are assimilated into a CGCM (e.g., Zhang et al. 2005; Sugiura et al. 2008).

In this study, we developed a data assimilation system in which ocean observation data are employed for constraining ocean fields of a CGCM, that is, a “quasi-coupled data assimilation system” as a prototype of a truly coupled data assimilation system. The system is called Multivariate Ocean Variational Estimation System–Coupled Version (MOVE-C) here. We expect that MOVE-C will provide us with a suitable four-dimensional coupled atmosphere–ocean dataset for analyzing climate variability because slow variations in the coupled atmosphere–ocean system tend to be controlled by the ocean field. We particularly anticipate that the dataset is going to be a powerful tool for clarifying the role of air–sea interaction in the climate system. There is also a possibility that seasonal forecasting will be improved by using the coupled atmosphere–ocean state simulated by MOVE-C as the initial condition for a CGCM.

We conducted a reanalysis experiment using this system and compared results with the Atmospheric Model Intercomparison Project (AMIP) run, a simulation run of the atmosphere model used in MOVE-C with observed daily SST as the oceanic boundary condition, for the period from 1980 to 2005. We found that MOVE-C Reanalysis (MOVE-C RA) improved precipitation fields over the AMIP run. In particular, the underestimated summer precipitation in the Philippine Sea is effectively improved in MOVE-C RA. This was surprising because the AMIP run seems to have a better SST field; SST in MOVE-C RA deviates from the observed SST, while the observed SST is directly passed to the atmosphere model in the AMIP run. To explore the reason for improved summer precipitation in the Philippine Sea, we analyzed the activity of tropical cyclones, the summer Southeast Asian monsoon trough, and the zonal Walker circulation.

We also investigate the relationship of the variations between SST and precipitation. Wang et al. (2004) indicated that the relationship between local summer rainfall and SST anomalies over the Philippine Sea, the South China Sea, and the Bay of Bengal is not reproduced adequately in simulations of atmosphere models forced by observed SST. They concluded that the deficiency causes poor estimation of the precipitation field. This deficiency can possibly be mitigated in MOVE-C RA because MOVE-C permits a small deviation of the SST field from observed data following the physics of the coupled model.

The rest of this paper is organized as follows. The configuration of MOVE-C RA and description of other experiments and datasets, including the AMIP run, are presented in section 2. In the first subsection of section 3, we describe the validation of the SST fields in MOVE-C RA. Improvements of precipitation, the monsoon trough, and the zonal Walker circulation in MOVE-C RA over the AMIP run are introduced in the following subsections. We state that the improvements stem from the reestablishment of the relationship between SST and precipitation in section 4. This study is summarized in section 5.

2. Experiments and data

a. MOVE-C reanalysis

The Multivariate Ocean Variational Estimation System–Coupled Version is a system in which ocean observation data are used for constraining ocean fields of a CGCM through a global ocean data assimilation scheme. The CGCM used in MOVE-C is the Japan Meteorological Agency (JMA)/Meteorological Research Institute (MRI)-CGCM (Yasuda et al. 2007). JMA has used it operationally for ENSO forecasting since March 2008. The atmospheric component is a general circulation model (GCM) developed in JMA and MRI (Mizuta et al. 2006; JMA 2007). It is a global spectral model with a resolution of TL95/L40; it has 40 levels, and the horizontal resolution corresponds to about 180 km. The ocean component is the MRI Community Ocean Model (MRI.COM), a multilevel GCM developed in MRI (Tsujino and Yasuda 2004; Ishikawa et al. 2005). It has a near-global domain within 75°S–75°N and 50 levels. The zonal grid spacing is 1°. The meridional grid spacing is 0.3° within 5°S–5°N and 1° poleward of 15°S and 15°N. Within 5°–15°N (S) it gradually increases from 0.3° to 1°. The coupling takes place every hour; the ocean component gives SST to the atmosphere component once an hour, while the atmosphere component provides hourly mean heat, momentum, and freshwater fluxes to the ocean component. SST and fluxes are exchanged without any adjustment.

The ocean data assimilation scheme is based on that in MOVE/MRI.COM-G (Usui et al. 2006), which JMA has used operationally since March 2008 for monitoring the equatorial Pacific ocean state and providing JMA/MRI-CGCM with ocean initial conditions for the ENSO forecast. In the scheme, temperature and salinity fields are analyzed by a multivariate three-dimensional variational data assimilation (3DVAR) method using coupled temperature–salinity empirical orthogonal functional decomposition (Fujii and Kamachi 2003; Fujii et al. 2005). The analysis field is reflected in the model field through incremental analysis updates (IAU) (Bloom et al. 1996). An online model-bias correction scheme using the one-step bias-correction algorithm (Balmaseda et al. 2007) is additionally applied in MOVE-C. This scheme enables one to estimate biases of model temperature and salinity fields while an assimilation run is performed and permits their slow variations.

The MOVE-C Reanalysis experiment is performed from 1940 to 2006. In the experiment, in situ ocean temperature and salinity profiles, satellite sea surface height (SSH) anomaly data, and observation-based gridded SST data are assimilated into the CGCM. The temperature and salinity profiles are collected from World Ocean Database 2001 (WOD01) (Conkright et al. 2002), the Global Temperature–Salinity Profile Program (GTSPP) database (Hamilton 1994), and the data of the Tropical Atmosphere Ocean/Triangle Trans-Ocean Buoy Network (TAO/TRITON) array (Hayes et al. 1991; McPhaden et al. 1998; Kuroda 2002). The SSH data is the along-track data from Ocean Topography Experiment (TOPEX)/Poseidon, Jason-1, European Remote Sensing satellites (ERS-1 and ERS-2), and environmental satellite multimission altimeters (ENVISAT), extracted from Ssalto/Duacs delayed-time multimission altimeter products (CLS 2004). The gridded SST data are the Centennial In-situ Observation-Based Estimates (COBE) of the variability of SSTs and marine meteorological variables (COBE-SST) compiled in JMA (Ishii et al. 2005).

It should be noted that IAU is applied in MOVE-C RA with an interval of one month. The calculation procedure is as follows. First, the CGCM is integrated from the beginning to the middle of a month. Second, the ocean field at the middle of the month is analyzed with data observed in the month, and increments are calculated from the difference between the predicted and analyzed fields. Third, the CGCM is integrated again through the month while adding the increments to the ocean field. Here, the increments are not changed in the month. A half-month tendency predicted by the CGCM is thus reflected in the analysis increments, while shorter time-scale variations are not constrained by the assimilation of ocean data. Actually, this careful handling seems to be essential for reconstructing the appropriate relationship between precipitation and SST, as discussed in section 4.

b. References and data processing

We compare the atmosphere field of MOVE-C RA with the AMIP run so as to examine what is improved on the simulation of the uncoupled atmosphere model. The AMIP run is a simulation run of the atmosphere model used in MOVE-C. Daily data of COBE-SST is employed as the ocean boundary condition of the atmosphere model.

We refer to the Japanese 25-yr Reanalysis (JRA-25) (Onogi et al. 2007) as a proxy of atmosphere fields in the real world (except precipitation). The original period of the dataset is from 1979 to 2004, but it is extended to 2006 by adding the product of JMA Climate Data Assimilation System (JCDAS). For precipitation, we use the Climate Prediction Center Merged Analysis of Precipitation (CMAP) (Xie and Arkin 1997).

We also present the monthly climatology of the CGCM free simulation run (CGCM run) using JMA/MRI-CGCM in the same configuration as MOVE-C for the comparison to show the effect of assimilating ocean data on the model climatology. The simulation is started from the analysis results of MOVE/MRI.COM-G and JRA-25 on 1 January 2000. The integration is performed for 101 years. Coupling is performed without any adjustment of SST and fluxes. The climatology is calculated from the last 26 years.

For SST and ocean fields, we refer to the Multivariate Ocean Variational Estimation System–Global Version Reanalysis 2007 (MOVE-G RA07), which is an ocean reanalysis using the ocean data assimilation part of MOVE-C, that is, MOVE-G/MRI.COM, with the same observations as MOVE-C RA except COBE-SST. A detailed description of MOVE-G RA07 can be found in the appendix. We also use COBE-SST as another reference of SST.

All datasets are first converted to monthly mean (or climatology) with a grid spacing of 2.5°. All figures and statistics are made or calculated from the monthly data, except for the daily atmosphere fields (Fig. 6), which are made from original model output. Climatologies and correlations are evaluated for 1980–2005 because JRA-25 is available from 1979 to 2006 (the first and last years are excluded for calculating lagged correlation coefficients).

3. Climate and variability in MOVE-C RA

In this section, we examine climate and variability in MOVE-C RA. First, we validate the SST field in MOVE-C RA by comparing it with COBE-SST and then examine the atmosphere field in the following subsections.

a. SST field

The monthly mean SST fields from MOVE-C RA, MOVE-G RA07, and the CGCM run are presented in Fig. 1, where the color shading represents the deviation from the monthly climatology of COBE-SST. This figure indicates that the global patterns of the climatological SST fields are realistically reconstructed in MOVE-C RA. In the CGCM run, SST is estimated to be higher than COBE-SST both in January and July in almost the whole area, whereas these biases are dramatically reduced in MOVE-C RA.

The deviation of MOVE-C RA in January is similar to that of MOVE-G RA07. Both fields have a cold bias in the eastern equatorial Pacific. This bias is caused by the overestimated upwelling in this region. They also share a warm bias in the southern part of the Southern Ocean and a cold bias north of it. These imply that the frontal structure of SST in the Southern Ocean is smoothed in both reanalyses. The warm bias in MOVE-C RA is larger than in MOVE-G RA07 because the circumpolar westerly wind is underestimated in MOVE-C RA. In the North Pacific, cold biases are dominant in both fields, although MOVE-C RA has larger biases because of the southward shift of the westerly wind maximum in MOVE-C RA.

The cold bias in the eastern equatorial Pacific also exists in July in both fields, although it is larger in MOVE-C RA. It also has a cold bias in the equatorial Atlantic. In contrast, the South Pacific and South Atlantic have warm biases in MOVE-G RA07, but these biases are eliminated in MOVE-C RA. Cold biases in the North Pacific and North Atlantic in MOVE-G RA07 are also mitigated in MOVE-C RA. Finally, the accuracy of the SST climatology in MOVE-C RA is much improved over the CGCM run and comparable with that in MOVE-G RA07.

Figure 2 depicts the distribution of anomaly correlation coefficients (ACCs) of SST between MOVE-C RA and COBE-SST, where the anomalies are calculated as the deviation from the monthly climatology of each dataset. The ACC exceeds 0.9 in the central and eastern equatorial Pacific, where ENSO is a major factor governing SST variation. Moreover, it exceeds 0.6 in a large part of the model region. This large ACC is achieved mainly by assimilating COBE-SST but is supplemented by assimilation of subsurface data as well. Thus, the variability of SST on time scales longer than a month is reproduced reasonably in MOVE-C RA.

We also validated the ocean heat content (OHC) fields in MOVE-C RA. The large OHC biases found in the CGCM run are effectively removed by assimilating ocean data in MOVE-C RA. Furthermore, MOVE-C reproduces the climatology and monthly variability of OHC as well as the current operational ocean data assimilation system in JMA (MOVE/MRI.COM-G), while shorter time-scale variations are not constrained by the data assimilation. OHC fields are discussed in detail in the appendix.

b. Precipitation field

In this and the following subsections, we mainly compare the atmospheric field in MOVE-C RA with the AMIP run. Observed SST itself is employed as the oceanic surface condition in the AMIP run, while in MOVE-C RA the ocean state, including SST, is basically adjusted to the atmospheric condition (thus, SST slightly deviates from the observed one) although ocean data including the observed SST are assimilated into it. This comparison, therefore, manifests the effect of the air–sea interaction on the simulated atmospheric field.

We depict the distribution of the monthly mean precipitation in MOVE-C RA, the AMIP run, the CGCM run, and CMAP in Fig. 3. In January, the AMIP run has too large a peak north of New Guinea Island, while precipitation in the South Pacific convergence zone (SPCZ) is underestimated. The CGCM run has a commonly recognized bias in CGCMs, that is, a double intertropical convergence zone (ITCZ) (e.g., Lin 2007): the southern and northern belts of large precipitation are extended zonally to the east more than CMAP. Both of these biases are mitigated in MOVE-C RA. In addition, the CGCM run has more precipitation than the AMIP run in the tropical Indian Ocean, and MOVE-C RA reflects this feature, resulting in an improved precipitation field there on the AMIP run. In contrast, the precipitation peak over the south of the African continent is overestimated in the CGCM run, but in MOVE-C RA the overestimate is mitigated as well as in the AMIP run.

In July, precipitation is considerably overestimated in the western part of the Bay of Bengal and underestimated around the Philippines in the AMIP run. The overestimate in the Bay of Bengal is mitigated and precipitation around the Philippines increases in MOVE-C RA. The precipitation field in July in the CGCM run has a bias like the double-ITCZ as well as in January. However, the southern band of the precipitation associated with SPCZ is restricted west of the international date line, and the bias is thus improved in MOVE-C RA. The (northern) ITCZ has a peak of precipitation at the eastern edge of the Pacific. This feature is reproduced in MOVE-C RA as well as in the CGCM run, although the peak is located in the central Pacific in the AMIP run and deviates from that in the CMAP. Precipitation over the equatorial Atlantic is overestimated both in the AMIP and CGCM runs. This error is also reduced in MOVE-C RA.

The ACC of precipitation with CMAP is compared between MOVE-C RA and the AMIP run for boreal winter (December–February), spring (March–May), summer (June–August), and fall (September–November) in Fig. 4. Here the ACC is calculated from the anomaly from the climatology of each season. In all seasons the ACC is relatively large in the central and eastern equatorial Pacific because the SST anomaly associated with ENSO mainly governs the variability of precipitation there. In MOVE-C RA, the large ACC area extends to the west more than in the AMIP run. MOVE-C thus reproduces the variation of precipitation in the western equatorial Pacific (around equator, 160°E) better than the AMIP run. In addition, MOVE-C RA exhibits an improvement over the AMIP run in the Philippine Sea in spring and summer. This improvement is examined in detail in the next subsection. In the Indian Ocean negative values of the ACC are noticeable in winter and spring in the AMIP run. This deficiency is also mitigated in MOVE-C RA.

In this subsection, we demonstrated that MOVE-C RA has improvements of the precipitation field over the CGCM and AMIP runs. Improvements over the AMIP run are particularly surprising because observed SST (COBE-SST) is directly passed to the atmosphere model in the AMIP run while SST in MOVE-C RA deviates from the observed SST. It is difficult to examine the mechanisms of all the improvements on the AMIP run here. We therefore focus our concentration on the precipitation in the western tropical North Pacific, especially in summer, since the precipitation is associated with the activity of tropical cyclones and may affect the weather around East Asia.

c. Activity of tropical cyclones and the summer monsoon trough

Figure 5 is a plot of the time series of monthly mean precipitation averaged in the area east of the Philippines (5°–20°N, 125°–150°E). From this figure, we find that MOVE-C improves the variance of precipitation in this area over the AMIP run. In particular, large precipitation in boreal summer is fairly underestimated in the AMIP run but is reproduced better in MOVE-C RA. The minima in the boreal winters of 1987 and 1993 in MOVE-C RA are also closer to CMAP. The rms difference (RMSD) between MOVE-C RA (the AMIP run) and CMAP is 2.37 (2.94) mm day−1. The RMSD for MOVE-C RA is smaller than that of the AMIP run with a significance level exceeding 95%. Thus, MOVE-C RA has a smaller error than the AMIP run on average.

Actually, the improvement of summer precipitation in this area in MOVE-C RA reflects the increase of tropical cyclones in the western tropical Pacific. Figure 6 presents the daily sea level pressure (SLP) and precipitation fields in July 1997 when the difference of precipitation between MOVE-C RA and the AMIP run is large in Fig. 5. In MOVE-C RA, we can find three low pressures corresponding to tropical cyclones. The first one is at 15°N, 130°E on 5 July, the second at 15°N, 145°E on 9 July, and the third at 22°N, 138°E on 29 July. These cyclones develop, and their central pressures decrease to less than 1000 hPa later. Substantial precipitation is then seen around those cyclones. In contrast, no tropical cyclone developed in the AMIP run, although a weak low pressure system is found at 22°N, 138°E on 25 July. Therefore, precipitation is relatively low in the Philippine Sea.

The increase of tropical cyclones in the boreal summer of 1997 is associated with the improved monsoon trough in MOVE-C RA. Figure 7 illustrates the SLP field and the wind field at 850 hPa averaged in the summer (June–August). In JRA-25, the area where SLP is less than 1008 hPa extends from the south of China to northeast of the Philippines and forms the monsoon trough. Southwest (southeast) wind prevails on the south (northeast) of the trough. Although the low pressure associated with the monsoon trough is also simulated in the AMIP run, the pressure is higher than in JRA-25. The sparse contours of SLP also imply that the horizontal pressure gradient is small. Southwest winds are hardly represented on the south of the trough, and the southeast wind is weakened northeast of it. The monsoon trough is thus not developed as well as in JRA-25. In MOVE-C RA, the horizontal pressure gradient is as large as in JRA-25, although the pressure is lower over the northeast of the Philippines. Southwest (southeast) wind is seen on the south (northeast) of the trough. Thus, the monsoon trough in the summer is better simulated in MOVE-C RA.

The improvement of the monsoon trough can also be found in the SLP climatology field in boreal summer (June–August, Fig. 8). In JRA-25, the low pressure area where the pressure is less than 1010 hPa extends from Southeast Asia to the east of the date line and forms the monsoon trough. This feature is well reconstructed in MOVE-C RA, although the pressure is slightly lower than in JRA-25. The low pressure area, however, is narrow and separated around 140°E in the AMIP run. This difference means the monsoon trough in the AMIP run is weaker than in JRA-25 and MOVE-C RA. This may be why fewer tropical cyclones were generated in the AMIP run.

The monsoon trough can be regarded as the eastern edge of the zonal Walker circulation and are therefore closely associated. Figure 8 also depicts the seasonal mean vertical shear of zonal winds between 850 and 200 hPa in the same season. This shear represents the strength of the zonal Walker circulation (e.g., Wang et al. 2003). In the AMIP run, the shear is smaller over the Indian Ocean and the Maritime Continent than in JRA-25. The shear exceeds 30 m s−1 in a large area of the northwestern Indian Ocean in JRA-25 but is limited to a small area between 50° and 60°E in the AMIP run. The area where the shear exceeds 20 m s−1 extends to 110°E in JRA-25, but it retreats to 90°E in the AMIP run. The area where the shear exceeds 10 m s−1 over the Maritime Continent is narrower in the AMIP run. These deficiencies are resolved in MOVE-C RA: the area where the shear exceeds 30 m s−1 is as large as in JRA-25, and the area where the shear exceeds 20 m s−1 extends to 120°E although the shear is larger around the Maritime Continent in MOVE-C RA. The zonal Walker circulation is therefore weak in the AMIP run but intensified and reconstructed more properly in MOVE-C RA. The lower westerly wind is thus amplified in the Indian Ocean and the western tropical Pacific, stimulating the monsoon trough in MOVE-C RA.

The interannual variability of the monsoon trough and the zonal Walker circulation is also improved in MOVE-C RA. To clearly demonstrate this improvement, we examine the yearly variation of the DU2 index proposed by Wang and Fan (1999) and the W–Y index proposed by Webster and Yang (1992). Both indices are calculated from the mean field of the zonal wind in boreal summer (June–August). The DU2 index is an anomaly associated with the difference between the zonal winds averaged in the square areas of 5°–15°N, 90°–130°E and of 22.5°–32.5°N, 110°–140°E from its climatology. It represents how well the monsoon trough developed in the year. The W–Y index is an anomaly related to the vertical shear of the zonal wind between 850 and 200 hPa averaged within 20°N–0°, 40°–120°E from its climatology. It represents how well the zonal Walker circulation is intensified during the year. We also define an index related to the precipitation in the area east of the Philippines (5°–20°N, 125°–150°E) in summer in the same manner: the index (EPP index, hereafter) is defined as an anomaly of the three-month mean precipitation in June–August in that area from its climatology.

The DU2 index calculated from JRA-25 tends toward high values before onsets of El Niños (e.g., 1981/82, 1985/86, 1990, 1997, 2001/02) and low values right after their terminations (e.g., 1983, 1988, 1998, 2003) as illustrated in Fig. 9a. The low values are caused by the anticyclones developing around the termination of El Niños (Wang et al. 2000, 2003). This yearly variation of the DU2 index is smaller in the AMIP run. In particular, the index is almost unchanged around 0 after 1995. The standard deviation (STD) of the index for the AMIP run is less than a half of that for JRA-25 (Table 1). This deficiency is improved in MOVE-C RA. The low values in 1998 and 2003 in JRA-25 are well estimated in MOVE-C RA. The high values in 1996, 2001, and 2004 are also recovered. The STD of the index for MOVE-C RA is comparable with that for JRA-25 (Table 1). The correlation coefficient of the index for MOVE-C RA with that for JRA-25 is higher than that for the AMIP run with a significance level exceeding 90% (Table 2). The yearly variation of the monsoon trough in the boreal summer is, thus, improved in MOVE-C RA.

Figure 9b indicates that the variation of the W–Y index in JRA-25 is loosely correlated with ENSO: the value tends to become low around El Niño periods. It then correlates with the DU2 index with a significance above 95% (Table 1), which demonstrates the connection of the zonal Walker circulation to the monsoon trough. However, it is difficult to find a connection between the variation of the W–Y index and ENSO in the AMIP run. The STD of the W–Y index is only 60% of that for JRA-25, and the correlation coefficient between the W–Y and DU2 indices is almost 0. In MOVE-C RA, the W–Y index is improved. For example, the low values in 1987 and 2003 and the high value in 2001 are better estimated. The index correlates with the DU2 index, and its STD is closer to that for JRA-25 (Table 1). The correlation with the index for JRA-25 is also higher than that for the AMIP run with a significance level above 90% (Table 2).

The EPP index for CMAP is well correlated with the DU2 index for JRA-25 (Fig. 9c, Table 1), from which we can reconfirm the close relation between the monsoon trough and precipitation in the Philippine Sea. There is also a correlation between the EPP and W–Y indices. These relations are reconstructed in MOVE-C RA although the STD of the EPP index for MOVE-C RA is smaller than that for CMAP. In contrast, there is no correlation between the EPP and W–Y indices in the AMIP run, although the EPP index is correlated with the DU2 index. Moreover, the STD of the EPP index is less than a half of that for CMAP. Comparing the correlation coefficient of the EPP index with that for CMAP, the estimation is better in MOVE-C RA with a significance level exceeding 90% (Table 2).

4. Discussion

In the previous section, we demonstrated that summer precipitation around the Philippine Sea is improved in MOVE-C RA and indicated that the improvement originates from the better reproduction of the zonal Walker circulation and the monsoon trough. We assume that the improvements stem from the reconstruction of the coupled variation between the near-surface ocean field and atmospheric convection in the tropical region in MOVE-C RA.

In the real world, high SST tends to stimulate convection, resulting in increased precipitation. There also exists a negative feedback mechanism: the increase of precipitation tends to cool SST because shortwave heating is reduced by cloud cover. The ocean mixing enhanced by the cyclone activity may also reduce SST. The change of precipitation in the tropical region then lags about one month behind SST because of this feedback, as described in Arakawa and Kitoh (2004).

To examine whether the negative feedback is activated in the AMIP run and MOVE-C RA, we present the map of the time lag of precipitation behind SST in those datasets, together with the equivalent map calculated from CMAP and COBE-SST, in Fig. 10. Areas where the significance level is less than 99% are blanked in these figures. From Fig. 10a, we can confirm that a one-month time lag exists between SST and precipitation in most parts of the Indian Ocean, western tropical Pacific, and northern tropical Atlantic.

However, this lag is not found in the AMIP run (Fig. 10c): SST and precipitation are basically correlated with no lag in most of the tropical region because the uncoupled atmosphere model lacks the negative feedback mechanism. In contrast, the negative feedback mechanism can be reestablished by the ocean model in MOVE-C, although it might be deformed by the ocean data assimilation, particularly by assimilating SST data. Figure 10b shows that a one-month lag of precipitation behind SST is found around the Indian Ocean and the western equatorial Pacific in MOVE-C RA, although the areas are smaller than in Fig. 10a. This lag indicates that the negative feedback mechanism works in MOVE-C.

This negative feedback has an important role in adjusting precipitation. In particular, the overestimated summer precipitation in the Bay of Bengal in the AMIP run (Fig. 3) is suppressed by the negative feedback in MOVE-C RA: the convection cools SST by reducing solar heating and enhancing ocean mixing, and the cooled SST, in turn, deactivates the convection in the coupled system. The reduction of the upward transport of the air mass by the convection in the western part of the Bay of Bengal increases the lower westerly wind over the eastern part of the Indian Ocean and the Maritime Continent, resulting in the improved monsoon trough (Fig. 7).

It also suppresses divergence in the upper troposphere over the Bay of Bengal. Figure 11 illustrates the velocity potential fields at 200 hPa in the boreal summer (June–August) of 1997 in the AMIP run and MOVE-C RA. In the AMIP run, the overestimated precipitation in the ocean east of India generates a spurious divergence maximum. Convective activity associated with tropical cyclones is then suppressed in the western tropical Pacific so as to compensate for the extra divergence. This divergence is weakened in MOVE-C RA, resulting in increased convective activity and the upper-layer divergence over the Philippine Sea. Finally, precipitation is increased there. It also intensifies the zonal contrast of the velocity potential in the upper troposphere and improves the zonal Walker circulation.

The negative feedback mechanism also causes decoupling between precipitation and SST. Figure 12 shows maps of the correlation coefficients between SST and precipitation in boreal summer (June–August). The correlation is positive in most of the Indian Ocean and western tropical Pacific in the AMIP run, which implies that SST variation strongly controls precipitation there. Wang et al. (2004) indicated that the positive correlation is also found in simulations of other atmosphere models forced by observed SST. In contrast, negative correlation is dominant in the western tropical Pacific in the map of the correlation coefficients calculated from observed data (CMAP and COBE-SST). Actually, summer precipitation in this region is likely to be controlled by nonlocal conditions such as the zonal Walker circulation and the monsoon trough rather than the local SST. Recent studies indicated that precipitation is also affected by westward propagation of the atmospheric Rossby waves (Wang et al. 2000, 2003) and the tropical Indian Ocean warming after El Niños (Xie et al. 2009). The negative correlation in the western equatorial Pacific implies that SST is inversely controlled by precipitation through the negative feedback: the enhanced (suppressed) convection induces an SST drop (rise). In addition, the correlation coefficient is around 0 for a large part of Indian Ocean. Precipitation is thus not controlled by SST there.

The spurious coupling between SST and precipitation in the AMIP run deforms the variation of precipitation, as discussed in Wang et al. (2004). For example, the vigorous convection decreases SST in the Philippine Sea in the summer of 1997 in the real world. The SST, given as a prescribed condition, automatically decreases without the convective activity in the AMIP run. The low SST suppresses precipitation there. The situation is opposite in the summer of 1998. High SST caused by drought in the real world increases precipitation in the AMIP run. This coupling between SST and precipitation is possibly why the W–Y index is not correlated with the DU2 and EPP indices in the AMIP run. A similar situation also occurred in the Indian Ocean.

This shortcoming is improved in the coupled model with ocean data assimilation (MOVE-C). Analysis increments are calculated from the difference between the ocean field of the half-month model prediction and the analysis. If suppressed (enhanced) convection is predicted by the model, SST is increased (decreased) at the end of the prediction, and the increments are decreased (increased). Thus, the increments reflect the effect of the convective activity, that is, the effect of the negative feedback mechanism from precipitation to SST. Therefore, the negative correlation between SST and precipitation is reproduced in MOVE-C RA (Fig. 12). The coupling of precipitation to SST is also weakened in the equatorial Indian Ocean and the northern part of the Bay of Bengal. Since MOVE-C recovers the monthly variation of SST with the assimilation of SST, the better distribution of the correlation coefficient implies that the variation of the precipitation is improved over the AMIP run.

Finally, the improvements in the atmospheric fields in MOVE-C RA are mostly because the negative feedback mechanism from precipitation to SST is reconstructed in MOVE-C RA, while it is destroyed in the AMIP run.

5. Summary

We developed MOVE-C, a data assimilation system in which ocean observation data are used to constrain the ocean part of a CGCM as a prototype of a truly coupled data assimilation system. This system is expected to reproduce climate fluctuations realistically because many of the slow variations in the coupled atmosphere–ocean system are surmised to be controlled by the ocean field. We also anticipate that MOVE-C can be a powerful tool for analyzing the effect of air–sea interaction because the interaction is explicitly estimated by the coupled model.

To explore the feasibility of MOVE-C, we conducted the MOVE-C Reanalysis (MOVE-C RA) experiment. In MOVE-C RA, SST and OHC biases found in the CGCM free simulation run (CGCM run) are effectively reduced, and the ocean climatology and variability is reproduced along with the current operational ocean data assimilation system in JMA (MOVE/MRI.COM-G). We then found that MOVE-C RA has an improved precipitation field over the AMIP run (a simulation of the atmosphere model in MOVE-C with observed daily SST) and the CGCM run. In particular, MOVE-C estimates precipitation around the Philippine Sea better than the AMIP run, especially in summer. Our analysis indicates that the improvement is caused by better representation of the monsoon trough and zonal Walker circulation in MOVE-C RA.

The improvements of the atmosphere fields in MOVE-C RA over the AMIP run are assumed to stem from the proper reproduction of the interaction between SST and precipitation. While warm (cold) SST tends to increase (decrease) precipitation, enhanced (suppressed) convection tends to induce a SST drop (rise) because of the cloud cover and ocean mixing. This negative feedback is lacking in the AMIP run. In contrast, the one-month lagged correlation of precipitation with SST in MOVE-C RA implies that this negative feedback works there. This feedback suppresses the overestimated precipitation in the western part of the Bay of Bengal found in the AMIP run, resulting in the intensification of the zonal Walker circulation and monsoon trough and promoting convective activity in the western tropical Pacific. The negative feedback also decouples SST and precipitation in MOVE-C RA, while coupling to SST deforms the change of precipitation in the AMIP run, as discussed in Wang et al. (2004). Simulating the ocean interior variation by the ocean model is thus essential for reproducing the climatological state and variability of the coupled atmosphere–ocean system. This conclusion is also consistent with Wang et al. (2005), who reported that CGCMs outperform uncoupled atmospheric models for forecasting of the summer monsoon.

This study has demonstrated that constraining ocean fields of a CGCM by assimilating ocean data is a potential way to reconstruct climate variability. It is also a candidate for the method of providing a coupled initial condition with a CGCM in seasonal and ENSO forecasting. In addition, it enables one to make an ocean reanalysis that does not depend on any atmospheric reanalysis data and, therefore, is not affected by artificial data gaps caused by transition of the atmospheric observing system. It may further improve atmosphere fields over atmospheric reanalyses in some aspects. For example, atmospheric reanalyses exhibit a spurious increasing trend in precipitation over the Indian Ocean, which is partly caused by the warming trend in the prescribed SST (Yamanaka 2008). This spurious trend is expected to be reduced in MOVE-C, which can appropriately deal with the air–sea interaction.

Assimilating ocean data alone, however, is not sufficient for reproducing the climate variability completely because not all climate phenomena are controlled by the ocean field alone. It should also be noted that state-of-the-art atmosphere models are not sophisticated enough to reflect the ocean variability correctly even if the ocean field is well reconstructed by assimilating ocean observation data. Inadequate reproduction of the atmosphere field may then cause degradation of the ocean field. MOVE-C is, therefore, not likely to be superior to the atmosphere and ocean reanalyses generated by uncoupled atmosphere and ocean models. It is thus essential to assimilate atmosphere observations additionally to ocean data for reconstructing climate variability more effectively in MOVE-C. It is also thought to provide better initial conditions for seasonal and ENSO forecasting. Therefore, our crucial target is to develop a truly coupled data assimilation system in which both atmosphere and ocean observations are assimilated into a CGCM.

Acknowledgments

We thank two reviewers for fruitful discussions about this paper. A part of this study was supported by the Grants-in-Aid for Science Research 19540469 from the Ministry of Education, Culture, Sports, Science and Technology, Japan.

REFERENCES

  • Arakawa, O., , and A. Kitoh, 2004: Comparison of local precipitation–SST relationship between the observation and a reanalysis dataset. Geophys. Res. Lett., 31 , L12206. doi:10.1029/2004GL020283.

    • Search Google Scholar
    • Export Citation
  • Balmaseda, M. A., , D. Dee, , A. Vialard, , and D. L. T. Anderson, 2007: A multivariate treatment of bias for sequential data assimilation: Application to the tropical oceans. Quart. J. Roy. Meteor. Soc., 133 , 167179.

    • Search Google Scholar
    • Export Citation
  • Bloom, S. C., , L. L. Takacs, , A. M. D. Silva, , and D. Ledvina, 1996: Data assimilation using incremental analysis updates. Mon. Wea. Rev., 124 , 12561271.

    • Search Google Scholar
    • Export Citation
  • CLS, 2004: SSALTO/DUACS user handbook: (M)SLA and (M)ADT near-real time and delayed time products. CLS-DOS-NT-04 103, Collecte Localisation Satellites (CLS), Toulouse, France, 42 pp.

    • Search Google Scholar
    • Export Citation
  • Conkright, M. E., and Coauthors, 2002: Introduction. Vol. 1, World Ocean Database 2001, NOAA Atlas NESDIS 42, 167 pp.

  • Fujii, Y., , and M. Kamachi, 2003: Three-dimensional analysis of temperature and salinity in the equatorial Pacific using a variational method with vertical coupled temperature-salinity empirical orthogonal function modes. J. Geophys. Res., 108 , 3297. doi:10.1029/2002JC001745.

    • Search Google Scholar
    • Export Citation
  • Fujii, Y., , S. Ishizaki, , and M. Kamachi, 2005: Application of nonlinear constraints in a three dimensional variational ocean analysis. J. Oceanogr., 61 , 655662.

    • Search Google Scholar
    • Export Citation
  • Hamilton, D., 1994: GTSPP builds an ocean temperature-salinity database. Earth Syst. Monit., 4 (4) 45.

  • Hayes, S. P., , L. J. Mangum, , J. Picaut, , A. Sumi, , and K. Takeuchi, 1991: TOGA-TAO: A moored array for real-time measurements in tropical Pacific Ocean. Bull. Amer. Meteor. Soc., 72 , 339347.

    • Search Google Scholar
    • Export Citation
  • Ishii, M., , A. Shouji, , S. Sugimoto, , and T. Matsumoto, 2005: Objective analyses of sea-surface temperature and marine meteorological variables for the 20th century using ICOADS and the Kobe collection. Int. J. Climatol., 25 , 865879.

    • Search Google Scholar
    • Export Citation
  • Ishikawa, I., , H. Tsujino, , M. Hirabara, , H. Nakano, , T. Yasuda, , and H. Ishizaki, 2005: Meteorological Research Institute Community Ocean Model (MRI.COM) manual. Tech. Rep. of the Meteorological Research Institute 47, Meteorological Research Institute, Tsukuba, Japan, 189 pp.

    • Search Google Scholar
    • Export Citation
  • Ishizaki, S., , S. Matsumoto, , Y. Fujii, , T. Yasuda, , and M. Kamachi, 2006: Correction of the zonal wind stress in the tropical Pacific using the ocean data assimilation system in Meteorological Research Institute. Proc. The Oceanographic Society of Japan 2006 Spring Meeting, Yokohama, Japan, The Oceanographic Society of Japan, 23.

    • Search Google Scholar
    • Export Citation
  • JMA, 2007: Outline of the operational numerical weather prediction at the Japan Meteorological Agency. Appendix to WMO Technical Progress Report on the Global Data-Processing and Forecasting System and Numerical Weather Prediction, Japan Meteorological Agency, Tokyo, Japan, 196 pp.

    • Search Google Scholar
    • Export Citation
  • Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77 , 437471.

  • Kanamitsu, M., and Coauthors, 2002: NCEP dynamical seasonal forecast system 2000. Bull. Amer. Meteor. Soc., 83 , 10191037.

  • Kuroda, Y., 2002: TRITON, present status and future plan. Report for the international workshop for review of the tropical moored buoy network, JAMSTEC, Yokosuka, Japan, 77 pp.

    • Search Google Scholar
    • Export Citation
  • Lin, J., 2007: The double-ITCZ problem in IPCC AR4 coupled GCMs: Ocean–atmosphere feedback analysis. J. Climate, 20 , 44974525.

  • Locarnini, R. A., , A. V. Mishonov, , J. I. Antonov, , T. P. Boyer, , and H. E. Garcia, 2006: Temperature. Vol. 1, World Ocean Atlas 2005, NOAA Atlas NESDIS 61, 182 pp.

    • Search Google Scholar
    • Export Citation
  • Luo, J-J., , S. Masson, , S. K. Behera, , and T. Yamagata, 2008: Extended ENSO predictions using a fully coupled ocean–atmospheric model. J. Climate, 21 , 8493.

    • Search Google Scholar
    • Export Citation
  • Madden, R. A., , and P. R. Julian, 1972: Description of global-scale circulation cells in the tropics with a 40–50 day period. J. Atmos. Sci., 29 , 11091123.

    • Search Google Scholar
    • Export Citation
  • McPhaden, M. J., and Coauthors, 1998: The Tropical Ocean-Global Atmosphere observing system: A decade of progress. J. Geophys. Res., 103 , (C7). 1416914240.

    • Search Google Scholar
    • Export Citation
  • Mizuta, R., and Coauthors, 2006: 20-km-mesh global climate simulations using JMA-GSM model—Mean climate states. J. Meteor. Soc. Japan, 84 , 165185.

    • Search Google Scholar
    • Export Citation
  • Onogi, K., and Coauthors, 2007: The JRA-25 Reanalysis. J. Meteor. Soc. Japan, 85 , 369432.

  • Stammer, D., 2007: Draft report of the first CLIVAR/GODAE meeting on ocean synthesis evaluation. International CLIVAR Project Office, 80 pp. [Available online at http://www.clivar.org/organization/gsop/synthesis/mit/synthesis_2.php].

    • Search Google Scholar
    • Export Citation
  • Sugiura, N., , T. Awaji, , S. Masuda, , T. Mochizuki, , T. Toyoda, , T. Miyama, , H. Igarashi, , and Y. Ishikawa, 2008: Development of a four-dimensional variational coupled data assimilation system for enhanced analysis and prediction of seasonal to interannual climate variations. J. Geophys. Res., 113 , C10017. doi:10.1029/2008JC004741.

    • Search Google Scholar
    • Export Citation
  • Tsujino, H., , and T. Yasuda, 2004: Formation and circulation of mode waters of the North Pacific in a high-resolution GCM. J. Phys. Oceanogr., 34 , 399415.

    • Search Google Scholar
    • Export Citation
  • Usui, N., , S. Ishizaki, , Y. Fujii, , H. Tsujino, , T. Yasuda, , and M. Kamachi, 2006: Meteorological Research Institute Multivariate Ocean Variational Estimation (MOVE) system: Some early results. Adv. Space Res., 37 , 806822.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , and Z. Fan, 1999: Choice of South Asian summer monsoon indices. Bull. Amer. Meteor. Soc., 80 , 629638.

  • Wang, B., , R. Wu, , and X. Fu, 2000: Pacific–East Asian teleconnection: How does ENSO affect East Asian climate? J. Climate, 13 , 15171536.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , R. Wu, , and T. Li, 2003: Atmosphere–warm ocean interaction and its impacts on Asian–Australian monsoon variation. J. Climate, 16 , 11951211.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , I-S. Kang, , and J-Y. Lee, 2004: Ensemble simulations of Asian–Australian monsoon variability by 11 AGCMs. J. Climate, 17 , 803818.

    • Search Google Scholar
    • Export Citation
  • Wang, B., , Q. Ding, , X. Fu, , I-S. Kang, , K. Jin, , J. Shukla, , and F. Doblas-Reyes, 2005: Fundamental challenge in simulation and prediction of summer monsoon rainfall. Geophys. Res. Lett., 32 , L15711. doi:10.1029/2005GL022734.

    • Search Google Scholar
    • Export Citation
  • Webster, P. J., , and S. Yang, 1992: Monsoon and ENSO: Selectively interactive systems. Quart. J. Roy. Meteor. Soc., 118 , 877926.

  • Xie, P., , and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78 , 25392558.

    • Search Google Scholar
    • Export Citation
  • Xie, S. P., , K. Hu, , J. Hafner, , H. Tokinaga, , Y. Du, , G. Huang, , and T. Sampe, 2009: Indian Ocean capacitor effect on Indo-Western Pacific climate during the summer following El Niño. J. Climate, 22 , 730747.

    • Search Google Scholar
    • Export Citation
  • Yamanaka, G., 2008: Discrepancies between observed and ocean general circulation model–simulated anomalies in recent SSTs of the tropical Indian Ocean caused by apparent trends in atmospheric reanalysis data. Geophys. Res. Lett., 35 , L18603. doi:10.1029/2008GL034737.

    • Search Google Scholar
    • Export Citation
  • Yasuda, T., , Y. Takaya, , C. Kobayashi, , M. Kamachi, , H. Kamahori, , and T. Ose, 2007: Asian monsoon predictability in JMA/MRI Seasonal Forecast System. CLIVAR Exchanges, No. 43, International CLIVAR Project Office, Southampton, United Kingdom, 18–24.

    • Search Google Scholar
    • Export Citation
  • Zhang, S., , M. J. Harrison, , A. T. Wittenberg, , and A. Rosati, 2005: Initialization of an ENSO forecast system using a parallelized ensemble filter. Mon. Wea. Rev., 133 , 31763201.

    • Search Google Scholar
    • Export Citation

APPENDIX

Ocean Heat Content in MOVE-C RA

In this appendix, we compare the monthly ocean heat content (OHC) fields of MOVE-C RA with those of MOVE-G RA07, the CGCM run, and World Ocean Atlas 2005 (WOA05) (Locarnini et al. 2006) to confirm that MOVE-C realistically reproduces the mean state and variability of the oceanic field. MOVE-G RA07 is an updated version of the global ocean reanalysis introduced in Usui et al. (2006). It is produced with MOVE/MRI.COM-G, the uncoupled global ocean data assimilation system in JMA. Here, surface fluxes from the atmosphere are based on National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis (NCEP R1) (Kalnay et al. 1996). The zonal momentum flux around the equatorial Pacific is adjusted so as to be balanced with the observed zonal pressure gradient by multiplying by a prescribed factor, 1.4 (Ishizaki et al. 2006). The same temperature and salinity profiles and SSH data as in MOVE-C RA, including SST data in WOD01 and GTSPP, are assimilated in MOVE-G RA07 except that COBE-SST is not employed. The interval of calculating analysis increments is one month and is also the same as in MOVE-C RA. The accuracy of ocean temperature and salinity fields in MOVE-G RA07 is as high as that of state-of-the-art ocean data assimilation products, as confirmed in the Climate Variability and Predictability/Global Synthesis and Observations Panel (CLIVAR/GSOP) project (Stammer 2007).

Figure A1 depicts the monthly climatology of vertically averaged temperature in 0–300 m in MOVE-C RA, MOVE-G RA07, and the CGCM run. We regard the averaged temperature as OHC in the upper ocean here. The color shading in this figure indicates the deviation from OHC calculated from WOA05. OHC as well as SST is estimated higher in most of the region in the CGCM run. However, the deviation is reduced in MOVE-C RA again.

It should be noted that the deviations in MOVE-C RA and MOVE-G RA07 have similar patterns in January and July. In January a cold deviation is dominant in the Pacific and South Atlantic. The eastern tropical Indian Ocean also has a cold deviation in both reanalyses. Warm deviations are found around the frontal zone in the Southern Ocean and south of the western boundary currents in the North Atlantic and North Pacific, that is, the Gulf Stream and Kuroshio. It should be noted that the climatologies of MOVE-C RA and MOVE-G RA07 do not necessarily correspond to the field of WOA05 because it is not climatology of the specific period but, rather, reconstructed with available data observed in the past. Regions with large horizontal gradients are apt to have particularly large deviations. The deviations around the Gulf Stream and Kuroshio further imply that those fronts, which are artificially smoothed in WOA05, are improved in both data assimilation results.

The deviation fields in July also have a cold deviation in a large part of the Pacific and warm deviations around the frontal zones in both reanalyses. They also have warm deviations in the eastern tropical Pacific. The only notable difference between the two reanalyses is that MOVE-C RA alone has a dipole-pattern bias in the tropical Indian Ocean in July. This pattern appears because the westerly wind stress given to the ocean in MOVE-C RA is stronger there than the forcing of MOVE-G RA07.

The monthly variation of OHC is also reconstructed in MOVE-C RA as well as in MOVE-G RA07 (Fig. A2). A large part of the model region has an ACC larger than 0.8. The ACC is relatively small in the Indian Ocean, eastern South Pacific, South Atlantic, and Southern Ocean. These correspond to the areas where there are fewer available observations. Figure A3 depicts the longitude–time sections of the OHC (0–300-m averaged temperature) fields in 2000–05 (this period is chosen for highlighting short variations) in MOVE-C RA and MOVE-G RA07. This figure demonstrates that the high-frequency variability of the subsurface temperature field is not estimated satisfactorily. This deficiency reflects the fact that the intraseasonal variability, including the disturbance of the Madden–Julian oscillation (Madden and Julian 1972), is not well represented in the atmospheric model of MOVE-C because of its coarse resolution. MOVE-C, however, reproduces the slow variability of the temperature field as well as in MOVE-G RA07.

Summarizing this appendix, there are some deficiencies in MOVE-C, but it reproduces the variability of the OHC field satisfactorily enough that the coupled ocean–atmosphere fluctuation in MOVE-C RA is worth evaluating.

Fig. 1.
Fig. 1.

Monthly climatology of SST (°C) in (left) January and (right) July for (top) MOVE-C RA, (middle) MOVE-G RA07, and (bottom) the CGCM run. Contours indicate the climatology, and color shading, the deviation from the climatology of COBE-SST.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 2.
Fig. 2.

ACC maps of SST in MOVE-C RA with COBE-SST.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 3.
Fig. 3.

Monthly climatology of precipitation (mm day−1) in (left) January and (right) July for (a) CMAP, (b) MOVE-C RA, (c) the AMIP run, and (d) the CGCM run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 4.
Fig. 4.

ACC maps of precipitation with CMAP in boreal winter, spring, summer, and fall in the tropical Pacific and Indian Ocean for (a) MOVE-C RA and (b) the AMIP run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 5.
Fig. 5.

Time series of precipitation (mm day−1) in the area east of the Philippines (5°–20°N, 125°–150°E) for MOVE-C RA (black solid line), AMIP run (gray solid line), and CMAP (black dashed line).

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 6.
Fig. 6.

Daily mean SLP (contour, hPa) and daily precipitation (color shading, mm day−1) every four days in July 1997 for (left) MOVE-C RA and (right) the AMIP run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 7.
Fig. 7.

SLP field (contour, hPa) and wind field at 850 hPa (arrow) averaged in the summer of 1997 for (a) JRA-25, (b) MOVE-C RA, and (c) the AMIP run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 8.
Fig. 8.

Climatological fields of SLP (contour, hPa) and zonal wind shear between 850 and 200 hPa (shading, m s−1) in summer for (a) JRA-25, (b) MOVE-C RA, and (c) the AMIP run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 9.
Fig. 9.

Yearly variations of (a) the DU2 index, (b) the W–Y index, and (c) the EPP index for MOVE-C RA (black solid line with open circles), the AMIP run (gray solid line with closed circle), and JRA-25 in (a) and (b) and CMAP in (c) (black dashed line with open square). “E” denotes an El Niño period defined by JMA.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 10.
Fig. 10.

Maps of time lags (month) of precipitation change behind SST for (a) CMAP and COBE-SST, (b) MOVE-C RA, and (c) the AMIP run. Positions where the significance level is less than 99% are blank.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 11.
Fig. 11.

Velocity potential fields (m2 s−1) at 200 hPa averaged in summer 1997 for MOVE-C RA (black contour) and the AMIP run (gray contour). Color shading represents the difference between MOVE-C RA and AMIP run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Fig. 12.
Fig. 12.

Maps of the correlation between SST and precipitation in summer (June–August) for (a) CMAP and COBE-SST, (b) MOVE-C RA, and (c) the AMIP run.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

i1520-0442-22-20-5541-fa01

Fig. A1. As in Fig. 1, but for OHC (°C): color shading represents the deviation of OHC calculated from WOA05.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

i1520-0442-22-20-5541-fa02

Fig. A2. Map of ACC between MOVE-C RA and MOVE-G RA07 for OHC.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

i1520-0442-22-20-5541-fa03

Fig. A3. Time–longitude section of monthly OHC anomaly (°C) at the equator in the Pacific between 2000 and 2005 for (left) MOVE-C RA and (right) MOVE-G RA07. Solid (dashed) lines indicate contours of positive (negative) values.

Citation: Journal of Climate 22, 20; 10.1175/2009JCLI2814.1

Table 1.

STDs of indices and correlations among indices for MOVE-C RA, the AMIP run, and JRA-25 (DU2 and W–Y indices) or CMAP (EPP index). Asterisks indicate that the level of significance is under 95%.

Table 1.
Table 2.

Correlation of indices for MOVE-C RA and the AMIP run with that for JRA-25 (DU2 and W–Y indices) or CMAP (EPP index), and the significance level for the hypothesis that the correlation coefficient for MOVE-C RA is larger than that for the AMIP run.

Table 2.
Save