• Arakawa, A., , and W. H. Shubert, 1974: Interaction of a cumulus ensemble with the large-scale environment, Part I. J. Atmos. Sci., 31, 674704, doi:10.1175/1520-0469(1974)031<0674:IOACCE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bechtold, P., , M. Köhler, , T. Jung, , F. Doblas-Reyes, , M. Leutbecher, , M. J. Rodwell, , F. Vitart, , and G. Balsamo, 2008: Advances in simulating atmospheric variability with the ECMWF model: From synoptic to decadal time-scales. Quart. J. Roy. Meteor. Soc., 134, 13371351, doi:10.1002/qj.289.

    • Search Google Scholar
    • Export Citation
  • Benedict, J. J., , and D. A. Randall, 2007: Observed characteristics of the MJO relative to maximum rainfall. J. Atmos. Sci., 64, 23322354, doi:10.1175/JAS3968.1.

    • Search Google Scholar
    • Export Citation
  • Benedict, J. J., , and D. A. Randall, 2009: Structure of the Madden–Julian oscillation in the superparameterized CAM. J. Atmos. Sci., 66, 32773296, doi:10.1175/2009JAS3030.1.

    • Search Google Scholar
    • Export Citation
  • Bernie, D. J., , E. Guilyardi, , G. Madec, , J. M. Slingo, , S. J. Woolnough, , and J. Cole, 2008: Impact of resolving the diurnal cycle in an ocean–atmosphere GCM. Part 2: A diurnally coupled CGCM. Climate Dyn., 31, 909925, doi:10.1007/s00382-008-0429-z.

    • Search Google Scholar
    • Export Citation
  • Fu, X., , and B. Wang, 2009: Critical roles of stratiform rainfall in sustaining the Madden–Julian Oscillation: GCM experiments. J. Climate, 22, 39393959, doi:10.1175/2009JCLI2610.1.

    • Search Google Scholar
    • Export Citation
  • Fu, X., , B. Yang, , Q. Bao, , and B. Wang, 2008: Sea surface temperature feedback extends the predictability of tropical intraseasonal oscillation. Mon. Wea. Rev., 136, 577597, doi:10.1175/2007MWR2172.1.

    • Search Google Scholar
    • Export Citation
  • Fu, X., , W. Wang, , J.-Y. Lee, , B. Wang, , K. Kikuchi, , J. Xu, , J. Li, , and S. Weaver, 2015: Distinctive roles of air–sea coupling on different MJO events: A new perspective revealed from the DYNAMO/CINDY field campaign. Mon. Wea. Rev., 143, 794–812, doi:10.1175/MWR-D-14-00221.1.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterization. Mon. Wea. Rev., 121, 764787, doi:10.1175/1520-0493(1993)121<0764:PEOAUB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Han, J., , and H. L. Pan, 2011: Revision of convection and vertical diffusion schemes in the NCEP Global Forecast System. Wea. Forecasting, 26, 520533, doi:10.1175/WAF-D-10-05038.1.

    • Search Google Scholar
    • Export Citation
  • Huang, B., , M. L’Heureux, , J. Lawrimore, , C. Liu, , H.-M. Zhang, , V. Banzon, , Z.-Z. Hu, , and A. Kumar, 2013: Why did large differences arise in the sea surface temperature datasets across the tropical Pacific during 2012? J. Atmos. Oceanic Technol., 30, 29442953, doi:10.1175/JTECH-D-13-00034.1.

    • Search Google Scholar
    • Export Citation
  • Hung, M.-P., , J.-L. Lin, , W. Wang, , D. Kim, , T. Shinoda, , and S. J. Weaver, 2013: MJO and convectively coupled equatorial waves simulated by CMIP5 climate models. J. Climate, 26, 61856124, doi:10.1175/JCLI-D-12-00541.1.

    • Search Google Scholar
    • Export Citation
  • Jiang, X., and Coauthors, 2009: Vertical heating structures associated with the MJO as characterized by TRMM estimates, ECMWF reanalyses, and forecasts: A case study during 1998/99 winter. J. Climate, 22, 60016020, doi:10.1175/2009JCLI3048.1.

    • Search Google Scholar
    • Export Citation
  • Joyce, R. J., , J. E. Janowiak, , P. A. Arkin, , and P. Xie, 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487503, doi:10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kemball-Cook, S., , B. Wang, , and X. Fu, 2002: Simulation of the intraseasonal oscillation in the ECHAM-4 model: The impact of coupling with an ocean model. J. Atmos. Sci., 59, 14331453, doi:10.1175/1520-0469(2002)059<1433:SOTIOI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kim, H.-M., , C. D. Hoyos, , P. J. Webster, , and I.-S. Kang, 2008: Sensitivity of MJO simulation and predictability to sea surface temperature variability. J. Climate, 21, 53045317, doi:10.1175/2008JCLI2078.1.

    • Search Google Scholar
    • Export Citation
  • Kim, H.-M., , P. J. Webster, , V. E. Toma, , and D. Kim, 2014: Predictability and prediction skill of the MJO in two operational forecasting systems. J. Climate, 27, 53645378, doi:10.1175/JCLI-D-13-00480.1.

    • Search Google Scholar
    • Export Citation
  • Klingaman, N. P., , P. M. Inness, , H. Weller, , and J. M. Slingo, 2008: The importance of high-frequency sea surface temperature variability to the intraseasonal oscillation of Indian monsoon rainfall. J. Climate, 21, 61196140, doi:10.1175/2008JCLI2329.1.

    • Search Google Scholar
    • Export Citation
  • Klingaman, N. P., , S. J. Woolnough, , H. Weller, , and J. M. Slingo, 2011: The impact of finer-resolution air–sea coupling on the intraseasonal oscillation of the Indian monsoon. J. Climate, 24, 24512468, doi:10.1175/2010JCLI3868.1.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., , L. Zhang, , and W. Wang, 2013: Sea surface temperature–precipitation relationship in different reanalyses. Mon. Wea. Rev., 141, 11181123, doi:10.1175/MWR-D-12-00214.1.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , and L. Peng, 1987: Origin of low frequency (intraseasonal) oscillations in the tropical atmosphere. Part I: The basic theory. J. Atmos. Sci., 44, 950972, doi:10.1175/1520-0469(1987)044<0950:OOLFOI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Liebmann, B., , and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 12751277.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., and Coauthors, 2006: Tropical intraseasonal variability in 14 IPCC AR4 climate models. Part I: Convective signals. J. Climate, 19, 26652690, doi:10.1175/JCLI3735.1.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., , B. Mapes, , M. Zhang, , and M. Newman, 2004: Stratiform precipitation, vertical heating profiles, and the Madden–Julian oscillation. J. Atmos. Sci., 61, 296309, doi:10.1175/1520-0469(2004)061<0296:SPVHPA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., , M.-I. Lee, , D. Kim, , I.-S. Kang, , and D. M. W. Frierson, 2008: The impacts of convective parameterization and moisture triggering on AGCM-simulated convectively coupled equatorial waves. J. Climate, 21, 883909, doi:10.1175/2007JCLI1790.1.

    • Search Google Scholar
    • Export Citation
  • Madden, R. A., , and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122, 814837, doi:10.1175/1520-0493(1994)122<0814:OOTDTO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Matthews, A. J., 2004: Atmospheric response to observed intraseasonal tropical sea surface temperature anomalies. Geophys. Res. Lett., 31, L14107, doi:10.1029/2004GL020474.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., , and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 9781002, doi:10.1175/1520-0493(1992)120<0978:RASAPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., , and M. J. Suarez, 1999: Documentation of version 2 of relaxed Arakawa–Schubert cumulus parameterization with convective downdrafts. NOAA Office Note 99-01, 44 pp.

  • Pan, H.-L., , and W.-S. Wu, 1995: Implementing a mass flux convection parameterization package for the NMC medium-range forecast model. NMC Office Note 409, 39 pp.

  • Reynolds, R. W., , N. A. Rayner, , T. M. Smoth, , D. C. Stokes, , and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 16091625, doi:10.1175/1520-0442(2002)015<1609:AIISAS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., , T. M. Smith, , C. Liu, , D. B. Chelton, , K. S. Casey, , and M. G. Schlax, 2007: Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 54735496, doi:10.1175/2007JCLI1824.1.

    • Search Google Scholar
    • Export Citation
  • Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc.,91, 1015–1057, doi:10.1175/2010BAMS3001.1.

  • Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 21852208, doi:10.1175/JCLI-D-12-00823.1.

  • Seo, K.-H., , and W. Wang, 2010: The Madden–Julian oscillation simulated in the NCEP Climate Forecast System Model: The importance of stratiform heating. J. Climate, 23, 47704793, doi:10.1175/2010JCLI2983.1.

    • Search Google Scholar
    • Export Citation
  • Shelly, A., , P. Xavier, , D. Copsey, , T. Johns, , J. M. Rodriguez, , S. Milton, , and N. Klingaman, 2014: Coupled versus uncoupled hindcast simulations of the Madden–Julian Oscillation in the Year of Tropical Convection. Geophys. Res. Lett., 41, 56705677, doi:10.1002/2013GL059062.

    • Search Google Scholar
    • Export Citation
  • Shinoda, T., , H. H. Hendon, , and J. Glick, 1998: Intraseasonal variability of surface fluxes and sea surface temperature in the tropical western Pacific and Indian Oceans. J. Climate, 11, 16851702, doi:10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Takahashi, M., 1987: A theory of slow phase speed of the intraseasonal oscillation using the wave-CISK. J. Meteor. Soc. Japan, 65, 4349.

    • Search Google Scholar
    • Export Citation
  • Thiébaux, J., , E. Rogers, , W. Wang, , and B. Katz, 2003: A new high-resolution blended real-time global sea surface temperature analysis. Bull. Amer. Meteor. Soc., 84, 645656, doi:10.1175/BAMS-84-5-645.

    • Search Google Scholar
    • Export Citation
  • Tian, B., , D. E. Waliser, , E. J. Fetzer, , B. H. Lambrigtsen, , Y. L. Yung, , and B. Wang, 2006: Vertical moist thermodynamic structure and spatial–temporal evolution of the MJO in AIRS observations. J. Atmos. Sci., 63, 24622485, doi:10.1175/JAS3782.1.

    • Search Google Scholar
    • Export Citation
  • Tiedtke, M., , W. A. Heckley, , and J. Slingo, 1988: Tropical forecasting at ECMWF: The influence of physical parametrization on the mean structure of forecasts and analyses. Quart. J. Roy. Meteor. Soc., 114, 639664, doi:10.1002/qj.49711448106.

    • Search Google Scholar
    • Export Citation
  • Tokioka, T., , K. Yamazaki, , A. Kitoh, , and T. Ose, 1988: The equatorial 30–60 day oscillation and the Arakawa–Schubert penetrative cumulus parameterization. J. Meteor. Soc. Japan, 66, 883901.

    • Search Google Scholar
    • Export Citation
  • Vitart, F., , S. Woolnough, , M. A. Balmaseda, , and A. M. Tompkins, 2007: Monthly forecasts of the Madden–Julian oscillation using a coupled GCM. Mon. Wea. Rev., 135, 27002715, doi:10.1175/MWR3415.1.

    • Search Google Scholar
    • Export Citation
  • Waliser, D. E., , K. M. Lau, , and J.-H. Kim, 1999: The influence of coupled SSTs on the Madden–Julian oscillation: A model perturbation experiment. J. Atmos. Sci., 56, 333358, doi:10.1175/1520-0469(1999)056<0333:TIOCSS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wang, W., , and M. Schlesinger, 1999: The dependence on convection parameterization of the tropical intraseasonal oscillation simulated by the UIUC 11-layer atmospheric GCM. J. Climate, 12, 14231457, doi:10.1175/1520-0442(1999)012<1423:TDOCPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wang, W., , M.-P. Hung, , S. J. Weaver, , A. Kumar, , and X. Fu, 2014: MJO prediction in the NCEP Climate Forecast System version 2. Climate Dyn., 42, 25092520, doi:10.1007/s00382-013-1806-9.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , C. L. Gentemann, , D. K. Smith, , and D. B. Chelton, 2000: Satellite measurements of sea-surface temperature through clouds. Science, 288, 847850, doi:10.1126/science.288.5467.847.

    • Search Google Scholar
    • Export Citation
  • Woolnough, S. J., , J. M. Slingo, , and B. J. Hoskins, 2000: The relationship between convection and sea surface temperature on intraseasonal timescales. J. Climate, 13, 20862104, doi:10.1175/1520-0442(2000)013<2086:TRBCAS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Woolnough, S. J., , F. Vitart, , and M. A. Balmaseda, 2007: The role of the ocean in the Madden–Julian oscillation: Implications for MJO prediction. Quart. J. Roy. Meteor. Soc., 133, 117128, doi:10.1002/qj.4.

    • Search Google Scholar
    • Export Citation
  • Wu, M., , S. Schubert, , I. Kang, , and D. Waliser, 2002: Forced and free intraseasonal variability over the South Asian monsoon region simulated by 10 AGCMs. J. Climate, 15, 28622880, doi:10.1175/1520-0442(2002)015<2862:FAFIVO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zeng, X., , M. Zhao, , R. E. Dickinson, , and Y. He, 1999: A multi-year hourly sea surface skin temperature dataset derived from the TOGA TAO bulk temperature and wind speed over the tropical Pacific. J. Geophys. Res., 104, 15251536, doi:10.1029/1998JC900060.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., , and M. Mu, 2005: Simulation of the Madden–Julian Oscillation in the NCAR CCM3 using a revised Zhang–McFarlane convection parameterization scheme. J. Climate, 18, 40464064, doi:10.1175/JCLI3508.1.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., , and X. Song, 2009: Interaction of deep and shallow convection is key to Madden–Julian Oscillation simulation. Geophys. Res. Lett., 36, L09708, doi:10.1029/2009GL037340.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    (a) Observed 10°S–10°N average OLR anomalies (W m−2). (b) As in (a), but for NCDC SST analysis (K). (c) As in (a), but for TMI SST (K). (d) Standard deviation for NCDC SST (blue curve) and TMI SST (red curve). (e) Lag correlation between OLR and SST anomalies for NCDC SST (blue curve) and TMI SST (red curve).

  • View in gallery

    Forecast at 12-day time of 10°S–10°N average OLR anomalies (W m−2) forced by climatology SSTs using (a) the SAS2 scheme, (b) the SAS scheme, and (c) the RAS scheme.

  • View in gallery

    As in Fig. 2, but for the forecast forced by NCDC SSTs.

  • View in gallery

    As in Fig. 2, but for the forecast forced by TMI SSTs.

  • View in gallery

    Statistics of 10°S–10°N average OLR anomalies from the forecast experiments: standard deviation (W m−2) for forecasts with (a) SAS2, (b) SAS, and (c) RAS; root-mean-square error (W m−2) for forecasts with (d) SAS2, (e) SAS, and (f) RAS; correlation with observations for forecasts with (g) SAS2, (h) SAS, and (i) RAS. Forecasts with climatology, NCDC, and TMI SSTs are plotted with solid black, red, and blue curves, respectively. The observed standard deviation is plotted with black dotted curve in (a),(b), and (c).

  • View in gallery

    Lag regression of the OLR anomalies against TMI SST anomalies (W m−2 K−1) as a function of forecast lead time (y axis) and the time that the OLR lags the SST (x axis). (a) Observation, (b) forecast with SAS2, (c) forecast with SAS, and (d) forecast with RAS.

  • View in gallery

    As in Fig. 6, but for surface net downward shortwave flux anomalies.

  • View in gallery

    As in Fig. 6, but for surface downward latent heat flux anomalies.

  • View in gallery

    Forecast precipitation anomalies (mm day−1) from 7 Nov 2011 with (a),(d),(g),(j) climatology SSTs; (b),(e),(h), NCDC SSTs; and (c),(f),(i), TMI SSTs using (a)–(c) SAS2, (d)–(f) SAS, and (g)–(i) RAS. SST anomalies are shown for (k) NCDC analysis and (l) TMI analysis.

  • View in gallery

    Evolution of total values of variables averaged over 10°S–10°N and 60°–80°E from the forecasts with SAS2 (red curves) and RAS (blue curves) schemes initialized from 7 Nov 2011. (a) Precipitation (mm day−1), (b) evaporation (mm day−1), (c) vertical column integrated moisture convergence (mm day−1), and (d) atmospheric precipitation water (mm). The TMI SST (K) is plotted in each panel with a black curve. The left y axis is used for all variables except for the SST, which follows the right y axis.

  • View in gallery

    Differences in the evolution in specific humidity (g kg−1) averaged over 10°S–10°N and 60°–80°E between the forecast with SAS2 scheme and that with RAS scheme initialized from 7 Nov 2011. Positive values means the forecast with RAS is wetter than that with SAS2.

  • View in gallery

    Convective heating rate (K day−1) averaged over 10°S–10°N and 60°–80°E in the forecast initialized from 7 Nov 2011. (a) Deep convection with SAS2, (b) deep convection with RAS, (c) differences in deep convection between SAS2 and RAS, (d) shallow convection with SAS2, (e) shallow convection with RAS, and (f) difference in shallow convection between SAS2 and RAS. The heating rate in (a),(b),(d), and (e) is plotted at −1, −0.5, 0.5, 1, 2, 4, 6, and 8 K day−1. The differences in (c) and (f) are plotted as SAS2 minus RAS at −2, −1.5, −1, −0.5, 0.5, 1, and 2 K day−1.

  • View in gallery

    Moistening rate (g kg−1 day−1) averaged over 10°S–10°N and 60°–80°E in the forecast initialized from 7 Nov 2011. (a) Deep convection with SAS2, (b) deep convection with RAS, (c) difference between (a) and (b), (d) shallow convection with SAS2, (e) shallow convection with RAS, (f) difference between (c) and (d), (g) large-scale condensation with SAS2, (h) large-scale condensation with RAS, (i) difference between (g) and (h), (j) vertical diffusion with SAS2, (k) vertical diffusion with RAS, and (l) difference between (j) and (k). The moistening rate in (a),(b),(d),(e),(g),(h),(j), and (k) is plotted at −1.2, −0.8, −0.4, −0.2, −0.1, 0.1, 0.2, 0.4, 0.8, and 1.2 g kg−1 day−1. The differences in (c),(f),(i), and (l) are plotted as SAS2 minus RAS at −0.8, −0.4, −0.2, −0.1, 0.1, 0.2, 0.4, and 0.8 g kg−1 day−1.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 21 21 7
PDF Downloads 6 6 3

What is the Role of the Sea Surface Temperature Uncertainty in the Prediction of Tropical Convection Associated with the MJO?

View More View Less
  • 1 Climate Prediction Center, NOAA/NWS/NCEP, College Park, Maryland
  • 2 International Pacific Research Center, School of Ocean and Earth Science and Technology, University of Hawai‘i at Mānoa, Honolulu, Hawaii
  • 3 Climate Prediction Center, NOAA/NWS/NCEP, College Park, Maryland
© Get Permissions
Full access

Abstract

This study investigated the influence of the uncertainty in the sea surface temperature (SST) on the representation of the intraseasonal rainfall variability associated with the Madden–Julian oscillation (MJO) and how this influence varies with convection parameterization. The study was motivated by the fact that there exist substantial differences in observational SST analyses, and by the possibility that lacking sufficient accuracy for SSTs in dynamical models may degrade the MJO simulation and prediction. Experiments for the DYNAMO intensive observing period were carried out using the NCEP atmospheric Global Forecast System (GFS) with three convection schemes forced by three SST specifications. The SST specifications included the widely used National Climatic Data Center (NCDC) daily SST analysis, the TRMM Microwave Imager (TMI) SST retrieval, and an SST climatology that only contains climatological seasonal cycle.

The experiments show that for all convection schemes, the advantage of using observed (TMI and NCDC) SSTs over the climatology SSTs can be seen as early as 5 days to 1 week after the start of the forecast. Further, the prediction with TMI SSTs was more skillful than that with the NCDC SSTs, indicating that the current level of SST uncertainties in the observational analyses can lead to large differences when they are used as the lower boundary conditions. The results suggest that the simulation and prediction can be improved with an atmosphere-only model forced by more accurate SSTs, or with a coupled atmosphere–ocean model that has a more realistic representation of the SST variability. Differences in the prediction among the convection schemes are also presented and discussed.

Corresponding author address: Wanqiu Wang, NOAA/Center for Weather and Climate Prediction, 5830 University Research Court, Room 3004, College Park, MD 20740. E-mail: wanqiu.wang@noaa.gov

Abstract

This study investigated the influence of the uncertainty in the sea surface temperature (SST) on the representation of the intraseasonal rainfall variability associated with the Madden–Julian oscillation (MJO) and how this influence varies with convection parameterization. The study was motivated by the fact that there exist substantial differences in observational SST analyses, and by the possibility that lacking sufficient accuracy for SSTs in dynamical models may degrade the MJO simulation and prediction. Experiments for the DYNAMO intensive observing period were carried out using the NCEP atmospheric Global Forecast System (GFS) with three convection schemes forced by three SST specifications. The SST specifications included the widely used National Climatic Data Center (NCDC) daily SST analysis, the TRMM Microwave Imager (TMI) SST retrieval, and an SST climatology that only contains climatological seasonal cycle.

The experiments show that for all convection schemes, the advantage of using observed (TMI and NCDC) SSTs over the climatology SSTs can be seen as early as 5 days to 1 week after the start of the forecast. Further, the prediction with TMI SSTs was more skillful than that with the NCDC SSTs, indicating that the current level of SST uncertainties in the observational analyses can lead to large differences when they are used as the lower boundary conditions. The results suggest that the simulation and prediction can be improved with an atmosphere-only model forced by more accurate SSTs, or with a coupled atmosphere–ocean model that has a more realistic representation of the SST variability. Differences in the prediction among the convection schemes are also presented and discussed.

Corresponding author address: Wanqiu Wang, NOAA/Center for Weather and Climate Prediction, 5830 University Research Court, Room 3004, College Park, MD 20740. E-mail: wanqiu.wang@noaa.gov

1. Introduction

Previous studies have shown that the development and evolution of the observed eastward-propagating Madden–Julian oscillation (MJO; Madden and Julian 1994) include contributions from internal atmospheric dynamics and atmosphere–ocean interactions. The atmospheric dynamics are a result of the interaction between the convection and large-scale circulation. In particular, the onset of the convection is preceded by a preconditioning of warm and moist lower troposphere (Tian et al. 2006; Benedict and Randall 2007). It also has been shown that the propagation speed of the MJO may depend on the vertical profile of the diabatic heating (Lau and Peng 1987; Takahashi 1987). In addition, the formation of stratiform precipitation is suggested to have an impact on the evolution of the MJO (Lin et al. 2004; Fu and Wang 2009; Jiang et al. 2009).

Numerical simulations have also demonstrated the importance of convection in the MJO dynamics. It has been shown that the simulated MJO depends on the criteria for the onset of the convection, such as the convection entrainment rate and critical relative humidity (Tokioka et al. 1988; Wang and Schlesinger 1999; Zhang and Mu 2005; Bechtold et al. 2008; Lin et al. 2008). Use of different convection schemes may also result in different levels of fidelity in the representation of the MJO in general circulation models (Seo and Wang 2010). The existence of shallow convection was also found to be important in the preconditioning of the development of the deep convection associated with the MJO (Zhang and Song 2009). These studies suggest that the simulated MJO can be affected by different aspects in the convection parameterization.

Another process that may affect the MJO is the air–sea interaction. Observational diagnoses have demonstrated the relationship among the MJO activity, surface heat fluxes, and the underlying sea surface temperature (SST; Woolnough et al. 2000; Shinoda et al. 1998; Kumar et al. 2013). Numerical studies have demonstrated that the spatial coherence and propagation of the MJO are much improved in the simulations with coupled atmosphere–ocean models compared to that with atmosphere-only models (Waliser et al. 1999; Kemball-Cook et al. 2002). The inclusion of air–sea coupling has been shown to extend the MJO predictability and enhance prediction skill of the tropical intraseasonal oscillation (Vitart et al. 2007; Fu et al. 2008; Shelly et al. 2014).

While the inclusion of air–sea coupling improves the simulation and prediction of the MJO, there still exist systematic errors in the MJO in coupled atmosphere–ocean models. The MJO produced in free simulations from most of the contemporary coupled atmosphere–ocean models is generally weaker and more stationary than that observed (Lin et al. 2006; Hung et al. 2013). Similar deficiencies are also found in the initialized MJO prediction from operational systems with amplitude that is too weak and propagation that is too slow for the MJO (Wang et al. 2014; Kim et al. 2014). The current operational MJO prediction is produced with forecast systems that adopt different treatments for the ocean. For example, the NCEP CFSv2 includes an interactive oceanic component starting from the beginning of the prediction (Wang et al. 2014) while the ECMWF monthly forecast system persists initial SST anomalies until day 10 after which an interactive ocean component is used (Kim et al. 2014).

The oceanic variable that directly interacts with the atmosphere is the SST. Improvements in the MJO simulation due to the inclusion of air–sea interaction suggest that the SST not only changes in response to atmospheric forcing but also feedbacks to influence the atmospheric circulation at the intraseasonal time scales. Accordingly, a realistic representation of the MJO may not only need the use of coupled atmosphere–ocean models but also require that SST variations associated with the MJO are correctly simulated. The assessment of the role of SSTs in the observed MJO variability and its use to validate coupled atmosphere–ocean models requires accurate SST observations.

Global observational SST analyses used to force the atmospheric models and to validate coupled models, however, may contain substantial uncertainties. One source of such uncertainties in SST analyses results from the analysis procedure, for example, how the bias in satellite retrieval is determined (Huang et al. 2013). Another source of the uncertainties in SST analyses may arise from the definition of the SST. There exist two types of SSTs that are used in climate and weather analyses. The first type is the conventional bulk SST, which represents the temperature of the top few meters of the ocean and is measured by in situ buoys and ships. Such a bulk-temperature definition has been adopted in some widely used SST analyses such as the NCEP optimum interpolation weekly analyses (Reynolds et al. 2002), NCEP 2DVAR daily analysis (Thiébaux et al. 2003), and NCDC optimum interpolation daily analysis (Reynolds et al. 2007). The second type is the skin SST, which represents the temperature at the surface and can be directly retrieved from satellite observations. The bulk SST and skin SST can be significantly different especially when the surface winds are weak and the sky is clear (Zeng et al. 1999). It has been argued that the use of conventional treatment of the SST as the mean bulk temperature may have underestimated the amplitude of the intraseasonal SST variability (Woolnough et al. 2007; Bernie et al. 2008; Klingaman et al. 2011). Most of the contemporary global coupled atmosphere–ocean models use a vertical resolution of 10 m or so for the upper ocean, which may be too coarse to produce a realistic SST variability associated with the MJO. It is possible that such a coarse vertical resolution for the upper ocean used in the global coupled atmosphere–ocean models could underestimate SST intraseasonal variability and adversely affect the simulation of the MJO in the long-term integration of the models.

As an example to show differences between different SST analyses, Fig. 1 compares SST anomalies from the NCDC daily analysis (Reynolds et al. 2007) (Fig. 1b) and Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI; Wentz et al. 2000) (Fig. 1c) during the DYNAMO campaign intensive observation period (IOP; 1 October 2011–15 January 2012). Both of SST datasets have been widely used in climate studies. Three eastward-propagating MJO events initiated in the Indian Ocean in October, November, and December 2011, respectively, were observed during this period as revealed in observed outgoing longwave radiation (OLR) anomalies (Fig. 1a). The TMI analysis showed clear positive SST anomalies (Fig. 1c) leading each of these three events. Such leading warm SST anomalies are less clear in the NCDC analysis (Fig. 1b). The November warm SST anomalies in NCDC analysis are not as well organized as those in the TMI dataset. Further, the comparison of the SST standard deviation during the DYNAMO IOP period indicates that the overall amplitude of the SSTs in NCDC analysis is weaker than that in TMI analysis (Fig. 1d).

Fig. 1.
Fig. 1.

(a) Observed 10°S–10°N average OLR anomalies (W m−2). (b) As in (a), but for NCDC SST analysis (K). (c) As in (a), but for TMI SST (K). (d) Standard deviation for NCDC SST (blue curve) and TMI SST (red curve). (e) Lag correlation between OLR and SST anomalies for NCDC SST (blue curve) and TMI SST (red curve).

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Is the current level of the SST uncertainties in observational analyses sufficient to make significant differences if they are used to force an initialized forecast? To what extent do the errors in MJO prediction depend on the uncertainties in the underlying SSTs? Can the prediction of the MJO with a coupled model be improved if the predicted SSTs became more accurate? Does the conclusion about the role of SSTs vary with the use of different convection parameterizations? These are the questions that are addressed in this study based on a suite of forecast experiments with an atmosphere-only model using different convection schemes and specifications of SSTs. Conclusions based on atmospheric model experiments will also shed light on the accuracy requirements for SST prediction in coupled models, and on the requirement for vertical discretization in the upper layers of the ocean component of the coupled models. Previous studies have shown that the response of the atmospheric models to SSTs depends on the sampling frequency of the SSTs with daily SSTs resulting in more realistic response compared to weekly mean and monthly mean SSTs (Fu et al. 2008; Kim et al. 2008; Klingaman et al. 2008). Daily SSTs will be used in this study. A few studies have also indicated the importance of the diurnal SST variations (Bernie et al. 2008), which will not be addressed.

The model and experiments are described in section 2. Analysis on the role of SSTs on MJO prediction is given in section 3 followed by a discussion in section 4. A summary and conclusions are provided in section 5.

2. The model and experiments

a. The model

The model used in this study is the 2011 version of the NCEP operational atmospheric GFS at T126 horizontal resolution with 64 vertical levels. The model physics are configured to be the same as the atmospheric component of NCEP CFSv2 (Saha et al. 2014) except that, in addition to the convection scheme used in CFSv2, two other built-in convection schemes are also used for the experiments. These convection schemes are the following:

  1. Simplified Arakawa–Schubert (SAS) cumulus convection (Pan and Wu 1995), which was used in the NCEP Climate Forecast System (CFS; Saha et al. 2014). The SAS scheme is based on Arakawa and Schubert (Arakawa and Shubert 1974) and simplified by Grell (1993) to consider only one cloud instead of a spectrum of clouds. Convection occurs when the cloud work function exceeds a certain threshold. A simple trigger is employed, which requires that the level of free convection must exist and must be within the distance of 150 hPa of the parcel starting level.
  2. Relaxed Arakawa–Schubert (RAS) cumulus convection in GFS is developed by Moorthi and Suarez (1992, 1999). The RAS scheme simplifies the entrainment relation and assumes that the normalized mass flux is a linear function of height rather than being exponential as in the original AS scheme. In addition, rather than requiring that “quasi equilibrium” of the cloud ensemble be achieved each time, the scheme only relaxes the ambient atmospheric state toward equilibrium.
  3. Simplified Arakawa–Schubert version 2 (SAS2) convection scheme (Han and Pan 2011), which has been used in the NCEP operational GFS since 2011. The SAS2 scheme is modified from its earlier version (SAS). Instead of using a fixed distance of 150 hPa, the convection trigger in SAS2 uses a distance range of 120–180 hPa in proportion to the large-scale vertical velocity. Unlike the old SAS scheme, the revised SAS scheme specifies finite entrainment and detrainment rates for heat, moisture, and momentum above the cloud base following Bechtold et al. (2008).

There are two shallow convection (SC) schemes. The first SC scheme uses a simple turbulent eddy diffusion approach with a specified eddy diffusivity profile for the transport of sensible heat and moisture within convectively unstable layers, following the procedure proposed by Tiedtke et al. (1988). This diffusion SC scheme does not generate precipitation and is used in combination with SAS and RAS for deep convection. The second shallow convection scheme is a bulk mass-flux parameterization scheme developed based on the simplified Arakawa–Schubert cumulus convection (Han and Pan 2011). This mass-flux SC scheme allows precipitation and is used in combination with SAS2 for deep convection.

b. The experiments

Initialized forecast experiments with GFS are carried out for the IOP of the DYNAMO field campaign (1 October 2011–15 January 2012) using three SST datasets. Two of the SST datasets are taken from observational analyses: one from the NCDC daily analysis (Reynolds et al. 2007) and the other from the TMI daily analysis (Wentz et al. 2000). Anomalies of SSTs in these two datasets are shown in Figs. 1b and 1c. The third SST condition is taken as the 1999–2010 climatology of NCDC analysis and thus only contains the seasonal cycle without any anomalous SSTs. Three convection schemes (SAS2, SAS, and RAS) are used for the experiments forced by three SST conditions (climatology, NCDC, and TMI), resulting in a total of nine forecast experiments (Table 1).

Table 1.

List of forecast experiments.

Table 1.

Each experiment consists of forecast runs initialized each day during the 107 days of IOP of the DYNAMO field campaign. For each initial date, four runs are made from 0000, 0600, 1200, and 1800 UTC and initial conditions at these four time levels are directly taken from the Climate Forecast System Reanalysis (CFSR; Saha et al. 2010). Each forecast run covers 31 target days. In our analysis, the definition of lead time is such that the lead time of 1 day (31 days) corresponds to day 1 (day 31) forecast.

c. Analysis method

The four daily forecast runs are used to form an ensemble and the ensemble mean is used for diagnoses. The performance of each experiment in predicting the observed variability is analyzed using the daily-mean total field as well as the intraseasonal anomaly. Using F(d, L) to represent the total daily-mean forecast field F from initial day d (1–107 days) for lead time L (1–31 days). The intraseasonal anomaly is defined as , where is a background representing seasonal and interannual variability. This background is computed as a second-order polynomial fit over 107 days for each lead time. This analysis focuses on the forecast of the convective activity as represented by the OLR. Forecast precipitation, evaporation, atmospheric moisture convergence, atmospheric precipitable water, and specific humidity are also used for additional diagnoses. Observational data used in this study include daily SST analyses from NCDC (Reynolds et al. 2007) and TMI (Wentz et al. 2000), OLR from the National Oceanic and Atmospheric Administration (NOAA) polar-orbiting series of satellites (Liebmann and Smith 1996), rainfall estimate from the CPC morphing technique (CMORPH) satellite retrieval (Joyce et al. 2004), and surface shortwave and latent heat fluxes from the CFSR (Saha et al. 2010).

3. Forecasts of the convection from the experiments

In this section, we present the results from the experiments to show the impact of SSTs and the dependence on convection parameterizations. The experiments are assessed by comparing equatorial OLR anomalies for a specific lead time (day 12) and statistics of the anomaly amplitude, errors, and correlation for the entire range of the lead time. Differences among the experiments are further discussed in the next section.

Forecasts of equatorial (10°S–10°N average) OLR anomalies at day 12 with climatological SSTs are shown in Fig. 2. The anomalies are too weak with all three convection schemes, with the experiment using SAS2 being the weakest. It is interesting to see that the experiments perform better for the October and December events with relatively stronger negative eastward-propagating OLR anomalies (indicating enhanced convection) than the November event. However, the zonal scale of the forecast anomalies is too large in late October because of the enhanced convection (negative OLR anomalies) over the Maritime Continent and far-western Pacific (100°–140°E), especially for experiments with SAS and RAS schemes. The November event in all experiments with climatology SSTs is barely discernable, although this event is the strongest and lasts longest among the three events. This is consistent with the results of Fu et al. (2015) who find that stronger air–sea coupling is involved in the November event than the October and December events.

Fig. 2.
Fig. 2.

Forecast at 12-day time of 10°S–10°N average OLR anomalies (W m−2) forced by climatology SSTs using (a) the SAS2 scheme, (b) the SAS scheme, and (c) the RAS scheme.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

When the NCDC SSTs are used, the evolution of OLR anomalies are improved with better organized eastward propagation (Fig. 3) compared the experiments with climatology SSTs (Fig. 2). The clearest improvement with NCDC daily SSTs is that the enhanced convection associated with the November event is better captured in experiments with all convection schemes. Another improvement is the enhanced unrealistic convection over the Maritime Continent and far-western Pacific (100°–140°E) in late October in the experiments with climatology SSTs (Fig. 2) is largely reduced in the experiments with NCDC experiments (Fig. 3). As in the climatology experiments, the SAS and RAS schemes produce stronger convection variability.

Fig. 3.
Fig. 3.

As in Fig. 2, but for the forecast forced by NCDC SSTs.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

The forecasts are further improved with TMI SST analysis with stronger amplitude of OLR anomalies compared to the experiments with NCDC SSTs (Figs. 3 and 4). Although the SAS2 and SAS experiments (Figs. 4a and 4b) are still too weak, the anomaly amplitude in RAS experiment (Fig. 4c) is quite comparable to the observed. The organization of the eastward propagation is also improved, especially for the October and November events.

Fig. 4.
Fig. 4.

As in Fig. 2, but for the forecast forced by TMI SSTs.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Forecast OLR anomaly standard deviation, root-mean-square error (RMSE), and correlation with the observation for the tropics (10°S–10°N) from the Indian Ocean to the western Pacific (50°–150°E) are used for a quantitative comparison among the experiments (Fig. 5). The statistics are calculated as a function of lead time (1–31 days). Since the period of initial dates is fixed from 1 October 2011 to 15 January 2012, the target period at different lead times varies with the lead time. For example, the target period for 1-day lead time is 1 October 2011–15 January 2012 and that for 31-day lead time is 31 October 2011–14 February 2012. Major features from the comparison among the experiments include the following:

  • Differences in the statistics among experiments become clear after the forecast integration of 5 days to 1 week. The statistics are similar among the experiments within the first few days and differences can be seen as early as in 5 days to 1 week. For example, within the first 5 days, the correlation skill is similar for all convection schemes and for all three SST datasets (climatology, NCDC, and TMI), indicating that the skill is primarily from the atmospheric initial state and the ocean surface condition is not important within the first few days. The impact of underlying SSTs is more clearly seen after the first week.
  • The anomalies are strongest with the TMI SST (Figs. 5a–c). For SAS2 and RAS experiments, the standard deviation is larger with NCDC SST than those with climatology SSTs. The SAS scheme produces comparable standard deviation with climatology SSTs and NCDC SSTs. The standard deviation with SAS2 and SAS schemes using all three types of SSTs is weaker than the observed for the entire range of lead time. For the experiments with RAS, the standard deviation is stronger than the observed for the first 9, 10, and 15 days, respectively, with the climatology, NCDC, and TMI SSTs, and weaker than the observed thereafter. Among convection schemes, the standard deviation is weakest with the SAS2 scheme and strongest with the RAS scheme.
  • The RMSE with the TMI SST is smaller than that with NCDC and climatology SSTs for SAS and RAS schemes (Figs. 5e and 5f). For SAS2, the RMSE is comparable among the three SSTs. The RMSE is similar with climatology and NCDC SSTs. Among convection schemes, the RMSE is smallest with the SAS scheme and largest with the RAS scheme. This dependence of RMSE on the convection scheme is due to the differences in the amplitude of the anomalies (Figs. 5a–c) with larger standard deviation generally corresponding to a larger RMSE.
  • The correlation skill is highest with the TMI SSTs and generally lowest with the climatology SSTs for all convection schemes (Figs. 5g–i). The only exception to this in the forecast after one week is that the skill is similar between the experiment with NCDC SST and that with climatology SST for the lead time from 19 to 26. Across the convection schemes, the correlation skill is similar among the convection schemes with the NCDC and climatology SSTs. When the TMI SST analysis is used, both RAS and SAS schemes have higher correlation skills than the SAS2 scheme. The forecast correlation skill with SAS and RAS is comparable for the first 15 days after which the skill with RAS becomes better. Table 2 gives a comparison among the experiments using the lead time when the correlation becomes less than 0.3. Such a lead time is about 8.5 days with climatology SST and is slightly increased to 9 days or so with NCDC SSTs. A larger increase is seen between the experiments with NCDC SST and those with TMI SSTs.
Fig. 5.
Fig. 5.

Statistics of 10°S–10°N average OLR anomalies from the forecast experiments: standard deviation (W m−2) for forecasts with (a) SAS2, (b) SAS, and (c) RAS; root-mean-square error (W m−2) for forecasts with (d) SAS2, (e) SAS, and (f) RAS; correlation with observations for forecasts with (g) SAS2, (h) SAS, and (i) RAS. Forecasts with climatology, NCDC, and TMI SSTs are plotted with solid black, red, and blue curves, respectively. The observed standard deviation is plotted with black dotted curve in (a),(b), and (c).

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Table 2.

The lead time when OLR correlation becomes less than 0.3.

Table 2.

4. Discussion

These results show that the representation of convection at the intraseasonal time scales depends on the underlying SSTs and that the level of the uncertainty in the current SST analyses can make significant differences in the atmospheric simulation when these SST analyses are used to force an atmospheric model. The stronger OLR amplitude and improved skill (smaller RMSE and higher correlation) using TMI SSTs compared to those using NCDC SSTs are due to the SST differences between NCDC and TMI analyses (Figs. 1b–e). The TMI SST anomalies are better organized, leading the eastward-propagating OLR anomalies, especially for the November event for which the NCDC SSTs are more loosely distributed in longitude and time (Fig. 1b) compared to TMI SSTs (Fig. 1c). In addition, the tropical SST intraseasonal variability in the TMI analysis is stronger than that in NCDC analysis throughout the Indian Ocean and far western Pacific (Fig. 1d). On average the SST standard deviation in TMI analysis is about 39% stronger than that in NCDC analysis. The TMI SST also shows stronger relationship with the observed OLR with a strongest correlation of −0.65 when the observed OLR lags the TMI SST by 9 days compared to the strongest correlation of −0.48 when the observed OLR lags the NCDC SST by 7 days (Fig. 1e). The numerical experiments in this study demonstrate that these differences between observational SST analyses can cause significant differences in the atmospheric response.

Another result from these experiments is that the differences in the atmospheric response to the SST uncertainties vary with the model physics. In particular, while all convection schemes produce improved convective anomalies with TMI SSTs compared to the use of NCDC and climatology SSTs, such an improvement in terms of RMSE and correlation is most significant with the RAS scheme. Next we discuss the variations in atmospheric response to SST anomalies with convection schemes by comparing the results from RAS and SAS2 schemes. Since the largest differences between convection schemes are found when TMI SSTs are used, we will focus on the experiments with TMI SSTs.

Local lag regression of the observed and forecast OLR against TMI SST is calculated to quantify the relationship these two variables as a function of forecast lead time and the time that the OLR lags the SST (Fig. 6). The calculation is based on spatial average over 10°S–10°N and 70°–90°E. Notice that for all forecast lead times the initial dates are always 1 October 2011–15 January 2012. The lead time determines what target period of the forecasted and the corresponding observed OLR is used, and the time OLR lagging SST determines the period of observed SSTs. For example, the regression value for 5-day forecast lead time and 3-day OLR lag time is computed using the forecast OLR for 5 October 2011–19 January 2012 (at the same 5-day forecast lead time) and the TMI SST for 2 October 2011–16 January 2012. The observed OLR is rearranged in a same structure as the forecast for a consistent comparison.

Fig. 6.
Fig. 6.

Lag regression of the OLR anomalies against TMI SST anomalies (W m−2 K−1) as a function of forecast lead time (y axis) and the time that the OLR lags the SST (x axis). (a) Observation, (b) forecast with SAS2, (c) forecast with SAS, and (d) forecast with RAS.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

The regression from the observations shows positive regression values near 0-day OLR lag time, indicating that the warm SST anomaly peak generally corresponds to the weakest convection activity (Fig. 6a). The strongest observed negative OLR anomalies lag the SST by about 8–9 days. There are small variations of the regression based on observation with lead time, which is due to the gradual change of target period with lead time. The regression in the forecast experiments shows similar variations with OLR lag time to the observed, with strongest convection at the 8–9-day lag. The amplitude in the forecast is comparable to the observed near the beginning of the forecast, indicating that when the atmospheric state is close to the nature, all schemes are capable of reproducing the observed precipitation variability. However, the regression amplitude generally decreases with lead time in forecasts with all schemes and the decrease is larger with the SAS2 scheme (Fig. 6b) compared to that with the RAS and SAS schemes. This suggests that convection variability in the model becomes less accurate with increasing lead time as the error in the initial atmospheric state grows and the error growth varies with the model physical parameterizations.

Surface fluxes in the forecast runs are important variables in the evolution of the MJO, as they would mediate SST variations if the atmospheric model were coupled to an ocean component. Lag regression of surface downward heat flux anomalies from the CFSR and forecast against TMI SST is shown in Fig. 7 for the shortwave flux and Fig. 8 for the latent heat flux. The strongest negative anomalies of the surface shortwave flux in the CFSR lag the SST by 9 days or so (Fig. 7a). The amplitude of the regression of the surface shortwave flux in the forecast runs decreases with the lead time and becomes weaker than that in the CFSR after the first 7, 15, and 12 days in the forecast runs with the SAS2, SAS, and RAS schemes (Fig. 7). These lag relationships between the SST and shortwave flux in the CFSR and forecast runs are consistent with that of the observed and forecasted OLR in Fig. 6.

Fig. 7.
Fig. 7.

As in Fig. 6, but for surface net downward shortwave flux anomalies.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Fig. 8.
Fig. 8.

As in Fig. 6, but for surface downward latent heat flux anomalies.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

The regression of surface latent heat flux in the CFSR shows a lag of about 10 days of the strongest evaporative cooling after warm SST anomalies (Fig. 8a). This lag is maintained in the forecast runs for the first week or so (Figs. 8b–d). After the first week, the regression becomes too week in the forecast with SAS2 scheme (Fig. 8b) and the lag becomes too short (about 4 days) in the forecast with SAS and RAS schemes (Figs. 8c and 8d).

Atmospheric responses to intraseasonal SST anomalies have been investigated in some previous studies based on integrations with atmospheric general circulation models (AGCMs) forced with observational SST estimates (Wu et al. 2002; Matthews 2004). Based on simulations with an ensemble of 10 AGCMs, Wu et al. (2002) showed that the SST-forced convection response in their simulations was approximately in phase with the underlying SST, which also implies an in-phase relationship between warm SST anomalies and negative anomalies of the downward surface shortwave flux due to the associated cloud impact on the radiation. A 4-day lag between warm SST and enhanced convection was found in the AGCM modeling study of Matthews (2004), which also showed a 4-day lag between warm SST anomalies and negative downward shortwave and latent heat fluxes.

Our analysis differs from the studies of Wu et al. (2002) and Matthews (2004) in that these previous studies were based on SST-forced long simulations that did not have observed atmospheric information from the initial conditions while our forecast runs are initialized from an observational reanalysis that essentially incorporates the atmospheric state of the observed coupled atmosphere–ocean system. Accordingly, the lags between SST and the OLR and surface heat fluxes in our forecast runs are overall closer to the observed than those found in the studies of Wu et al. (2002) and Matthews (2004), especially for the first week in the forecast runs. However, the short 4-day lag between SST and the following evaporative cooling after the first week in the forecasts with SAS and RAS schemes (Figs. 8c and 8d) represents a bias of the model. The reason for this bias is not clear and requires further investigation in future studies.

The use of forecasts initialized from the same observational reanalysis allows an examination of the differences in the evolution of the related atmospheric fields. To investigate how the forecasts with different SSTs and convection schemes diverge, Fig. 9 compares 10°S–10°N average precipitation anomalies in the forecasts initialized from 7 November (Figs. 9a–i). A corresponding estimate from the CMORPH satellite retrieval (Joyce et al. 2004) is shown in Fig. 9j for an observational reference. The forecast target period from this initial date covers the initiation of the convection and the subsequent evolution of the November MJO event (Fig. 1a). In both NCDC and TMI SST analyses, warm SST anomalies started to develop in the western Indian Ocean after the initialization of the forecast, which expanded eastward and strengthened with an increasing lead time (Figs. 9k and 9l). The TMI anomalies are generally stronger in the Indian Ocean and in particular, the TMI shows negative anomalies over most of the Indian Ocean after 23 November, which are barely seen in the NCDC analysis except for some negative anomalies near 60° and 100°E.

Fig. 9.
Fig. 9.

Forecast precipitation anomalies (mm day−1) from 7 Nov 2011 with (a),(d),(g),(j) climatology SSTs; (b),(e),(h), NCDC SSTs; and (c),(f),(i), TMI SSTs using (a)–(c) SAS2, (d)–(f) SAS, and (g)–(i) RAS. SST anomalies are shown for (k) NCDC analysis and (l) TMI analysis.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Evolutions of rainfall anomalies in all forecast experiments within the first week are similar to the CMORPH estimate regardless which SST datasets or which convection schemes are used. After the first 10 days, precipitation anomalies with climatology SSTs are either very weak (Fig. 9a with the SAS2 scheme) or primarily stationary to the west of 60°E (Fig. 9d with the SAS scheme and Fig. 9g with the RAS scheme). The use of NCDC SSTs results in enhanced and more sustained precipitation anomalies, but the anomalies are largely localized near 70°E with the SAS2 scheme or near 60°E with the SAS and RAS schemes without clear eastward propagation (Figs. 9b,e,h).

With TMI SSTs, the forecasted rainfall anomalies (Figs. 9c,f,i) become more comparable to the CMORPH estimate with generally stronger amplitude and better organized eastward propagation compared to the forecasts with NCDC SSTs. There exist two differences in the precipitation anomalies associated with the November event between the forecasts with SAS2 (Fig. 9c) and the forecasts with SAS and RAS (Figs. 9f and 9i). One difference is that the enhanced precipitation anomalies are produced earlier in the SAS2 scheme than those in the SAS and RAS schemes. The SAS2 scheme started producing weak positive precipitation anomalies near 60°E on 11 November, which remain stationary until 15 November and expand eastward thereafter. Both SAS and RAS schemes produced precipitation anomalies after 12 November that expanded and propagated continuously. The other difference is that the amplitude of the precipitation anomalies in the SAS and RAS schemes is much larger than that in the SAS2 scheme. It is noted that the amplitude in RAS scheme is much larger than that in SAS scheme.

To further understand the precipitation differences between convection schemes, we analyze temporal evolution of moisture budgets averaged over 10°S–10°N and 60°–80°E from the experiments with TMI SSTs, where the enhanced precipitation anomalies are relatively strong in all convection schemes (Figs. 9c,f,i). Figure 10 shows a comparison between SAS2 and RAS schemes in precipitation (Prec), evaporation (Evap), moisture convergence integrated over the entire atmospheric column (Convq), and precipitable water (Pwat) for the forecast initialized from 7 November 2011. The temporal moisture variation is largely a balance between Prec and Convq. Evap is less changeable in time compared to Prec and Convq. A clear difference between SAS2 and RAS is that Pwat with the RAS scheme is consistently larger than that with SAS2 after 9 November, indicating an overall drier atmosphere in the SAS2 experiments than RAS experiments. Figure 11 shows that development of such a consistently drier atmosphere with SAS2 scheme is probably associated with the higher precipitation rate with SAS2 scheme during the first five days (7–11 November) than that with RAS scheme (Fig. 10a). Vertical distribution of the specific humidity shows that the dryness with SAS2 scheme relative to RAS scheme is mostly in the lower troposphere above the planetary boundary layer. Previous studies have shown that a moist lower troposphere preconditioning is critical for the development of deep convection (Benedict and Randall 2009; Zhang and Song 2009). The dryness of the lower troposphere in the SAS2 forecast may result in less intense convection and is possibly responsible for the forecast of weaker rainfall variability associated with the MJO.

Fig. 10.
Fig. 10.

Evolution of total values of variables averaged over 10°S–10°N and 60°–80°E from the forecasts with SAS2 (red curves) and RAS (blue curves) schemes initialized from 7 Nov 2011. (a) Precipitation (mm day−1), (b) evaporation (mm day−1), (c) vertical column integrated moisture convergence (mm day−1), and (d) atmospheric precipitation water (mm). The TMI SST (K) is plotted in each panel with a black curve. The left y axis is used for all variables except for the SST, which follows the right y axis.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Fig. 11.
Fig. 11.

Differences in the evolution in specific humidity (g kg−1) averaged over 10°S–10°N and 60°–80°E between the forecast with SAS2 scheme and that with RAS scheme initialized from 7 Nov 2011. Positive values means the forecast with RAS is wetter than that with SAS2.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

To investigate the impact of the use of different convection schemes, we compare the heating and moistening rate averaged over 10°S–10°N and 60°–80°E due to deep convection and shallow convection (SC). Convective heating rate in the forecasts with the SAS2 and RAS schemes initialized from 7 November 2011 is shown in Fig. 12 for the first 11 days. From 7 to 12 November when the SST started to warm up (Fig. 10), the SAS2 deep convection generated stronger heating than the RAS scheme, especially in the middle to upper troposphere (Figs. 12a–c), consistent with the larger rainfall rate in SAS2 than that in RAS (Fig. 10a). As the SST continued to warm up after 12 November (Fig. 10), the RAS convective heating became more intense starting from lower troposphere and expanding to middle and upper troposphere with time. The SAS2 SC constantly produced a warming near 900 mb (1 mb = 1 hPa) while the SC scheme used with RAS produced a heating near 900 mb as well as a cooling above 800 mb, resulting in a stronger warming in SAS2 between 700 and 800 mb (Figs. 12d–f).

Fig. 12.
Fig. 12.

Convective heating rate (K day−1) averaged over 10°S–10°N and 60°–80°E in the forecast initialized from 7 Nov 2011. (a) Deep convection with SAS2, (b) deep convection with RAS, (c) differences in deep convection between SAS2 and RAS, (d) shallow convection with SAS2, (e) shallow convection with RAS, and (f) difference in shallow convection between SAS2 and RAS. The heating rate in (a),(b),(d), and (e) is plotted at −1, −0.5, 0.5, 1, 2, 4, 6, and 8 K day−1. The differences in (c) and (f) are plotted as SAS2 minus RAS at −2, −1.5, −1, −0.5, 0.5, 1, and 2 K day−1.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

Consistent with the stronger heating during the first few days, the SAS2 deep convection induced larger drying from 7 to 12 November in the middle and upper troposphere between 600 and 200 mb (Figs. 13a–c). There existed a stronger drying near 800 mb in the SAS2 forecast during the first 9 days (from 7 to 15 November). After 12 November, the drying in the SAS2 forecast became weaker around 750 mb before 15 November and in most of the middle and upper troposphere after 15 November, indicating weaker convection in the SAS2 forecast than that in the RAS forecast with RAS scheme after the warming up of the SST (Fig. 10). The differences between SAS2 and RAS forecasts in Fig. 13c also showed a stronger moistening (drying) near 900 mb (the surface) in SAS2.

Fig. 13.
Fig. 13.

Moistening rate (g kg−1 day−1) averaged over 10°S–10°N and 60°–80°E in the forecast initialized from 7 Nov 2011. (a) Deep convection with SAS2, (b) deep convection with RAS, (c) difference between (a) and (b), (d) shallow convection with SAS2, (e) shallow convection with RAS, (f) difference between (c) and (d), (g) large-scale condensation with SAS2, (h) large-scale condensation with RAS, (i) difference between (g) and (h), (j) vertical diffusion with SAS2, (k) vertical diffusion with RAS, and (l) difference between (j) and (k). The moistening rate in (a),(b),(d),(e),(g),(h),(j), and (k) is plotted at −1.2, −0.8, −0.4, −0.2, −0.1, 0.1, 0.2, 0.4, 0.8, and 1.2 g kg−1 day−1. The differences in (c),(f),(i), and (l) are plotted as SAS2 minus RAS at −0.8, −0.4, −0.2, −0.1, 0.1, 0.2, 0.4, and 0.8 g kg−1 day−1.

Citation: Monthly Weather Review 143, 8; 10.1175/MWR-D-14-00385.1

SC moistening showed a similar pattern in SAS2 and RAS forecasts with a drying (moistening) below (above) 900 mb (Figs. 13d and 13e), with the moistening rate in the SAS2 forecast being weaker than that in the RAS forecast (Fig. 13f). Further, there was an enhancement in the moistening after 11 November in RAS forecast, possibly in association with the warming SST (Fig. 10). Such an enhancement did not exist in the SAS2 forecast. Similar to the differences in deep convective moistening (Fig. 13c), the SC in SAS2 forecast produced a stronger moistening (drying) near 900 mb (the surface).

The use of different convection schemes also led to differences in other components in the water balance such as the large-scale condensation or moistening associated with grid-scale reevaporation (Figs. 13g–i), and vertical diffusion (Figs. 13j–l). The stronger drying in the SAS2 forecast in the middle and upper troposphere associated with the deep convection before 15 November (Fig. 13c) was largely balanced by the stronger large-scale moistening (Fig. 13i). The stronger moistening (drying) near 900 mb (the surface) associated with both deep convection and SC in the SAS2 forecast (Figs. 13c,d) was mostly a balance for the vertical diffusive moistening (Fig. 13j). Comparison between the difference in moistening rate in Fig. 13 and the difference in the specific humidity in Fig. 11 suggests that the drier lower troposphere in the SAS2 forecast was probably caused by the constantly weaker moistening in the lower troposphere between 700 and 800 mb associated with the SC (Fig. 13f) as well the stronger drying near 800 mb during the first several days associated with the deep convection (Fig. 13c).

While these results relate the differences in the MJO forecast with different convection schemes to the differences in the atmospheric specific humidity, it is difficult to further pinpoint the exact part of the convection schemes that cause the difference without additional experiments to test the impacts of individual components and parameterizations. Unlike studies that focus on the impact of a particular parameter within specific convection schemes, such as the study by Wang and Schlesinger (1999) who investigated the role of a relative specific humidity criterion for convection onset and the study of Tokioka et al. (1988) who examined the role of entrainment rate, convections schemes used in this study include different components and parameters. Finding out the parameters or components that are responsible requires further experiments, which is beyond the scope of this study.

5. Summary and conclusions

This study investigated the impact of uncertainties in the underlying sea surface temperatures (SSTs) on the representation of the intraseasonal rainfall variability associated with the Madden–Julian oscillation (MJO) in an atmospheric model and how such impact varies with convection parameterization. The study was motivated by the fact that there exist substantial differences in observational SST analyses and by the possibility that lacking sufficient accuracy for SSTs in dynamical models may degrade the MJO simulation and prediction. It is important to understand how important these differences are if the observational SST analyses are used to force atmosphere-only models or to validate the SSTs from coupled atmosphere–ocean models. Since the MJO simulation is dependent on the atmospheric model physics, especially the convection parameterization, it is also useful to investigate if the SST impact depends on the convection scheme.

Initialized forecast experiments with the NCEP atmospheric Global Forecast System (GFS) model were carried out to test two widely used daily SST analyses—the NCDC analysis and TMI analysis—as well as an SST climatology, which did not contain any variability except for the seasonal cycle. Three convection schemes were used, including the revised simplified Arakawa–Schubert (SAS2) scheme currently used in the NCEP operational medium-range weather forecast, the original SAS scheme currently used in the NCEP operational seasonal climate prediction, and the Relaxed Arakawa–Schubert (RAS) scheme. Retrospective forecasts are made for the DYNAMO intensive observation period 1 October 2011–15 January 2012, which cover three MJO events initiated in October, November, and December 2011.

The comparison among the experiments show that the advantage of using observed (TMI and NCDC) SSTs over the climatology SSTs can be seen as early as 5 days to 1 week after the start of the forecast. Further, the prediction with TMI SSTs was more skillful than that with the NCDC SSTs, indicating the current level of SST uncertainties in the observational analyses can make large difference when they are used as the lower boundary conditions for the simulation and prediction of the MJO. It is probable that the simulation and prediction can be improved with an atmosphere-only model forced with more accurate SSTs, or with a coupled atmosphere–ocean model with a more realistic representation of the variability in the nature.

Among the convection schemes, the difference was least clear with the climatology SSTs, which give similar anomaly correlations, and most clear with the TMI SSTs, which resulted in higher anomaly correlation skill with the RAS and SAS schemes than the SAS2 scheme. The SAS2 scheme also produced weakest convection variability than the other two convection schemes. Consistently, the OLR–SST lagged regression was strongest in the SAS and RAS experiments compared to the SAS2 experiment. The weak convection activity and low forecast skill in SAS2 was probably related to its lower troposphere being too dry and leading to less intense convection activity. The SAS and RAS schemes also lead to stronger lagged regression between SST and surface shortwave heat flux than the SAS2 scheme. However, there is a bias in the lag between warm SST and the subsequent evaporative cooling in the forecast after the first week, which is about 4 days and is too short compared to the estimate of 10 days based on the observational data.

There are two implications of the results from this study: 1) an accurate SST observation is critical for the test of the physics and configuration of atmospheric models and, further, for verification of the SST from coupled atmosphere–ocean models; and 2) a reliable forecast of the MJO not only requires the use of coupled models but also requires that the coupled model realistically reproduce the phase and amplitude of the SST. Many contemporary coupled atmosphere–ocean models use a coarse vertical resolution (~10 m) for the upper ocean, which is probably not sufficient to resolve the intraseasonal SST variability. Improvements of the model toward a better SST representation may include the use of a higher vertical resolution as well as enhanced parameterizations for the vertical mixing processes in the upper ocean.

This study has shown that the impacts of the SST depend on the model physics. In particular, the SAS2 results in a weaker response to SST anomalies. Such a weaker response with the SAS2 scheme is linked to the drier lower troposphere due to a persistent weaker shallow convection moistening between 700 and 800 mb as well as a stronger drying near 800 mb during the first several days of the forecast. However, it should be emphasized that this study is not intended to rank which convection parameterization is the best scheme since the MJO is one of the phenomena in the coupled atmosphere–ocean system. A scheme that produces a better simulation of the MJO may not necessarily be superior in representing other aspects of the weather and climate.

Acknowledgments

We greatly appreciate the helpful internal reviews by Drs. Q. Zhang and Z.-Z. Hu. The authors also thank the two anonymous reviewers for their constructive comments and suggestions. Meng-Pai Hung gratefully acknowledges the financial support given by the Earth System Science Organization, Ministry of Earth Sciences, government of India, to conduct this research under the Monsoon Mission.

REFERENCES

  • Arakawa, A., , and W. H. Shubert, 1974: Interaction of a cumulus ensemble with the large-scale environment, Part I. J. Atmos. Sci., 31, 674704, doi:10.1175/1520-0469(1974)031<0674:IOACCE>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Bechtold, P., , M. Köhler, , T. Jung, , F. Doblas-Reyes, , M. Leutbecher, , M. J. Rodwell, , F. Vitart, , and G. Balsamo, 2008: Advances in simulating atmospheric variability with the ECMWF model: From synoptic to decadal time-scales. Quart. J. Roy. Meteor. Soc., 134, 13371351, doi:10.1002/qj.289.

    • Search Google Scholar
    • Export Citation
  • Benedict, J. J., , and D. A. Randall, 2007: Observed characteristics of the MJO relative to maximum rainfall. J. Atmos. Sci., 64, 23322354, doi:10.1175/JAS3968.1.

    • Search Google Scholar
    • Export Citation
  • Benedict, J. J., , and D. A. Randall, 2009: Structure of the Madden–Julian oscillation in the superparameterized CAM. J. Atmos. Sci., 66, 32773296, doi:10.1175/2009JAS3030.1.

    • Search Google Scholar
    • Export Citation
  • Bernie, D. J., , E. Guilyardi, , G. Madec, , J. M. Slingo, , S. J. Woolnough, , and J. Cole, 2008: Impact of resolving the diurnal cycle in an ocean–atmosphere GCM. Part 2: A diurnally coupled CGCM. Climate Dyn., 31, 909925, doi:10.1007/s00382-008-0429-z.

    • Search Google Scholar
    • Export Citation
  • Fu, X., , and B. Wang, 2009: Critical roles of stratiform rainfall in sustaining the Madden–Julian Oscillation: GCM experiments. J. Climate, 22, 39393959, doi:10.1175/2009JCLI2610.1.

    • Search Google Scholar
    • Export Citation
  • Fu, X., , B. Yang, , Q. Bao, , and B. Wang, 2008: Sea surface temperature feedback extends the predictability of tropical intraseasonal oscillation. Mon. Wea. Rev., 136, 577597, doi:10.1175/2007MWR2172.1.

    • Search Google Scholar
    • Export Citation
  • Fu, X., , W. Wang, , J.-Y. Lee, , B. Wang, , K. Kikuchi, , J. Xu, , J. Li, , and S. Weaver, 2015: Distinctive roles of air–sea coupling on different MJO events: A new perspective revealed from the DYNAMO/CINDY field campaign. Mon. Wea. Rev., 143, 794–812, doi:10.1175/MWR-D-14-00221.1.

    • Search Google Scholar
    • Export Citation
  • Grell, G. A., 1993: Prognostic evaluation of assumptions used by cumulus parameterization. Mon. Wea. Rev., 121, 764787, doi:10.1175/1520-0493(1993)121<0764:PEOAUB>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Han, J., , and H. L. Pan, 2011: Revision of convection and vertical diffusion schemes in the NCEP Global Forecast System. Wea. Forecasting, 26, 520533, doi:10.1175/WAF-D-10-05038.1.

    • Search Google Scholar
    • Export Citation
  • Huang, B., , M. L’Heureux, , J. Lawrimore, , C. Liu, , H.-M. Zhang, , V. Banzon, , Z.-Z. Hu, , and A. Kumar, 2013: Why did large differences arise in the sea surface temperature datasets across the tropical Pacific during 2012? J. Atmos. Oceanic Technol., 30, 29442953, doi:10.1175/JTECH-D-13-00034.1.

    • Search Google Scholar
    • Export Citation
  • Hung, M.-P., , J.-L. Lin, , W. Wang, , D. Kim, , T. Shinoda, , and S. J. Weaver, 2013: MJO and convectively coupled equatorial waves simulated by CMIP5 climate models. J. Climate, 26, 61856124, doi:10.1175/JCLI-D-12-00541.1.

    • Search Google Scholar
    • Export Citation
  • Jiang, X., and Coauthors, 2009: Vertical heating structures associated with the MJO as characterized by TRMM estimates, ECMWF reanalyses, and forecasts: A case study during 1998/99 winter. J. Climate, 22, 60016020, doi:10.1175/2009JCLI3048.1.

    • Search Google Scholar
    • Export Citation
  • Joyce, R. J., , J. E. Janowiak, , P. A. Arkin, , and P. Xie, 2004: CMORPH: A method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeor., 5, 487503, doi:10.1175/1525-7541(2004)005<0487:CAMTPG>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kemball-Cook, S., , B. Wang, , and X. Fu, 2002: Simulation of the intraseasonal oscillation in the ECHAM-4 model: The impact of coupling with an ocean model. J. Atmos. Sci., 59, 14331453, doi:10.1175/1520-0469(2002)059<1433:SOTIOI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Kim, H.-M., , C. D. Hoyos, , P. J. Webster, , and I.-S. Kang, 2008: Sensitivity of MJO simulation and predictability to sea surface temperature variability. J. Climate, 21, 53045317, doi:10.1175/2008JCLI2078.1.

    • Search Google Scholar
    • Export Citation
  • Kim, H.-M., , P. J. Webster, , V. E. Toma, , and D. Kim, 2014: Predictability and prediction skill of the MJO in two operational forecasting systems. J. Climate, 27, 53645378, doi:10.1175/JCLI-D-13-00480.1.

    • Search Google Scholar
    • Export Citation
  • Klingaman, N. P., , P. M. Inness, , H. Weller, , and J. M. Slingo, 2008: The importance of high-frequency sea surface temperature variability to the intraseasonal oscillation of Indian monsoon rainfall. J. Climate, 21, 61196140, doi:10.1175/2008JCLI2329.1.

    • Search Google Scholar
    • Export Citation
  • Klingaman, N. P., , S. J. Woolnough, , H. Weller, , and J. M. Slingo, 2011: The impact of finer-resolution air–sea coupling on the intraseasonal oscillation of the Indian monsoon. J. Climate, 24, 24512468, doi:10.1175/2010JCLI3868.1.

    • Search Google Scholar
    • Export Citation
  • Kumar, A., , L. Zhang, , and W. Wang, 2013: Sea surface temperature–precipitation relationship in different reanalyses. Mon. Wea. Rev., 141, 11181123, doi:10.1175/MWR-D-12-00214.1.

    • Search Google Scholar
    • Export Citation
  • Lau, K. M., , and L. Peng, 1987: Origin of low frequency (intraseasonal) oscillations in the tropical atmosphere. Part I: The basic theory. J. Atmos. Sci., 44, 950972, doi:10.1175/1520-0469(1987)044<0950:OOLFOI>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Liebmann, B., , and C. A. Smith, 1996: Description of a complete (interpolated) outgoing longwave radiation dataset. Bull. Amer. Meteor. Soc., 77, 12751277.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., and Coauthors, 2006: Tropical intraseasonal variability in 14 IPCC AR4 climate models. Part I: Convective signals. J. Climate, 19, 26652690, doi:10.1175/JCLI3735.1.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., , B. Mapes, , M. Zhang, , and M. Newman, 2004: Stratiform precipitation, vertical heating profiles, and the Madden–Julian oscillation. J. Atmos. Sci., 61, 296309, doi:10.1175/1520-0469(2004)061<0296:SPVHPA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Lin, J.-L., , M.-I. Lee, , D. Kim, , I.-S. Kang, , and D. M. W. Frierson, 2008: The impacts of convective parameterization and moisture triggering on AGCM-simulated convectively coupled equatorial waves. J. Climate, 21, 883909, doi:10.1175/2007JCLI1790.1.

    • Search Google Scholar
    • Export Citation
  • Madden, R. A., , and P. R. Julian, 1994: Observations of the 40–50-day tropical oscillation—A review. Mon. Wea. Rev., 122, 814837, doi:10.1175/1520-0493(1994)122<0814:OOTDTO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Matthews, A. J., 2004: Atmospheric response to observed intraseasonal tropical sea surface temperature anomalies. Geophys. Res. Lett., 31, L14107, doi:10.1029/2004GL020474.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., , and M. J. Suarez, 1992: Relaxed Arakawa–Schubert: A parameterization of moist convection for general circulation models. Mon. Wea. Rev., 120, 9781002, doi:10.1175/1520-0493(1992)120<0978:RASAPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Moorthi, S., , and M. J. Suarez, 1999: Documentation of version 2 of relaxed Arakawa–Schubert cumulus parameterization with convective downdrafts. NOAA Office Note 99-01, 44 pp.

  • Pan, H.-L., , and W.-S. Wu, 1995: Implementing a mass flux convection parameterization package for the NMC medium-range forecast model. NMC Office Note 409, 39 pp.

  • Reynolds, R. W., , N. A. Rayner, , T. M. Smoth, , D. C. Stokes, , and W. Wang, 2002: An improved in situ and satellite SST analysis for climate. J. Climate, 15, 16091625, doi:10.1175/1520-0442(2002)015<1609:AIISAS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Reynolds, R. W., , T. M. Smith, , C. Liu, , D. B. Chelton, , K. S. Casey, , and M. G. Schlax, 2007: Daily high-resolution blended analyses for sea surface temperature. J. Climate, 20, 54735496, doi:10.1175/2007JCLI1824.1.

    • Search Google Scholar
    • Export Citation
  • Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc.,91, 1015–1057, doi:10.1175/2010BAMS3001.1.

  • Saha, S., and Coauthors, 2014: The NCEP Climate Forecast System version 2. J. Climate, 27, 21852208, doi:10.1175/JCLI-D-12-00823.1.

  • Seo, K.-H., , and W. Wang, 2010: The Madden–Julian oscillation simulated in the NCEP Climate Forecast System Model: The importance of stratiform heating. J. Climate, 23, 47704793, doi:10.1175/2010JCLI2983.1.

    • Search Google Scholar
    • Export Citation
  • Shelly, A., , P. Xavier, , D. Copsey, , T. Johns, , J. M. Rodriguez, , S. Milton, , and N. Klingaman, 2014: Coupled versus uncoupled hindcast simulations of the Madden–Julian Oscillation in the Year of Tropical Convection. Geophys. Res. Lett., 41, 56705677, doi:10.1002/2013GL059062.

    • Search Google Scholar
    • Export Citation
  • Shinoda, T., , H. H. Hendon, , and J. Glick, 1998: Intraseasonal variability of surface fluxes and sea surface temperature in the tropical western Pacific and Indian Oceans. J. Climate, 11, 16851702, doi:10.1175/1520-0442(1998)011<1685:IVOSFA>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Takahashi, M., 1987: A theory of slow phase speed of the intraseasonal oscillation using the wave-CISK. J. Meteor. Soc. Japan, 65, 4349.

    • Search Google Scholar
    • Export Citation
  • Thiébaux, J., , E. Rogers, , W. Wang, , and B. Katz, 2003: A new high-resolution blended real-time global sea surface temperature analysis. Bull. Amer. Meteor. Soc., 84, 645656, doi:10.1175/BAMS-84-5-645.

    • Search Google Scholar
    • Export Citation
  • Tian, B., , D. E. Waliser, , E. J. Fetzer, , B. H. Lambrigtsen, , Y. L. Yung, , and B. Wang, 2006: Vertical moist thermodynamic structure and spatial–temporal evolution of the MJO in AIRS observations. J. Atmos. Sci., 63, 24622485, doi:10.1175/JAS3782.1.

    • Search Google Scholar
    • Export Citation
  • Tiedtke, M., , W. A. Heckley, , and J. Slingo, 1988: Tropical forecasting at ECMWF: The influence of physical parametrization on the mean structure of forecasts and analyses. Quart. J. Roy. Meteor. Soc., 114, 639664, doi:10.1002/qj.49711448106.

    • Search Google Scholar
    • Export Citation
  • Tokioka, T., , K. Yamazaki, , A. Kitoh, , and T. Ose, 1988: The equatorial 30–60 day oscillation and the Arakawa–Schubert penetrative cumulus parameterization. J. Meteor. Soc. Japan, 66, 883901.

    • Search Google Scholar
    • Export Citation
  • Vitart, F., , S. Woolnough, , M. A. Balmaseda, , and A. M. Tompkins, 2007: Monthly forecasts of the Madden–Julian oscillation using a coupled GCM. Mon. Wea. Rev., 135, 27002715, doi:10.1175/MWR3415.1.

    • Search Google Scholar
    • Export Citation
  • Waliser, D. E., , K. M. Lau, , and J.-H. Kim, 1999: The influence of coupled SSTs on the Madden–Julian oscillation: A model perturbation experiment. J. Atmos. Sci., 56, 333358, doi:10.1175/1520-0469(1999)056<0333:TIOCSS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wang, W., , and M. Schlesinger, 1999: The dependence on convection parameterization of the tropical intraseasonal oscillation simulated by the UIUC 11-layer atmospheric GCM. J. Climate, 12, 14231457, doi:10.1175/1520-0442(1999)012<1423:TDOCPO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Wang, W., , M.-P. Hung, , S. J. Weaver, , A. Kumar, , and X. Fu, 2014: MJO prediction in the NCEP Climate Forecast System version 2. Climate Dyn., 42, 25092520, doi:10.1007/s00382-013-1806-9.

    • Search Google Scholar
    • Export Citation
  • Wentz, F. J., , C. L. Gentemann, , D. K. Smith, , and D. B. Chelton, 2000: Satellite measurements of sea-surface temperature through clouds. Science, 288, 847850, doi:10.1126/science.288.5467.847.

    • Search Google Scholar
    • Export Citation
  • Woolnough, S. J., , J. M. Slingo, , and B. J. Hoskins, 2000: The relationship between convection and sea surface temperature on intraseasonal timescales. J. Climate, 13, 20862104, doi:10.1175/1520-0442(2000)013<2086:TRBCAS>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Woolnough, S. J., , F. Vitart, , and M. A. Balmaseda, 2007: The role of the ocean in the Madden–Julian oscillation: Implications for MJO prediction. Quart. J. Roy. Meteor. Soc., 133, 117128, doi:10.1002/qj.4.

    • Search Google Scholar
    • Export Citation
  • Wu, M., , S. Schubert, , I. Kang, , and D. Waliser, 2002: Forced and free intraseasonal variability over the South Asian monsoon region simulated by 10 AGCMs. J. Climate, 15, 28622880, doi:10.1175/1520-0442(2002)015<2862:FAFIVO>2.0.CO;2.

    • Search Google Scholar
    • Export Citation
  • Zeng, X., , M. Zhao, , R. E. Dickinson, , and Y. He, 1999: A multi-year hourly sea surface skin temperature dataset derived from the TOGA TAO bulk temperature and wind speed over the tropical Pacific. J. Geophys. Res., 104, 15251536, doi:10.1029/1998JC900060.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., , and M. Mu, 2005: Simulation of the Madden–Julian Oscillation in the NCAR CCM3 using a revised Zhang–McFarlane convection parameterization scheme. J. Climate, 18, 40464064, doi:10.1175/JCLI3508.1.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., , and X. Song, 2009: Interaction of deep and shallow convection is key to Madden–Julian Oscillation simulation. Geophys. Res. Lett., 36, L09708, doi:10.1029/2009GL037340.

    • Search Google Scholar
    • Export Citation
Save