• Adams, D. K., and A. C. Comrie, 1997: The North American monsoon. Bull. Amer. Meteor. Soc., 78 , 21972213.

  • Adams, J. L., and D. J. Stensrud, 2007: Impact of tropical easterly waves on the North American monsoon. J. Climate, 20 , 12191238.

  • Arritt, R. W., D. C. Goerling, and C. J. Anderson, 2000: The North American monsoon system in the Hadley Centre coupled ocean–atmosphere GCM. Geophys. Res. Lett., 27 , 565568.

    • Search Google Scholar
    • Export Citation
  • Barlow, M., and D. Salstein, 2006: Summertime influence of the Madden-Julian Oscillation on daily rainfall over Mexico and Central America. Geophys. Res. Lett., 33 .L21708, doi:10.1029/2006GL027738.

    • Search Google Scholar
    • Export Citation
  • Barlow, M., S. Nigam, and E. H. Berbery, 1998: Evolution of the North American monsoon system. J. Climate, 11 , 22382257.

  • Berbery, E. H., 2001: Mesoscale moisture analysis of the North American monsoon. J. Climate, 14 , 121137.

  • Betts, A. K., 1986: A new convective adjustment scheme. Part I: Observational and theoretical basis. Quart. J. Roy. Meteor. Soc., 112 , 677691.

    • Search Google Scholar
    • Export Citation
  • Bougeault, P., 1985: A simple parameterization of the large-scale effects of cumulus convection. Mon. Wea. Rev., 113 , 21082121.

  • Carleton, A. M., D. A. Carpenter, and P. J. Weser, 1990: Mechanisms of interannual variability of the southwest United States summer rainfall maximum. J. Climate, 3 , 9991015.

    • Search Google Scholar
    • Export Citation
  • Castro, C. L., T. B. McKee, and R. A. Pielke Sr., 2001: The relationship of the North American monsoon to tropical and North Pacific sea surface temperatures as revealed by observational analyses. J. Climate, 14 , 44494473.

    • Search Google Scholar
    • Export Citation
  • Collier, J. C., and G. J. Zhang, 2007: Effects of increased horizontal resolution on simulation of the North American monsoon in the NCAR CAM3: An evaluation based on surface, satellite, and reanalysis data. J. Climate, 20 , 18431861.

    • Search Google Scholar
    • Export Citation
  • Collins, W. D., and Coauthors, 2006: The Community Climate System Model version 3 (CCSM3). J. Climate, 19 , 21222143.

  • Del Genio, A. D., and M-S. Yao, 1993: Efficient cumulus parameterization for long-term climate studies: The GISS scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 181–184.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., and Coauthors, 2006: GFDL’s CM2 global coupled climate models. Part I: Formulation and simulation characteristics. J. Climate, 19 , 643674.

    • Search Google Scholar
    • Export Citation
  • de Szoeke, S. P., and S-P. Xie, 2008: The tropical eastern Pacific seasonal cycle: Assessment of errors and mechanisms in IPCC AR4 coupled ocean–atmosphere general circulation models. J. Climate, 21 , 25732590.

    • Search Google Scholar
    • Export Citation
  • Diansky, N. A., and E. M. Volodin, 2002: Simulation of present-day climate with a coupled atmosphere-ocean general circulation model. Izv. Atmos. Oceanic Phys., 38 , 732747.

    • Search Google Scholar
    • Export Citation
  • Douglas, M. W., R. A. Maddox, K. Howard, and S. Reyes, 1993: The Mexican monsoon. J. Climate, 6 , 16651677.

  • Emanuel, K. A., 1991: A scheme for representing cumulus convection in large-scale models. J. Atmos. Sci., 48 , 23132329.

  • Emori, S., T. Nozawa, A. Numaguti, and I. Uno, 2001: Importance of cumulus parameterization for precipitation simulation over East Asia in June. J. Meteor. Soc. Japan, 79 , 939947.

    • Search Google Scholar
    • Export Citation
  • Fawcett, P. J., J. R. Stalker, and D. S. Gutzler, 2002: Multistage moisture transport into the interior of northern Mexico during the North American summer monsoon. Geophys. Res. Lett., 29 .2094, doi:10.1029/2002GL015693.

    • Search Google Scholar
    • Export Citation
  • Flato, G. M., and G. J. Boer, 2001: Warming asymmetry in climate change simulations. Geophys. Res. Lett., 28 , 195198.

  • Fuller, R. D., and D. J. Stensrud, 2000: The relationship between tropical easterly waves and surges over the Gulf of California during the North American monsoon. Mon. Wea. Rev., 128 , 29832989.

    • Search Google Scholar
    • Export Citation
  • Gates, W. L., and Coauthors, 1999: An overview of the results of the Atmospheric Model Intercomparison Project (AMIP I). Bull. Amer. Meteor. Soc., 80 , 2955.

    • Search Google Scholar
    • Export Citation
  • Gochis, D. J., W. J. Shuttleworth, and Z-L. Yang, 2002: Sensitivity of the modeled North American monsoon regional climate to convective parameterization. Mon. Wea. Rev., 130 , 12821298.

    • Search Google Scholar
    • Export Citation
  • Gochis, D. J., A. Jimenez, C. J. Watts, J. Garatuza-Payan, and W. J. Shuttleworth, 2004: Analysis of 2002 and 2003 warm-season precipitation from the North American Monsoon Experiment event rain gauge network. Mon. Wea. Rev., 132 , 29382953.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability-dependent closure. Mon. Wea. Rev., 118 , 14831506.

    • Search Google Scholar
    • Export Citation
  • Gutzler, D. S., and J. W. Preston, 1997: Evidence for a relationship between spring snow cover in North American and summer rainfall in New Mexico. Geophys. Res. Lett., 24 , 22072210.

    • Search Google Scholar
    • Export Citation
  • Gutzler, D. S., and Coauthors, 2005: The North American Monsoon Model Assessment Project: Integrating numerical modeling into a field-based process study. Bull. Amer. Meteor. Soc., 86 , 14231429.

    • Search Google Scholar
    • Export Citation
  • Harrington, J. A., R. S. Cerveny, and R. C. Balling, 1992: Impact of the Southern Oscillation on the North American monsoon. Phys. Geogr., 13 , 318330.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., and W. Shi, 2000: Dominant factors responsible for interannual variability of the summer monsoon in the southwestern United States. J. Climate, 13 , 759775.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., and D. Gochis, 2007: Synthesis of Results from the North American Monsoon Experiment (NAME) Process Study. J. Climate, 20 , 16011607.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., Y. Yao, E. S. Yarosh, J. E. Janowiak, and K. C. Mo, 1997a: Influence of the Great Plains low-level jet on summertime precipitation and moisture transport over the central United States. J. Climate, 10 , 481507.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., Y. Yao, and X. L. Wang, 1997b: Influence of the North American monsoon system on the U.S. summer precipitation regime. J. Climate, 10 , 26002622.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., K. C. Mo, and Y. Yao, 1998: Interannual variability of the U.S. summer precipitation regime with emphasis on the southwestern monsoon. J. Climate, 11 , 25822606.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., Y. Chen, and A. V. Douglas, 1999: Interannual variability of the North American warm season precipitation regime. J. Climate, 12 , 653680.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., W. Shi, and E. Yarosh, 2000: Improved United States precipitation quality control system and analysis. NCEP/Climate Prediction Center Atlas Number 7, 40 pp. [Available from Climate Prediction Center, World Weather Building, Room 605, Camp Springs, MD 20746.].

  • Higgins, R. W., and Coauthors, 2006: The NAME 2004 field campaign and modeling strategy. Bull. Amer. Meteor. Soc., 87 , 7994.

  • Hu, Q., and S. Feng, 2002: Interannual rainfall variations in the North American summer monsoon region: 1900–98. J. Climate, 15 , 11891202.

    • Search Google Scholar
    • Export Citation
  • Ishii, M., M. Kimoto, K. Sakamoto, and S. Iwasaki, 2006: Steric sea level changes estimated from historical ocean subsurface temperature and salinity analyses. J. Oceanogr., 62 , 155170.

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

  • Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305 , 11381140.

  • Lee, M-I., and Coauthors, 2007: Sensitivity to horizontal resolution in the AGCM simulations of warm season diurnal cycle of precipitation over the United States and Northern Mexico. J. Climate, 20 , 18621881.

    • Search Google Scholar
    • Export Citation
  • Liang, X-Z., L. Li, K. E. Kunkel, M. Ting, and J. X. L. Wang, 2004: Regional climate model simulation of U.S. precipitation during 1982–2002. Part I: Annual cycle. J. Climate, 17 , 35103528.

    • Search Google Scholar
    • Export Citation
  • Liang, X-Z., M. Xu, K. E. Kunkel, G. A. Grell, and J. Kain, 2007: Regional climate model simulation of U.S.–Mexico summer precipitation using the optimal ensemble of two cumulus parameterizations. J. Climate, 20 , 52015207.

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

  • Lin, J-L., B. E. Mapes, K. M. Weickmann, G. N. Kiladis, S. D. Schubert, M. J. Suarez, J. T. Bacmeister, and M-I. Lee, 2008: North American monsoon and convectively coupled equatorial waves simulated by IPCC AR4 coupled GCMs. J. Climate, 21 , 29192937.

    • Search Google Scholar
    • Export Citation
  • Livezey, R. E., and W. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111 , 4659.

    • Search Google Scholar
    • Export Citation
  • Lorenz, D. J., and D. L. Hartmann, 2006: The effect of the MJO on the North American monsoon. J. Climate, 19 , 333343.

  • Meehl, G. A., G. J. Boer, C. Covey, M. Latif, and R. J. Stouffer, 2000: The Coupled Model Intercomparison Project (CMIP). Bull. Amer. Meteor. Soc., 81 , 313318.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., J. M. Arblaster, D. M. Lawrence, A. Seth, E. K. Schneider, B. P. Kirtman, and D. Min, 2006: Monsoon regimes in the CCSM3. J. Climate, 19 , 24822495.

    • Search Google Scholar
    • Export Citation
  • Mitchell, T., and P. Jones, 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25 , 693712.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., J. E. Schemm, H. M. H. Juang, R. W. Higgins, and Y. Song, 2005: Impact of model resolution on the prediction of summer precipitation over the United States and Mexico. J. Climate, 18 , 39103927.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., J. E. Schemm, H. Kim, and W. R. Higgins, 2006: Influence of initial conditions on summer precipitation simulations over the United States and Mexico. J. Climate, 19 , 36403658.

    • 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.

    • Search Google Scholar
    • Export Citation
  • NAME, cited. 2005: NAME modeling and data assimilation: A strategic overview. [Available online at http://www.eol.ucar.edu/projects/name/documentation/NAME_model.html.].

  • Nordeng, T. E., 1994: Extended versions of the convective parameterization scheme at ECMWF and their impact on the mean and transient activity of the model in the tropics. Tech. Memo. 206, European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 41 pp.

  • Pan, D-M., and D. A. Randall, 1998: A cumulus parameterization with a prognostic closure. Quart. J. Roy. Meteor. Soc., 124 , 949981.

  • Reyes, S., and A. Mejía-Trejo, 1991: Tropical perturbations in the eastern Pacific and the precipitation field over northwestern Mexico in relation to the ENSO phenomenon. Int. J. Climatol., 11 , 515528.

    • Search Google Scholar
    • Export Citation
  • Schmitz, J. T., and S. Mullen, 1996: Water vapor transport associated with the summertime North American monsoon as depicted by ECMWF analyses. J. Climate, 9 , 16211634.

    • Search Google Scholar
    • Export Citation
  • Stensrud, D. J., R. L. Gall, S. L. Mullen, and K. W. Howard, 1995: Model climatology of the Mexican monsoon. J. Climate, 8 , 17751794.

    • Search Google Scholar
    • Export Citation
  • Stensrud, D. J., R. L. Gall, and M. K. Nordquist, 1997: Surges over the Gulf of California during the Mexican monsoon. Mon. Wea. Rev., 125 , 417437.

    • Search Google Scholar
    • Export Citation
  • Tang, M., and E. R. Reiter, 1984: Plateau monsoons of the Northern Hemisphere: A comparison between North America and Tibet. Mon. Wea. Rev., 112 , 617637.

    • Search Google Scholar
    • Export Citation
  • Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117 , 17791800.

    • Search Google Scholar
    • Export Citation
  • Ting, M., X-Z. Liang, J. Pan, and H-P. Huang, 2005: North American summer precipitation with and without topography in CCM3 and the MM5-based regional climate model (CMM5). Proc. 2005 WRF/MM5 User’s Workshop, Boulder, CO, National Center for Atmospheric Research, 15 pp.

  • 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
  • Vera, C., and Coauthors, 2006: Toward a unified view of the American monsoon systems. J. Climate, 19 , 49775000.

  • Waliser, D., and Coauthors, 2006: The experimental MJO prediction project. Bull. Amer. Meteor. Soc., 87 , 425431.

  • Washington, W. M., and Coauthors, 2000: Parallel climate model (PCM) control and transient simulations. Climate Dyn., 16 , 755774.

  • Wu, X., L. Deng, X. Song, G. Vettoretti, W. R. Peltier, and G. J. Zhang, 2007: Impact of a modified convective scheme on the Madden-Julian Oscillation and El Niño–Southern Oscillation in a coupled climate model. Geophys. Res. Lett., 34 .L16823, doi:10.1029/2007GL030637.

    • Search Google Scholar
    • Export Citation
  • Xu, J., X. Gao, J. Shuttleworth, S. Sorooshian, and E. Small, 2004: Model climatology of the North American monsoon onset period during 1980–2001. J. Climate, 17 , 38923906.

    • Search Google Scholar
    • Export Citation
  • Yukimoto, S., and Coauthors, 2001: The new Meteorological Research Institute coupled GCM (MRI-CGCM2)—Model climate and variability. Pap. Meteor. Geophys., 51 , 4788.

    • Search Google Scholar
    • Export Citation
  • Yu, B., and J. M. Wallace, 2000: The principal mode of interannual variability of the North American monsoon system. J. Climate, 13 , 27942800.

    • Search Google Scholar
    • Export Citation
  • Zhang, C., 2005: Madden-Julian Oscillation. Rev. Geophys., 43 .RG2003, doi:10.1029/2004RG000158.

  • Zhang, C., M. Dong, S. Gualdi, H. H. Hendon, E. D. Maloney, A. Marshall, K. R. Sperber, and W. Wang, 2006: Simulations of the Madden-Julian Oscillation in four pairs of coupled and uncoupled global models. Climate Dyn., 27 , 573592. doi:10.1007/s00382-006-0148-2.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the CCC-GCM. Atmos.–Ocean, 3 , 407446.

    • Search Google Scholar
    • Export Citation
  • Zou, C-Z., and W. Zheng, 2004: Simulation of diurnal patterns of summer precipitation in the North American monsoon: An assessment using TRMM. Geophys. Res. Lett., 31 .L07105, doi:10.1029/2004GL019415.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    The major study domain. Outlined are the six key regions stated in the text, including the NAM core for the precipitation index (solid trapezoid), the northwest and southeast flanks of the core (dotted rectangles) for the metric of correlations with 200-hPa meridional wind (V200) in June/September and July/August, and the eastern tropical Pacific and the Gulf of Mexico (dashed rectangles) for the metric of correlations with SSTs in June and July. Inserted is a blowup map showing the summer average distribution of monthly teleconnection correlations of precipitation interannual variations during 1961–90 at every point with the mean of a small area (little square) in the central Mexico; correlations greater than 0.60 (0.35) are hatched (shaded).

  • View in gallery

    Interannual variations during 1948–2004 of the seasonal (June–September) mean precipitation (mm day−1) averaged over the NAM core region from three datasets.

  • View in gallery

    The phase lags (month, hollow bars) and rms errors (mm day−1, solid bars) of all models simulated from observed NAM core precipitation annual cycles. Each simulation is labeled as the modeling institution name, followed by a dot plus an identification for a specific model, and then by an underscore plus a number for every run. In particular, ncar.c and .p denote CCSM and PCM; giss.r and .h GISS ER and EH; and gfdl.0 and 0.1GFDL CM2.0 and CM2.1.

  • View in gallery

    The NAM core precipitation (mm day−1) annual cycle observed and modeled by seven distinct groups. The legend lists observations (OBS) (thick solid), CGCM/SST (thick dashed) for the AMIP run using that CGCM forced with prescribed SSTs, and model names (thin curves with various patterns). Multiple realizations, if any, are depicted by a same curve pattern.

  • View in gallery

    (a), (c), (e), (g) Geographic distributions of July minus June and (b), (d), (f), (h) September minus August 30-yr (1961–90) mean differences in precipitation [mm day−1: (a)–(d)] and 200-hPa wind [m s−1: (e)–(h)] from (left) obs or NRA reanalysis and (right) simulated by the MRI model. Contour intervals are 1 mm day−1 for precipitation and 2 m s−1 for wind speed. Wind direction is depicted by thin streamlines with arrowheads.

  • View in gallery

    Geographic distributions of the monthly interannual correlations (×10) of the NAM core precipitation with pointwise 200-hPa (a)–(d) meridional V200 and (e)–(h) zonal U200 wind components observed during 1961–90. Outlined are the two key centers with statistically significant high V200 correlations that distinguish the regimes of June and September (monsoon transitions) from July and August (peak monsoon). Contour intervals are 1 unit. Shaded areas denote where correlations are statistically significant at the 95% confidence level by the Monte Carlo test (see text). The areas in (a)–(h) account for 7.8%, 8.1%, 5.7%, 20.3%, 10.3%, 21.6%, 0.0%, and 0.3% of the entire domain, respectively. Those having greater than 5% areas with significant correlations are considered to be of field significance.

  • View in gallery

    Interannual variations during 1961–90 of the normalized NAM rainfall anomalies for each month (Jun, Jul, Aug, Sep) and the whole summer average. The respective five strongest wet and dry years are marked by upward and downward arrows, except for those marked by (×) that are not accompanied by the apparent cyclonic and anticyclonic anomalies (see text). For comparison, the wet/dry years based on the summer average are marked by (+/−).

  • View in gallery

    Geographic distributions of June to September monthly and summer mean (JJAS) composites over the five strongest (left) wet and (middle) dry NAM years and (right) their differences for 200-hPa wind observed during 1961–90. The composite is constructed for each month and the whole summer, and thus may include a list of different years between June and September (see Fig. 7). Contour intervals are 3 (2) m s−1 for the wet or dry (their difference) composite wind speed. Wind direction is depicted by thin streamlines with arrowheads.

  • View in gallery

    Monthly interannual correlations of the NAM core precipitation with the 200-hPa meridional wind averaged over the observed connectivity centers specified in Fig. 6. Shown are the correlations observed (black bar) and simulated by the MRI model (gray bar) and all other CGCMs (hollow bar). Simulation labels follow the convention for Fig. 3.

  • View in gallery

    Geographic distributions of (a)–(d) June to September monthly and (e) summer JJAS mean interannual correlations of the NAM core precipitation with pointwise surface temperature (SST over oceans and 2-m air temperature over land) observed during 1961–90. Outlined are the two key centers (eastern Pacific and Gulf of Mexico) with statistically significant high correlations in June and July. Contour intervals are 2 units: zero contour eliminated. Shading areas denote where correlations are statistically significant at the 95% confidence level by the Monte Carlo test (see text). The areas in (a)–(e) account for 12.7%, 12.1%, 3.5%, 0.7%, 10.7% of the entire domain, respectively. Those having greater than 5% areas with significant correlations are considered to be of field significance. For a reference, the climatological summer mean SST distribution is also shown in (f) at contour intervals of 2°C with >28°C shaded.

  • View in gallery

    The June, July and summer June–September mean (JJAS) interannual correlations of the NAM core precipitation with the SST gradient between the averages over the observed connectivity centers in the eastern Pacific and the Gulf of Mexico specified respectively in Figs. 10a,b,e. Shown are the correlations observed (black bar) and simulated by the MRI (gray bar) and all other CGCMs (hollow bar). Simulation labels follow the convention for Fig. 3.

  • View in gallery

    (a)–(h) The SST (°C) annual cycle averaged over the eastern Pacific (see Fig. 10e for area specification) as observed and modeled by 17 CGCMs: models are grouped as in Fig. 4. The legend lists OBS (thick solid) for observations and model names (thin curves with various patterns). Multiple realizations, if any, are depicted by a same curve pattern. Also shown in (h) are the observed SSTs over the eastern Pacific (EP) and the Gulf of Mexico (GM) as well as their gradient (EP − GM, scaled on right).

  • View in gallery

    The SST gradients between the eastern Pacific and Gulf of Mexico (see Fig. 10e for the area specification) as observed (OBS) and simulated by 17 CGCMs for individual months June–September and the summer average (JJAS). Simulation labels follow the convention for Fig. 3 except that multiple realizations, if any, are averaged to present only the ensemble mean for each model.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 125 41 0
PDF Downloads 63 28 1

Do CGCMs Simulate the North American Monsoon Precipitation Seasonal–Interannual Variability?

View More View Less
  • 1 Illinois State Water Survey, University of Illinois at Urbana–Champaign, Urbana, Illinois
  • | 2 Lamont-Doherty Earth Observatory, Columbia University, New York, New York
  • | 3 National Oceanic and Atmospheric Administration/Air Resources Laboratory, Silver Spring, Maryland
Full access

Abstract

This study uses the most recent simulations from all available fully coupled atmosphere–ocean general circulation models (CGCMs) to investigate whether the North American monsoon (NAM) precipitation seasonal–interannual variations are simulated and, if so, whether the key underlying physical mechanisms are correctly represented. This is facilitated by first identifying key centers where observed large-scale circulation fields and sea surface temperatures (SSTs) are significantly correlated with the NAM precipitation averages over the core region (central–northwest Mexico) and then examining if the modeled and observed patterns agree.

Two new findings result from the analysis of observed NAM interannual variations. First, precipitation exhibits significantly high positive (negative) correlations with 200-hPa meridional wind centered to the northwest (southeast) of the core region in June and September (July and August). As such, wet conditions are associated with strong anomalous southerly upper-level flow on the northwest flank during the monsoon onset and retreat, but with anomalous northerly flow on the southeast flank, during the peak of the monsoon. They are often identified with a cyclonic circulation anomaly pattern over the central Great Plains for the July–August peak monsoon and, reversely, an anticyclonic anomaly pattern centered over the northern (southern) Great Plains for the June (September) transition. Second, wet NAM conditions in June and July are also connected with a SST pattern of positive anomalies in the eastern Pacific and negative anomalies in the Gulf of Mexico, acting to reduce the climatological mean gradient between the two oceans. This pattern suggests a possible surface gradient forcing that favors a westward extension of the North Atlantic subtropical ridge. These two observed features connected to the NAM serve as the metric for quantitative evaluation of the model performance in simulating the important NAM precipitation mechanisms.

Out of 17 CGCMs, only the Meteorological Research Institute (MRI) model with a medium resolution consistently captures the observed NAM precipitation annual cycle (having a realistic amplitude and no phase shift) as well as interannual covariations with the planetary circulation patterns (having the correct sign and comparably high magnitude of correlation) throughout the summer. For the metric of correlations with 200-hPa meridional wind, there is general agreement among all CGCMs with observations for June and September. This may indicate that large-scale forcings dominate interannual variability for the monsoon onset and retreat, while variability of the peak of the monsoon in July and August may be largely influenced by local processes that are more challenging for CGCMs to resolve. For the metric of correlations with SSTs, good agreement is found only in June. These results suggest that the NAM precipitation interannual variability may likely be determined by large-scale circulation anomalies, while its predictability based on remote signals such as SSTs may not be sufficiently robust to be well captured by current CGCMs, with the exception of the June monsoon onset which is potentially more predictable.

Corresponding author address: Dr. Xin-Zhong Liang, Illinois State Water Survey, University of Illinois at Urbana–Champaign, 2204 Griffith Dr., Champaign, IL 61820-7495. Email: xliang@uiuc.edu

Abstract

This study uses the most recent simulations from all available fully coupled atmosphere–ocean general circulation models (CGCMs) to investigate whether the North American monsoon (NAM) precipitation seasonal–interannual variations are simulated and, if so, whether the key underlying physical mechanisms are correctly represented. This is facilitated by first identifying key centers where observed large-scale circulation fields and sea surface temperatures (SSTs) are significantly correlated with the NAM precipitation averages over the core region (central–northwest Mexico) and then examining if the modeled and observed patterns agree.

Two new findings result from the analysis of observed NAM interannual variations. First, precipitation exhibits significantly high positive (negative) correlations with 200-hPa meridional wind centered to the northwest (southeast) of the core region in June and September (July and August). As such, wet conditions are associated with strong anomalous southerly upper-level flow on the northwest flank during the monsoon onset and retreat, but with anomalous northerly flow on the southeast flank, during the peak of the monsoon. They are often identified with a cyclonic circulation anomaly pattern over the central Great Plains for the July–August peak monsoon and, reversely, an anticyclonic anomaly pattern centered over the northern (southern) Great Plains for the June (September) transition. Second, wet NAM conditions in June and July are also connected with a SST pattern of positive anomalies in the eastern Pacific and negative anomalies in the Gulf of Mexico, acting to reduce the climatological mean gradient between the two oceans. This pattern suggests a possible surface gradient forcing that favors a westward extension of the North Atlantic subtropical ridge. These two observed features connected to the NAM serve as the metric for quantitative evaluation of the model performance in simulating the important NAM precipitation mechanisms.

Out of 17 CGCMs, only the Meteorological Research Institute (MRI) model with a medium resolution consistently captures the observed NAM precipitation annual cycle (having a realistic amplitude and no phase shift) as well as interannual covariations with the planetary circulation patterns (having the correct sign and comparably high magnitude of correlation) throughout the summer. For the metric of correlations with 200-hPa meridional wind, there is general agreement among all CGCMs with observations for June and September. This may indicate that large-scale forcings dominate interannual variability for the monsoon onset and retreat, while variability of the peak of the monsoon in July and August may be largely influenced by local processes that are more challenging for CGCMs to resolve. For the metric of correlations with SSTs, good agreement is found only in June. These results suggest that the NAM precipitation interannual variability may likely be determined by large-scale circulation anomalies, while its predictability based on remote signals such as SSTs may not be sufficiently robust to be well captured by current CGCMs, with the exception of the June monsoon onset which is potentially more predictable.

Corresponding author address: Dr. Xin-Zhong Liang, Illinois State Water Survey, University of Illinois at Urbana–Champaign, 2204 Griffith Dr., Champaign, IL 61820-7495. Email: xliang@uiuc.edu

1. Introduction

The North American monsoon (NAM) is a major challenge to global and regional climate modeling (Gutzler et al. 2005; Higgins et al. 2006). In particular, the NAM precipitation, including its annual cycle and interannual variability, is most difficult to predict owing to the dominant contribution from moist convective processes and their development, organization, and maintenance mechanisms in the presence of complex terrain and land–sea contrasts. The NAM system reflects the integration of a wide range of air–sea–land interactions at local to planetary scales (Tang and Reiter 1984; Douglas et al. 1993; Schmitz and Mullen 1996; Adams and Comrie 1997; Higgins et al. 1997b; Gutzler and Preston 1997; Stensrud et al. 1997; Barlow et al. 1998; Fuller and Stensrud 2000; Yu and Wallace 2000; Berbery 2001; Castro et al. 2001; Hu and Feng 2002; Fawcett et al. 2002; Koster et al. 2004; Gochis et al. 2004; Lorenz and Hartmann 2006; Barlow and Salstein 2006; Vera et al. 2006; Adams and Stensrud 2007; Higgins and Gochis 2007). These include large-scale orographic dynamic and thermodynamic modulations on the atmospheric circulation; teleconnections with tropical and North Pacific sea surface temperatures (SSTs) and remote land surface conditions; and effects of the low-level jets (LLJs) and accompanying moisture transport or surges from the Gulfs of California and Mexico, the Madden–Julian oscillation (MJO), and the tropical easterly waves (TEWs). A complete, accurate representation of these processes is currently not possible, leading to a common failure by global general circulation models (GCMs) and mesoscale regional climate models (RCMs) in simulating many key characteristics of NAM precipitation spatial and temporal variations [NAME 2005; see also the Eighth North American Monsoon Experiment SWG Meeting (http://www.eol.ucar.edu/projects/name/science_planning/SWG8) and the Ninth North American Monsoon Experiment SWG Meeting (http://www.eol.ucar.edu/projects/name/science_planning/SWG9)].

While MJO and TEWs remain difficult to simulate in general (Zhang 2005; Zhang et al. 2006; Waliser et al. 2006; Wu et al. 2007; Lin et al. 2008), the large-scale orographic effects (ORO) and teleconnections (TEL) are supposed to be resolvable by GCMs, and the LLJs should be better represented by RCMs. The relative contributions of these phenomena on the NAM precipitation seasonal–interannual variability, however, are not known. If the MJO/TEWs are dominant, then all current models would fail; if LLJs are critical, then only RCMs may succeed (Stensrud et al. 1995; Zou and Zheng 2004; Xu et al. 2004; Ting et al. 2005; Liang et al. 2007); in both cases, no current GCM could resolve the NAM. On the other hand, if ORO/TEL play a central role, certain GCMs may capture major observed signals and, likely, their associated physical processes (Arritt et al. 2000; Meehl et al. 2006). In this case, we can then ask questions concerning the GCM ability to simulate the NAM variability. For example, if the SST forcing is a determining factor (Carleton et al. 1990; Reyes and Mejía-Trejo 1991; Harrington et al. 1992; Higgins et al. 1998, 1999; Higgins and Shi 2000; Castro et al. 2001), we should see important differences as the ocean component of a GCM is changed; if the model resolution is a crucial factor (Mo et al. 2005; Collier and Zhang 2007; Lee et al. 2007), we would expect a refinement in grid spacing of a GCM or between GCMs to significantly improve the simulation. This study uses the most recent, complete suite of simulations from all available fully coupled atmosphere–ocean CGCMs to address these issues.

In preparation for the Fourth Assessment Report, the Intergovernmental Panel on Climate Change (IPCC) has engaged worldwide CGCM modeling centers to produce historical climate simulations of the twentieth century (1901–99). These newly available simulations provide a unique opportunity to investigate the current NAM modeling capability. The objectives of this study are twofold: to determine whether the present generation of CGCMs, as driven by modern estimates of time-varying anthropogenic plus natural external forcings, can reproduce key observed features of the NAM precipitation seasonal–interannual variability and to conduct comparisons between models and against observations to gain insight into the underlying causal mechanisms for success or failure. Some CGCMs have multiple realizations, differing only in initial conditions; the spread between them measures how natural variability plays a role. Some CGCMs include runs with different oceanic GCMs; their contrasting results indicate the sensitivity to the sea surface modeling. In this regard, we also use integrations from AGCMs where SSTs are prescribed as observations during 1979–2002. Comparing simulations by the same GCM when SSTs are observed versus predicted by different interactive oceanic GCMs enables a determination of the importance of the air–sea interactions and the external forcings. Comparison between GCMs can identify systematic biases or consistent dependencies, such as on model resolution.

The NAM precipitation variability is characterized by two distinct regimes over the southwest United States and central–northwest Mexico (Hu and Feng 2002) and by a clear northward progression identified with different monthly circulation changes during June–September (Douglas et al. 1993; Adams and Comrie 1997; Higgins et al. 1997a). Our analyses are thus based primarily on monthly means. Note that the NAM produces over twice as much summer precipitation amount and interannual variability in central–northwest Mexico than in the southwest United States (3.16 versus 1.22 and 0.45 versus 0.20 mm day−1). It is also known that the NAM is most spatially consistent over central–northwest Mexico with greater variability to the north (e.g., Adams and Comrie 1997). Thus, the focus is on the core monsoon region, encompassing central–northwest Mexico and a small portion of Texas (Fig. 1). This region was defined by a correlation analysis to locate where observed precipitation interannual variations are highly coherent. Examined are June–September monthly telecorrelation patterns of precipitation year-to-year variations over the domain with those at a select small area (6 × 6 30-km grids) from the central Mexico, while only the average of these patterns is shown as a blowup insertion in Fig. 1. Clearly the spatial coherence or teleconnection of precipitation variability within the core monsoon region is excellent, having correlations generally exceeding 0.6.

A major focus of the present study is to seek the signatures of the NAM precipitation variability in SSTs and planetary circulation fields, which are more likely resolvable by GCMs. To facilitate this, we first identify key centers where observed SSTs and circulation fields are significantly correlated with the NAM precipitation averages over the core region and then examine if the modeled and observed patterns agree. Our goal is to answer whether the current CGCMs have the ability to simulate the NAM precipitation seasonal–interannual variability and, if so, whether the key underlying physical mechanisms are correctly represented.

2. Model simulations and observations

Two sources of GCM climate simulations are used. The primary one, denoted as 20C3M, is the historical “climate of the twentieth century” simulated by 16 separate CGCMs from 13 different modeling groups. The basic model information is given in Table 1 (more details are available online at http://www-pcmdi.llnl.gov/ipcc/model_documentation/ipcc_model_ documentation.php). Here the CGCMs were driven by estimated historical forcings of anthropogenic and natural changes, including greenhouse gases, solar variations, volcanic eruptions, land use, ozone, halocarbons, sulfate, and other aerosols. Following the Coupled Model Intercomparison Project (CMIP) (Meehl et al. 2000), the modeling groups were free to use their own best estimates of these forcings. All simulations with documentation included the key forcings from greenhouse gases and sulfate emissions, although other forcings were not common and the direct and indirect effects of sulfate aerosols were not consistent either. This study analyzes the 20C3M simulations in the 30-yr period 1961–90 and, when comparing their temporal correspondences with observations, identifies the systematic signature, if any, of these external forcings on the NAM variability. Multiple realizations with the same forcing but different initial conditions were available for 10 CGCMs, resulting in a total of 51 separate simulations. The spread between multiple realizations of a same model defines the uncertainty due to internal variability in isolating the externally forced signals. Since each CGCM contains an interactive ocean component that predicts rather than specifies ocean surface conditions, the resulting SSTs can differ among models and from observations. Diagnosis is thus conducted to determine if the climate biases in the SSTs and NAM precipitation are physically related.

The other source is the historical climate of the period 1979–2002 simulated by 13 separate AGCMs under the protocol of the Atmospheric Model Intercomparison Project (AMIP) (Gates et al. 1999), where the same “perfect” ocean surface conditions, as specified by the global distributions of the observed monthly mean SST and sea ice evolutions, were incorporated to determine the heat, moisture, and momentum fluxes that drive the atmosphere. Each model also used the identical atmospheric CO2 concentration (345 ppm) and solar constant (1365 W m−2), while specification of the land surface and inclusion of the radiative effects of other greenhouse gases varied. Although a few models had multiple realizations with different initial conditions, this study only uses the first run of the ensemble. There are four modeling groups, Geophysical Fluid Dynamics Laboratory (GFDL), Goddard Institute for Space Studies (GISS), Meteorological Research Institute (MRI), and National Center for Atmospheric Research (NCAR), that have both 20C3M and AMIP simulations using the same AGCMs. This study only utilizes these four groups’ AMIP results and compares them with their respective 20C3M simulations to demonstrate the important role of SST forcings or air–sea interactions on the NAM precipitation annual cycle.

The observational data come from several sources. Over the land area of the United States and Mexico, precipitation data are from the analysis of Higgins et al. (2000) at a grid spacing of 0.25° and 1°, respectively; surface air temperature (T2m) data are from the 0.5° analysis of Mitchell and Jones (2005); both were based on station measurements. Over the global oceans, SST data are from the 1° analysis of Ishii et al. (2006). For circulation fields (wind, height, sea level pressure), we use the 2.5° National Centers for Environmental Prediction (NCEP)–NCAR reanalysis (NRA) (Kalnay et al. 1996). All data cover the entire period from 1961 to 1990 analyzed in this study.

Because of the various spatial resolutions in the GCM simulations and observations, direct comparisons must be based on a common grid mesh that is most representative of the suite of data used and the regional climate characteristics to be studied. GCM simulations use grids ranging from 1° to 5° (Table 1), while observations have data resolutions as fine as 1° over oceans and 0.25° over land. Our strategy is to map GCM simulations through bilinear spatial interpolation onto the finer observational data grid. Such mapping does not enhance the data resolution in quality, but is simply a data manipulation for easy comparison between models and with observations as well as RCM downscaling results to be presented in forthcoming papers. Precipitation is the key variable of interest and has a smaller scale of variability than others; thus we adopt the RCM 30-km grid over a domain including the United States, Mexico, and adjacent oceans [Liang et al. (2004); also see Fig. 1] to retain the observed details. For interpretation of the underlying physical processes, this same grid mesh is chosen for mapping the circulation fields. To explore the surface memory impact on the NAM precipitation variability, the oceanic SSTs and land T2m data are combined to produce a complete distribution of “surface” temperatures at the 0.5° grid over the globe.

Unless otherwise specifically stated, 1961–90 and 1979–2002 are chosen as the reference periods for the 20C3M and AMIP simulations, respectively, to construct the annual cycle (i.e., average over the entire duration) and interannual anomalies (i.e., departures from the average). The same periods apply in any comparison with observations. In the case of interannual variability, monthly (instead of summer mean) series are used to capture the characteristics of the monsoon progression.

This study relies on correlation analysis to identify key climate signals and understand underlying physical processes. Statistical significance of correlation coefficients is determined through the following Monte Carlo technique to take into account the effects of number and interdependence (Livezey and Chen 1983). In the case of NAM precipitation correlations between observations and simulations, the observed time series is randomly reordered and then used with the modeled series to calculate a correlation coefficient. In the case of correlations between NAM precipitation and pointwise climate variables (e.g., wind, SSTs), observed or modeled, the NAM precipitation time series is randomly reordered and then used to calculate a correlation coefficient. As such the spatial structures in the climate variables are preserved. This is repeated 5000 times and the resulting distribution of correlation coefficients determined the 95% level of confidence thresholds. Values exceeding such thresholds are defined here statistically significant. This definition is applied throughout the paper. The impact of the autocorrelation on the statistical significance is examined by repeating the Monte Carlo calculations using chunks of two consecutive years in the random reordering. For all of the fields presented below, such impact is small and does not affect the major centers (locations and areas) of statistically significant correlations.

3. Annual cycle of precipitation

Figure 2 compares seasonal (June through September) mean precipitation averaged over the core monsoon region during 1948–2004 from three different datasets: Higgins et al. (2000), Mitchell and Jones (2005), and the Mexico 18 climate diversions’ data archived by Arthur V. Douglas (http://flare.creighton.edu/douglas/data_sets.htm). The precipitation values from Higgins et al. are considerably lower than those in the other two datasets prior to 1960 and after 1990. The reasons for these discrepancies are not clear. However, the good agreement between the three datasets for 1961–90 provides high confidence in the validity of the values during this period. Therefore, the observational climatology and subsequent comparisons with CGCM or AMIP simulations are based on this period, using the precipitation data of Higgins et al. Seasonal mean precipitation during 1961–90 (Fig. 2) generally varies between 3 and 4 mm day−1, with drought in 1969 (2.5 mm day−1) and floods in 1984 and 1990 (4.8 mm day−1).

The annual cycle of the NAM core regional precipitation exhibits a minimum in early spring, followed by a rapid rise in late spring, peaking in July. There is a rapid decline in the fall to a secondary minimum in November. Figure 3 compares two characteristics between the modeled and observed annual cycle: the root-mean-square (rms) error and the phase lag (in number of months) yielding the maximum correlation, both calculations based on 12 monthly climatological values. The rms errors vary from about 0.46 to 2.23 mm day−1. The lowest rms errors (0.46–0.64 mm day−1) are for the MRI. This model also shows no shift in the phase. Most other models produce larger rms errors and/or phase shifts of 1–2 months, about equally divided between earlier and later than observed.

Figure 4 compares with observations the CGCM simulated NAM core region precipitation annual cycle and illustrates the sensitivity to model configurations from the same group. The annual cycle for the five MRI simulations (Fig. 4a) exhibits very good agreement with observations in terms of both the amplitude and phase of the cycle, except for some overestimations in August and winter; the sensitivity to initial conditions is relatively small, which applies to all other CGCMs. For both NCAR models (Fig. 4b), the monsoon onset is early and the intensity is higher than observed; as compared with the Parallel Climate Model (PCM) (Washington et al. 2000), the latest Community Climate System Model (CCSM) (Collins et al. 2006) with higher resolution and updated physics representation does reduce the amplitude error by half (see also Meehl et al. 2006). For the Model for Interdisciplinary Research on Climate (MIROC), the high resolution produces a better simulation of the phase, but the magnitude is more accurate in the medium resolution (Fig. 4c). The GISS Model E-R (ER) and E-H (EH) have the same atmosphere but different ocean models; the ER simulates a monsoon with somewhat early onset but similar magnitude as observed; by contrast, the EH does not produce a monsoon, with precipitation evenly distributed throughout the year (Fig. 4d). There are several formulation differences between the GFDL Climate Model versions 2.0 (CM2.0) and 2.1 (CM2.1), including the dynamical core of the atmosphere model, cloud tuning, and land scheme (Delworth et al. 2006); the CM2.0 produces a late monsoon with the correct amplitude; the CM2.1 monsoon peaks later than observed and has a too strong magnitude (Fig. 4e). The Canadian Centre for Climate Modelling and Analysis (CCCMA), Commonwealth Scientific and Industrial Research Organisation (CSIRO), Max Planck Institute (MPI), and Met Office (UKMO) all correctly simulate the phase (Fig. 4f), whereas the Centre National de Recherches Météorologiques (CNRM), Institute of Atmospheric Physics (IAP), Institute of Numerical Mathematics (INM), and L’Institut Pierre-Simon Laplace (IPSL) all exhibit large phase errors (Fig. 4g); both groups of these models are identified with different degrees of amplitude errors.

Also shown in Fig. 4 are the AMIP simulations corresponding to four CGCMs where the oceanic component is prescribed with observed SST variations. For the MRI, the prescription of SSTs has little effect on the NAM precipitation throughout the annual cycle (Fig. 4a). Similarly, for the CCSM, the SST prescription causes small precipitation change in summer, albeit reducing the positive biases from October through April (Fig. 4b). On the other hand, for the GISS, in addition to the dramatic contrast between the EH and ER (differing only in the oceanic models) results discussed above, the AMIP run produces another largely different annual cycle with a correct monsoon onset but substantially underestimated precipitation amount in July–September and still wetter winter (Fig. 4d). Lin (2007) showed that the GISS atmospheric model produces certain incorrect air–sea feedbacks but, when coupled with different oceanic models, ER tends to cancel the errors while EH fails. Likewise, for the GFDL, the AMIP result, based on the same atmospheric model of the CM2.1, is much closer to CM2.0 in May–August and overall more realistic in the phase and magnitude than both of the coupled models (Fig. 4e). These comparisons depict a mixed picture: the MRI and CCSM simulations show little change in the NAM precipitation annual cycle between the predicted and prescribed SSTs, suggesting that these CGCMs may have captured the responsible oceanic processes whereas the GISS and GFDL results indicate the critical role of SST forcings and/or air–sea interactions on the NAM cycle, implying that further improvements may be needed for their coupled oceanic models.

A comparison between the observed monsoon climate and the MRI run 3, the simulation most closely matching the observed precipitation annual cycle, indicates that the model captures important features of the climate processes associated with the NAM. The observed rapid increase in precipitation from June to July and the decrease from August to September (Figs. 5a,b) are simulated rather well, though the central core is displaced slightly to the east and extended insufficiently to the north (Figs. 5c,d). The observed switch from westerlies in June to easterlies in July at 200 hPa and reversal from August to September (Figs. 5e,f) are reproduced in both magnitude and position by the MRI (Figs. 5g,h). On the other hand, the observed precipitation over the southern Great Plains exhibits an out-of-phase relationship with the NAM from June to July and in a reverse tendency from August to September (Figs. 5a,b; Higgins et al. 1997b), whereas the MRI fails to simulate this feature as identified with stronger 200-hPa northerly and southerly flows, respectively.

4. Interannual variability and associated regional climate circulation

An immediate question is whether the observed NAM precipitation interannual variability can be simulated by the current GCMs. Temporal correlations of the observed NAM core precipitation with all 20C3M simulations are, in general, not statistically significant for all summer months and are about equally divided between negative and positive values except in July, which is mostly negative. This lack of time correspondence, indicating that the historical forcings of anthropogenic and natural changes are not the determining factor for the observed NAM core precipitation interannual variability, is not surprising since the changes in external forcing are slow compared to interannual variability. However, similar temporal correlations with four AMIP simulations (GFDL, GISS, MRI, CCSM) are also not statistically significant, having mostly negative coefficients except in June with all positive and one significant value of +0.55 by the CCSM. The corresponding correlations based on summer averages for both 20C3M and AMIP simulations are in the range of −0.29 to +0.20, all statistically not significant. As such, the specification of the observed global SST variations does not capture the NAM interannual variability with one possible exception for the June monsoon onset. This is in contrast with the result of the annual cycle, where SST forcings and/or air–sea interactions may play a certain role as simulated by the GISS and GFDL (Figs. 4d,e).

Given the general failure in direct correspondence with the observed NAM core precipitation evolution, the next question is whether the GCMs can reproduce the underlying physical processes. To address this issue, we first identify the key observational signatures that link the NAM precipitation with the large-scale circulation patterns and then evaluate the GCMs’ ability to depict these patterns. As such, the 1961–90 time series of precipitation averaged in the core monsoon region was correlated with geographic distributions of wind at 850, 500, and 200 hPa, sea level pressure, and surface temperature, for each month of summer (June–September). The most robust correlations were found with the meridional wind component at 200 hPa (Figs. 6a–d). There are high positive correlations centered over the northwest of the core monsoon region in June and September. This indicates that wet conditions are significantly correlated with anomalously strong southerly upper flow for the monsoon transition (onset and retreat) periods. During the peak monsoon period, July and August, the sign of the correlations are reversed (wet conditions associated with anomalous northerly flow) and the center of strongest correlation shifted to the southeast of the core monsoon region.

Meanwhile, significant negative correlations (Figs. 6e,f) are found with the 200-hPa zonal wind component across the southern United States in June and the central United States in July. Along with the meridional component, these correlations identify the NAM wet conditions with an anticyclonic circulation pattern in June and, reversely, a cyclonic pattern in July and dry conditions with an opposite pattern. In contrast, correlations with the zonal wind in August and September are generally weak (Figs. 6g,h). As discussed below, the cyclonic or anticyclonic features are more distinguishable in June and July than in August and September.

The substantially different circulation correlation patterns in July/August versus June/September disqualify the use of averages over the whole summer. In fact, the magnitude of precipitation correlations with the 200-hPa wind based on June–September means is generally less than 0.3, and hence not statistically significant. This may be one reason why previous studies have not identified such distinct precipitation–flow coupling patterns outlined above.

A better understanding of the actual circulation patterns associated with these correlations was determined by compositing the five wettest and driest years, separately for each summer month. Figure 7 illustrates interannual variations during 1961–90 of the normalized NAM precipitation anomalies for each month and the whole summer average. No single year was identified with systematic wet or dry conditions throughout the summer months. The year 1984 was extremely (very) wet for June (July) but quite dry for September, while 1990 was extremely (very) wet for July (August, September) but nearly normal for June. When averaged over summer, 1984 was the record wet year, followed by 1990. In contrast, 1982 had the driest summer, peaking in August, while 1969 was ranked second with severe drought in June, August, and September. The NAM precipitation autocorrelations between any pair of June, July, August, and September are all not statistically significant, having 0.21 for June/July and within ±0.10 for the rest. Therefore, it is desirable to diagnose interannual variability, including correlation and composite analyses, by individual months.

The 200-hPa wind distributions for wet and dry years and the differences between the two (Fig. 8) were compared with similar composite maps (not shown) for 850-hPa wind, 500-hPa height, and sea level pressure. During the transition month of June, wet conditions presumably reflect an earlier onset of the monsoon; these are associated with an early northward retreat and weakening of the upper-tropospheric westerlies and an early strengthening and westward extension of the North Atlantic subtropical surface high pressure system, bringing in moist lower-level southeasterly winds (not shown). Wet conditions in July are associated with a strong ridge at 200 hPa centered over the northwest part of the core monsoon region, probably reflecting the strong, warm core convective systems characteristic of active monsoons. When July conditions are dry, this ridge is shifted to the northeast of the core area. Wet conditions are also associated with anomalously strong southerly flow at 850 hPa over the monsoon area, an anomalously strong Atlantic subtropical ridge at the surface that enhances the southerly advection of moisture, and anomalously low 500-hPa heights probably reflecting greater convective instability. Wet conditions in August somewhat resemble the July features; this includes anomalous southerly flow at 850 hPa and below normal 500-hPa heights; at 200 hPa the location of the ridge is similar during wet and dry conditions, but the ridge is stronger in wet years; however, there is little difference in the strength of the Atlantic subtropical surface high during wet and dry years. Wet conditions in September are associated with an eastward shift of the 200-hPa ridge and an anomalous 500-hPa trough above the core; these conditions may provide steering of eastern Pacific tropical cyclones into the area and may also reflect very early season baroclinic systems.

The above wet NAM conditions and teleconnection features are often identified with a cyclonic circulation anomaly pattern over the central Great Plains for the July–August peak monsoon and, reversely, an anticyclonic anomaly pattern centered over the northern (southern) Great Plains for the June (September) transition. To our knowledge, this distinction of the cyclonic from anticyclonic anomalies has not been made previously. Most often, anticyclonic anomalies were described as a dominant feature at the upper level. This notion may arise from several reasons. First the definition for wet/dry NAM conditions differs. For example, Higgins et al. (1998) mainly concerned the southwestern monsoon, using an index of precipitation averaged over 32°–36°N, 112.5°–107.5°W almost entirely outside of our core area, while Higgins et al. (1999) separated the NAM into three subregions, Arizona–New Mexico (>31°N), northwest Mexico (31°–25°N), and southwest Mexico (<25°N), with the later two areas overlapping our definition (21°–31°N). On the other hand, Yu and Wallace (2000) focused on a much broader domain (5°–35°N, 80°–125°W), encompassing not only the continental regions with pronounced monsoonal precipitation signature but also adjacent oceans to the south of Mexico and Central America. As such the list of anomalous wet/dry years for the composite differ between these studies and from ours (Fig. 7). Second, all of these three studies presented the results for the bulk of summer average. As discussed above, this average masks the actual signals that change dramatically from month to month. For a comparison, the composites based on the June–September means using our core definition are also illustrated in Fig. 8. The wet minus dry pattern is now characterized by the cyclonic (over the southwest United States) and anticyclonic (across the eastern–southern United States and Mexico) anomaly structure, inclined toward the June/September feature. This pattern is similar to the result of Higgins et al. (1999) for the southwest Mexico wet–dry conditions (their Fig. 20).

The statistical robustness of the results shown in Fig. 8 is not readily assessable. As a simple solution, we examine the anomaly patterns in individual wet or dry years for each month of June to September. It is found that the principal anomaly signals are manifested in the respective month for individual wet or dry conditions. In most cases, the 200-hPa wind circulation anomalies are distinguishable over those areas where correlations are statistically significant, as indicated in Fig. 6. Among the 40 monthly wet or dry cases, 33 are correctly accompanied by the cyclonic or anticyclonic patterns, as discussed above, although noticeable differences exist in the center location and magnitude. For the remaining seven cases, marked by (×) in Fig. 7, five occur in August and September when correlations with zonal wind component are generally not significant (Figs. 6g,h). Note that August 1982 was observed with the strongest westerly jet stream anomaly of the record, dominating the composite.

The MRI (run 3) simulation (not shown) reproduces, in general, many of the observed features, discussed above, including the early retreat of the westerlies in June, the enhanced ridge at 200 hPa in July and August, the eastward shift of the 200-hPa ridge in September, the anomalous southerly flow at 850 hPa, and westward extension of the Atlantic subtropical high, as well as their frequently accompanying cyclonic (July, August) and anticyclonic (June, September) 200-hPa circulation anomaly patterns. The result indicates that the MRI realistically captures the NAM precipitation annual cycle and interannual covariations with the key atmospheric circulation patterns.

The correlation between precipitation in the core monsoon region and the meridional wind component at 200 hPa in the boxes centered on the regions of highest observed correlations in Fig. 6 was used as one metric of the CGCMs model performance in simulating the key NAM circulation features. As a reference, a correlation coefficient with a magnitude exceeding 0.35 is approximately equivalent to the statistical significance at the 95% confidence level by the Monte Carol test (see section 2). For June (Fig. 9a), a large majority of the models exhibit positive correlations, similar to observed, and 24 of the 51 simulations have coefficients greater than +0.35, compared to +0.44 in observations. For both July (Fig. 9b) and August (Fig. 9c), most models have negative correlations, as observed, but the magnitude of the coefficients are in most cases considerably less than observations (around −0.6); only 14 in July and 17 in August simulations have statistically significant correlations (<−0.35). For September (Fig. 9d), the observed positive correlations are again reproduced in most models; 26 simulations have values exceeding +0.35, although only one (MRI run 5) is as high as +0.68 in observations. Considering all results, only the MRI simulates correlations with the correct sign and comparably high magnitude in all four months. The better agreement among all CGCMs with observations for June and September may reflect the general large-scale circulation determination of the interannual variability for monsoon transitions, while variability of the peak monsoon in July and August may also be largely influenced by local processes (such as stability, moisture recycling, and flow interactions with topography), which are more challenging for coarse-resolution global models.

One may argue that modeled correlation patterns may still exist between the NAM core precipitation and 200-hPa meridional wind component interannual variations, but the “action” centers are shifted. Given that the teleconnection patterns cover broad areas, small shifts of these centers are not supposed to significantly change the correlation metric shown in Fig. 9, whereas large shifts are considered here as a model failure. One may also argue that most GCMs fail to reproduce the observed annual cycle phase and magnitude (Figs. 3 and 4), and thus the June–September average would be more representative of the model overall feature. However, as discussed above, observations did not exhibit an obvious teleconnection pattern when using the summer average, and hence there is no point to evaluate the model ability in this context. In addition, Fig. 6 reveals strong intraseasonal variations of the circulation patterns that are phase locked to the monsoon transition, that is, different circulation patterns in the onset and withdrawal phases. It is not surprising for the GCMs simulating the monsoon onset shifted by 1–2 months (Fig. 2) to produce poor precipitation–circulation correlations (Fig. 9). It would be more desirable to conduct the correlation analysis based on the monsoon transition phases than calendar months. This would require the use of daily data, warranting future investigation.

The observed precipitation in the core monsoon region shows broad areas of sizeable positive correlations with SSTs in the eastern Pacific during June and July, with a displacement farther north in July (Fig. 10). There is also an area of negative correlations in the Gulf of Mexico during June, but weaker in July. The result indicates that wet NAM core conditions are associated with a pattern of positive SST anomalies in the eastern Pacific and negative SST anomalies in the Gulf of Mexico, acting to reduce the climatological mean (Fig. 10f) gradient between the two oceans. Note that the June NAM core precipitation correlations with the SST anomalies in the eastern Pacific (+0.39) and the Gulf of Mexico (−0.43) are enhanced for their gradient (+0.58). A similar enhancement occurs in July. This gradient pattern is consistent with possible forcing on the sea level pressure distribution which would tend to favor a westward extension of the North Atlantic subtropical ridge. Correlations are much weaker in August and September.

Previous studies have not emphasized the association of NAM precipitation interannual variations with the gradient of SST anomalies between the eastern Pacific and Gulf of Mexico. The signature however appeared in the July–August–September mean wet–dry SST anomaly composite of Higgins et al. (1998) for the southwest U. S. monsoon (their Fig. 15c). For a comparison, the observed interannual correlations of the NAM precipitation with SSTs based on the June–September averages are shown in Fig. 10e. This summer mean pattern is inclined toward the mixture of the June and July features, and still depicts the statistically significant role of the SST gradient.

Several important land signatures exist over North America for abnormal NAM precipitation (Fig. 10). It is not surprising that precipitation interannual variations are negatively correlated with surface temperature in the core monsoon region throughout June–September; significant negative correlations are confined to the core area in the transition periods (June and September) but expanded to cover the entire Mexico and Texas during the peak monsoon, especially strong in July. More importantly, wet June conditions are remotely associated with warmer surface temperature anomalies over most of the United States (with significant correlations centered in the southern Great Plains and the Midwest) and southern central Canada. In contrast, wet July conditions are identified with two separate remote warmer temperature centers: northwest United States and northeast United States–southeast Canada. In August, these two centers become weaker and confined to the coastal areas. In September, a single positive correlation center exists in the southeast United States–Gulf of Mexico. However, most of correlations in August and September are not statistically significant, and the physical processes responsible for the June and July land surface temperature patterns are not known.

The MRI (run 3) simulation (not shown) captures the key characteristics of the June precipitation–temperature correlation patterns, including negative centers in the core region and the Gulf of Mexico and positive centers over most of the United States and in the eastern Pacific. For the other three summer months, the negative center around the core monsoon region is well reproduced, but all remote centers are largely displaced.

The correlation of precipitation in the core monsoon region with the SST gradient between the observed positive center in the eastern Pacific and negative center in the Gulf of Mexico (see the boxes in Fig. 10) was used as one metric of the CGCMs model performance in simulating the key NAM physical mechanisms. Model correlations were calculated for June and July. For June (Fig. 11a), correlations are positive for a majority of the models, similar to observed; however, the magnitudes are mostly underestimated by a considerable margin; only 13 of 51 simulations have correlations greater than +0.35, compared to +0.58 in observations. For July (Fig. 11b), 20 model simulations have an incorrect sign, while 12 others have values less than +0.20, compared to +0.49 in observations. The corresponding correlations for the MRI (run 3) are +0.44 (weaker than observed) in June and only +0.14 (not statistically significant) in July.

Interestingly, the CSIRO is the single model that well reproduces both June and July correlations of precipitation in the core monsoon region with the SST gradient between the eastern Pacific and Gulf of Mexico, the coefficients of +0.58 and +0.60 being comparable to observed. This model simulates the NAM precipitation annual cycle with no phase lag but a much larger amount than observed (Fig. 4f). More troublesome is that the CSIRO generally fails to simulate the observed strong link between precipitation and the 200-hPa meridional wind component (Fig. 9), with much weaker (except August) correlation coefficients of +0.18 (June), −0.34 (July), −0.64 (August), and +0.45 (September). These results are less physically consistent than those of the MRI. In addition, the NAM core precipitation link to remote SSTs in observations and among CGCMs (Fig. 11) is relatively less coherent compared with its link to the 200-hPa meridional wind component (Fig. 9). This suggests that the NAM core precipitation interannual variability may likely be determined by large-scale circulation anomalies, while its predictability based on remote signals such as SSTs may not be sufficiently robust and not well captured by current CGCMs.

Since observations did exhibit a significant pattern in the summer average, the same metric based on the June–September mean SST gradient is also illustrated in Fig. 11c. Here the area for the eastern Pacific center (see the outline in Fig. 10e) is expanded to cover both the June and July action centers, while that for the Gulf of Mexico remains the same. Again, the result is inclined toward a mixture of the June and July features, with small contributions from August and September. Only 10 of 51 simulations have correlations above +0.35, compared to +0.55 in observations. As in June and July, the GISS remains the worst model in capturing this summer mean relationship, with small or even significantly negative correlations. The failure is indicative of incorrect representation of air–sea feedbacks in the GISS atmospheric model (Lin 2007). In addition, IAP (run 3) and PCM (run 1) produces the highest summer mean correlations with the SST gradient, respectively of 0.51 and 0.50. These two models happen to very poorly simulate the NAM precipitation annual cycle (Fig. 4).

For the metric of correlations with both 200-hPa meridional wind (Fig. 9) and SSTs (Fig. 11), model results are sensitive to initial conditions, indicating that internal variability provides an important contribution to the NAM precipitation interannual variations and associated circulation patterns; hence the ensemble prediction is desired. Impacts of initial conditions, mainly due to differences in land surface states, on the U.S.–Mexico precipitation interannual variability were also demonstrated by Mo et al. (2006). This sensitivity, however, does not significantly change the key results for specific models discussed above.

5. Possible reasons for model climate biases

It is imperative, but extremely difficult, to understand the exact physical processes and underlying mechanisms that cause CGCM climate biases and differences among the models. In particular, we wish to explain why the MRI is so much better than all other CGCMs in simulating the NAM precipitation annual cycle. In addition to the discussion in sections 3 and 4, we have searched for other possible causes from the synthesis of all model simulations as compared with observations. First, the biases of precipitation in the NAM core region from all 20C3M simulations were compared against CGCM atmospheric or oceanic model horizontal or vertical resolution; no systematic link was found for each month and the average of the summer months as well as the rms error or phase lag of the full annual cycle. For a particular model, such as the CCSM and MIROC, higher resolution may improve certain limited aspects but worsen others. In fact, the MRI has only a medium resolution but most realistically simulates the NAM precipitation annual cycle and associated circulation patterns, much better than the high-end CCSM and MIROC. Second, there appears no direct correspondence of the NAM precipitation biases to specific physics formulations. For example, a critical modeling component is the parameterization for deep convection (e.g., Gochis et al. 2002; Liang et al. 2007), with a total of nine schemes among 16 CGCMs differing in convective closure, trigger function, and cloud model (Table 1); none of these schemes nor a specific characteristic can differentiate the biases and intermodel differences as shown in Figs. 3 and 4.

Another potentially important factor is the flux correction applied to prevent ocean models from serious climate drifts. Most CGCMs are freely coupled without flux correction. There are three exceptions: the MRI incorporates flux correction for heat, water, and momentum (Yukimoto et al. 2001); the CCCMA applies that for heat and water (Flato and Boer 2001); and the INM only accounts for water (Diansky and Volodin 2002). None of these models with flux correction was found to be more realistic than others in simulating the intertropical convergence zone precipitation and SST annual cycle over the tropical eastern Pacific (de Szoeke and Xie 2008). Although the MRI depicts an almost perfect annual cycle of the NAM core precipitation, the CCCMA underestimates largely the amplitude, whereas the INM produces one of the worst simulations (Fig. 4). As such, the application of oceanic flux correction may not be the major source responsible for the MRI superiority.

A credible NAM simulation may require a reasonable representation of the SST annual cycle distribution over the adjacent oceans. Figure 12 compares observed and simulated SST annual cycles averaged over the eastern Pacific covering both the June and July action centers (see Fig. 10e for the area specification). Among 16 CGCMs, the MRI most realistically simulates the SST annual cycle, especially in the summer months, having errors less than 1°C. It is followed by the CSIRO, CNRM, UKMO, and CCCMA, with summer positive errors of 1°–2°C. All other models substantially overestimate summer SST by 3°–4°C; some (GISS-ER, MPI, IAP, IPSL) are more systematic throughout the year, while others (GISS-EH, MIROC, GFDL, INM, PCM) warm more in summer than in winter, enhancing the seasonal amplitude; one exception is the CCSM, which produces warm biases most severe in spring, making seasonal transition from winter to summer very smooth. The general warm SST biases (except for MRI) may result from the common failure in modeling the prevailing stratus–SST feedbacks over the region (Lin 2007).

For SSTs over the Gulf of Mexico (see Fig. 10e for area specification), the MPI and MIROC best reproduce the whole annual cycle, followed by CCSM and MRI, while the PCM becomes the worst mainly because of its substantial underestimation in winter. Note that, in observations (Fig. 12h), the SSTs over the Gulf of Mexico peak in August, about one month lead and 2°–6°C warmer than those over the eastern Pacific; the difference between the two regions is strongest in June and weakest in February. Figure 13 compares observed and simulated SST gradients between the eastern Pacific and Gulf of Mexico for the whole summer and individual months. The UKMO and MRI most accurately simulate the gradient, having respectively ∼1°C stronger and weaker than observed. The PCM is the worst, followed by the GISS-ER, with underestimation by a factor of 3 or more. All other models also underestimate the gradient by a range of 25%–70%.

As compared with Fig. 4, the best SST performer MRI is also the one that most realistically simulates the NAM core precipitation annual cycle, while the followers (CSIRO, UKMO, CCCMA) happen to be the ones that produce no phase lag but large amplitude biases in the precipitation cycle. In this regard, accurate simulation of the eastern Pacific SST annual cycle, and perhaps also its gradient to the Gulf of Mexico, seems critical for the NAM precipitation modeling. This however raises many questions, for example, why the CNRM (good in SSTs) is so poor in precipitation, why the CCSM and GISS do not resolve the precipitation problems even when forced by global observational SSTs, and why UKMO significantly underestimates summer NAM precipitation amount and its interannual coupling with the 200-hPa meridional wind component (Fig. 9). Therefore other atmospheric and air–sea interactive factors must also play a central role.

6. Conclusions

This study presents a comprehensive assessment of the ability of current CGCMs in simulating the NAM core precipitation seasonal–interannual variability and associated circulation patterns and physical mechanisms. Major conclusions are drawn below separately for the annual cycle and interannual variation, followed by a discussion on future investigation.

a. Annual cycle

Of the 16 CGCMs, the MRI model most realistically simulates the NAM core precipitation annual cycle, with smallest rms error and no phase shift, and also captures the important features of the climate processes identified with the cycle. In particular, the observed rapid precipitation increase and the associated upper wind switch from June to July and reversal from August to September are reproduced in both magnitude and position. Most other models produce larger rms errors and/or phase shifts of 1–2 months, about equally divided between earlier and later than observed.

Various analyses were conducted to understand possible causes for CGCM biases in simulating the NAM core precipitation annual cycle. First, they cannot be explained by internal variability due to perturbations in initialization. Second, they have no systematic link to model horizontal or vertical resolution of the atmospheric or oceanic components. For a particular model, higher resolution may improve certain, but limited, aspects. Third, there appears no direct correspondence of the NAM precipitation biases to specific physics formulation, for example, deep convection parameterization or oceanic flux correction. Fourth, the impacts from the representation of SST forcing and/or air–sea interaction are mixed: some models produce little response to predicted versus prescribed SST while others yield significant differences as the oceanic component is changed. Accurate simulation of the eastern Pacific SST annual cycle, and perhaps also its gradient to the Gulf of Mexico, seems critical for the NAM precipitation modeling, but other atmospheric factors and air–sea interaction must also play a central role.

b. Interannual variability

Observational data diagnoses revealed two important characteristics of the NAM precipitation interannual variations, which have not been identified previously. First, there are significant positive (negative) correlations with 200-hPa meridional wind centered over the northwest (southeast) of the core region in June and September (July and August). These correlation patterns reflect coherent large-scale circulation changes, which differ as the monsoon progresses. Wet conditions are typically associated with 1) an early northward retreat and weakening of the upper-tropospheric westerlies and an early strengthening and westward extension of the North Atlantic subtropical high during the June onset, 2) a strong ridge at 200 hPa centered over the northwest part of the core (depicting active warm convection) during the July and August monsoon peak, and 3) an eastward shift of the 200-hPa ridge and an anomalous 500-hPa trough above the core (steering eastern Pacific tropical cyclones into the area) during the September retreat. They are often identified with a cyclonic circulation anomaly pattern over the central Great Plains for the July–August peak monsoon and, reversely, an anticyclonic anomaly pattern centered over the northern (southern) Great Plains for the June (September) transition. These features are reversed for dry conditions. Second, wet NAM conditions in June and July are also remotely connected with a distinct SST pattern of positive anomalies in the eastern Pacific and negative anomalies in the Gulf of Mexico, acting to reduce the climatological mean gradient between the two oceans. The anomaly correlation with the gradient is greater than that with either center. Accompanying this pattern is a westward extension of the North Atlantic subtropical ridge. Dry conditions are linked with anomalies of opposite signs. Correlations are very weak for August and September. The above two observed connected features can serve as the metric to quantify the model performance in simulating the key large-scale circulation features that are closely associated with the NAM precipitation variations.

By the performance metric of correlations with the 200-hPa meridional wind, only the MRI model consistently reproduces the observed relationships with the correct sign and realistically high magnitude throughout June to September. Meanwhile, there is a general agreement among all CGCMs with observations for June and September, reflecting that large-scale circulation features are identified with the interannual variability for the monsoon onset and retreat. In contrast, substantial differences are found between models and with observations for the peak monsoon in July and August, depicting strong influences from local processes that are not correctly resolved by current CGCMs. For the metric of correlations with SST, only June shows agreement with observations by a majority of the models but with considerably underestimated magnitudes. These results suggest that the NAM precipitation interannual variability may likely be determined by large-scale circulation anomalies, while its predictability based on remote signals like SSTs will probably be very limited.

Further analyses disclosed the complexity of NAM precipitation interannual variability. First, the performance metric of correlations with both the 200-hPa meridional wind and SST is sensitive to initial conditions, indicating that internal variability has an important contribution; hence, the ensemble prediction is desired. Second, the historical forcing of anthropogenic and natural changes is not the determinant factor for the observed precipitation evolution. Third, the specification of realistic global SST distributions does not capture the actual precipitation evolution, with one possible exception for the June monsoon onset. These results suggest a great challenge for accurate NAM prediction by the current GCMs, coupled or uncoupled.

c. Future investigation

This study demonstrates that at least one CGCM, the MRI with a medium resolution, is able to realistically simulate the NAM precipitation annual cycle and interannual variability and capture their key circulation patterns and physical mechanisms. Hence, accurate depiction of MJO, TEWs, and LLJs is not the determinant factor for a successful simulation of the NAM seasonal–interannual variations. On the other hand, it also shows the great complexity of the NAM precipitation system and the substantial uncertainty in predicting its interannual variability. Even the best model, MRI, can only marginally represent the observed teleconnection pattern with opposite SST anomalies between the eastern Pacific and the Gulf of Mexico for the June onset but totally fails in July. No statistically robust SST signal is found for August and September from both observations and current CGCM simulations. It is most desirable to use a CGCM like the MRI with the best performance to further investigate the physical processes and underlying mechanisms that govern the NAM variability at various time scales. Sensitivity experiments using such a model can be conducted to better understand the relative roles and their scale dependencies of MJO, TEWs, and LLJs versus those processes discovered in this study. These include changes in model physics representation (especially for convection and land–air–sea interactions), grid resolution and initialization, or perturbations in atmospheric circulation or surface characteristics over the key observed correlation centers. It is equally important that similar sensitivity studies are performed by using models, like CCSM and CM2.1, with significant biases to determine whether the aforementioned changes can improve the model performance. We may further distinguish the relative contributions between the planetary to large-scale forcings, if any, and regional to local processes by the downscaling approach where RCMs are driven by CGCM simulations with or without modifications. These topics will be the focus for our future studies.

Acknowledgments

We thank PCMDI and the modeling groups for the 20C3M and AMIP model simulations and Arthur V. Douglas for the precipitation data over Mexico. We appreciate two anonymous reviewers for their constructive comments and suggestions. We acknowledge NOAA/ESRL/GSD and NCSA/UIUC for the supercomputing support. The research was partially supported by the National Oceanic and Atmospheric Administration Award NA04OAR4310162. The views expressed are those of the authors and do not necessarily reflect those of the sponsoring agencies or the Illinois State Water Survey.

REFERENCES

  • Adams, D. K., and A. C. Comrie, 1997: The North American monsoon. Bull. Amer. Meteor. Soc., 78 , 21972213.

  • Adams, J. L., and D. J. Stensrud, 2007: Impact of tropical easterly waves on the North American monsoon. J. Climate, 20 , 12191238.

  • Arritt, R. W., D. C. Goerling, and C. J. Anderson, 2000: The North American monsoon system in the Hadley Centre coupled ocean–atmosphere GCM. Geophys. Res. Lett., 27 , 565568.

    • Search Google Scholar
    • Export Citation
  • Barlow, M., and D. Salstein, 2006: Summertime influence of the Madden-Julian Oscillation on daily rainfall over Mexico and Central America. Geophys. Res. Lett., 33 .L21708, doi:10.1029/2006GL027738.

    • Search Google Scholar
    • Export Citation
  • Barlow, M., S. Nigam, and E. H. Berbery, 1998: Evolution of the North American monsoon system. J. Climate, 11 , 22382257.

  • Berbery, E. H., 2001: Mesoscale moisture analysis of the North American monsoon. J. Climate, 14 , 121137.

  • Betts, A. K., 1986: A new convective adjustment scheme. Part I: Observational and theoretical basis. Quart. J. Roy. Meteor. Soc., 112 , 677691.

    • Search Google Scholar
    • Export Citation
  • Bougeault, P., 1985: A simple parameterization of the large-scale effects of cumulus convection. Mon. Wea. Rev., 113 , 21082121.

  • Carleton, A. M., D. A. Carpenter, and P. J. Weser, 1990: Mechanisms of interannual variability of the southwest United States summer rainfall maximum. J. Climate, 3 , 9991015.

    • Search Google Scholar
    • Export Citation
  • Castro, C. L., T. B. McKee, and R. A. Pielke Sr., 2001: The relationship of the North American monsoon to tropical and North Pacific sea surface temperatures as revealed by observational analyses. J. Climate, 14 , 44494473.

    • Search Google Scholar
    • Export Citation
  • Collier, J. C., and G. J. Zhang, 2007: Effects of increased horizontal resolution on simulation of the North American monsoon in the NCAR CAM3: An evaluation based on surface, satellite, and reanalysis data. J. Climate, 20 , 18431861.

    • Search Google Scholar
    • Export Citation
  • Collins, W. D., and Coauthors, 2006: The Community Climate System Model version 3 (CCSM3). J. Climate, 19 , 21222143.

  • Del Genio, A. D., and M-S. Yao, 1993: Efficient cumulus parameterization for long-term climate studies: The GISS scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 181–184.

    • Search Google Scholar
    • Export Citation
  • Delworth, T. L., and Coauthors, 2006: GFDL’s CM2 global coupled climate models. Part I: Formulation and simulation characteristics. J. Climate, 19 , 643674.

    • Search Google Scholar
    • Export Citation
  • de Szoeke, S. P., and S-P. Xie, 2008: The tropical eastern Pacific seasonal cycle: Assessment of errors and mechanisms in IPCC AR4 coupled ocean–atmosphere general circulation models. J. Climate, 21 , 25732590.

    • Search Google Scholar
    • Export Citation
  • Diansky, N. A., and E. M. Volodin, 2002: Simulation of present-day climate with a coupled atmosphere-ocean general circulation model. Izv. Atmos. Oceanic Phys., 38 , 732747.

    • Search Google Scholar
    • Export Citation
  • Douglas, M. W., R. A. Maddox, K. Howard, and S. Reyes, 1993: The Mexican monsoon. J. Climate, 6 , 16651677.

  • Emanuel, K. A., 1991: A scheme for representing cumulus convection in large-scale models. J. Atmos. Sci., 48 , 23132329.

  • Emori, S., T. Nozawa, A. Numaguti, and I. Uno, 2001: Importance of cumulus parameterization for precipitation simulation over East Asia in June. J. Meteor. Soc. Japan, 79 , 939947.

    • Search Google Scholar
    • Export Citation
  • Fawcett, P. J., J. R. Stalker, and D. S. Gutzler, 2002: Multistage moisture transport into the interior of northern Mexico during the North American summer monsoon. Geophys. Res. Lett., 29 .2094, doi:10.1029/2002GL015693.

    • Search Google Scholar
    • Export Citation
  • Flato, G. M., and G. J. Boer, 2001: Warming asymmetry in climate change simulations. Geophys. Res. Lett., 28 , 195198.

  • Fuller, R. D., and D. J. Stensrud, 2000: The relationship between tropical easterly waves and surges over the Gulf of California during the North American monsoon. Mon. Wea. Rev., 128 , 29832989.

    • Search Google Scholar
    • Export Citation
  • Gates, W. L., and Coauthors, 1999: An overview of the results of the Atmospheric Model Intercomparison Project (AMIP I). Bull. Amer. Meteor. Soc., 80 , 2955.

    • Search Google Scholar
    • Export Citation
  • Gochis, D. J., W. J. Shuttleworth, and Z-L. Yang, 2002: Sensitivity of the modeled North American monsoon regional climate to convective parameterization. Mon. Wea. Rev., 130 , 12821298.

    • Search Google Scholar
    • Export Citation
  • Gochis, D. J., A. Jimenez, C. J. Watts, J. Garatuza-Payan, and W. J. Shuttleworth, 2004: Analysis of 2002 and 2003 warm-season precipitation from the North American Monsoon Experiment event rain gauge network. Mon. Wea. Rev., 132 , 29382953.

    • Search Google Scholar
    • Export Citation
  • Gregory, D., and P. R. Rowntree, 1990: A mass flux convection scheme with representation of cloud ensemble characteristics and stability-dependent closure. Mon. Wea. Rev., 118 , 14831506.

    • Search Google Scholar
    • Export Citation
  • Gutzler, D. S., and J. W. Preston, 1997: Evidence for a relationship between spring snow cover in North American and summer rainfall in New Mexico. Geophys. Res. Lett., 24 , 22072210.

    • Search Google Scholar
    • Export Citation
  • Gutzler, D. S., and Coauthors, 2005: The North American Monsoon Model Assessment Project: Integrating numerical modeling into a field-based process study. Bull. Amer. Meteor. Soc., 86 , 14231429.

    • Search Google Scholar
    • Export Citation
  • Harrington, J. A., R. S. Cerveny, and R. C. Balling, 1992: Impact of the Southern Oscillation on the North American monsoon. Phys. Geogr., 13 , 318330.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., and W. Shi, 2000: Dominant factors responsible for interannual variability of the summer monsoon in the southwestern United States. J. Climate, 13 , 759775.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., and D. Gochis, 2007: Synthesis of Results from the North American Monsoon Experiment (NAME) Process Study. J. Climate, 20 , 16011607.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., Y. Yao, E. S. Yarosh, J. E. Janowiak, and K. C. Mo, 1997a: Influence of the Great Plains low-level jet on summertime precipitation and moisture transport over the central United States. J. Climate, 10 , 481507.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., Y. Yao, and X. L. Wang, 1997b: Influence of the North American monsoon system on the U.S. summer precipitation regime. J. Climate, 10 , 26002622.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., K. C. Mo, and Y. Yao, 1998: Interannual variability of the U.S. summer precipitation regime with emphasis on the southwestern monsoon. J. Climate, 11 , 25822606.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., Y. Chen, and A. V. Douglas, 1999: Interannual variability of the North American warm season precipitation regime. J. Climate, 12 , 653680.

    • Search Google Scholar
    • Export Citation
  • Higgins, R. W., W. Shi, and E. Yarosh, 2000: Improved United States precipitation quality control system and analysis. NCEP/Climate Prediction Center Atlas Number 7, 40 pp. [Available from Climate Prediction Center, World Weather Building, Room 605, Camp Springs, MD 20746.].

  • Higgins, R. W., and Coauthors, 2006: The NAME 2004 field campaign and modeling strategy. Bull. Amer. Meteor. Soc., 87 , 7994.

  • Hu, Q., and S. Feng, 2002: Interannual rainfall variations in the North American summer monsoon region: 1900–98. J. Climate, 15 , 11891202.

    • Search Google Scholar
    • Export Citation
  • Ishii, M., M. Kimoto, K. Sakamoto, and S. Iwasaki, 2006: Steric sea level changes estimated from historical ocean subsurface temperature and salinity analyses. J. Oceanogr., 62 , 155170.

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

  • Koster, R. D., and Coauthors, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305 , 11381140.

  • Lee, M-I., and Coauthors, 2007: Sensitivity to horizontal resolution in the AGCM simulations of warm season diurnal cycle of precipitation over the United States and Northern Mexico. J. Climate, 20 , 18621881.

    • Search Google Scholar
    • Export Citation
  • Liang, X-Z., L. Li, K. E. Kunkel, M. Ting, and J. X. L. Wang, 2004: Regional climate model simulation of U.S. precipitation during 1982–2002. Part I: Annual cycle. J. Climate, 17 , 35103528.

    • Search Google Scholar
    • Export Citation
  • Liang, X-Z., M. Xu, K. E. Kunkel, G. A. Grell, and J. Kain, 2007: Regional climate model simulation of U.S.–Mexico summer precipitation using the optimal ensemble of two cumulus parameterizations. J. Climate, 20 , 52015207.

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

  • Lin, J-L., B. E. Mapes, K. M. Weickmann, G. N. Kiladis, S. D. Schubert, M. J. Suarez, J. T. Bacmeister, and M-I. Lee, 2008: North American monsoon and convectively coupled equatorial waves simulated by IPCC AR4 coupled GCMs. J. Climate, 21 , 29192937.

    • Search Google Scholar
    • Export Citation
  • Livezey, R. E., and W. Chen, 1983: Statistical field significance and its determination by Monte Carlo techniques. Mon. Wea. Rev., 111 , 4659.

    • Search Google Scholar
    • Export Citation
  • Lorenz, D. J., and D. L. Hartmann, 2006: The effect of the MJO on the North American monsoon. J. Climate, 19 , 333343.

  • Meehl, G. A., G. J. Boer, C. Covey, M. Latif, and R. J. Stouffer, 2000: The Coupled Model Intercomparison Project (CMIP). Bull. Amer. Meteor. Soc., 81 , 313318.

    • Search Google Scholar
    • Export Citation
  • Meehl, G. A., J. M. Arblaster, D. M. Lawrence, A. Seth, E. K. Schneider, B. P. Kirtman, and D. Min, 2006: Monsoon regimes in the CCSM3. J. Climate, 19 , 24822495.

    • Search Google Scholar
    • Export Citation
  • Mitchell, T., and P. Jones, 2005: An improved method of constructing a database of monthly climate observations and associated high-resolution grids. Int. J. Climatol., 25 , 693712.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., J. E. Schemm, H. M. H. Juang, R. W. Higgins, and Y. Song, 2005: Impact of model resolution on the prediction of summer precipitation over the United States and Mexico. J. Climate, 18 , 39103927.

    • Search Google Scholar
    • Export Citation
  • Mo, K. C., J. E. Schemm, H. Kim, and W. R. Higgins, 2006: Influence of initial conditions on summer precipitation simulations over the United States and Mexico. J. Climate, 19 , 36403658.

    • 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.

    • Search Google Scholar
    • Export Citation
  • NAME, cited. 2005: NAME modeling and data assimilation: A strategic overview. [Available online at http://www.eol.ucar.edu/projects/name/documentation/NAME_model.html.].

  • Nordeng, T. E., 1994: Extended versions of the convective parameterization scheme at ECMWF and their impact on the mean and transient activity of the model in the tropics. Tech. Memo. 206, European Centre for Medium-Range Weather Forecasts, Reading, United Kingdom, 41 pp.

  • Pan, D-M., and D. A. Randall, 1998: A cumulus parameterization with a prognostic closure. Quart. J. Roy. Meteor. Soc., 124 , 949981.

  • Reyes, S., and A. Mejía-Trejo, 1991: Tropical perturbations in the eastern Pacific and the precipitation field over northwestern Mexico in relation to the ENSO phenomenon. Int. J. Climatol., 11 , 515528.

    • Search Google Scholar
    • Export Citation
  • Schmitz, J. T., and S. Mullen, 1996: Water vapor transport associated with the summertime North American monsoon as depicted by ECMWF analyses. J. Climate, 9 , 16211634.

    • Search Google Scholar
    • Export Citation
  • Stensrud, D. J., R. L. Gall, S. L. Mullen, and K. W. Howard, 1995: Model climatology of the Mexican monsoon. J. Climate, 8 , 17751794.

    • Search Google Scholar
    • Export Citation
  • Stensrud, D. J., R. L. Gall, and M. K. Nordquist, 1997: Surges over the Gulf of California during the Mexican monsoon. Mon. Wea. Rev., 125 , 417437.

    • Search Google Scholar
    • Export Citation
  • Tang, M., and E. R. Reiter, 1984: Plateau monsoons of the Northern Hemisphere: A comparison between North America and Tibet. Mon. Wea. Rev., 112 , 617637.

    • Search Google Scholar
    • Export Citation
  • Tiedtke, M., 1989: A comprehensive mass flux scheme for cumulus parameterization in large-scale models. Mon. Wea. Rev., 117 , 17791800.

    • Search Google Scholar
    • Export Citation
  • Ting, M., X-Z. Liang, J. Pan, and H-P. Huang, 2005: North American summer precipitation with and without topography in CCM3 and the MM5-based regional climate model (CMM5). Proc. 2005 WRF/MM5 User’s Workshop, Boulder, CO, National Center for Atmospheric Research, 15 pp.

  • 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
  • Vera, C., and Coauthors, 2006: Toward a unified view of the American monsoon systems. J. Climate, 19 , 49775000.

  • Waliser, D., and Coauthors, 2006: The experimental MJO prediction project. Bull. Amer. Meteor. Soc., 87 , 425431.

  • Washington, W. M., and Coauthors, 2000: Parallel climate model (PCM) control and transient simulations. Climate Dyn., 16 , 755774.

  • Wu, X., L. Deng, X. Song, G. Vettoretti, W. R. Peltier, and G. J. Zhang, 2007: Impact of a modified convective scheme on the Madden-Julian Oscillation and El Niño–Southern Oscillation in a coupled climate model. Geophys. Res. Lett., 34 .L16823, doi:10.1029/2007GL030637.

    • Search Google Scholar
    • Export Citation
  • Xu, J., X. Gao, J. Shuttleworth, S. Sorooshian, and E. Small, 2004: Model climatology of the North American monsoon onset period during 1980–2001. J. Climate, 17 , 38923906.

    • Search Google Scholar
    • Export Citation
  • Yukimoto, S., and Coauthors, 2001: The new Meteorological Research Institute coupled GCM (MRI-CGCM2)—Model climate and variability. Pap. Meteor. Geophys., 51 , 4788.

    • Search Google Scholar
    • Export Citation
  • Yu, B., and J. M. Wallace, 2000: The principal mode of interannual variability of the North American monsoon system. J. Climate, 13 , 27942800.

    • Search Google Scholar
    • Export Citation
  • Zhang, C., 2005: Madden-Julian Oscillation. Rev. Geophys., 43 .RG2003, doi:10.1029/2004RG000158.

  • Zhang, C., M. Dong, S. Gualdi, H. H. Hendon, E. D. Maloney, A. Marshall, K. R. Sperber, and W. Wang, 2006: Simulations of the Madden-Julian Oscillation in four pairs of coupled and uncoupled global models. Climate Dyn., 27 , 573592. doi:10.1007/s00382-006-0148-2.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., and N. A. McFarlane, 1995: Sensitivity of climate simulations to the parameterization of cumulus convection in the CCC-GCM. Atmos.–Ocean, 3 , 407446.

    • Search Google Scholar
    • Export Citation
  • Zou, C-Z., and W. Zheng, 2004: Simulation of diurnal patterns of summer precipitation in the North American monsoon: An assessment using TRMM. Geophys. Res. Lett., 31 .L07105, doi:10.1029/2004GL019415.

    • Search Google Scholar
    • Export Citation

Fig. 1.
Fig. 1.

The major study domain. Outlined are the six key regions stated in the text, including the NAM core for the precipitation index (solid trapezoid), the northwest and southeast flanks of the core (dotted rectangles) for the metric of correlations with 200-hPa meridional wind (V200) in June/September and July/August, and the eastern tropical Pacific and the Gulf of Mexico (dashed rectangles) for the metric of correlations with SSTs in June and July. Inserted is a blowup map showing the summer average distribution of monthly teleconnection correlations of precipitation interannual variations during 1961–90 at every point with the mean of a small area (little square) in the central Mexico; correlations greater than 0.60 (0.35) are hatched (shaded).

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 2.
Fig. 2.

Interannual variations during 1948–2004 of the seasonal (June–September) mean precipitation (mm day−1) averaged over the NAM core region from three datasets.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 3.
Fig. 3.

The phase lags (month, hollow bars) and rms errors (mm day−1, solid bars) of all models simulated from observed NAM core precipitation annual cycles. Each simulation is labeled as the modeling institution name, followed by a dot plus an identification for a specific model, and then by an underscore plus a number for every run. In particular, ncar.c and .p denote CCSM and PCM; giss.r and .h GISS ER and EH; and gfdl.0 and 0.1GFDL CM2.0 and CM2.1.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 4.
Fig. 4.

The NAM core precipitation (mm day−1) annual cycle observed and modeled by seven distinct groups. The legend lists observations (OBS) (thick solid), CGCM/SST (thick dashed) for the AMIP run using that CGCM forced with prescribed SSTs, and model names (thin curves with various patterns). Multiple realizations, if any, are depicted by a same curve pattern.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 5.
Fig. 5.

(a), (c), (e), (g) Geographic distributions of July minus June and (b), (d), (f), (h) September minus August 30-yr (1961–90) mean differences in precipitation [mm day−1: (a)–(d)] and 200-hPa wind [m s−1: (e)–(h)] from (left) obs or NRA reanalysis and (right) simulated by the MRI model. Contour intervals are 1 mm day−1 for precipitation and 2 m s−1 for wind speed. Wind direction is depicted by thin streamlines with arrowheads.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 6.
Fig. 6.

Geographic distributions of the monthly interannual correlations (×10) of the NAM core precipitation with pointwise 200-hPa (a)–(d) meridional V200 and (e)–(h) zonal U200 wind components observed during 1961–90. Outlined are the two key centers with statistically significant high V200 correlations that distinguish the regimes of June and September (monsoon transitions) from July and August (peak monsoon). Contour intervals are 1 unit. Shaded areas denote where correlations are statistically significant at the 95% confidence level by the Monte Carlo test (see text). The areas in (a)–(h) account for 7.8%, 8.1%, 5.7%, 20.3%, 10.3%, 21.6%, 0.0%, and 0.3% of the entire domain, respectively. Those having greater than 5% areas with significant correlations are considered to be of field significance.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 7.
Fig. 7.

Interannual variations during 1961–90 of the normalized NAM rainfall anomalies for each month (Jun, Jul, Aug, Sep) and the whole summer average. The respective five strongest wet and dry years are marked by upward and downward arrows, except for those marked by (×) that are not accompanied by the apparent cyclonic and anticyclonic anomalies (see text). For comparison, the wet/dry years based on the summer average are marked by (+/−).

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 8.
Fig. 8.

Geographic distributions of June to September monthly and summer mean (JJAS) composites over the five strongest (left) wet and (middle) dry NAM years and (right) their differences for 200-hPa wind observed during 1961–90. The composite is constructed for each month and the whole summer, and thus may include a list of different years between June and September (see Fig. 7). Contour intervals are 3 (2) m s−1 for the wet or dry (their difference) composite wind speed. Wind direction is depicted by thin streamlines with arrowheads.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 9.
Fig. 9.

Monthly interannual correlations of the NAM core precipitation with the 200-hPa meridional wind averaged over the observed connectivity centers specified in Fig. 6. Shown are the correlations observed (black bar) and simulated by the MRI model (gray bar) and all other CGCMs (hollow bar). Simulation labels follow the convention for Fig. 3.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 10.
Fig. 10.

Geographic distributions of (a)–(d) June to September monthly and (e) summer JJAS mean interannual correlations of the NAM core precipitation with pointwise surface temperature (SST over oceans and 2-m air temperature over land) observed during 1961–90. Outlined are the two key centers (eastern Pacific and Gulf of Mexico) with statistically significant high correlations in June and July. Contour intervals are 2 units: zero contour eliminated. Shading areas denote where correlations are statistically significant at the 95% confidence level by the Monte Carlo test (see text). The areas in (a)–(e) account for 12.7%, 12.1%, 3.5%, 0.7%, 10.7% of the entire domain, respectively. Those having greater than 5% areas with significant correlations are considered to be of field significance. For a reference, the climatological summer mean SST distribution is also shown in (f) at contour intervals of 2°C with >28°C shaded.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 11.
Fig. 11.

The June, July and summer June–September mean (JJAS) interannual correlations of the NAM core precipitation with the SST gradient between the averages over the observed connectivity centers in the eastern Pacific and the Gulf of Mexico specified respectively in Figs. 10a,b,e. Shown are the correlations observed (black bar) and simulated by the MRI (gray bar) and all other CGCMs (hollow bar). Simulation labels follow the convention for Fig. 3.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 12.
Fig. 12.

(a)–(h) The SST (°C) annual cycle averaged over the eastern Pacific (see Fig. 10e for area specification) as observed and modeled by 17 CGCMs: models are grouped as in Fig. 4. The legend lists OBS (thick solid) for observations and model names (thin curves with various patterns). Multiple realizations, if any, are depicted by a same curve pattern. Also shown in (h) are the observed SSTs over the eastern Pacific (EP) and the Gulf of Mexico (GM) as well as their gradient (EP − GM, scaled on right).

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Fig. 13.
Fig. 13.

The SST gradients between the eastern Pacific and Gulf of Mexico (see Fig. 10e for the area specification) as observed (OBS) and simulated by 17 CGCMs for individual months June–September and the summer average (JJAS). Simulation labels follow the convention for Fig. 3 except that multiple realizations, if any, are averaged to present only the ensemble mean for each model.

Citation: Journal of Climate 21, 17; 10.1175/2008JCLI2174.1

Table 1.

CGCM information, including model acronym, modeling group, atmospheric (AGCM) and oceanic (OGCM) model resolution, number of runs from different initial conditions, convection parameterization, and oceanic flux correction.

Table 1.
Save