• Allan, R. P., and B. J. Soden, 2008: Atmospheric warming and the amplification of precipitation extremes. Science, 321, 14811484.

  • Arakawa, A., J.-H. Jung, and C.-M. Wu, 2011: Toward unification of the multiscale modelling of the atmosphere. Atmos. Chem. Phys., 11, 37313742, doi:10.5194/acp-11-3731-2011.

    • Search Google Scholar
    • Export Citation
  • Berg, P., C. Moseley, and J. O. Haerter, 2013: Strong increase in convective precipitation in response to higher temperatures. Nat. Geosci., 6, 181185, doi:10.1038/NGEO1731.

    • Search Google Scholar
    • Export Citation
  • Boyle, J., and S. Klein, 2010: Impact of horizontal resolution on climate model forecasts of tropical precipitation and diabatic heating for the TWP-ICE period. J. Geophys. Res., 115, D23113, doi:10.1029/2010JD014262.

    • Search Google Scholar
    • Export Citation
  • Buizza, R., 2010: Horizontal resolution impact on short- and long-range forecast error. Quart. J. Roy. Meteor. Soc., 136, 10201035.

  • Cheng, A., K.-M. Xu, and B. Stevens, 2010: Effects of resolution on the simulation of boundary-layer clouds and the partition of kinetic energy to subgrid scales. J. Adv. Model. Earth Syst.,2 (1), doi:10.3894/JAMES.2010.2.3.

  • Coumou, C., and S. Rahmstorf, 2012: A decade of weather extremes. Nat. Geosci., 2, 491496, doi:10.1038/NCLIMATE1452.

  • Evans, K. J., P. H. Lauritzen, S. K. Mishra, R. B. Neale, M. A. Taylor, and J. J. Tribbia, 2013: AMIP simulation with the CAM4 spectral element dynamical core. J. Climate, 26, 689709.

    • Search Google Scholar
    • Export Citation
  • Hack, J. J., 1994: Parameterization of moist convection in the National Center for Atmospheric Research Community Climate Model (CCM2). J. Geophys. Res., 99 (D3), 55515568.

    • Search Google Scholar
    • Export Citation
  • Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 56865699.

  • Landu, K., L. R. Leung, S. Hagos, V. Vinoj, S. A. Rauscher, T. Ringler, and M. Taylor, 2014: The dependence of ITCZ structure on model resolution and dynamical core in aquaplanet simulations. J. Climate, 27, 23752385.

    • Search Google Scholar
    • Export Citation
  • Lean, H. W., and P. A. Clark, 2003: The effects of changing resolution on mesoscale modelling of line convection and slantwise circulations in FASTEX IOP16. Quart. J. Roy. Meteor. Soc., 129, 22552278.

    • Search Google Scholar
    • Export Citation
  • Leung, L. R., T. Ringler, W. D. Collins, M. Taylor, and M. Ashfaq, 2013: A hierarchical evaluation of regional climate simulations. Eos, Trans. Amer. Geophys. Union, 94, 297298, doi:10.1002/2013EO340001.

    • Search Google Scholar
    • Export Citation
  • Li, F., W. D. Collins, M. F. Wehner, D. L. Williamson, and J. G. Olson, 2011a: Response of precipitation extremes to idealized global warming in an aqua-planet climate model: Towards a robust projection across different horizontal resolutions. Tellus, 63A, 876883.

    • Search Google Scholar
    • Export Citation
  • Li, F., W. D. Collins, M. F. Wehner, D. L. Williamson, J. G. Olson, and C. Algieri, 2011b: Impact of horizontal resolution on simulation of precipitation extremes in an aqua-planet version of Community Atmospheric Model (CAM3). Tellus, 63A, 884892.

    • Search Google Scholar
    • Export Citation
  • Mason, P. J., and A. R. Brown, 1999: On subgrid models and filter operations in large eddy simulations. J. Atmos. Sci., 56, 21012114.

    • Search Google Scholar
    • Export Citation
  • McComiskey, A., and G. Feingold, 2012: The scale problem in quantifying aerosol indirect effects. Atmos. Chem. Phys., 12, 10311049, doi:10.5194/acp-12-1031-2012.

    • Search Google Scholar
    • Export Citation
  • Neale, R. B., and B. J. Hoskins, 2000: A standard test for AGCMs including their physical parametrizations: I: The proposal. Atmos. Sci. Lett., 1, 101107.

    • Search Google Scholar
    • Export Citation
  • Neale, R. B., and Coauthors, 2010: Description of the NCAR Community Atmosphere Model (CAM 4.0). Tech. Rep. NCAR/TN-485+STR, 212 pp.

  • O’Brien, T., F. Li, W. D. Collins, S. A. Rauscher, T. D. Ringler, M. Taylor, S. M. Hagos, and L. R. Leung, 2013: Observed scaling in clouds and precipitation and scale incognizance in regional to global atmospheric models. J. Climate, 26, 93139333.

    • Search Google Scholar
    • Export Citation
  • O’Gorman, P. A., and T. Schneider, 2009a: The physical basis for increases in precipitation extremes in simulations of 21st-century climate change. Proc. Natl. Acad. Sci. USA, 106, 14 77314 777.

    • Search Google Scholar
    • Export Citation
  • O’Gorman, P. A., and T. Schneider, 2009b: Scaling of precipitation extremes over a wide range of climates simulated with an idealized GCM. J. Climate, 22, 56765685.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, Solomon et al., Eds., Cambridge University Press, 589–662.

  • Rasch, P. J., and J. E. Kristjánsson, 1998: A comparison of the CCM3 model climate using diagnosed and predicted condensate parameterizations. J. Climate, 11, 15871614.

    • Search Google Scholar
    • Export Citation
  • Rauscher, S. A., T. D. Ringler, W. C. Skamarock, and A. A. Mirin, 2013: Exploring a global multiresolution modeling approach using aquaplanet simulations. J. Climate, 26, 24322452.

    • Search Google Scholar
    • Export Citation
  • Seager, R., and G. A. Vecchi, 2010: Greenhouse warming and the 21st century hydroclimate of southwestern North America. Proc. Natl. Acad. Sci. USA, 107, 21 27721 282.

    • Search Google Scholar
    • Export Citation
  • Seager, R., N. Naik, and L. Vogel, 2012: Does global warming cause intensified interannual hydroclimate variability? J. Climate, 25, 33553372.

    • Search Google Scholar
    • Export Citation
  • Simmons, A. J., and D. M. Burridge, 1981: An energy and angular-momentum conserving vertical finite difference scheme and hybrid vertical coordinates. Mon. Wea. Rev., 109, 758766.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., and Coauthors, 2010: Dreary state of precipitation in global models. J. Geophys. Res., 115, D24211, doi:10.1029/2010JD014532.

    • Search Google Scholar
    • Export Citation
  • Sugiyama, M., H. Shiogama, and S. Emori, 2010: Precipitation extreme changes exceeding moisture content increases in MIROC and IPCC climate models. Proc. Natl. Acad. Sci. USA, 107, 571575.

    • Search Google Scholar
    • Export Citation
  • Taylor, M. A., and A. Fournier, 2010: A compatible and conservative spectral finite element method on unstructured grids. J. Comput. Phys., 229, 58795895.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and C. J. Guillemot, 1995: Evaluation of the global atmospheric moisture budget as seen from analysis. J. Climate, 8, 22552272.

    • Search Google Scholar
    • Export Citation
  • Williamson, D. L., 1999: Convergence of atmospheric simulations with increasing horizontal resolution and fixed forcing scales. Tellus, 51A, 663673.

    • Search Google Scholar
    • Export Citation
  • Williamson, D. L., 2008: Convergence of aqua-planet simulations with increasing resolution in the Community Atmospheric Model, version 3. Tellus, 60A, 848862.

    • Search Google Scholar
    • Export Citation
  • Williamson, D. L., 2013: The effect of time steps and time-scales on parametrization suites. Quart. J. Roy. Meteor. Soc., 139, 548560.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., and N. A. McFarlane, 1995: Role of convective-scale momentum transport in climate simulation. J. Geophys. Res., 100 (D1), 14171426.

    • Search Google Scholar
    • Export Citation
  • Zhang, M., W. Lin, C. S. Bretherton, J. J. Hack, and P. J. Rasch, 2003: A modified formulation of fractional stratiform condensation rate in the NCAR Community Atmospheric Model CAM2. J. Geophys. Res., 108, 4035, doi:10.1029/2002JD002523.

    • Search Google Scholar
    • Export Citation
  • View in gallery
    Fig. 1.

    Latitudinal variations of mean and extreme (≥99.9th percentile of both wet and dry conditions) precipitation with the (top) native grid resolution at different temporal data frequencies (6 hourly, 12 hourly, 24 hourly, and every 5 day) and (bottom) for the simulation at 240-km grid resolution. Because of the relatively small magnitudes, the mean values are scaled by a factor of 5 (long dashed lines).

  • View in gallery
    Fig. 2.

    Meridional variations of mean moisture budget terms based on simulations using the MPAS-A dynamical core with (a) 240-, (b) 120-, (c) 60-, and (d) 30-km native grid resolution. The moisture budget terms include precipitation (P, black dashed), surface evaporation (E, yellow dashed), advective moisture convergence (div, blue solid), and PE (purple, solid).

  • View in gallery
    Fig. 3.

    Meridional variations of moisture budget terms under extreme precipitation conditions based on simulations with the MPAS-A dynamical core and with (a) 240-, (b) 120-, (c) 60-, and (d) 30-km horizontal grid resolutions. Zonal, meridional, and vertical advective moisture convergence during extremes, {−u∂q/∂x}, {−υ∂q/∂y}, and {−ω∂q/∂p}, are shown in solid brown, blue, and green lines along with the corresponding total advective moisture convergence (sum, yellow solid) and extreme precipitation (P, black dashed). Note that symbols “∂” and “δ” were used interchangeably.

  • View in gallery
    Fig. 4.

    Vertical profiles of zonal (−u∂q/∂x, brown), meridional (−υ∂q/∂y, blue), and vertical (−ω∂q/∂p, green) advective moisture convergence for simulations at their native grid resolutions of 30- (dashed–dotted–dotted), 60- (dotted), 120- (dashed), and 240-km (solid) with the MPAS-A dynamical core for latitude bands of (a) 5°S–5°N and (b) 25°–35° in both hemispheres. Note that symbols “∂” and “δ” were used interchangeably.

  • View in gallery
    Fig. 5.

    The breakdown of vertical advective moisture convergence during extremes into corresponding mean flow, {−[ω][q]/∂p} (brown solid), and two eddies-related terms, {−ω∂q′/∂p} (yellow dotted) and {−ω[q]/∂p} (green dashed), for the MPAS-A simulations at four native grid resolutions of (a) 240, (b) 120, (c) 60, and (d) 30 km. The total vertical advective moisture convergence (sum) and precipitation (P) during extremes are also shown as black dashed and dotted lines, respectively. Note that symbols “∂” and “δ” were used interchangeably.

  • View in gallery
    Fig. 6.

    Vertical profiles of mean flow and eddy moisture transport for simulations at the native grid resolutions of 30 (dashed–dotted), 60 (dotted), 120 (dashed), and 240 km (solid) with the MPAS-A dynamical core for the latitude bands of (a) 5°S–5°N and (b) 25°–35° (including both hemispheres). The mean flow and eddies terms are too small and are omitted. Note that symbol “∂” is equivalent to “δ” in Eq. (2).

  • View in gallery
    Fig. 7.

    Estimated differences in vertical advective convergence [Δ(ωδq/δp), green] between simulations at fine resolutions (30, 60, and 120 km) and that at the coarse resolution (240 km) during extremes and the contributions to these differences by changes in dynamics [Δ(ω)δq/δp, blue], thermodynamics [ωΔ(δq/δp), yellow], and their covariance changes [Δ(ω)Δ(δq/δp), black] relative to the coarse resolution (240 km). The estimates in this figure are based on simulations with the MPAS-A dynamical core and at their native grid resolution.

  • View in gallery
    Fig. 8.

    Variations of mean and extreme precipitation with grid resolution assessed at the same grid scale for simulations with the (top) MPAS-A and (bottom) HOMME dynamical cores. The simulations conducted at fine resolutions have been aggregated (area-weighted averaged) to the coarse-resolution grids (240 km for MPAS-A and 220 km for HOMME). The mean values are scaled by a factor of 5 (long dashed lines) because of their relatively small magnitudes.

  • View in gallery
    Fig. 9.

    Differences of extreme precipitation between fine (120, 60, and 30 km for MPAS-A and 110, 55, and 28 km for HOMME) and coarse resolutions (240 km for MPAS-A and 220 km for HOMME) when assessed at the same grid using both (a),(c) 6-hourly and (b),(d) daily data. The simulations conducted at fine resolutions have been aggregated (area-weighted averaged) to the coarse-resolution grids (240 km for MPAS-A and 220 km for HOMME) before calculating the differences. The gray area indicates that the 95% confidence interval within which the two fine-resolution simulations are not statistically different.

  • View in gallery
    Fig. 10.

    Contributions of dynamics [Δ(ω)δq/δp, blue], thermodynamics [ωΔ(δq/δp), yellow], and their covariance changes [Δ(ω)Δ(δq/δp), black] to the scale dependency of vertical advective convergence [Δ(ωδq/δp), green] for simulations with the (a),(b) MPAS-A and (c),(d) HOMME dynamical cores. The changes are estimated as differences relative to the coarse resolution (240 km for MPAS-A and 220 km for HOMME) at the same grid scale.

  • View in gallery
    Fig. 11.

    Resolution dependency of PDFs of −1ω (at 600 hPa) for the (a),(b) 5°S–5°N and (c),(d) 25°–35° (including both hemispheres) latitude bands. Colors indicate the native grid resolutions in which the simulations were conducted. In the PDF, the different line styles indicate the coarse grid scale to which the fine-resolution simulations have been aggregated, and the corresponding vertical line indicates the 99.9th percentile of the negative ω.

  • View in gallery
    Fig. 12.

    (a),(d) The mean negative ω and (b),(c),(e),(f) the same grid-scale differences (ΔGS) in the 99.9th percentile of negative ω on 600 hPa for (top) 5°S–5°N and (bottom) 25°–35° (including both hemispheres) latitude bands based on simulations with the MPAS-A or HOMME dynamical cores. The x axis indicates the native grid resolution the simulations were conducted in, and the solid lines, dashed lines, and filled circles indicate the grid scales of 240, 120, and 60 km for MPAS and 220, 110, and 55 km for HOMME, respectively. The fine-resolution simulations were aggregated to for the grid-scale comparison.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 279 105 12
PDF Downloads 213 76 7

Atmospheric Moisture Budget and Spatial Resolution Dependence of Precipitation Extremes in Aquaplanet Simulations

Qing YangPacific Northwest National Laboratory, Richland, Washington

Search for other papers by Qing Yang in
Current site
Google Scholar
PubMed
Close
,
L. Ruby LeungPacific Northwest National Laboratory, Richland, Washington

Search for other papers by L. Ruby Leung in
Current site
Google Scholar
PubMed
Close
,
Sara A. RauscherLos Alamos National Laboratory, Los Alamos, New Mexico

Search for other papers by Sara A. Rauscher in
Current site
Google Scholar
PubMed
Close
,
Todd D. RinglerLos Alamos National Laboratory, Los Alamos, New Mexico

Search for other papers by Todd D. Ringler in
Current site
Google Scholar
PubMed
Close
, and
Mark A. TaylorSandia National Laboratories, Albuquerque, New Mexico

Search for other papers by Mark A. Taylor in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

This study investigates the moisture budgets and resolution dependency of precipitation extremes in an aquaplanet framework based on the Community Atmosphere Model, version 4 (CAM4). Moisture budgets from simulations using two different dynamical cores, the Model for Prediction Across Scales-Atmosphere (MPAS-A) and High Order Method Modeling Environment (HOMME), but the same physics parameterizations suggest that during precipitation extremes the intensity of precipitation is approximately balanced by the vertical advective moisture transport. The resolution dependency in extremes from simulations at their native grid resolution originates from that of vertical moisture transport, which is mainly explained by changes in dynamics (related to vertical velocity ω) with resolution. When assessed at the same grid scale by area-weighted averaging the fine-resolution simulations to the coarse grids, simulations with either dynamical core still demonstrate resolution dependency in extreme precipitation with no convergence over the tropics, but convergence occurs at a wide range of latitudes over the extratropics. The use of lower temporal frequency data (i.e., daily vs 6 hourly) reduces the resolution dependency. Although thermodynamic (moisture) changes become significant in offsetting the effect of dynamics when assessed at the same grid scale, especially over the extratropics, changes in dynamics with resolution are still large and explain most of the resolution dependency during extremes. This suggests that the effects of subgrid-scale variability of ω and vertical moisture transport during extremes are not adequately parameterized by the model at coarse resolution. The aquaplanet framework and analysis described in this study provide an important metric for assessing sensitivities of cloud parameterizations to spatial resolution and dynamical cores under extreme conditions.

Corresponding author address: Qing Yang, 902 Battelle Boulevard, P.O. Box 999, MSIN K9-30, Richland, WA 99352. E-mail: qing.yang@pnnl.gov

Abstract

This study investigates the moisture budgets and resolution dependency of precipitation extremes in an aquaplanet framework based on the Community Atmosphere Model, version 4 (CAM4). Moisture budgets from simulations using two different dynamical cores, the Model for Prediction Across Scales-Atmosphere (MPAS-A) and High Order Method Modeling Environment (HOMME), but the same physics parameterizations suggest that during precipitation extremes the intensity of precipitation is approximately balanced by the vertical advective moisture transport. The resolution dependency in extremes from simulations at their native grid resolution originates from that of vertical moisture transport, which is mainly explained by changes in dynamics (related to vertical velocity ω) with resolution. When assessed at the same grid scale by area-weighted averaging the fine-resolution simulations to the coarse grids, simulations with either dynamical core still demonstrate resolution dependency in extreme precipitation with no convergence over the tropics, but convergence occurs at a wide range of latitudes over the extratropics. The use of lower temporal frequency data (i.e., daily vs 6 hourly) reduces the resolution dependency. Although thermodynamic (moisture) changes become significant in offsetting the effect of dynamics when assessed at the same grid scale, especially over the extratropics, changes in dynamics with resolution are still large and explain most of the resolution dependency during extremes. This suggests that the effects of subgrid-scale variability of ω and vertical moisture transport during extremes are not adequately parameterized by the model at coarse resolution. The aquaplanet framework and analysis described in this study provide an important metric for assessing sensitivities of cloud parameterizations to spatial resolution and dynamical cores under extreme conditions.

Corresponding author address: Qing Yang, 902 Battelle Boulevard, P.O. Box 999, MSIN K9-30, Richland, WA 99352. E-mail: qing.yang@pnnl.gov
Keywords: Climate models

1. Introduction

Increasing numbers of record-breaking and devastating hydrological extremes in the past decade (Coumou and Rahmstorf 2012) has raised concerns that such extremes may have intensified in a warmer world (Allan and Soden 2008). Climate models have been used to attribute the causes of precipitation extremes (PEs) and to project the frequency and intensity of these events in the future. However, a known bias in global climate models is the prevalence of drizzle and underprediction of heavy precipitation partly as a result of the coarse spatial resolution of the models (Stephens et al. 2010). Also climate models are known to exhibit varying degrees of dependency on parameterization and model resolution (e.g., Boyle and Klein 2010; Williamson 1999). Hence, significant uncertainties exist in future PEs projected by an ensemble of climate models (Randall et al. 2007), although they generally project intensification of the hydrological cycle, with larger extremes in both floods and droughts (Held and Soden 2006).

Precipitation is difficult to characterize because of its high variability across nearly all temporal and spatial scales (Berg et al. 2013). Precipitation extremes are associated with low-frequency variability in the flow field that occasionally results in large anomalous horizontal moisture flux convergence or divergence (Held and Soden 2006). Although PEs can be defined rather unambiguously using statistical measures, the physical and dynamical mechanisms associated with PEs vary and are still not fully understood. Furthermore, global climate model (GCM) parameterizations are typically developed and tested more extensively to reproduce large-scale climatological-mean conditions. Williamson (2013) found that using typical parameter values of convective time scales (e.g., 1–2 h) assumed in shallow and convective parameterizations can cause unrealistic development of gridpoint storms producing heavy precipitation in high-resolution GCM simulations. In addition, mass-flux parameterizations (e.g., Zhang and McFarlane 1995) assume that the fractional area covered by convective clouds in a grid cell is far less than 1. The validity of this assumption becomes questionable as GCMs are applied at finer resolutions, making it necessary for parameterizations to adapt to a wider range of model resolutions (Arakawa et al. 2011), with implications to simulate the full spectrum of precipitation intensity.

With increasing high-performance computational resources, high-resolution modeling provides a promising path to improving simulations of PE. However, more research is needed to understand the sensitivity of PE to model resolutions and physics parameterizations. Williamson (1999) showed that physics parameterizations, not dynamics, are responsible for the sensitivity of model results to horizontal resolution in his GCM simulations. Li et al. (2011a,b) used an aquaplanet version of Community Atmospheric Model, version 3 (CAM3), and found that PEs are more sensitive to horizontal resolution than the mean precipitation, suggesting that the effects of horizontal resolution must be taken into account in developing robust projections of PEs. Without improvement in physics representation, increasing resolution alone may only provide limited improvement in climate models skill (Buizza 2010). Understanding the resolution dependency of PEs is an important first step toward achieving scalable and robust simulations.

This study aims to characterize the resolution dependency of extreme precipitation intensity and the variation of this dependency with dynamical cores and to estimate the relative contributions of thermodynamics and dynamics to that resolution dependency. The study is conducted using an aquaplanet framework that simplifies the analyses and interpretations in the absence of complex land–ocean contrast and land surface heterogeneity. Detailed moisture budget analyses have been performed to compare simulations produced by the Community Atmosphere Model, version 4 (CAM4) (Neale et al. 2010), with two different dynamical cores applied at a range of spatial resolution, all using the same physics parameterizations.

The rest of this paper is organized as follows: Section 2 describes the model, the aquaplanet experiments, and the simulation outputs. Section 3 discusses the analytical methods including the definition of extremes, the moisture budget equation, and the breakdowns of moisture transport to mean flow and eddy components. The characteristics of the resolution dependency in PEs and attribution of the resolution dependency are presented in section 4. The last section (section 5) summarizes the main conclusions from this study.

2. Model description and experiments

As part of the collaborative Robust Regional Modeling project (Leung et al. 2013) supported by the U.S. Department of Energy (DOE), a set of aquaplanet simulations have been conducted using the CAM4 physics package and three different dynamical cores: Model for Prediction Across Scales-Atmosphere (MPAS-A) (Rauscher et al. 2013), High Order Method Modeling Environment (HOMME) (Taylor and Fournier 2010; Evans et al. 2013), and Spectral Eulerian (EUL). Detailed model description and configuration can be found in Rauscher et al. (2013), O’Brien et al. (2013), and Leung et al. (2013). This study primarily uses simulations from the newer dynamical-core MPAS-A, with some comparison with HOMME, to provide insights on the impacts of dynamical cores on PE.

Briefly, in CAM4 the convective parameterization includes both a deep and a shallow moisture convection schemes. The shallow moist convective scheme by Hack (1994), as described in Neale et al. (2010), is used to simulate shallow and midlevel convections. The modified deep convection scheme of Zhang and McFarlane (1995) based on a plume ensemble approach in which convective-scale updrafts occur when the convective available potential energy (CAPE) exceeds a threshold value in the lower troposphere (Neale et al. 2010) is used to simulate deep convective cloud and precipitation.

Nonconvective cloud processes are parameterized by the Rasch–Kristjánsson scheme (Rasch and Kristjánsson 1998) with improved treatment of condensation and evaporation as described in Zhang et al. (2003). The Rasch–Kristjánsson scheme is a bulk microphysical scheme with both macro and micro components. It represents prognostic condensate with exchange between condensate and vapor and the associated temperature change and includes a bulk representation of conversion from condensate to precipitation (Neale et al. 2010).

The simulations used the default configuration in CAM4 consisting of 26 vertical levels in a sigma vertical coordinate with varying horizontal resolutions. Identical physics parameterizations with no adjustment for dynamical cores or spatial resolutions were used in all simulations. A set of four experiments was conducted for each dynamical core with varying resolutions (equivalent 240, 120, 60, and 30 km for MPAS-A; 220, 110, 55, 28 km for HOMME; and T42, T85, T170, and T340 for EUL). The CAM4 physics time step was set to be 10 min for all experiments.

In the aquaplanet experiments, the prescribed sea surface temperature (SST) is symmetric about the equator and decreases from a peak value of 27°C at the equator to 0°C above 60° latitude (Φ) according to the following equation in Neale and Hoskins (2000):
eq1

Solar radiation is fixed to equinox solar conditions and is symmetric in the two hemispheres. Thus, seasonal and interannual variations are not represented in these aquaplanet experiments.

The analysis in this study mainly uses the archived 6-hourly data, although daily data were used to test the consistency between the scale dependency of extremes determined based on 6-hourly data and that achieved using lower temporal frequency data. The simulations with the MPAS-A dynamical core uses hexagonal grids, and they have been remapped to the corresponding finite-volume (FV) grid with equivalent resolution using conservative remapping described in Rauscher et al. (2013) to facilitate the analysis. Similar mapping to the FV grid has also been conducted for simulations with the HOMME dynamical core. Each experiment was 5-yr long and the first 6 months were considered the spinup period and excluded from the analysis. The moisture budget analysis (section 4b) is based on the 6-hourly data from the third year simulation, while the rest of the scale dependency study includes 6-hourly and daily model outputs from the full 4.5 yr of simulations. All extreme statistics provided are “mean statistics” (time and zonal averaged) for extremes, and our definition of extremes leads to 210 extreme events per latitude per year when assessed at the coarse scale (240 km) using 6-hourly data; basic statistical inferences using both the mean and variability of the extremes suggest that 1 yr of data is sufficient for statistical robustness (with ≤10% error at a 5% significance level) because of the zonal, hemispheric, and seasonal symmetry of the aquaplanet simulations.

3. Methodology

Precipitation extremes are defined, similar to O’Gorman and Schneider (2009a), as 6-hourly precipitation that exceeds the 99.9th percentile, including both wet and dry cases in a 1-yr simulation. Considering the large meridional variability in precipitation, extremes are determined by latitude. In zonally symmetric aquaplanet simulations with no seasonality, stable statistics can be obtained for extreme precipitation using all 6-hourly precipitation samples from all grid points for each latitude band in both the Northern and Southern Hemispheres.

Assuming conservation of water vapor and ignoring ice formation, the moisture budget equation can be written in a vertically integrated flux form:
e1
The left-hand side of Eq. (1) includes precipitation P and evaporation from the surface E, while the right-hand side consists of, from left to right, column integrations of local tendency, zonal advection of specific humidity, meridional advection of specific humidity, and vertical advection of specific humidity. The total of the last three terms is called total advective convergence and is denoted by TOTDIV in the main text. Related variables and constants are specific humidity q, water density ρw, gravitational acceleration g, zonal velocity component u, meridional velocity component υ, vertical velocity at pressure coordinate ω, time t, and surface pressure ps. Tendency and advective terms were calculated on 20 pressure surfaces before the column integration.
For simplicity, throughout the rest of the manuscript curly braces { } are used to represent the linear operator consisting of a definite integral and multiplication by a constant that is the inverse product of ρw and g, that is, for any variable A:
eq2
Monthly averages are used to represent the mean fields as a common convention (e.g., Seager et al. 2012; Trenberth and Guillemot 1995). Since seasonality is not represented in the aquaplanet simulations, annual means are used for all the large-scale means for a grid point. The transient eddy quantities are estimated as the deviations of the 6-hourly values from the mean. Thus, a variable such as ω can be written as a mean field component and an eddy component :
eq3
Each advective term can be further divided into an eddy term and a mean flow term. The large-scale mean is estimated for each grid box using its annual mean. The advective vertical moisture convergence term can be divided into four terms:
e2
The horizontal moisture convergence terms can be broken down into mean flow and eddies components in a similar manner.

Equation (1) is used to describe the moisture budgets for the mean conditions as well as conditions under the 99.9th percentile extremes. In the latter, the 99.9th percentile extreme events are identified based on 6-hourly precipitation data and the various terms in Eq. (1) are calculated for each event before time and space averaging. Approximately 10° latitude bands were selected to investigate the vertical characteristics of the moisture transport in more detail.

4. Results and discussion

a. Sensitivity of extremes at native grid resolution

Here analysis is performed using model outputs at the native grid resolution of the simulations (i.e., the grid resolution at which the simulations were performed). As shown in Fig. 1, the means of the precipitation extremes show large variations with the native grid resolution and temporal resolution of the data (6 hourly, 12 hourly, daily, or every 5 days) over most of the latitude ranges. The variation in extremes at the native grid resolution is much larger than that of the mean values (dashed lines in Fig. 1). Depending on the dynamical core used, the intensities of precipitation extremes vary considerably, especially over the tropics, and the extremes from the HOMME and EUL simulations show stronger variation with resolution than those from MPAS-A at their native grid (figure not shown). It is also interesting to note that for MPAS-A, increasing grid resolution to 30 km leads to double peaks at the tropics in both mean and extreme precipitation. This feature corresponding to a double intertropical convergence zone (ITCZ) has been analyzed and discussed by Landu et al. (2014) and seems to be unique to MPAS-A despite the fact that the same physics and similar grid resolutions were applied to HOMME and EUL.

Fig. 1.
Fig. 1.

Latitudinal variations of mean and extreme (≥99.9th percentile of both wet and dry conditions) precipitation with the (top) native grid resolution at different temporal data frequencies (6 hourly, 12 hourly, 24 hourly, and every 5 day) and (bottom) for the simulation at 240-km grid resolution. Because of the relatively small magnitudes, the mean values are scaled by a factor of 5 (long dashed lines).

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

Motivated by the large dependency of extremes on the native grid resolution and temporal frequency of the data, we first investigate the relative contribution of different moisture budget terms to extremes using data on the native grid from MPAS-A (section 4b) to provide insights on processes that are responsible for the resolution and temporal frequency dependency. Next, we explore whether the precipitation extremes converge with increasing grid resolution and the sensitivity of the results to temporal resolution when the extreme precipitation is compared at the same grid resolution using 6-hourly and daily data (section 4c). Both simulations with the MPAS-A and HOMME dynamical cores are included in the analysis in section 4c to determine if the results are consistent among simulations with different dynamical cores. Finally, the resolution dependency of extremes is further investigated from budget analysis and the grid-scale and subgrid-scale (SGS) variability of ω.

b. Moisture budget

1) Mean versus extreme conditions

The analysis of moisture budget was carried out using 6-hourly model outputs at the native grid resolution of each simulation. Averaged over the long term, the difference between mean P and E (PE) is roughly balanced by the divergence of the time-averaged, column-integrated moisture flux in the simulated aquaplanet atmosphere (Fig. 2), which is consistent with findings by others based on global reanalysis or global climate simulations (Trenberth and Guillemot 1995; Seager and Vecchi 2010). As shown in Fig. 2, the latitudinal variation and magnitudes of PE are similar to those of the total advective convergence, although PE is substantially more negative than TOTDIV over the subtropical regions. For the mean conditions, E is a significant term compared to P at all latitudes, and the tendency term is negligible (not shown).

Fig. 2.
Fig. 2.

Meridional variations of mean moisture budget terms based on simulations using the MPAS-A dynamical core with (a) 240-, (b) 120-, (c) 60-, and (d) 30-km native grid resolution. The moisture budget terms include precipitation (P, black dashed), surface evaporation (E, yellow dashed), advective moisture convergence (div, blue solid), and PE (purple, solid).

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

During extremes, E is nonnegligible but comparably small (<8% of P), so the intensity of the simulated extreme precipitation is approximately balanced by the corresponding advective moisture convergence in our aquaplanet simulations as shown in Fig. 3 for MPAS-A. The magnitudes of the advective moisture convergence typically agree with the corresponding extreme precipitation intensity within several percent. For example, in the MPAS-A simulations, the ratios of TOTDIV to P during extremes are 0.97–0.98 at 5°N–5°S and 0.95–1.0 at 25°–35° in latitude (including both hemispheres), respectively. Since the extreme defined in this study is also a measure of zonal and temporal variability, the small E compared to P during extremes suggests that the dominant spatial and temporal variability in PE originates from P, not E, consistent with the results by Trenberth and Guillemot (1995). The typical several percent differences between P and TOTDIV indicate that the strength of the simulated precipitation extremes is approximately balanced by the corresponding total advective moisture convergence as follows:
e3
Thus, the resolution dependency in precipitation extreme could be investigated from the scale dependency of the advective moisture convergence. All analyses described below focus on the contributions of different moisture budget terms to precipitation extremes based on the aquaplanet simulations with the MPAS-A dynamical core.
Fig. 3.
Fig. 3.

Meridional variations of moisture budget terms under extreme precipitation conditions based on simulations with the MPAS-A dynamical core and with (a) 240-, (b) 120-, (c) 60-, and (d) 30-km horizontal grid resolutions. Zonal, meridional, and vertical advective moisture convergence during extremes, {−u∂q/∂x}, {−υ∂q/∂y}, and {−ω∂q/∂p}, are shown in solid brown, blue, and green lines along with the corresponding total advective moisture convergence (sum, yellow solid) and extreme precipitation (P, black dashed). Note that symbols “∂” and “δ” were used interchangeably.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

2) Horizontal versus vertical advection

It is worth noting that the horizontal advective moisture convergence is very small (≤2% of total moisture convergence) compared to its vertical counterpart over the tropics, suggesting that the total advective convergence can be approximated by the vertical moisture transport {−ω∂q/∂p} over this region (Fig. 3). However, the same rule does not apply over the extratropics, where horizontal advective moisture convergence contributes up to approximately 20% of TOTDIV. Zonal advective convergence {−u∂q/∂x} is negligible at most latitude ranges except near around 30° in latitude, where negative values indicating local drying associated with the subtropical high pressure system are important, with magnitudes of roughly several percent (~7%–10%) of TOTDIV with MPAS-A.

The relative importance of moisture transport in the zonal, meridional, and vertical directions and their variations with the native grid resolution at different vertical layers are illustrated for two selected latitudinal bands (Fig. 4). A 5°S–5°N band represents the tropical region where only vertical moisture transport is important, and the 25°–35° band (including both hemispheres) represents the extratropical regions where zonal, meridional, and vertical moisture advective transport are all nonnegligible. Consistent with the previous results, the moisture transport over both latitude bands is dominated by the vertical moisture transport (Fig. 4), which shows a strong sensitivity to the model’s native grid resolution. The vertical moisture transport peaks at the midatmosphere (~600 and ~700 hPa for the tropical and subtropical bands, respectively), where its sensitivity to the native grid resolution is also the largest. Zonal and meridional transport is negligible over the tropics, but over the extratropical band, zonal and meridional transport also plays a role especially in the lower atmosphere. Over this region, zonal moisture transport −u∂q/∂x contributes to the drying of the atmosphere below approximately 650 hPa, and this term increases slightly in magnitude with increasing native grid resolution below approximately 800 hPa. The meridional moisture transport over the extratropical band has magnitudes comparable to the vertical moisture transport near the surface layer (p > 900 hPa), but decreases with height and becomes fairly small above 400 hPa. The meridional transport −υ∂q/∂y over this region shows a slight converging tendency with similar values at high resolution. Over both tropical and extratropical bands, the variation of TOTDIV during extremes with the native grid resolution is dominated by the corresponding variation in vertical moisture transport that maximizes in the midtroposphere, where the dependency on the native grid resolution is also the greatest.

Fig. 4.
Fig. 4.

Vertical profiles of zonal (−u∂q/∂x, brown), meridional (−υ∂q/∂y, blue), and vertical (−ω∂q/∂p, green) advective moisture convergence for simulations at their native grid resolutions of 30- (dashed–dotted–dotted), 60- (dotted), 120- (dashed), and 240-km (solid) with the MPAS-A dynamical core for latitude bands of (a) 5°S–5°N and (b) 25°–35° in both hemispheres. Note that symbols “∂” and “δ” were used interchangeably.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

3) Mean fields versus transient eddies

The vertical advective convergence is further partitioned into mean fields and transient eddy components according to Eq. (2) (Fig. 5 for MPAS-A) to estimate the relative contributions of mean circulation and eddies to extremes. Over both the tropics and extratropics, the eddy transport of mean moisture {−ω[q]/∂p} is the dominant term, followed by eddy transport of eddy moisture {−ω∂q′/∂p}. While these two terms are comparable in magnitude over the extratropics (e.g., 76.4 versus 53.5 mm day−1 over 30°–35°N at 30 km with MPAS-A), the former is several times higher (e.g., 377.8 versus 76.4 mm day−1 over 5°S–5°N at 30-km resolution with MPAS-A) than the latter over the tropics. The mean flow term {−[ω][q]/∂p} is not significant (<0.8 mm day−1 with MPAS-A) at all latitudes except over a narrow band centered at the equator where it shows converging tendency with increasing native grid resolution in MPAS-A (11.5, 13.5, 14.4, and 14.5 mm day−1 for 240, 120, 60, and 30 km). The third term of Eq. (2) {−[ω]∂q′/∂p} is negligible (<0.05 mm day−1 over the tropics and extratropics). Thus, the large variation of the vertical advective moisture transport at the native grid resolution is mainly coming from that of the two eddy terms {−ω[q]/∂p} and {−ω∂q′/∂p}.

Fig. 5.
Fig. 5.

The breakdown of vertical advective moisture convergence during extremes into corresponding mean flow, {−[ω][q]/∂p} (brown solid), and two eddies-related terms, {−ω∂q′/∂p} (yellow dotted) and {−ω[q]/∂p} (green dashed), for the MPAS-A simulations at four native grid resolutions of (a) 240, (b) 120, (c) 60, and (d) 30 km. The total vertical advective moisture convergence (sum) and precipitation (P) during extremes are also shown as black dashed and dotted lines, respectively. Note that symbols “∂” and “δ” were used interchangeably.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

Similar breakdowns of zonal and meridional advective convergence found that zonal-mean flow transport of eddy moisture {−[u]∂q′/∂x} and meridional eddy transport of eddy moisture {−υ∂q′/∂y} (Fig. 6) are only significant over the extratropical subtropical band during extremes where {−[u]∂q′/∂x} slightly dries the storm at most vertical levels, while {−υ∂q′/∂y} along with vertical moisture transport supply moisture for the extreme precipitation over this region. In the lower atmosphere (p > ~850 hPa), the vertical eddy transport of eddy moisture −ω∂q′/∂p contributes negatively to P during extremes over both the tropics and extratropics. Although both instantaneous q and large-scale [q] decreases with height, q′ peaks at around 850–875 hPa vertically (figure not shown). This is likely to result from low-level convergence, especially that at the zonal direction, which brings drier and colder air from the outer region to the convective center. As a result of the reverse vertical gradients of q′ and the strong upward eddy motion, −ω∂q′/∂p dries the lower level and contributes negatively to precipitation intensity during extremes. Over the tropics, the source of moisture, its associated energy, and the resolution dependency of the moisture transport at the native grid resolution are dominated by eddy transport of mean moisture −ω[q]/∂p. However, over the extratropics, this is true only for the lower altitudes (p > ~700 hPa) while at higher altitudes (p < ~700 hPa), the resolution dependency in the moisture transport is dominated by the eddy transport of eddy moisture −ω∂q′/∂p. At the midlayer (~700 < p < ~800 hPa), both eddy terms are comparable in terms of their contributions to the total moisture convergence and its resolution dependency.

Fig. 6.
Fig. 6.

Vertical profiles of mean flow and eddy moisture transport for simulations at the native grid resolutions of 30 (dashed–dotted), 60 (dotted), 120 (dashed), and 240 km (solid) with the MPAS-A dynamical core for the latitude bands of (a) 5°S–5°N and (b) 25°–35° (including both hemispheres). The mean flow and eddies terms are too small and are omitted. Note that symbol “∂” is equivalent to “δ” in Eq. (2).

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

In profiles of moisture and ω (not shown), it is indicated that the resolution dependency of ω assessed at the native grid resolution is large, which is passed on to the vertical moisture transport terms. The vertical moisture structures in mean and eddy quantities act as an averaging kernel, which determines the relative contribution from different levels to the overall scale dependency and magnitude of the column-integrated moisture transport assessed at the native grid resolution. Over the tropics, the mean atmosphere is already very humid (~13 g kg−1 near surface), and [q] decreases very sharply with the increasing altitude. As a result, the vertical gradients of [q] are much larger than that of q′, which in addition to the more significant drying at the lower atmosphere associated with the reverse vertical moisture gradients results in larger magnitude and larger resolution dependency of {−ω[q]/∂p} than {−ω∂q′/∂p} over the tropics. Over the extratropics, however, the mean atmosphere above approximately 700 hPa is much drier and has a small vertical gradient compared to that under extreme conditions, which explains the dominance of {−ω∂q′/∂p} over {−ω[q]/∂p} in magnitudes and dependency on the native grid resolution for the upper atmosphere over this region.

4) Dynamics versus thermodynamics

Previous analysis has identified the relative role of mean circulation and eddies in the moisture transport and characterized their variation with the native grid resolution at different vertical layers under extreme conditions. An important question to ask is how much of the variation with the native grid resolution is because of changes in dynamics versus thermodynamics. The dependency on the native grid resolution is represented by the difference Δ between the fine-resolution and coarse-resolution (240 km for MPAS-A) simulations. The resolution dependency of moisture transport is estimated as follows:
e4
where the three terms on the right-hand side represent changes resulting from dynamic (ω), thermodynamic (q), and covariance changes. Note that for simplicity, the negative sign is ignored in the vertical moisture transport term in Eq. (4). As shown in Fig. 7, the profiles of Δ(ωδq/δp) and Δ(ω)δq/δp under extreme conditions are close to each other with typical agreement within 10% in magnitude at most vertical layers. This indicates that the variation of vertical moisture transport with the native grid resolution is mainly explained by the changes in dynamics (ω) with native grid resolution at most altitude ranges.
Fig. 7.
Fig. 7.

Estimated differences in vertical advective convergence [Δ(ωδq/δp), green] between simulations at fine resolutions (30, 60, and 120 km) and that at the coarse resolution (240 km) during extremes and the contributions to these differences by changes in dynamics [Δ(ω)δq/δp, blue], thermodynamics [ωΔ(δq/δp), yellow], and their covariance changes [Δ(ω)Δ(δq/δp), black] relative to the coarse resolution (240 km). The estimates in this figure are based on simulations with the MPAS-A dynamical core and at their native grid resolution.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

c. Scale dependency at the same grid scale

Because coarse-resolution simulations cannot produce the fine-scale features resolvable by the high-resolution simulations, a more useful analysis of resolution dependency is to compare simulations aggregated to the same spatial resolution. For such an analysis, the resolution dependency is represented by the difference Δ between the fine- and coarse-resolution simulations after the fine-resolution simulations have been aggregated or upscaled to the coarse-resolution grid using area-weighted averaging. We define convergence as the time-averaged zonal mean of extremes in simulations performed at different grid resolutions but aggregated to the same coarse resolution are not statistically different (at a 5% significance level) from each other. That is, simulations performed at a range of grid resolutions do not exhibit a resolution dependency when they are aggregated to the same grid resolution for comparison. Figure 8 shows the zonal-averaged mean and extreme precipitation for simulations performed at 30-, 60-, 120-, and 240-km grid resolution, all compared at the 240-km grid resolution for MPAS-A, and similarly for HOMME. Interestingly, the double ITCZ feature simulated at 30-km resolution simulation with MPAS-A disappears or becomes inconspicuous after upscaling to the coarse grid, which indicates the association of this feature with small-scale variability. Neither mean nor extreme precipitation shows convergence over the tropics, and the differences between high- and low-resolution simulations do not vary monotonically with grid resolution. This is seen more clearly for extremes in Fig. 9 that show the difference Δ between higher-resolution simulations aggregated to 240-km grid resolution and the coarse-resolution simulation performed at 240-km grid resolution. There is no convergence of extremes over the tropics (i.e., Δ is statistically different from zero) in MPAS-A at 6-hourly and daily time scales. The same is true when assessed at 120- or 60-km grid scales for the MPAS-A simulations using 6-hourly, 12-hourly, or daily data (not shown). Similarly, simulations with HOMME do not converge at 6-hourly or daily time scales over the tropics.

Fig. 8.
Fig. 8.

Variations of mean and extreme precipitation with grid resolution assessed at the same grid scale for simulations with the (top) MPAS-A and (bottom) HOMME dynamical cores. The simulations conducted at fine resolutions have been aggregated (area-weighted averaged) to the coarse-resolution grids (240 km for MPAS-A and 220 km for HOMME). The mean values are scaled by a factor of 5 (long dashed lines) because of their relatively small magnitudes.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

Fig. 9.
Fig. 9.

Differences of extreme precipitation between fine (120, 60, and 30 km for MPAS-A and 110, 55, and 28 km for HOMME) and coarse resolutions (240 km for MPAS-A and 220 km for HOMME) when assessed at the same grid using both (a),(c) 6-hourly and (b),(d) daily data. The simulations conducted at fine resolutions have been aggregated (area-weighted averaged) to the coarse-resolution grids (240 km for MPAS-A and 220 km for HOMME) before calculating the differences. The gray area indicates that the 95% confidence interval within which the two fine-resolution simulations are not statistically different.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

Over the extratropics, for both dynamical cores, Fig. 9 shows that the simulations do not converge when compared at the coarse grid scale except for a small latitude range (~20°–25°, including both hemispheres) for MPAS-A. The magnitudes of extremes from the coarse-resolution simulations (blue lines in Fig. 8) are much lower than that from the finer grid resolutions, and there is a poleward shift of the extratropical peak (~30° in latitude) from coarse to fine resolutions. The poleward shift is even more apparent when the simulations are compared at their native grid resolution (Fig. 1). The scale dependency is reduced when coarser temporal resolution data are used over the extratropical regions. As shown in Fig. 9, at the daily time scale, the differences between high- and low-resolution simulations become negligible so the four simulations performed at grid resolutions ranging from 30 to 240 km essentially converged at a wide range of latitudes (e.g., 50°–65° for MPAS-A). When assessed at the 120-km scale using 6-hourly data, simulations at 60- and 30-km resolution with the MPAS-A converge over 20°–50° in latitude (figure not shown). When assessed at the 120-km scale using daily data, simulations with MPAS-A converge in the latitude range of 15°–60° (figure not shown). The scale dependency of simulations with HOMME over the extratropics region is similar but generally larger than that with MPAS-A (Fig. 9). There is no converging tendency over the high latitudes (>70°) for simulations with both dynamical cores. We note, however, that near the poles, sampling becomes an issue as the number of grid cells becomes much smaller.

The dynamical and thermodynamical contributions to the scale dependency of extremes compared at the same grid scale are shown in Fig. 10 for the tropical and extratropical bands. The contributions from thermodynamic changes are generally consistent between MPAS-A and HOMME. When assessed at the same coarse grid scale, thermodynamic (moisture) changes become significant in offsetting the dynamic effect (ω), especially over the extratropics, although contributions from dynamic changes to the scale dependency is still larger and explain most of the scale dependency of the vertical advective moisture convergence. Contributions of thermodynamic changes are negative within most vertical layers over both the tropics and extratropics for both dynamical cores, thus reducing the overall scale dependency. The offsetting thermodynamic effects are associated with the decrease in specific humidity with resolution for both mean and extreme conditions with either dynamical core. The contributions of dynamic changes to scale dependency are positive except over the tropics for MPAS-A where their contributions are negative.

Fig. 10.
Fig. 10.

Contributions of dynamics [Δ(ω)δq/δp, blue], thermodynamics [ωΔ(δq/δp), yellow], and their covariance changes [Δ(ω)Δ(δq/δp), black] to the scale dependency of vertical advective convergence [Δ(ωδq/δp), green] for simulations with the (a),(b) MPAS-A and (c),(d) HOMME dynamical cores. The changes are estimated as differences relative to the coarse resolution (240 km for MPAS-A and 220 km for HOMME) at the same grid scale.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

Over the tropics, with the MPAS-A dynamical core, both dynamical changes and thermodynamical changes become more negative with increasing resolution (Fig. 10a), so the extremes do not converge. Although thermodynamical changes offset some of the scale dependency caused by the dynamics with HOMME over the tropics, the scale dependency from dynamics contributions is several times larger in magnitude than that over the extratropics (note the different scales used in the x axis for Figs. 10c,d). The scale dependency remains too large for the simulations to converge even when assessed using daily data.

At the upper atmosphere (<~300 hPa) in the 58-km simulation with HOMME, the resolution dependency is mainly explained by changes in thermodynamics over both latitudinal bands, which is an exception. This is likely as a result of the higher convective cloud tops simulated at 58-km resolution with HOMME. A slightly more humid atmosphere (especially above ~700 hPa) is seen in the 58-km simulation with HOMME under both mean and extreme conditions, and the more humid atmosphere allows storms to reach higher tops. The differences in thermodynamics in the HOMME simulations at the upper atmosphere are likely influenced by nonlinear feedbacks from large changes in dynamics, which require further investigation.

The dominant contributions of dynamics changes to scale dependency of the vertical moisture transport during extremes warrant further examination of the sensitivity of ω with resolution. As indicated by the vertical velocity spectra in Lean and Clark (2003), the vertical velocity in a convective system consists of variability at scales much smaller than the finest grid spacing used in our experiments. When upscaled to a coarse grid scale through averaging, signals/variances at scales smaller than the analysis scale are reduced at coarser resolution, so the mismatch of process and analysis scales often leads to difficulty in producing reliable statistics from aggregating areal data (McComiskey and Feingold 2012). While data aggregation does not alter the mean values, it does impact extreme intensities since extremes represent events with high temporal and spatial variability.

The differences in ω in the fine () and coarse () simulations at their respective native grid resolution can be expressed as a SGS component not resolvable by the coarse simulation, and the difference between the simulations when compared at the same grid scale (ΔGS):
eq4
e5
eq5
where the subgrid variability can be estimated as the difference between the statistics of the simulations aggregated to the coarse scale (denoted as fine2coarse in the subscript) and simulations at the native grid resolution (denoted by fine_n in the subscript). This approach of estimating subgrid-scale variability is a variation of the approach used by Cheng et al. (2010) and Mason and Brown (1999), in which a running-mean operator on both horizontal directions was used. Note that although the negative sign is not indicated, the subscript p here refers to the 99.9th percentile of negative ω (equivalent to 0.1th percentile of ω) for consistency with the use of high percentiles in the precipitation extreme definition. As an example, denotes the 99.9th percentile of the negative ω (upward motion) that has been aggregated (area-weighted averaged) from the 30-km native grid resolution to the 240-km coarse grid scale, and it represents the resolved 240-km grid-scale variability of the fine-resolution simulation. Thus, the SGS and ΔGS for fine-resolution simulation (30 km) at 240-km grid scale are expressed as below:
eq6
eq7
The difference ΔGS is the difference between fine (e.g., 30 km) and coarse (e.g., 240 km) simulations after aggregating the fine resolution (e.g., 30 km) to the coarse (e.g., 240 km) grid scale. When assessing the scale dependency at the native grids, SGS dominates, but when assessing the scale dependency at the same grid scale, the difference is smaller as ΔGS is generally much smaller than SGS.

The probability density functions (PDFs) of the simulated ω at their native grid (e.g., the red dashed, green dotted, and blue solid curves in Fig. 11 corresponding to PDFs of ω from simulations performed at 60/58, 120/110, and 240/220 km, respectively) become increasingly broader accompanied by more extreme tail values with the increase in native grid resolution. The differences of PDFs at the same coarse grid scale [e.g., (black dashed) versus (red dashed) in Figs. 11a and 11c for MPAS-A in the tropics and extratropics, respectively] are relatively small. This indicates that the resolution dependency of the extreme vertical velocity at the native grid resolution can be largely explained by the SGS variability, although the contributions from ΔGS are not negligible.

Fig. 11.
Fig. 11.

Resolution dependency of PDFs of −1ω (at 600 hPa) for the (a),(b) 5°S–5°N and (c),(d) 25°–35° (including both hemispheres) latitude bands. Colors indicate the native grid resolutions in which the simulations were conducted. In the PDF, the different line styles indicate the coarse grid scale to which the fine-resolution simulations have been aggregated, and the corresponding vertical line indicates the 99.9th percentile of the negative ω.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

Following the above analysis, the sensitivity of extreme vertical velocity to spatial grid spacing is illustrated quantitatively in Fig. 12, along with those of the mean vertical velocity. With both MPAS-A and HOMME, the mean negative ω at 600 hPa is positive (upward) and in general increases with increasing resolution over both tropical and extratropical bands and show monotonic converging tendency over extratropical regions.

Fig. 12.
Fig. 12.

(a),(d) The mean negative ω and (b),(c),(e),(f) the same grid-scale differences (ΔGS) in the 99.9th percentile of negative ω on 600 hPa for (top) 5°S–5°N and (bottom) 25°–35° (including both hemispheres) latitude bands based on simulations with the MPAS-A or HOMME dynamical cores. The x axis indicates the native grid resolution the simulations were conducted in, and the solid lines, dashed lines, and filled circles indicate the grid scales of 240, 120, and 60 km for MPAS and 220, 110, and 55 km for HOMME, respectively. The fine-resolution simulations were aggregated to for the grid-scale comparison.

Citation: Journal of Climate 27, 10; 10.1175/JCLI-D-13-00468.1

The sensitivity of extreme vertical velocity to resolution is different from that of the means. The 99.9th percentiles of the negative at the native grid resolution increase drastically and almost linearly as the resolution doubles (not shown), but this scale dependency of ω at the native grid resolution is dominated by the SGS variability, as inferred from Fig. 11. Differences in when compared at the same grid scale (ΔGS) (e.g., ) indicate the resolution dependency after accounting for the subgrid-scale variability resolved by increasing resolution. The ΔGS, although small, explains the scale dependency of vertical moisture transport simulated at different resolutions but compared at the same grid scale (Fig. 10). At 600 hPa, the vertical level with a maximum of scale dependency from Fig. 10, the increasingly negative ΔGS with increasing resolution (solid line in Fig. 12b) explains the nonconvergence of vertical moisture transport over the tropics with MPAS-A. Note that the unique negative ΔGS in MPAS-A is likely associated with the unique double ITCZ feature in the fine-resolution simulation and they are likely to share the same causes. Over the extratropical band, there are some reductions of ΔGS with increasing resolution (e.g., solid line in Fig. 12e) with MPAS-A, and the magnitude of ΔGS is small enough to allow convergence at a certain grid scale using either 6-hourly or daily data (e.g., 60-km grid scale with 6-hourly data and 240-km grid scale with daily data). This is consistent with the previous analysis of MPAS-A that over the extratropics thermodynamic (moisture) changes become significant in offsetting the scale dependency due to dynamics (ω), which lead to the convergence at certain grid scales using daily or 6-hourly data. For both tropical and extratropical bands, the magnitudes of ΔGS are several times larger with HOMME compared with MPAS-A. The relatively large ΔGS in HOMME (note the difference in scale in the y axis) translates to larger-scale dependency in vertical moisture transport than MPAS-A. The generally positive ΔGS (except over the tropics with MPAS-A) indicates that the effects of subgrid-scale variability at the grid scales must be adequately parameterized in the model at coarse resolution to improve “scalability” or consistency in extreme vertical velocity when assessed at the same grid scale.

The intensities of precipitation extremes and the corresponding high percentile vertical motions are much higher in HOMME than MPAS-A over the tropics. This could be explained partly by the 20% higher effective resolution with HOMME compared to MPAS-A at similar grid resolution (figure not shown). Comparison of the kinetic energy spectra from MPAS-A and HOMME reveals that MPAS-A is slightly more dissipative than HOMME (figure not shown); however, the differences are relatively small and unlikely to be the main reason for the large differences from MPAS-A regarding extreme precipitation over the tropics. A 1-yr HOMME test run replacing the vertical discretization approach from Simmons and Burridge (1981) used in the HOMME simulations analyzed in this study by the vertical Lagrangian method reduces the mean precipitation simulated by 15% near the equator, which brings the HOMME simulations closer to MPAS-A in terms of the mean precipitation values over the tropics. However, changing the vertical discretization approach in HOMME appears to have little impacts on the extreme precipitation. Further investigation is required to identify the main reason leading to the large differences between HOMME and MPAS-A simulations when the same CAM4 physics package is used.

5. Summary and conclusions

In this study, we examine the resolution dependency of precipitation extremes in climate models. This has implications for projecting changes in climate variability and extremes in the future. A set of aquaplanet simulations produced by a global model using the same physics parameterizations but different dynamical cores and horizontal grid resolutions were used in this investigation. Strong increases of precipitation extremes with native grid resolutions are seen over both the tropics and extratropics. The magnitude of this dependency also varies with dynamical cores, especially over the tropics. Moisture budget estimates based on the aquaplanet simulations with MPAS-A suggest that during precipitation extremes, surface precipitation is approximately balanced by the advective moisture convergence with evaporation contributing an insignificant amount, although evaporation plays a much larger role in providing moisture for the mean precipitation. This is particularly true over the tropics, where the vertical advection of moisture explains nearly all of the surface precipitation during precipitation extremes. Over the extratropics, where the contribution of vertical moisture transport still dominates the total moisture transport, meridional advection also contributes.

The variation of moisture transport with the native grid resolution is mainly as a result of the eddy transport of mean moisture {−ω[q]/∂p} and the eddy transport of eddy moisture {−ω∂q′/∂p}. At most vertical levels over the tropics and in the lower atmosphere over the extratropics, the interaction between eddies and the mean moisture field {−ω[q]/∂p} dominates the contribution to precipitation extremes. Over the extratropics, the source of moisture and its associated energy are dominated by eddy transport {−ω∂q′/∂p} at the mid- and upper troposphere. With both the MPAS-A and HOMME dynamical cores, the resolution dependency of the vertical advective moisture convergence during extremes is mainly explained by dynamical changes (related to ω). Although the vertical gradients of moisture act like averaging kernels that determine the relative contributions of the changes in ω at different vertical levels to the overall scale dependency. The scale dependency of extremes is also assessed by aggregating (using area-weighted averages) the fine-resolution simulations to the coarse-resolution grids for comparison. The extreme precipitation does not converge with resolution over the tropics for simulations with either MPAS-A or HOMME when all simulations are compared at 240/220 km (and also at finer grid scales) with either 6-hourly or daily data. With either MPAS-A or HOMME, over the extratropics, extremes do not converge at the coarse grid scale (240 km for MPAS-A and 220 km for HOMME) at most latitudes. However, this scale dependency is reduced when coarser temporal resolution data are used, and convergence in extremes is found with daily data over a wide range of latitudes in the extratropics. In addition, there is a poleward shift in the extratropical peak of the extreme precipitation (~30° in latitude) from coarse to fine resolutions, which is very clear when comparing simulations at their native grid resolution and still apparent even when comparing simulations that are aggregated to the coarse scale.

When assessed at the same grid scale through aggregation of the finer-resolution simulations, thermodynamic (moisture) changes become significant in offsetting the effect of dynamics (ω), especially over the extratropics, although dynamics still show a more dominant impact on the resolution dependency than thermodynamics over most vertical layers, especially in the midtroposphere where the resolution dependency in vertical moisture transport is the largest.

This study shows that extreme precipitation in aquaplanet simulations using the Community Atmosphere Model (CAM) demonstrates spatial resolution and dynamical-core dependency, especially over the tropics, despite the fact that the same physics parameterizations were used. Simulations with the HOMME dynamical core shows higher variability in the vertical motion, leading to higher extreme precipitation intensities and a resolution dependency beyond what can be explained by a 20% higher effective resolution in HOMME compared to MPAS-A, which requires further investigation. Differences in simulations compared at the same grid scale suggest that the effects of subgrid variability in ω and vertical moisture transport during extreme conditions are not adequately parameterized at coarse resolutions in both MPAS-A and HOMME and more so for the latter.

Li et al. (2011b) have also identified ω as the source of resolution dependency of extreme precipitation in aquaplanet simulations using simple correlation coefficients to attribute the likely causes. This study quantifies the relationship between ω and extreme precipitation using a detailed moisture budget equation, with an added focus on comparing how the relationship varies with dynamical cores. Sugiyama et al. (2010) showed that changes of PE under global warming are strongly linked to the magnitude and profile of vertical velocity in the tropics. GCM projections of future PEs have been found to be very sensitive to model resolution and parameterizations (O’Gorman and Schneider 2009b). Hence, the resolution and dynamical-core dependency of extreme precipitation has important implications to the prediction of future changes in extreme precipitation.

Williamson (2008) discussed the behavior of extreme precipitation in the equatorial region (10°S–10°N) in the form of the probability density function using simulations with the CAM finite-volume dynamical core. Our analysis extends his study by including results from different dynamical cores. In light of the sensitivity of extreme precipitation and vertical velocity to dynamical cores, in addition to model resolution, future studies should further investigate how dynamical cores interact with model resolution and physics parameterization to influence extreme precipitation. Future development in convective parameterizations needs to address not only spatial-resolution dependency but also dynamical-core dependency. Convective parameterizations that incorporate probability density functions of ω might better capture the variability of vertical motion during extremes and improve consistency of extreme precipitation intensity simulated with different dynamical cores at a different spatial resolution. The aquaplanet framework and analysis described in this study provide an important metric for assessing the sensitivity of cloud parameterizations to spatial resolution as well as dynamical cores under extreme conditions.

Acknowledgments

We thank Samson Hagos and Kiranmayi Landu for the helpful discussion and the internal review of this manuscript. This research was supported by the Office of Science of the U.S. Department of Energy as part of the Regional and Global Climate Modeling program. The Pacific Northwest National Laboratory is operated for DOE by Battelle Memorial Institute under Contract DE-AC05-76RL01830.

REFERENCES

  • Allan, R. P., and B. J. Soden, 2008: Atmospheric warming and the amplification of precipitation extremes. Science, 321, 14811484.

  • Arakawa, A., J.-H. Jung, and C.-M. Wu, 2011: Toward unification of the multiscale modelling of the atmosphere. Atmos. Chem. Phys., 11, 37313742, doi:10.5194/acp-11-3731-2011.

    • Search Google Scholar
    • Export Citation
  • Berg, P., C. Moseley, and J. O. Haerter, 2013: Strong increase in convective precipitation in response to higher temperatures. Nat. Geosci., 6, 181185, doi:10.1038/NGEO1731.

    • Search Google Scholar
    • Export Citation
  • Boyle, J., and S. Klein, 2010: Impact of horizontal resolution on climate model forecasts of tropical precipitation and diabatic heating for the TWP-ICE period. J. Geophys. Res., 115, D23113, doi:10.1029/2010JD014262.

    • Search Google Scholar
    • Export Citation
  • Buizza, R., 2010: Horizontal resolution impact on short- and long-range forecast error. Quart. J. Roy. Meteor. Soc., 136, 10201035.

  • Cheng, A., K.-M. Xu, and B. Stevens, 2010: Effects of resolution on the simulation of boundary-layer clouds and the partition of kinetic energy to subgrid scales. J. Adv. Model. Earth Syst.,2 (1), doi:10.3894/JAMES.2010.2.3.

  • Coumou, C., and S. Rahmstorf, 2012: A decade of weather extremes. Nat. Geosci., 2, 491496, doi:10.1038/NCLIMATE1452.

  • Evans, K. J., P. H. Lauritzen, S. K. Mishra, R. B. Neale, M. A. Taylor, and J. J. Tribbia, 2013: AMIP simulation with the CAM4 spectral element dynamical core. J. Climate, 26, 689709.

    • Search Google Scholar
    • Export Citation
  • Hack, J. J., 1994: Parameterization of moist convection in the National Center for Atmospheric Research Community Climate Model (CCM2). J. Geophys. Res., 99 (D3), 55515568.

    • Search Google Scholar
    • Export Citation
  • Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 56865699.

  • Landu, K., L. R. Leung, S. Hagos, V. Vinoj, S. A. Rauscher, T. Ringler, and M. Taylor, 2014: The dependence of ITCZ structure on model resolution and dynamical core in aquaplanet simulations. J. Climate, 27, 23752385.

    • Search Google Scholar
    • Export Citation
  • Lean, H. W., and P. A. Clark, 2003: The effects of changing resolution on mesoscale modelling of line convection and slantwise circulations in FASTEX IOP16. Quart. J. Roy. Meteor. Soc., 129, 22552278.

    • Search Google Scholar
    • Export Citation
  • Leung, L. R., T. Ringler, W. D. Collins, M. Taylor, and M. Ashfaq, 2013: A hierarchical evaluation of regional climate simulations. Eos, Trans. Amer. Geophys. Union, 94, 297298, doi:10.1002/2013EO340001.

    • Search Google Scholar
    • Export Citation
  • Li, F., W. D. Collins, M. F. Wehner, D. L. Williamson, and J. G. Olson, 2011a: Response of precipitation extremes to idealized global warming in an aqua-planet climate model: Towards a robust projection across different horizontal resolutions. Tellus, 63A, 876883.

    • Search Google Scholar
    • Export Citation
  • Li, F., W. D. Collins, M. F. Wehner, D. L. Williamson, J. G. Olson, and C. Algieri, 2011b: Impact of horizontal resolution on simulation of precipitation extremes in an aqua-planet version of Community Atmospheric Model (CAM3). Tellus, 63A, 884892.

    • Search Google Scholar
    • Export Citation
  • Mason, P. J., and A. R. Brown, 1999: On subgrid models and filter operations in large eddy simulations. J. Atmos. Sci., 56, 21012114.

    • Search Google Scholar
    • Export Citation
  • McComiskey, A., and G. Feingold, 2012: The scale problem in quantifying aerosol indirect effects. Atmos. Chem. Phys., 12, 10311049, doi:10.5194/acp-12-1031-2012.

    • Search Google Scholar
    • Export Citation
  • Neale, R. B., and B. J. Hoskins, 2000: A standard test for AGCMs including their physical parametrizations: I: The proposal. Atmos. Sci. Lett., 1, 101107.

    • Search Google Scholar
    • Export Citation
  • Neale, R. B., and Coauthors, 2010: Description of the NCAR Community Atmosphere Model (CAM 4.0). Tech. Rep. NCAR/TN-485+STR, 212 pp.

  • O’Brien, T., F. Li, W. D. Collins, S. A. Rauscher, T. D. Ringler, M. Taylor, S. M. Hagos, and L. R. Leung, 2013: Observed scaling in clouds and precipitation and scale incognizance in regional to global atmospheric models. J. Climate, 26, 93139333.

    • Search Google Scholar
    • Export Citation
  • O’Gorman, P. A., and T. Schneider, 2009a: The physical basis for increases in precipitation extremes in simulations of 21st-century climate change. Proc. Natl. Acad. Sci. USA, 106, 14 77314 777.

    • Search Google Scholar
    • Export Citation
  • O’Gorman, P. A., and T. Schneider, 2009b: Scaling of precipitation extremes over a wide range of climates simulated with an idealized GCM. J. Climate, 22, 56765685.

    • Search Google Scholar
    • Export Citation
  • Randall, D. A., and Coauthors, 2007: Climate models and their evaluation. Climate Change 2007: The Physical Science Basis, Solomon et al., Eds., Cambridge University Press, 589–662.

  • Rasch, P. J., and J. E. Kristjánsson, 1998: A comparison of the CCM3 model climate using diagnosed and predicted condensate parameterizations. J. Climate, 11, 15871614.

    • Search Google Scholar
    • Export Citation
  • Rauscher, S. A., T. D. Ringler, W. C. Skamarock, and A. A. Mirin, 2013: Exploring a global multiresolution modeling approach using aquaplanet simulations. J. Climate, 26, 24322452.

    • Search Google Scholar
    • Export Citation
  • Seager, R., and G. A. Vecchi, 2010: Greenhouse warming and the 21st century hydroclimate of southwestern North America. Proc. Natl. Acad. Sci. USA, 107, 21 27721 282.

    • Search Google Scholar
    • Export Citation
  • Seager, R., N. Naik, and L. Vogel, 2012: Does global warming cause intensified interannual hydroclimate variability? J. Climate, 25, 33553372.

    • Search Google Scholar
    • Export Citation
  • Simmons, A. J., and D. M. Burridge, 1981: An energy and angular-momentum conserving vertical finite difference scheme and hybrid vertical coordinates. Mon. Wea. Rev., 109, 758766.

    • Search Google Scholar
    • Export Citation
  • Stephens, G. L., and Coauthors, 2010: Dreary state of precipitation in global models. J. Geophys. Res., 115, D24211, doi:10.1029/2010JD014532.

    • Search Google Scholar
    • Export Citation
  • Sugiyama, M., H. Shiogama, and S. Emori, 2010: Precipitation extreme changes exceeding moisture content increases in MIROC and IPCC climate models. Proc. Natl. Acad. Sci. USA, 107, 571575.

    • Search Google Scholar
    • Export Citation
  • Taylor, M. A., and A. Fournier, 2010: A compatible and conservative spectral finite element method on unstructured grids. J. Comput. Phys., 229, 58795895.

    • Search Google Scholar
    • Export Citation
  • Trenberth, K. E., and C. J. Guillemot, 1995: Evaluation of the global atmospheric moisture budget as seen from analysis. J. Climate, 8, 22552272.

    • Search Google Scholar
    • Export Citation
  • Williamson, D. L., 1999: Convergence of atmospheric simulations with increasing horizontal resolution and fixed forcing scales. Tellus, 51A, 663673.

    • Search Google Scholar
    • Export Citation
  • Williamson, D. L., 2008: Convergence of aqua-planet simulations with increasing resolution in the Community Atmospheric Model, version 3. Tellus, 60A, 848862.

    • Search Google Scholar
    • Export Citation
  • Williamson, D. L., 2013: The effect of time steps and time-scales on parametrization suites. Quart. J. Roy. Meteor. Soc., 139, 548560.

    • Search Google Scholar
    • Export Citation
  • Zhang, G. J., and N. A. McFarlane, 1995: Role of convective-scale momentum transport in climate simulation. J. Geophys. Res., 100 (D1), 14171426.

    • Search Google Scholar
    • Export Citation
  • Zhang, M., W. Lin, C. S. Bretherton, J. J. Hack, and P. J. Rasch, 2003: A modified formulation of fractional stratiform condensation rate in the NCAR Community Atmospheric Model CAM2. J. Geophys. Res., 108, 4035, doi:10.1029/2002JD002523.

    • Search Google Scholar
    • Export Citation
Save