Testing the Anelastic Convective Entity Model against a Cloud-Resolving Model for a Case of Elevated Convection

Technical Summary: Testing the Anelastic Convective Entity Model against a Cloud-Resolving Model for a Case of Elevated Convection

Problem Statement
Global Climate Models (GCMs) struggle to accurately simulate the initiation of nocturnal convection over continents, a phenomenon frequently associated with strong Convective Inhibition (CIN) caused by surface inversions. Traditional convective parameterizations rely on parcel theory, which assumes an infinitesimal parcel lifted from the surface. This framework suggests that overcoming strong surface CIN requires an initial updraft velocity ($w_c$) comparable to $\sqrt{2|CIN|}$, often exceeding 10 m s$^{-1}$. However, observations (e.g., from the GoAmazon campaign) show that deep convection frequently initiates under much weaker initial vertical velocities, often originating from elevated layers rather than the surface. The discrepancy suggests that traditional parameterizations fail to capture key physical mechanisms, specifically the nonlocal pressure effects and adaptive inflow dynamics inherent to finite-size disturbances.

Methodology
The study evaluates the Anelastic Convective Entity (ACE) model, a time-dependent, intermediate-complexity model designed to represent convective processes in global models, against a high-resolution benchmark.

• Benchmark Models:
  • VVM (Vector Vorticity Model): A full-physics, 3-D cloud-resolving model (CRM) using anelastic vorticity equations. Simulations were run at 500 m horizontal resolution (with higher-resolution tests at 100 m) in a 128 km $\times$ 128 km domain.
  • SPCAM (Superparameterized Community Atmosphere Model): Used to compare ACE against embedded 2-D CRMs (SAM) within a GCM framework.
• Experimental Design:
  • Case Study: A nighttime sounding from the GoAmazon campaign (22 Nov 2014) featuring a strong surface inversion and a near-saturated, high-equivalent potential temperature ($\theta_e$) layer just above the inversion.
  • Initiation Strategy: Instead of imposing buoyancy anomalies, experiments utilized weak initial vertical velocity fields ($w$) derived from short, idealized warm-bubble simulations. This isolates the role of nonlocal dynamics in triggering convection.
  • Scenarios:
    1. CTRL: Using the standard GoAmazon sounding.
    2. RH-reduced: A modified sounding where relative humidity was reduced in the elevated near-saturated layer to create a small but non-zero CIN, testing sensitivity to marginal conditions.
• Comparative Scope: The study compares ACE, VVM, and SPCAM/embedded SAM across a range of initial $w$ values (0.06 to 1.55 m s$^{-1}$).

Key Contributions and Results

1. Nonlocal Pressure Effects and Elevated Initiation:

   • Both VVM and ACE demonstrated that convection can initiate with initial vertical velocities far below the threshold predicted by surface-based parcel theory ($w_c \approx 10$ m s$^{-1}$).
   • A finite-size buoyancy anomaly (even near the surface) generates a nonlocal pressure perturbation that instantly accelerates flow over a vertical scale comparable to the horizontal dimension of the disturbance (3–4 km).
   • This mechanism allows the updraft to bypass surface CIN. Instead of lifting air from the surface, the flow adapts to entrain air from the conditionally unstable layer near the top of the inversion (approx. 1.2–1.5 km), where $\theta_e$ is high and CIN is negligible.

2. Adaptive Inflow and Buoyancy Co-evolution:

   • The models revealed a co-evolution of buoyancy and vertical mass flux. As the updraft develops, dynamic entrainment preferentially draws in high-$\theta_e$ air from the elevated inversion layer.
   • Near-surface negative buoyancy (CIN) was observed to be transient and oscillatory, with magnitudes ($\sim -0.04$ m s$^{-2}$) significantly weaker than parcel-theory estimates. The updraft does not need to "overcome" a static CIN barrier; rather, the flow structure adapts to avoid it.
   • Once a threshold $w$ is exceeded, the timing and intensity of precipitation become largely insensitive to further increases in initial velocity, indicating a saturation behavior.

3. ACE vs. VVM Performance:

   • ACE successfully captured the key qualitative features of the VVM benchmark, including the elevated initiation, the adaptive inflow from the inversion top, and the oscillatory near-surface buoyancy.
   • Quantitative differences existed (e.g., ACE showed a wider updraft core and delayed deepening due to simplified microphysics and resolution), but ACE reproduced the physical evolution of elevated convection more accurately than the embedded CRMs in SPCAM.

4. Limitations of Embedded CRMs (SPCAM/SAM):

   • The 2-D CRMs embedded in SPCAM exhibited multi-peaked precipitation and a delayed transition to deep convection compared to VVM and ACE.
   • This behavior was attributed to interactions between internal wave dynamics and the periodic boundary conditions of the 2-D domain, suggesting that standard superparameterization approaches may still struggle to represent the specific nuances of elevated initiation without further adjustment.

5. Sensitivity to Humidity (RH-reduced Case):

   • In the RH-reduced experiment, where the elevated layer was slightly subsaturated, convection initiation became highly sensitive to the initial velocity magnitude.
   • Initiation occurred only when the initial $w$ was sufficient to allow buoyancy oscillations to induce isobaric mixing, leading to localized saturation and latent heating.
   • Neither ACE nor SPCAM/embedded CRMs could initiate convection in this marginal RH-reduced case, highlighting a limitation in capturing the subtle mixing processes required when the environment is not pre-conditioned for saturation.

Significance and Claims
The paper argues that the failure of traditional GCM parameterizations to simulate nocturnal convection stems from their reliance on infinitesimal parcel theory, which neglects nonlocal pressure effects. The study demonstrates that:

• Surface CIN is not the primary limiting factor for elevated convection initiation; rather, the ability of the flow to adaptively draw mass from conditionally unstable elevated layers is the critical mechanism.
• Time-dependent, nonlocal representations (like ACE) are necessary to capture these dynamics. The ACE model, despite its intermediate complexity, offers a more physically consistent framework for representing convection in GCMs than traditional plume models or current embedded CRM approaches in superparameterization.
• The findings support the hypothesis that improving convective triggering in GCMs requires incorporating nonlocal pressure-mediated effects and adaptive inflow, potentially resolving biases in the diurnal cycle of precipitation and nocturnal convection over land.

The authors conclude that while ACE shows promise, further work is needed to refine its implementation for marginal cases involving complex inhibition and to integrate these nonlocal dynamics into full GCM parameterizations.