Most global climate models with convective parameterization have trouble in simulating the observed diurnal cycle of convection. Maximum precipitation usually happens too early during local summertime, especially over land. Observational analyses indicate that deep convection over land cannot keep pace with rapid variations in convective available potential energy (CAPE), which is largely controlled by boundary layer forcing. In this study, a new convective closure in which shallow and deep convection interact strongly, out of equilibrium, is implemented in atmosphere-only and ocean-atmosphere coupled models developed at the NOAA Geophysical Fluid Dynamics Laboratory (GFDL). The diurnal cycles of convection in both simulations are significantly improved without altering their mean states. These improvements in the diurnal cycle of these climate models are consistent with those obtained by Peter Bechtold and colleagues in the ECMWF Integrated Forecasting System. The new closure shifts maximum precipitation over land later by about three hours. Compared to satellite observations, the diurnal phase biases are reduced by half. Shallow convection to some extent equilibrates rapid changes in the boundary layer at sub-diurnal time scales. Future model improvement will focus on the remaining biases, which may be further reduced by including stochastic entrainment and cold pools.

How to cite: Zhang, B., Donner, L., Zhao, M., and Tan, Z.: Improved Diurnal Cycle in GFDL Earth System Models with Non-Equilibrium Convection, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-33, https://doi.org/10.5194/egusphere-egu24-33, 2024.

During the Australian fire season 2019/2020, an unprecedented amount of smoke aerosol was not only released, but also transported upwards and injected into the tropopause region by so-called pyro-cumulonimbus clouds (pyroCb). The resulting lower stratospheric aerosol loads in early 2020 were comparable to those of the largest volcanic eruptions of the twentieth century. PyroCbs have been identified as the main pathway for biomass burning aerosol into the stratosphere. To study the phenomenon of PyroCbs, simulations of the so-called Australian New Year Super Outbreak are performed with the numerical weather model ICON. Simulations were run in a nested, limited area mode setup, with the smallest domain reaching down to a horizontal grid spacing of 500 m. Within the domain, an idealised fire perturbation was applied for which an additional constant surface sensible heat and water vapour flux was introduced to represent the thermodynamical impacts of the fire. Simulations with this setup were successful in producing fire-induced deep convection with subsequent smoke injection into the lower stratosphere. Preliminary sensitivity experiments show a high sensitivity of the PyroCb properties to initial and boundary conditions. We can show, that especially water vapour emissions, which would originate from evaporating surface water as well as from combustion of organic materials, have a decisive, enhancing impact on the pyro-convection. Moreover, besides the fire intensity, the plume characteristics and smoke injection heights are also closely linked to the background meteorology, in particular. In the long term, the goal is to incorporate the effects of extreme biomass burning emission into large scale climate simulations by taking into account PyroCb activity. However, this will require a very deep understanding of wildfire triggered convection and PyroCb dynamics.  

How to cite: Müller, J., Senf, F., and Tegen, I.: Modelling the formation of an extreme Australian pyro-convection event and its sensitivities, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-3972, https://doi.org/10.5194/egusphere-egu24-3972, 2024.

The elimination of rain evaporation in the planetary boundary layer (PBL) has been found to lead to convective self-aggregation (CSA) even without radiative feedback (frequently referred to as “moisture memory aggregation”), but the precise mechanisms underlying this phenomenon remain unclear. We conducted cloud-resolving simulations with two domain sizes (L = 128 and 256 km; Δx = 1 and 4 km) with homogenised radiation and progressively reduced rain evaporation in the PBL by multiplying it with a factor 𝛼 = [1.0, 0.8, 0.6, 0.4, 0.2, 0]. Surprisingly, self aggregation only occurred when rain evaporation was almost completely removed (𝛼 ≈ 0). Similar to conventional radiatively-driven aggregation (RDA), a shallow circulation that leads to an upgradient moist static energy transport is present, but in this case it is the additional convective heating resulting from the reduction of evaporative cooling in the moist patch that triggers this circulation, thereafter a dry subsidence intrusion into the PBL in the dry patch takes over and intensifies aggregation. Hence, this type of aggregation should be more appropriately referred to as “convectively-driven aggregation” (CDA). Contrary to RDA, in CDA temperature and moisture anomalies oppose each other in their buoyancy effects, hence explaining the need for near-zero 𝛼 values: only when rain evaporation is almost completely removed can the additional heating trigger aggregation. Lastly, we found radiative cooling and not cold pools to be the leading cause of the domain size dependence of CDA. Runs with similar amounts of cold pools aggregate in the large but not small domain due to stronger radiative cooling rates and concomitant broadening of the range of precipitable water in the larger domain. 

How to cite: Hwong, Y.-L. and Muller, C.: The Unreasonable Efficiency of Total Rain Evaporation Removal in Triggering Convective Self-Aggregation, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-4139, https://doi.org/10.5194/egusphere-egu24-4139, 2024.

Convection frequently initiates in proximity to cold fronts during the European warm-season and can also be associated with hazards such as flooding, rain, and hail. Despite this, the frequency and underlying processes that drive such events are not well-understood. To understand the typical nature, frequency and forcing mechanisms of convection depending on the region relative to the front, automatic front detection methods, a convective cell detection and tracking dataset (KONRAD), and lightning data are combined between 2007–2016.

The climatology shows that convective cells are most frequent in Germany marginally ahead of the surface front. Furthermore, the 700 hPa frontal line marks the minimum frequency of convection and a shift in regime between cells with a strong diurnal cycle on the cold-side of the 700 hPa front and a weakened diurnal cycle on the warm-side of the 700 hPa front. The results are consistent for lightning data on a sub-European domain. Given cell detection ahead of the surface front, cells are up to 3 times more likely to be associated with a mesocyclone compared to non-cold-frontal cells in Germany. Cells with 55 dBZ cores are over 1.5 times more likely.

To unravel the complex relationships between different predictor variables and the probability of convection a logistic regression model is developed. Feature importance techniques are utilised to understand which variables carry the most importance depending on the region relative to the front. We find solar heating carries more importance towards the model’s predictive power behind the 700 hPa front than ahead of the 700 hPa front. The opposite is true for the elevation term, which acts as a proxy for the influence of orography on convective initiation. By giving the model information on the number of surrounding grid points associated with convection, a proxy for cell interactions, the most skill is added near the surface front.

These results are an important step towards a deeper understanding of the underlying processes that drive cold-frontal convection and improved forecasting.

How to cite: Pacey, G., Pfahl, S., Schielicke, L., and Wapler, K.: Climatology, characteristics and forcing mechanisms of warm-season cold-frontal convection in Europe, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5759, https://doi.org/10.5194/egusphere-egu24-5759, 2024.

Spontaneous aggregation of deep convection is a common feature of idealized numerical simulations of the tropical atmosphere in Radiative-Convective Equilibrium (RCE). However, at coarse grid resolution where deep convection is not fully resolved, the occurrence of this phenomenon is highly sensitive to subgrid-scale processes. This study investigates the role of mixing and entrainment, provided by either the turbulence model or the implicit numerical dissipation, in this phenomenon. The results of two different models, WRF and SAM, have been analysed and compared using different configurations by varying the turbulence models, initial conditions, and horizontal spatial resolution. At a coarse grid resolution of 3 km, the occurrence of Convective Self-Aggregation (CSA) is prevented in models with low numerical diffusivity due to the removal of turbulent mixing, while it is preserved in models with high numerical diffusivity. When refining the horizontal grid resolution to 1 km, which reduces the implicit numerical dissipation, CSA can only be achieved by increasing explicit turbulent mixing. Even with a small amount of shallow clouds, CSA was found to occur in this case. Therefore, this study suggests that the sensitivity of CSA to horizontal grid resolution is not primarily due to the corresponding decrease in shallow clouds. It has been found that the amplitude of initial humidity perturbations introduced by convection in the free troposphere is regulated by turbulent mixing and dissipation at small scales. The size and strength of humidity perturbations in the free troposphere that can destabilize the RCE state increase with greater dissipation at small scales.

How to cite: Silvestri, L., Saraceni, M., and Bongioannini Cerlini, P.: Numerical diffusion and turbulent mixing in convective self-aggregation, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-5959, https://doi.org/10.5194/egusphere-egu24-5959, 2024.

Improving weather and climate prediction cannot avoid accurately representing organized convection, as its convective and stratiform components distinctly reshape large-scale circulations via redistributing momentum and heat. For latent heating, the stratiform heating in organized convection shifts to higher altitudes compared to convective regions, presenting a significant challenge for representation in models across scales. The Multiscale Coherent Structural Parameterization (MCSP), introduced by Moncrieff et al. (2017), offers a promising solution by generating the top-heavy profile from convective heating in slantwise layer overturning scenarios. As part of the MCS: PRIME project, the PRIME-MCSP implementation by Zhang et al. (submitted, 2024) couples MCSP with the CoMorph-A convection scheme in the UK Met Office Unified Model with the following improvements: 1) CoMorph permits unstable air to rise from any height, diverging from the conventional CAPE trigger for deep convection, thereby enhancing continuity and facilitating storm tracking. 2) We activate MCSP selectively for deep mixed-phase clouds, recognizing the limited ability of shallow clouds to produce a stratiform component. 3) We configure the global model runs to include both a fixed convective-stratiform heating fraction and a fraction proportional to cloud top temperature.

MCS tracks in ensembles of weather runs show that PRIME-MCSP suppresses cloud deepening and reduces precipitation areas by dampening low-level updrafts. 20-year climate simulations show that PRIME-MCSP improves the precipitation seasonal cycle over the Indian Ocean, while increasing the warm-season wet bias over the Western Pacific. Additionally, PRIME-MCSP intensifies the Madden Julian Oscillation (MJO). The model run using a variable convective-stratiform fraction more accurately represents the MJO frequency and aligns better with reanalysis. Future plans focus on the stochastic representation of stratiform effects, steered by insights from data assimilation increments.

How to cite: Zhang, Z., Christensen, H., Muetzelfeldt, M., Woollings, T., Plant, B., Stirling, A., Whitall, M., Moncrieff, M., and Chen, C.-C.: Improving and Assessing Organized Convection Parameterization in the Unified Model, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-8367, https://doi.org/10.5194/egusphere-egu24-8367, 2024.

Representing land-ocean heterogeneity via convective
adjustment timescale
Bidyut Goswami¹ , Andrea Polesello¹ , Caroline Muller¹ .
¹Department of Earth Science, Institute of Science and Technology Austria, Klosterneuburg, Austria
January 2024

Abstract

The time needed by deep convection to bring the atmosphere back to equilibrium
is called convective adjustment timescale or simply adjustment timescale, typically
denoted by τ . In the Community Atmospheric Model version 6 (CAM6), convection
is parameterized through the Zhang-McFarlan scheme [1], where CAPE undergoes
an exponential consumption, of which τ is the time constant. τ is a tunable pa-
rameter in CAM6 and it has a default value of 1 hour, worldwide, on both ocean
and land. Albeit, there is no justified reason why one adjustment timescale value
should work over land and ocean both. Continental and oceanic convection is dif-
ferent in terms of the vigor of updraft and hence can have different durations.[2, 3]
So it is logical to investigate the prescription of two different convective adjustment
timescales for land (τ_L ) and ocean (τ_L ). To understand the impact of representing
land-ocean heterogeneity via τ , we investigated CAM climate simulations for two
different convective adjustment timescales for land and ocean in contrast to having
one value globally.
Following a comparative analysis of 5-year-long climate simulations, we find
τ_O =4hr and τ_L =1hr to yield the best results. In particular, we obtain a better
description of the Madden-Julian Oscillation (MJO). Although these τ values were
chosen empirically and require further tuning, the conclusion of our finding remains
the same, which is, to use two different τ values for land and ocean.
References
[1] G. Zhang and N. A. McFarlane, “Sensitivity of climate simulations to the parameterization of
cumulus convection in the canadian climate centre general circulation model,” Atmosphere-
Ocean, vol. 33, no. 3, pp. 407–446, 1995.
[2] C. Lucas, E. J. Zipser, and M. A. Lemone, “Vertical Velocity in Oceanic Convection off
Tropical Australia,” Journal of the Atmospheric Sciences, vol. 51, pp. 3183–3193, 11 1994.
[3] R. Roca, T. Fiolleau, and D. Bouniol, “A Simple Model of the Life Cycle of Mesoscale
Convective Systems Cloud Shield in the Tropics,” Journal of Climate, vol. 30, pp. 4283–
4298, 6 2017.

How to cite: Polesello, A., Goswami, B. B., and Muller, C.: Representing land-ocean heterogeneity via convective adjustment timescale, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-9561, https://doi.org/10.5194/egusphere-egu24-9561, 2024.

Within a diurnal cycle, the transition from shallow to deep convection takes several hours, despite having large environmental instability at the onset of shallow convection. During this period, the cloud environment remains rather steady, while the convection exhibits a rapid development. Properly predicting the timing of this rapid shallow-to-deep transition within a diurnal cycle is still a major shortcoming of weather and climate models that employ the so-called mass-flux parameterization of atmospheric convection, as they typically predict the onset of deep convection too early, not allowing for a gradual convective deepening. In this work, it is argued that the problem of correctly representing the diurnal cycle of deep convection comes from the fundamental assumptions of the mass-flux formulation, in which it is considered that the clouds, represented by steady-state plumes, only interact with a spatially homogeneous environment. However, in the rapid shallow-to-deep transition, the convection still requires several hours to deepen, even if the environment remains steady, so some interactions must be missing. Here, a conceptual model for cloud development is introduced, in which a cloud is formed due to the sum of water transport from the boundary layer by multiple updrafts during its life-time, allowing for cloud-cloud interactions. This process captures local preconditioning, in which the clouds themself provide favorable conditions for the development of subsequent updrafts. It is also argued that the cold pools act as a reinforcement of this process, organizing the updrafts, and thus, allowing for a greater degree of local preconditioning. Based on this new conceptual model, it is argued that the shallow-to-deep transition can be seen as a predator-prey problem, in which the cloud population at the cloud base acts as prey, while the surface precipitation rate acts as predators. This simple predator-prey model is then tested against an idealized large-eddy simulation, showing that indeed, the rapid shallow-to-deep transition of atmospheric convection exhibits predator-prey characteristics. Moreover, it is shown how easily the simple predator-prey model can be implemented in current mass-flux schemes, leading to improved representation of deep convection within a diurnal cycle. Overall, this suggests that better representing the spatial organisation of clouds can lead to improvements in the timing of cloud and precipitation properties, thanks to a better convective memory.

How to cite: Vraciu, C.-V., Savre, J., and Colin, M.: Predator-prey characteristics of the rapid shallow-to-deep transition of atmospheric convection, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-10474, https://doi.org/10.5194/egusphere-egu24-10474, 2024.

Deep convection over Amazonia can manifest in various forms, from scattered convective cells to mesoscale organizations like squall lines and cloud clusters. This diversity significantly influences vertical convective transport, impacting not only large-scale circulation but also the poorly understood cycle of gases and aerosols emitted by the forest. Monitoring convective systems over Amazonia during the CAFE-Brazil field campaign (Dec 2022-Jan 2023) involved the HALO aircraft and the ATTO-Campina ground site, employing meteorological, aerosol, and chemical measurements.

On January 18, a 500-km wide mesoscale system dissipated, giving rise to new convective cells initially disorganized and later organized into a large squall line. This event was measured by ATTO-Campina and HALO during the local afternoon. To understand the processes driving organizational changes and their impact on transport, 24-hour simulations with the Meso-NH model were conducted over an 800-km wide domain, ranging from horizontal resolutions of 1600 m down to 200 m, ultimately resulting in large-eddy simulations.

The simulations revealed a strong resolution sensitivity in mesoscale convective organization, with a distinct emergence of squall lines at the finest resolutions only. Surprisingly, at fine resolution, organized convection exhibited larger transport due to increased updraft size, rather than intensity. Cloud cluster organization exhibited a delayed onset compared to convective cell organization, aligning with expectations. Ongoing investigations are currently focusing on gravity waves and cold pools to better understand their impact on convective organization.

How to cite: Dauhut, T., Gonthier, H., Viala, B., and Haytzmann, G.: Organization of convection over Amazonia and its impact on transport, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-10894, https://doi.org/10.5194/egusphere-egu24-10894, 2024.

The launch of the joint ESA JAXA Earth Cloud Aerosol and Radiation Explorer (EarthCARE) mission (May, 2024) marks the beginning of a new era of spaceborne radar measurements that target atmospheric convection. In addition to the EarthCARE mission that features the first Cloud Profiling Radar (CPR) with Doppler capability, NASA’s Investigation of Convective Updrafts (INCUS) and Atmosphere Observing System (AOS) missions aim to provide unique observations of convective dynamics. Prior to this upcoming decade of the study of atmospheric convection from space, the CloudSat CPR collected remarkable data of convective cores over a period of 15 years. Despite its high frequency that results in significant attenuation and multiple scattering effects, the 94-GHz CloudSat CPR offers a relatively small footprint (compared to the TRMM/GPM radar footprint of 5 km) and collocated radar-radiometer (passive) brightness temperatures (Tb). Here, we propose a refined deep convective core (DCC) identification scheme by first selecting the CPR profiles with continuous echoes between below 2 and above 10 km. The 10-dBZ echo top height is also required to exceed 10 km and located within 2 km from cloud top. Additionally, profiles with stratiform precipitation flags in the CloudSat products are not included in the analysis.

We investigated the CloudSat observations from 2006 to 2019 globally and also with a focus over 4 convective basins where model simulations are performed by the NASA’s INCUS science team. The four deep convection basins are Amazon, Congo, Philippines, and Western Pacific, which represent a decent spectrum of atmospheric environments. It is found that the DCCs over the Congo basin are featured with larger size and likely more intensified updrafts, while the Western Pacific is characterized with finer-scale cores. The analysis shows that the DCCs with size below 5 km predominate, implying the narrow cores can be under detected by the large-footprint radars such as GPM. The distinct depressions of 94-GHz Tb due to the presence of high-density ice particles lend complementary information on DCC classifications. In addition, multiple scattering can be a confounding factor in interpretating the CPR measurements within deep convective clouds. Our preliminary calculations suggest the impact of multiple scattering becomes significant at ~2.5 km from radar cloud top on average and is subject to the DCC updraft intensity. Moreover, profiles of 94-GHz radar reflectivity and Tb are forward calculated from the high-resolution model simulation outputs to understand the constraints that such observations can afford on key measures such as convective mass fluxes.

How to cite: Xu, Z., Kollias, P., Battaglia, A., Treserras, B. P., and Marinescu, P.: Investigating Deep Convective Cores Combining CloudSat Observations and Model Simulations , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-13870, https://doi.org/10.5194/egusphere-egu24-13870, 2024.

Data-driven approaches using machine learning to parameterizing model physical processes in Earth System Models have been actively explored in recent years. Deep-learning-based convection parameterization is one such example. While significant progress has been made in emulating convection using neural networks (NN), serious roadblocks remain, including generalization of the NN-based scheme trained on model data from current climate to future climate and integration instability when it is implemented into the model for long-term integrations. This study uses a deep residual convolutional network to emulate convection simulated by a superparameterized global climate model (GCM). The NN uses the current environmental state variables and advection tendencies, as well as the history of convection to predict the GCM grid-scale temperature and moisture tendencies, cloud liquid and ice water contents from moist physics processes. Independent offline tests show that the NN-based scheme has extremely high prediction accuracy for all output variables considered. In addition, the scheme trained on data in the current climate generalizes well to a warmer climate with +4K sea surface temperature in an offline test, with high prediction accuracy as well. Further tests on different aspects of the NN architecture are performed to understand what factors are responsible for its generalization ability to a warmer climate. We are also able to perform multi-year integrations, without encountering any integration instability, when the scheme is implemented into the NCAR CAM5. The details will be presented at the meeting.

How to cite: Zhang, G., Han, Y., and Wang, Y.: Using Deep Learning for Convection Parameterization, EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-14599, https://doi.org/10.5194/egusphere-egu24-14599, 2024.

Mesoscale convective systems are the building block of tropical precipitation, as more than 40% of global precipitation and more than 80% of extreme rainrates are produced by these organized systems. However, when investigating the sensitivity of global rain extremes, the behavior and morphology of organized storm systems are typically ignored and corresponding dynamics are instead interpreted using the textbook framework of a convecting parcel. Indeed, despite rich observational and case studies describing the internal dynamics and structures of MCSs, no conceptual framework exist to this day to bridge the gap between global hydrologic sensitivity and MCS behavior.

This work introduces new approaches to link extreme precipitation rates in the tropics to the occurrence, internal dynamics and lifecycle of individual MCSs. Individual storms are idenditifed based on by the Lagrangian tracking algorithm TOOCAN which tracks storm anvils over their lifecycle, and which has been applied to satellite observations and to global storm resolving models in the DYAMOND experiment. We first use this rich dataset to develop a numerical interface that maps the occurrence of extreme precipitation rate onto the MCS cloud shield. We then introduce a novel conceptual framework to decompose the sensitivity of precipitation extremes to the change in storm occurrence and change in internal dynamics within this cloud shield. 

Results are threefold. We demonstrate a robust phasing in the timing of global extreme rainrates within the storm lifecycle, robustly occurring at 25-30% of the storm's lifetime for the models and regions analyzed. The analytical decomposition confirms that in a given climate state, variability in the heaviest rainrates across regions mostly occur through changes in MCS frequency, rather than changes in their efficiency at producing rain. We finally argue that the sensitivity of extremes to climate state may occur through both a change in occurrence and a change in internal MCS dynamics.

How to cite: Fildier, B., Carenso, M., Roca, R., and Fiolleau, T.: Mapping km-scale global extreme rainfall onto mesoscale convective systems lifecycle, frequency and dynamics , EGU General Assembly 2024, Vienna, Austria, 14–19 Apr 2024, EGU24-16951, https://doi.org/10.5194/egusphere-egu24-16951, 2024.