This study develops a physics-informed neural network (PINN) model to efficiently generate high-fidelity local circulation patterns in Taiwan while preserving an accurate representation of the physical relationship between generated local turbulence and upstream synoptic flow regimes. Large ensemble semi-realistic simulations were conducted using a high-resolution (2 km) TaiwanVVM model, in which the critical characteristics of the various synoptic flow regimes are carefully selected to concentrate on the effects of local circulation variations. A variational autoencoder (VAE) was constructed to capture essential representations of local circulation scenarios associated with the lee vortices by training on the ensemble dataset. The VAE’s latent space effectively encapsulated synoptic flow regimes as controlling factors, aligning with the physical understanding of Taiwan’s local circulation. The critical transition of flow regimes under the southeasterly synoptic flow regimes is also well represented in the VAE’s latent space, indicating that the VAE can learn the nonlinear characteristics of the multiscale interaction of the lee vortex. We further construct a reduced-order model for local circulation predictions under diverse synoptic conditions. This physics-informed VAE ensures the accurate prediction of the nonlinear characteristics of multiscale interactions between synoptic flows and the turbulence induced by topography, thereby accelerating the assessment of the local circulation responses under various climate change scenarios.

This study applied representation learning algorithms to satellite images and evaluated the learned latent spaces with classifications of various weather events. The algorithms investigated include the classical linear transformation, i.e., principal component analysis (PCA), state-of-the-art deep learning method, i.e., convolutional autoencoder (CAE), and a residual network pre-trained with large image datasets (PT). The experiment results indicated that the latent space learned by CAE consistently showed higher threat scores for all classification tasks. The classifications with PCA yielded high hit rates but also high false-alarm rates. In addition, the PT performed exceptionally well at recognizing tropical cyclones but was inferior in other tasks. Further experiments suggested that representations learned from higher-resolution datasets are superior in all classification tasks for deep-learning algorithms, i.e., CAE and PT. We also found that smaller latent space sizes had little impact on the classification task’s hit rate. Still, a latent space dimension smaller than 128 caused a significantly higher false-alarm rate. Though the CAE can learn latent spaces effectively and efficiently, the interpretation of the learned representation lacks direct connections to physical attributions. Therefore, developing a physics-informed version of CAE can be a promising outlook for the current work.

This study develops an explainable variational autoencoder (VAE) framework to efficiently generate high-fidelity local circulation patterns in Taiwan, ensuring an accurate representation of the physical relationship between generated local circulation and upstream synoptic flow regimes. Large ensemble semi-realistic simulations were conducted using a high-resolution (2 km) model, TaiwanVVM, where critical characteristics of various synoptic flow regimes were carefully selected to focus on the effects of local circulation variations. The VAE was constructed to capture essential representations of local circulation scenarios associated with the lee vortices by training on the ensemble dataset. The VAE’s latent space effectively captures the synoptic flow regimes as controlling factors, aligning with the physical understanding of Taiwan’s local circulation dynamics. The critical transition of flow regimes under the influence of southeasterly synoptic flow regimes is also well represented in the VAE’s latent space.This indicates that the VAE can learn the nonlinear characteristics of the multiscale interactions involving the lee vortex. The latent space within VAE can serve as a reduced-order model for predicting local circulation using synoptic wind speed and direction. This explainable VAE ensures the accurate predictions of the nonlinear characteristics of multiscale interactions between synoptic flows and the local circulation induced by topography, thereby accelerating the assessments under various climate change scenarios.

The anelastic theory of effective buoyancy has been generalized to include the effects of momentum flux convergence. Mediated by the nonlocal perturbation pressure, the dynamics tends to average over details of the forcing, yielding acceleration robust to small-scale variations of the flow. Here we demonstrate in a large-eddy simulation (LES) with a 100-m horizontal grid spacing that including the anelastic nonlocal dynamics can help capture the mean evolution of convection without fully resolving the fine-scale coherent turbulent structures embedded in the flow. Instances of convection in the LES are identified. For these, the buoyancy and dynamic contributions to the vertical momentum tendency are separately diagnosed. The diagnoses show that buoyancy is the leading effect in the vertical acceleration while strongly interacting with the vertical momentum flux convergence. In comparison, the influence of the horizontal momentum flux convergence on the vertical motion are substantially weaker. The sensitivity resulting from averaging over fine-scale features are quantified. For deep-convective cases, these contributions at the cloud scale ($\sim8$ km) exhibit a robustness—as measured in a root-mean-square sense—to horizontally smoothing out turbulent features of scales $\lesssim3$ km. As expected, such scales depend on the size of the convective element of interest, while dynamic contributions tend to be more susceptible to horizontal smoothing than does the buoyancy contribution. By verifying a key attribute of the pressure-mediated dynamics in an LES, results here lend support to simplifying the representation of moist convection under the anelastic nonlocal framework for global climate models and storm-resolving simulations.

Observations suggest tropical convection intensifies when aerosol concentrations enhance, but quantitative estimations of this effect remain highly uncertain. Leading theories for explaining the influence of aerosol concentrations on tropical convection are based on the dynamical response of convection to changes in cloud microphysics, neglecting possible changes in the environment. In recent years, global convection-permitting models (GCPM) have been developed to circumvent problems arising from imposing artificial scale separation on physical processes associated with deep convection. Here, we use a GCPM to investigate how enhanced concentrations of aerosols that act as cloud condensate nuclei (CCN) impact tropical convection features by modulating the convection-circulation interaction. Results from a pair of idealized non-rotating radiative-convective equilibrium simulations show that the enhanced CCN concentration leads to weaker large-scale circulation, the closeness of deep convective systems to the moist cluster edges, and more mid-level cloud water at an equilibrium state in which convective self-aggregation occurred. Correspondingly, the enhanced CCN concentration modulates how the diabatic processes that support or oppose convective aggregation maintain the aggregated state at equilibrium. Overall, the enhanced CCN concentration facilitates the development of deep convection in a drier environment but reduces the large-scale instability and the convection intensity. Our results emphasize the importance of allowing atmospheric phenomena to evolve continuously across spatial and temporal scales in simulations when investigating the response of tropical convection to changes in cloud microphysics.

Observations suggest tropical convection intensifies when aerosol concentrations enhance, but quantitative estimations of this effect remain highly uncertain. Leading theories for explaining the intensification are based on the dynamical response of convection to changes in cloud microphysics independently from possible changes in the environment. Here, we provide a new perspective on aerosol indirect effects on tropical convection by using a global convection-permitting model that realistically simulates convection-circulation interaction. Simulations of radiative-convective equilibrium show that pollution facilitates the development of deep convection in a drier environment, but cloud condensates are more likely to be exported from moist clusters to dry areas, impeding the large-scale moisture-convection feedback and limiting the intensity of maximum precipitation (30 vs. 47 mm h\textsuperscript{-1}). Our results emphasize the importance of allowing atmospheric phenomena to evolve continuously across spatial and temporal scales in simulations when investigating the response of tropical convection to changes in cloud microphysics.

The present study objectively classify the convective cloud objects detected by the space-borne CloudSat radar over the Asian-Australian monsoon region using the hierarchical agglomerative clustering algorithm. Based on key properties representing the morphological features and convective intensity of the systems, five distinct convective cloud regimes are derived. The unique Coastal-Intense regime exhibits the most expansive horizontal scales (> 1000 km), high convective strength, the strongest cloud radiative effects, the highest probability of extreme rainfall, and a significant coupling with the sharp onset of the Asian summer monsoon circulation. Secondly, the Coastal regime illustrates smaller but also highly organized coastal convections, with the strongest convective strength. Less than 10% of the systems in the CI and Coastal regimes overlap with the tropical cyclones. The rest three regimes mark the less organized convection at various life cycle stages mainly over the land areas, with small seasonal variation in their occurrence.

Two cloud-resolving models (SCALE and VVM) take different pathways toward convective self-aggregation (CSA) in the radiative-convective equilibrium simulations, although they have similar domain-averaged properties. Analyses in the moisture space show that radiative cooling in the dry area mainly drives CSA in SCALE, while subsidence triggered by the convection in the moist region dominates in VVM. The change in the convective structures is found on the isentropic diagram in VVM, but this transition is unclear in SCALE. The object-based analysis provides that the convective systems with larger sizes are rare in SCALE. In contrast, large-size convective systems frequently develop in the moist region as CSA evolves in VVM. The large-size systems can efficiently drive the circulation between the dry and moist areas. The different pathways to CSA are associated with the transition of convective structures, which provides a new insight to understand CSA among cloud-resolving models based on the perspective of the mechanism.

Two cloud-resolving models (SCALE and VVM) take different pathways toward convective self-aggregation under a radiative-convective equilibrium condition, although the model setups are the same. The physical processes driving the development of self-aggregation in SCALE and VVM, respectively, correspond to the mechanisms for self-aggregation in cold and warm climate states discussed in a previous study. Analyses in the moisture space show that radiative cooling in the dry area mainly drives the aggregation in SCALE, while subsidence induced by the convection in the moist region dominates in VVM. The transition of the convective circulation from convective scale to mesoscale is found on the isentropic diagram in VVM, but this transition is unclear in SCALE. The convective clouds with larger sizes are rare in SCALE. The frequent occurrence of larger convective clouds efficiently drives self-aggregation from the moist area in VVM. These results provide a new insight to understand convective self-aggregation among models.

This study introduces a framework to evaluate convective aggregation (CA) under radiative-convective equilibrium simulations using the vector vorticity equation cloud-resolving model (VVM) coupled to a mixed-layer slab ocean. The framework introduces the competing effects between the convection-SST feedback (CSF) and the moisture-convection feedback (MCF) by modifying the initial SST gradient and the mixed layer depth. Examples of applying this framework are demonstrated by comparing simulations with different microphysics schemes. The convectional five-category scheme (VVM-Lin) and the predicted particle properties scheme (P3) are examined by the matrix formed by these two factors. A clear bifurcation of the aggregated/non-aggregated states can be identified in both sets of experiments. The change of the bifurcation between two sets of experiments suggests that the P3 experiments tend to develop CA due to the stronger MCF. The budget analysis of spatial frozen moist static energy variance is applied to quantify the stronger MCF in the P3 simulations. Convective systems in P3 simulations develop more organized structures, which enhances MCF and leads to CA. The proposed framework provides a reconciled view of the process-based evaluation of CA among cloud-resolving models that use different dynamics and physical parameterizations.

In this study, the Superparameterized Community Atmosphere Model (SPCAM) is used to simulate tropical cyclones (TCs) using the hindcast approach. Three hindcast experiments are conducted with 32-km (D32), 128-km (D128), and 1024-km (D1024) horizontal scales in the sub-grid cloud-resolving models (CRMs). The results show that D1024 produces reasonable TCs compared with the reanalysis data. It is 3.42 TCs per 10 days for the reanalysis data, while there are 8.07, 4.88, and 3.73 for D32, D128, and D1024. The bias of overestimating TC numbers grows with decreasing CRM scale. The D32 experiment also produces stronger TCs with a higher precipitation rate and wind speed. The bias is highly related to the efficiency of adjusting convective instability in the sub-grid CRM. The D32 exhibits higher column-integrated water vapor under warm conditions compared with D1024, indicating its inefficiency in removing water vapor by the weaker convective mass fluxes in the small CRM scale. That is, the vertical transport of convection in a smaller horizontal scale will be restricted by stronger subsidence because CRM columns for compensating are limited. The distribution of accumulated convective instability is broader and more frequent in D32. As a result, large-scale precipitating systems tend to develop in D32, leading to a higher probability of TC genesis. This study highlights the importance of sub-grid configuration when estimating TC activities using SPCAM.

This study investigates the effects of microphysical processes on the precipitation spectrum in a strongly forced environment using a vector vorticity cloud-resolving model (VVM). Experiments are performed under imposed advective cooling and moistening with two microphysics parameterizations: predicted particle properties scheme (P3) and Lin scheme (VVM-Lin). Even though the domain-averaged precipitation is similar in two experiments, P3 exhibits stronger extreme precipitation in the spectrum compared with VVM-Lin. Changes in convective structures are responsible for such a difference. Using the isentropic analyses, we identify that in P3, stronger convective updrafts take place in the high frozen equivalent potential temperature regime where air parcels rarely reach. This is caused by the reduced melting of rimed ice particles for energic parcels. Through defining convective core clouds, the relation between the convective structure on the isentropic diagram and the extreme precipitation can be identified. The shifts toward extreme intensity in the precipitation spectrum suggest that the microphysical processes have significant impacts on the extreme precipitation by the convective core clouds. The treatment of microphysics has significant impacts on the convective structures and then alter the probability of extreme events under the strongly forced environment.