Thunderscapes: Simulating the Dynamics of Mesoscale Convective System

TL;DR

Thunderscapes introduces a physically based framework for visually realistic mesoscale convective systems by fusing Grabowski-style cloud microphysics with hydrometeor electrification into an artist-friendly, GPU-accelerated pipeline compatible with Houdini. The model tracks six state fields $q_v$, $q_c$, $q_p$, $\theta$, $\rho$, and $\mathbf{u}$, integrating cloud microphysics, electrification, lightning, atmospheric background, and fluid dynamics to simulate diverse thunderstorm types and lightning phenomena. Validation against meteorological data and comparisons with Stormscapes demonstrate improved cloud structure, especially at higher altitudes, and more physically consistent lightning dynamics, supporting production-ready realism and potential extension to broader mesoscale phenomena. The work lays groundwork for richer microphysical detail, expanded storm categories, and adaptive grid strategies to balance fidelity and performance in visual effects applications.

Abstract

A Mesoscale Convective System (MCS) is a collection of thunderstorms operating as a unified system, showcasing nature's untamed power. They represent a phenomenon widely referenced in both the natural sciences and the visual effects (VFX) industries.However, in computer graphics, visually accurate simulation of MCS dynamics remains a significant challenge due to the inherent complexity of atmospheric microphysical processes.To achieve a high level of visual quality while ensuring practical performance, we introduce Thunderscapes, the first physically based simulation framework for visually realistic MCS tailored to graphical applications.Our model integrates mesoscale cloud microphysics with hydrometeor electrification processes to simulate thunderstorm development and lightning flashes. By capturing various thunderstorm types and their associated lightning activities, Thunderscapes demonstrates the versatility and physical accuracy of the proposed approach.

Figures (6)

• Figure 1: Schematic illustration of our MCS life cycle, encompassing the interplay and feedback among key modules. Detailed explanations are provided in Section \ref{['sec:Method']}.(I) Cloud Microphysics: Describes the phase transitions of hydrometeors during MCS development.(II) Hydrometeor Electrification: Explains the accumulation of charge density resulting from the collision and coalescence of hydrometeors.(III) Lightning Discharge: Highlights the process of lightning channel formation and propagation, triggered when static charge exceeds a predefined electric field threshold, leading to the neutralization of hydrometeors along the lightning channel.(IV) Atmospheric Background: Introduces vertical buoyancy forces to the system driven by temperature variations caused by hydrometeor phase transitions.(V) Fluid Dynamics: Drives the system's temporal evolution and introduces forces influencing the dynamics of the MCS life cycle.
• Figure 2: Comparison between simulation results (a-d) and realistic observations (e-h) for four storm types: single cell, multicell, squall line, and supercell. This layout highlights the structural similarities between our simulated storms and their real-world counterparts. Additionally, the last row (i-l) illustrates the ground properties (including the temperature field and vapor field initialized with a consistent pattern based on a flat heightfield) associated with each thunderstorm type.
• Figure 3: Simulation inspired by real-world weather events:(a) Florida’s Biscayne National Park, (b) New Mexico’s Chiricahua Mountains, (c) Japan’s Fuji-Hakone-Izu National Park, (d) California’s Monterey Bay. These visualizations highlight the geographical diversity and meteorological phenomena captured in our simulation framework.Additionally, the last row (e-h) illustrates the ground properties (including the temperature field and vapor field, both initialized with a consistent pattern derived from a specific regional heightfield) associated with each region.
• Figure 4: Visualization comparing cloud fraction models and simulation results for four storm types: single cell, multi-cell, squall line, and supercell. Top row: Cloud fraction models, where orange represents our method, and blue represents the method of Hädrich et al. hadrich2020stormscapes. Middle row: Results generated using Hädrich et al. hadrich2020stormscapes. Bottom row: Results generated by our method. The visualization highlights differences in cloud fraction structures and thunderstorm characteristics captured by the two approaches.
• Figure 5: Comparison of MCS across four distinct regions: Florida, New Mexico, Japan, and California. Top row: Cloud coverage evolution over a 24-hour period during thunderstorm events. Middle row: Cloud fraction analysis, illustrating the structure and distribution of clouds at the maximum development height of thunderstorms. Bottom row: Lightning activity evolution, capturing the variations in lightning activity as thunderstorms develop. The figures highlight the unique patterns shaped by regional environmental conditions.