M.Sc. Dissertation Proposal

Vision and Motivation

The realistic simulation of clouds in synthetic environments has always been a hot research topic in computer graphics. However, simulation of clouds on a computer is a complicated task since they have physics-dependent dynamic shapes that evolve over time. The leading idea of this work is to combine both computer graphics and meteorology standpoints to simulate and render different types of weather clouds and, eventually, raining, and thunderstorms over time, as needed in flight simulators, virtual reality, computer games, and movies. As described in [2017-Duarte], simulating and rendering weather clouds over time is a challenging research topic, but it was already achieved during in-house experiments, in 2017. Nevertheless, this research exclusively targeted cumulus clouds, not other types of weather clouds.

State-of-the-Art

Clouds, water, fire, and smoke are natural phenomena, i.e., they are found in nature. However, because of their dynamics, they are not easily simulated and replicated on computer. In meteorology, the simulation procedures mainly aim at weather forecasting over the planet, not to visually reproduce clouds in a realistic manner on computer screen, as usual in computer graphics. But, this is computationally expensive because cloud dynamics are regulated by physically-based equations, i.e., Navier-Stokes equations. Therefore, these accurate physically-based methods do not fulfil the real-time requirements of computer graphics applications [2015-Barbosa]. In fact, to achieve real-time rates, many methods in computer graphics for the simulation of cloud formation are procedural, i.e., essentially clouds are modeled as static entities.

Research Methodology

We will take advantage of thermodynamic diagrams, also called SkewT/LogP (Skewed Temperature and Logarithmic Pressure) diagrams, to quickly solve partial differential equations that model the flow of clouds as fluids into the atmosphere. Our technique is data-driven because atmospheric data (also called atmospheric soundings) are obtained from weather forecasting agencies around the planet [1979-Whiton] (see, for example, http://www.woweather.com). This research work aims to generalize our technique based on SkewT/LogP diagrams (i.e., temperature/pressure diagrams) introduced in [2017-Duarte] to generate other types of clouds (e.g., cirrus, nimbus, mountain clouds, and so forth), and in the limit to create weather cloud agglomerates that change over time. We also intend to introduce deep learning techniques to predict the weather conditions from time series and their implications in the dynamics of cloud agglomerates, and this is something that has never been tried before in computer graphics. However, predicting the weather in meteorology through time series has been already tried before [2023-Chen].

References

[1979-Whiton] Whiton, R. and Overall, J. (1979): The Use of the skew- T/LogP diagram in analysis and forecasting. Air Weather Service, Technical Report AWS/TRA-79/006, Scott Air Force Base, Illinois, USA.
[2015-Barbosa] Barbosa, F., Welton, C., Dobashi, Y. and Yamamoto, T. (2015): Adaptive cloud simulation using position-based fluids. Computer Animation and Virtual Worlds 26(3-4): 367-375.
https://doi.org/10.1002/cav.1657
[2017-Duarte] Duarte, R.P. and Gomes, A.J.P. (2017): Real-time simulation of cumulus clouds through SkewT/LogP diagrams. Computers & Graphics 67(2017): 103-114.
https://doi.org/10.1016/j.cag.2017.06.005
[2023-Chen] Chen, L.; Han, B.; Wang, X.; Zhao, J.; Yang, W.; Yang, Z. (2023): Machine Learning Methods in Weather and Climate Applications: A Survey. Applied Sciences 13(21) 12019: 1-36.
https://doi.org/10.3390/app132112019