GUST: A GPU-Accelerated 3D Urban Surface Temperature Model Using Monte Carlo Ray Tracing and Random Walk Simulation

The increasing urban heat resulting from climate change and urbanization poses significant risks to public health and weakens urban resilience. Physics-based urban surface temperature models are crucial for understanding the primary drivers of urban heat and developing effective mitigation strategies. An urban surface temperature model (GUST) has been developed on a GPU (Graphics Processing Units) platform to accurately simulate three-dimensional urban surface temperatures. The model uses the Monte Carlo method to compute radiative exchanges between urban surfaces through a reverse ray tracing algorithm and solves coupled heat transfer processes (conduction, radiation, and convection) via a random walking algorithm. GPU-based parallel computing accelerates the Monte Carlo simulations, enhancing computational efficiency.

This model is first validated against the scaled outdoor experiment SOMUCH, which provides high spatial and temporal resolution data. The model demonstrates strong accuracy in predicting surface temperatures. To understand the factors influencing model accuracy, the surface energy balance is analyzed. The results show that longwave radiative exchanges between urban surfaces are a significant factor, while convective heat transfer plays a relatively smaller role. Lastly, the model is applied to simulate temperatures on complex urban surfaces, showcasing its applicability to real-world urban configurations. In conclusion, the validated and highly parallelized GUST model offers an efficient and accurate approach to simulating urban surface temperatures, providing valuable insights into the causes of urban heat and supporting the development of effective urban heat mitigation strategies.