Project No: 16307323

Title: Estimating Tropical Cyclone Changes due to Global Warming with Smart Dynamical Downscaling and Convection-Permitting Simulations

Principal Investigator: Prof. Xiaoming SHI

Co-Investigator: Prof. Johnny CHAN


Tropical cyclones (TCs) have substantial societal impacts in coastal regions where they make their landfall. Global warming can potentially alter TC characteristics in many ways, such as changing their spatial distributions, occurrence frequency, moving speed, and intensity. Quantifying TC sensitivities to climate change is challenging because global simulations and high resolutions are needed at the same time, making the computational cost unaffordable to individual researchers. The former is necessary for predicting large-scale circulation changes while resolving meso- to small-scale processes like convection and turbulence requires the latter. Studies by modeling centers have pushed the resolution of global climate simulations to 14 km in recent years, and significant advances have been achieved regarding TC responses to global warming. However, substantial uncertainties still exist regarding TC characteristics under climate change. Here we propose using a smart dynamical downscaling (SDD) approach to assess how intense tropical cyclones may respond to global warming. The SDD approach trains a deep learning (DL) model to predict surface variables, such as wind and precipitation, at unresolved scales (e.g., 1 km) based on coarse-resolution (e.g., 100 km) large-scale circulation data. Carefully trained, the DL model can be applied to global climate simulation data to help rank circulation patterns and identify those time slices that can produce intense tropical cyclones. Then kilometer-scale resolution, also known as the convection-permitting resolution, can be used for downscaling simulations of those selected time slices. In SDD, computational resources requirement becomes less demanding, but convection can be resolved confidently with sufficient resolutions. This project will apply the SDD approach to the climate simulations under the SSP5-8.5 warming scenario from 2020 to 2100. Many kilometer-scale resolution simulations of potential TC cases will be conducted, and the 20 most intense TCs making landfall in China's coastal regions for each 10-year subperiod will be kept for further analysis. We will evaluate how those severe TCs respond to global warming in terms of their spatial distribution, translation speed, maximum sustained wind, and precipitation. We will analyze the connections between those TC characteristics changes and other dynamical and physical processes, such as large-scale circulation patterns, surface enthalpy flux, and cloud-radiation feedback, aiming to provide a physical understanding of the relevant changes. Potential discrepancies between our predictions and previous studies will be quantified and discussed.