Deep Learning for Downscaling Tropical Cyclone Rainfall to Hazard-Relevant Spatial Scales

Key Points:

· We find that the Wasserstein Generative Adversarial Network can be used to downscale tropical cyclone rainfall

· The Wasserstein Generative Adversarial Network reproduces the fine-scale spatial structure of tropical cyclones both visually and in its power spectra

· The Wasserstein Generative Adversarial Network is able to extrapolate to storms more extreme than seen in training

Deep Learning for Downscaling Tropical Cyclone Rainfall to Hazard-Relevant Spatial Scales

Abstract Flooding, driven in part by intense rainfall, is the leading cause of mortality and damages from the most intense tropical cyclones. With rainfall from tropical cyclones set to increase under anthropogenic climate change, it is critical to accurately estimate extreme rainfall to better support short-term and long-term resilience efforts. While high-resolution climate models capture tropical cyclone statistics better than low-resolution models, they are computationally expensive. This leads to a trade-off between capturing tropical cyclone features accurately, and generating large enough simulation data sets to sufficiently sample high-impact, low-probability events. Downscaling can assist by predicting high-resolution features from relatively cheap, low-resolution models. Here, we develop and evaluate a set of three deep learning models for downscaling tropical cyclone rainfall to hazard-relevant spatial scales. We use rainfall from the Multi-Source Weighted-Ensemble Precipitation observational product at a coarsened resolution of approximately one hundred kilometers, and apply our downscaling model to reproduce the original resolution of approximately ten kilometers. We find that the Wasserstein Generative Adversarial Network is able to capture realistic spatial structures and power spectra and performs the best overall, with mean biases within five percent of observations. We also show that the model can perform well at extrapolating to the most extreme storms, which were not used in training.

Plain Language Summary Tropical cyclones are often associated with intense winds, but it is actually their associated rainfall and flooding that cause the majority of mortality and damages. A warmer atmosphere is able to hold more water vapor and therefore we expect to see increases in rainfall from tropical cyclones with global warming. To better support resilience efforts, it is critical to model current and future tropical cyclone rainfall, but climate models at standard resolution struggle to do this accurately. Running climate models at very high resolution produces better results, though this requires significant computational resources meaning that fewer high-impact, low-probability tropical cyclones can be generated. Other methods, called downscaling models, are used to provide a computationally cheaper alternative by generating high-resolution tropical cyclone-specific data rather than an entire global climate data set. In this study, we develop a set of deep learning models which can generate high-resolution rainfall data from low-resolution rainfall data. To do this, we train our models on data from observational data sets that have data for the period nineteen seventy-nine to twenty twenty. We find that the Wasserstein Generative Adversarial Network performs the best over the metrics studied and is able to reproduce the most extreme storms that were not used in training.