A comparative study of stochastic and deep generative models for multisite precipitation synthesis
This work addresses the need for better weather generation models to simulate future climate scenarios, but it is incremental as it focuses on comparative evaluation rather than introducing new methods.
The study compared classical stochastic weather generators (IBMWeathergen, RGeneratePrec) with deep generative models (GAN, VAE) for multisite precipitation synthesis, providing preliminary results to guide improvements in deep learning architectures for this task.
Future climate change scenarios are usually hypothesized using simulations from weather generators. However, there only a few works comparing and evaluating promising deep learning models for weather generation against classical approaches. This study shows preliminary results making such evaluations for the multisite precipitation synthesis task. We compared two open-source weather generators: IBMWeathergen (an extension of the Weathergen library) and RGeneratePrec, and two deep generative models: GAN and VAE, on a variety of metrics. Our preliminary results can serve as a guide for improving the design of deep learning architectures and algorithms for the multisite precipitation synthesis task.