Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Precipitation Nowcasting Using Physics Informed Discriminator Generative Models

Junzhe Yin, Cristian Meo, Ankush Roy, Zeineh Bou Cher, Yanbo Wang, Ruben Imhoff, Remko Uijlenhoet, Justin Dauwels

TL;DR

This model draws inspiration from the novel Physics-Informed Discriminator GAN (PID-GAN) formulation, directly integrating physics-based supervision within the adversarial learning framework, and outperforms numerical and SOTA deep generative models in terms of precipitation nowcasting downstream metrics.

Abstract

Nowcasting leverages real-time atmospheric conditions to forecast weather over short periods. State-of-the-art models, including PySTEPS, encounter difficulties in accurately forecasting extreme weather events because of their unpredictable distribution patterns. In this study, we design a physics-informed neural network to perform precipitation nowcasting using the precipitation and meteorological data from the Royal Netherlands Meteorological Institute (KNMI). This model draws inspiration from the novel Physics-Informed Discriminator GAN (PID-GAN) formulation, directly integrating physics-based supervision within the adversarial learning framework. The proposed model adopts a GAN structure, featuring a Vector Quantization Generative Adversarial Network (VQ-GAN) and a Transformer as the generator, with a temporal discriminator serving as the discriminator. Our findings demonstrate that the PID-GAN model outperforms numerical and SOTA deep generative models in terms of precipitation nowcasting downstream metrics.

Precipitation Nowcasting Using Physics Informed Discriminator Generative Models

TL;DR

This model draws inspiration from the novel Physics-Informed Discriminator GAN (PID-GAN) formulation, directly integrating physics-based supervision within the adversarial learning framework, and outperforms numerical and SOTA deep generative models in terms of precipitation nowcasting downstream metrics.

Abstract

Nowcasting leverages real-time atmospheric conditions to forecast weather over short periods. State-of-the-art models, including PySTEPS, encounter difficulties in accurately forecasting extreme weather events because of their unpredictable distribution patterns. In this study, we design a physics-informed neural network to perform precipitation nowcasting using the precipitation and meteorological data from the Royal Netherlands Meteorological Institute (KNMI). This model draws inspiration from the novel Physics-Informed Discriminator GAN (PID-GAN) formulation, directly integrating physics-based supervision within the adversarial learning framework. The proposed model adopts a GAN structure, featuring a Vector Quantization Generative Adversarial Network (VQ-GAN) and a Transformer as the generator, with a temporal discriminator serving as the discriminator. Our findings demonstrate that the PID-GAN model outperforms numerical and SOTA deep generative models in terms of precipitation nowcasting downstream metrics.
Paper Structure (10 sections, 9 equations, 2 figures, 1 table)

This paper contains 10 sections, 9 equations, 2 figures, 1 table.

Table of Contents

  1. INTRODUCTION
  2. Dataset
  3. Problem Formulation
  4. Related Works
  5. Methodology
  6. PID-GAN: Model Architecture
  7. Generator
  8. Physics-Informed Discriminator: PID-GAN
  9. Experiments
  10. Conclusion

Figures (2)

  • Figure 1: The proposed PID-GAN model structure. C means concatenation and $\eta$ represent $\eta_k = e^{-\lambda \mathcal{R}^{(k)}(x,\hat{x})}$, refering to the equation of the physics consistency score.
  • Figure 2: The precision-recall curves for detecting extreme events over 3 hours at 12 Dutch catchments. Every point from right to left represents a different precipitation threshold $(0.5$ to $10 \text{mm}/3\text{h})$ for prediction and a fixed threshold for ground truth by definition of extreme eventsbi2023nowcasting.