ISSN 1000-0526
CN 11-2282/P
CN 11-2282/P
Using Generative Adversarial Networks to Improve Heavy Precipitation Nowcasting in the Jianghuai Area
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
In this paper, the Spectrum GAN (SGAN) and the Multiscale GAN (MSG) are proposed by optimizing the Generative Adversarial Networks (GAN). Both of them are then combined with PhyDNet to form two deep learning based nowcasting methods for heavy precipitation in the Jianghuai region, namely PhySGAN and PhyMSG, which can provide precipitation forecasts for the next 3 hours. Through the forecast skill score, the performance evaluation in "complex scenarios" and the analysis of typical application examples, the performance of the two methods in the short-term heavy precipitation (precipitation ≥ 20 mm/h) forecast in Jiangsu during the flood season in 2024 is analyzed, and the following conclusions are obtained: (1) The TS of short-term heavy precipitation within 3 hours in different verification periods of PhySGAN and PhyMSG are significantly improved compared with those of the basic experiment PhyDNet and the general GAN experiment PhyGAN. The two new methods correct the low frequency problem of short-term heavy precipitation forecasts of PhyDNet and PhyGAN, so that the TS increases with the increase of the forecast lead time, thereby effectively extending the nowcasting lead time of short-term heavy precipitation.(2) Judging from the forecast performance shown by each method in "complex scenarios", deep learning can reflect the evolution of the generation and dissipation of heavy precipitation compared with traditional extrapolation methods. PhySGAN and PhyMSG show better forecast performance than PhyDNet and PhyGAN. The former has a better ability to depict local details such as the shape and intensity of heavy precipitation, while the latter has a better representation of the overall contour and position of the heavy precipitation rain band. (3) Combined with the application of typical heavy precipitation cases during the flood season, it is found that both PhySGAN and PhyMSG can forecast the precipitation enhancement process in advance in both systematic heavy precipitation and local heavy precipitation cases, which effectively guides the early warning before disasters. In addition, PhyMSG has a certain indicative effect on extreme rainfall intensities above 50 mm/h, while PhySGAN can better reflect the changes in the shape and position of the rain band.
Keywords:
Project Supported:
Clc Number:
Mobile website
