A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation
Abstract Anticipating climate impacts and risks in present or future climates requires predicting the statistics of high‐impact weather events at fine‐scales. Direct numerical simulations of fine‐scale weather are computationally too expensive for many applications. While deterministic‐based (deep‐l...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Geophysical Union (AGU)
2025-01-01
|
| Series: | Journal of Advances in Modeling Earth Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1029/2024MS004668 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832583490088992768 |
|---|---|
| author | Neelesh Rampal Peter B. Gibson Steven Sherwood Gab Abramowitz Sanaa Hobeichi |
| author_facet | Neelesh Rampal Peter B. Gibson Steven Sherwood Gab Abramowitz Sanaa Hobeichi |
| author_sort | Neelesh Rampal |
| collection | DOAJ |
| description | Abstract Anticipating climate impacts and risks in present or future climates requires predicting the statistics of high‐impact weather events at fine‐scales. Direct numerical simulations of fine‐scale weather are computationally too expensive for many applications. While deterministic‐based (deep‐learning or statistical) downscaling of low‐resolution climate simulations are several orders of magnitude faster than direct numerical simulations, it suffers from several limitations. These limitations include the tendency to regress to the mean, which produces excessively smooth predictions and underestimates the magnitude of extreme events. They also fail to preserve statistical measures that are key for climate research. We use a conditional GAN (cGAN) architecture to downscale daily precipitation as a Regional Climate Model (RCM) emulator. The cGAN generates plausible residuals on top of the predictable expectation state produced by a deterministic deep learning algorithm. The skill of cGANs is highly sensitive to a hyperparameter known as the weight of the adversarial loss (λadv), where the value of λadv required for accurate results varies with season and performance metric, casting doubt on the reliability of cGANs as usually implemented. However, by applying a simple intensity constraint to the loss function, it is possible to obtain reliable performance results across λadv spanning two orders of magnitude. CGANs are considerably more skillful in capturing climatological statistics, including the distribution and spatial characteristics of extreme events. With this modification, we expect cGANs to be readily transferable to other applications and time periods, making them a useful weather generator for representing extreme event statistics in present and future climates. |
| format | Article |
| id | doaj-art-925236889e6341f7860c4b3db4a1d935 |
| institution | Kabale University |
| issn | 1942-2466 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | American Geophysical Union (AGU) |
| record_format | Article |
| series | Journal of Advances in Modeling Earth Systems |
| spelling | doaj-art-925236889e6341f7860c4b3db4a1d9352025-01-28T13:21:09ZengAmerican Geophysical Union (AGU)Journal of Advances in Modeling Earth Systems1942-24662025-01-01171n/an/a10.1029/2024MS004668A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather GenerationNeelesh Rampal0Peter B. Gibson1Steven Sherwood2Gab Abramowitz3Sanaa Hobeichi4National Institute of Water and Atmospheric Research Auckland New ZealandNational Institute of Water and Atmospheric Research Wellington New ZealandClimate Change Research Centre & ARC Centre of Excellence for Climate Extremes University of New South Wales Sydney NSW AustraliaClimate Change Research Centre & ARC Centre of Excellence for Climate Extremes University of New South Wales Sydney NSW AustraliaClimate Change Research Centre & ARC Centre of Excellence for Climate Extremes University of New South Wales Sydney NSW AustraliaAbstract Anticipating climate impacts and risks in present or future climates requires predicting the statistics of high‐impact weather events at fine‐scales. Direct numerical simulations of fine‐scale weather are computationally too expensive for many applications. While deterministic‐based (deep‐learning or statistical) downscaling of low‐resolution climate simulations are several orders of magnitude faster than direct numerical simulations, it suffers from several limitations. These limitations include the tendency to regress to the mean, which produces excessively smooth predictions and underestimates the magnitude of extreme events. They also fail to preserve statistical measures that are key for climate research. We use a conditional GAN (cGAN) architecture to downscale daily precipitation as a Regional Climate Model (RCM) emulator. The cGAN generates plausible residuals on top of the predictable expectation state produced by a deterministic deep learning algorithm. The skill of cGANs is highly sensitive to a hyperparameter known as the weight of the adversarial loss (λadv), where the value of λadv required for accurate results varies with season and performance metric, casting doubt on the reliability of cGANs as usually implemented. However, by applying a simple intensity constraint to the loss function, it is possible to obtain reliable performance results across λadv spanning two orders of magnitude. CGANs are considerably more skillful in capturing climatological statistics, including the distribution and spatial characteristics of extreme events. With this modification, we expect cGANs to be readily transferable to other applications and time periods, making them a useful weather generator for representing extreme event statistics in present and future climates.https://doi.org/10.1029/2024MS004668generative adversarial networksclimate downscalingregional climate modelingdeep learningstatistical downscalingclimate projections |
| spellingShingle | Neelesh Rampal Peter B. Gibson Steven Sherwood Gab Abramowitz Sanaa Hobeichi A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation Journal of Advances in Modeling Earth Systems generative adversarial networks climate downscaling regional climate modeling deep learning statistical downscaling climate projections |
| title | A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation |
| title_full | A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation |
| title_fullStr | A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation |
| title_full_unstemmed | A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation |
| title_short | A Reliable Generative Adversarial Network Approach for Climate Downscaling and Weather Generation |
| title_sort | reliable generative adversarial network approach for climate downscaling and weather generation |
| topic | generative adversarial networks climate downscaling regional climate modeling deep learning statistical downscaling climate projections |
| url | https://doi.org/10.1029/2024MS004668 |
| work_keys_str_mv | AT neeleshrampal areliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT peterbgibson areliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT stevensherwood areliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT gababramowitz areliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT sanaahobeichi areliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT neeleshrampal reliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT peterbgibson reliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT stevensherwood reliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT gababramowitz reliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration AT sanaahobeichi reliablegenerativeadversarialnetworkapproachforclimatedownscalingandweathergeneration |