Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator
Abstract Downscaling methods are critical in efficiently generating high‐resolution atmospheric data. However, state‐of‐the‐art statistical or dynamical downscaling techniques either suffer from the high computational cost of running a physical model or require high‐resolution data to develop a down...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Geophysical Union (AGU)
2023-07-01
|
| Series: | Journal of Advances in Modeling Earth Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1029/2023MS003800 |
| _version_ | 1797647722804674560 |
|---|---|
| author | Peishi Jiang Zhao Yang Jiali Wang Chenfu Huang Pengfei Xue T. C. Chakraborty Xingyuan Chen Yun Qian |
| author_facet | Peishi Jiang Zhao Yang Jiali Wang Chenfu Huang Pengfei Xue T. C. Chakraborty Xingyuan Chen Yun Qian |
| author_sort | Peishi Jiang |
| collection | DOAJ |
| description | Abstract Downscaling methods are critical in efficiently generating high‐resolution atmospheric data. However, state‐of‐the‐art statistical or dynamical downscaling techniques either suffer from the high computational cost of running a physical model or require high‐resolution data to develop a downscaling tool. Here, we demonstrate a recently proposed zero‐shot super‐resolution method, the Fourier neural operator (FNO), to efficiently perform downscaling without the need for high‐resolution data. Because the FNO learns dynamics in Fourier space, FNO is a resolution‐invariant emulator; it can be trained at a coarse resolution and produces emulation at any high resolution. We applied FNO to downscale a 4‐km resolution Weather Research and Forecasting (WRF) Model simulation of near‐surface heat‐related variables over the Great Lakes region. The FNO is driven by the atmospheric forcings and topographic features used in the WRF model at the same resolution. We incorporated a physics‐constrained loss in FNO by using the Clausius–Clapeyron relation to better constrain the relations among the emulated states. Trained on merely 600 WRF snapshots at 4‐km resolution, the FNO shows comparable performance with a widely‐used convolutional network, U‐Net, achieving averaged modified Kling–Gupta Efficiency of 0.88 and 0.94 on the test data set for temperature and pressure, respectively. We then employed the FNO to produce 1‐km emulations to reproduce the fine climate features. Further, by taking the WRF simulation as ground truth, we show consistent performances at the two resolutions, suggesting the reliability of FNO in producing high‐resolution dynamics. Our study demonstrates the potential of using FNO for zero‐shot super‐resolution in generating first‐order estimation on atmospheric modeling. |
| first_indexed | 2024-03-11T15:21:36Z |
| format | Article |
| id | doaj.art-f0fcf3b37d4e4155b272fb7739e7da29 |
| institution | Directory Open Access Journal |
| issn | 1942-2466 |
| language | English |
| last_indexed | 2024-03-11T15:21:36Z |
| publishDate | 2023-07-01 |
| publisher | American Geophysical Union (AGU) |
| record_format | Article |
| series | Journal of Advances in Modeling Earth Systems |
| spelling | doaj.art-f0fcf3b37d4e4155b272fb7739e7da292023-10-28T13:31:25ZengAmerican Geophysical Union (AGU)Journal of Advances in Modeling Earth Systems1942-24662023-07-01157n/an/a10.1029/2023MS003800Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural OperatorPeishi Jiang0Zhao Yang1Jiali Wang2Chenfu Huang3Pengfei Xue4T. C. Chakraborty5Xingyuan Chen6Yun Qian7Pacific Northwest National Laboratory Richland WA USAPacific Northwest National Laboratory Richland WA USAArgonne National Laboratory Lemont IL USAMichigan Technological University Houghton MI USAArgonne National Laboratory Lemont IL USAPacific Northwest National Laboratory Richland WA USAPacific Northwest National Laboratory Richland WA USAPacific Northwest National Laboratory Richland WA USAAbstract Downscaling methods are critical in efficiently generating high‐resolution atmospheric data. However, state‐of‐the‐art statistical or dynamical downscaling techniques either suffer from the high computational cost of running a physical model or require high‐resolution data to develop a downscaling tool. Here, we demonstrate a recently proposed zero‐shot super‐resolution method, the Fourier neural operator (FNO), to efficiently perform downscaling without the need for high‐resolution data. Because the FNO learns dynamics in Fourier space, FNO is a resolution‐invariant emulator; it can be trained at a coarse resolution and produces emulation at any high resolution. We applied FNO to downscale a 4‐km resolution Weather Research and Forecasting (WRF) Model simulation of near‐surface heat‐related variables over the Great Lakes region. The FNO is driven by the atmospheric forcings and topographic features used in the WRF model at the same resolution. We incorporated a physics‐constrained loss in FNO by using the Clausius–Clapeyron relation to better constrain the relations among the emulated states. Trained on merely 600 WRF snapshots at 4‐km resolution, the FNO shows comparable performance with a widely‐used convolutional network, U‐Net, achieving averaged modified Kling–Gupta Efficiency of 0.88 and 0.94 on the test data set for temperature and pressure, respectively. We then employed the FNO to produce 1‐km emulations to reproduce the fine climate features. Further, by taking the WRF simulation as ground truth, we show consistent performances at the two resolutions, suggesting the reliability of FNO in producing high‐resolution dynamics. Our study demonstrates the potential of using FNO for zero‐shot super‐resolution in generating first‐order estimation on atmospheric modeling.https://doi.org/10.1029/2023MS003800zero‐shot super resolutionthe Fourier neural operatorWRFregional climate modelingnear‐surface heat‐exposure estimates |
| spellingShingle | Peishi Jiang Zhao Yang Jiali Wang Chenfu Huang Pengfei Xue T. C. Chakraborty Xingyuan Chen Yun Qian Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator Journal of Advances in Modeling Earth Systems zero‐shot super resolution the Fourier neural operator WRF regional climate modeling near‐surface heat‐exposure estimates |
| title | Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator |
| title_full | Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator |
| title_fullStr | Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator |
| title_full_unstemmed | Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator |
| title_short | Efficient Super‐Resolution of Near‐Surface Climate Modeling Using the Fourier Neural Operator |
| title_sort | efficient super resolution of near surface climate modeling using the fourier neural operator |
| topic | zero‐shot super resolution the Fourier neural operator WRF regional climate modeling near‐surface heat‐exposure estimates |
| url | https://doi.org/10.1029/2023MS003800 |
| work_keys_str_mv | AT peishijiang efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT zhaoyang efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT jialiwang efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT chenfuhuang efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT pengfeixue efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT tcchakraborty efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT xingyuanchen efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator AT yunqian efficientsuperresolutionofnearsurfaceclimatemodelingusingthefourierneuraloperator |