English
 
Privacy Policy Disclaimer
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT
Add to Basket
 Local TagsRelease HistoryDetailsSummary
  Improving climate models using corrective machine learning

Clark, S., Kwa, A., Watt-Meyer, O., Henn, B., McGibbon, J., Perkins, A., Brenowitz, N., Bretherton, C., Harris, L. (2023): Improving climate models using corrective machine learning, XXVIII General Assembly of the International Union of Geodesy and Geophysics (IUGG) (Berlin 2023).
https://doi.org/10.57757/IUGG23-2145

Item is

 

show all hide all

Basic

show hide
Item Permalink: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5018669 Version Permalink: https://gfzpublic.gfz-potsdam.de/pubman/item/item_5018669_4
Genre: Conference Paper

Files

show Files

Locators

show

Creators

show
hide
 Creators:
Clark, Spencer1, Author
Kwa, Anna1, Author
Watt-Meyer, Oliver1, Author
Henn, Brian1, Author
McGibbon, Jeremy1, Author
Perkins, Andre1, Author
Brenowitz, Noah1, Author
Bretherton, Christopher1, Author
Harris, Lucas1, Author
Affiliations:
1IUGG 2023, General Assemblies, 1 General, International Union of Geodesy and Geophysics (IUGG), External Organizations, ou_5011304              

Content

show
hide
Free keywords: -
 Abstract: AI2, partnering with GFDL, has developed a corrective machine learning (ML) methodology to improve weather forecast skill and reduce climate biases in a computationally efficient coarse-grid climate model. The corrective ML is trained by nudging the 3D temperature, humidity and wind fields forecast by the coarse-grid model to a time-dependent global reference and learning the ‘nudging tendencies’ as a function of the column state of the model. The reference can be a reanalysis (for present-climate simulation) or a finer-grid version of the same model that may be more trustworthy across a range of climates. The ML is interpreted as a correction to the combined physics parameterizations of the coarse-grid model. We trained two versions of corrective ML for a 200 km grid version of the US weather forecast model, FV3GFS. First, we train on two-year simulations with a finer 25 km grid version of FV3GFS in four climates specified using sea-surface temperature (SST) offsets of -4K, 0, 4K and 8K from a present-day climatological annual cycle. The ML reduces spatial biases of annual-mean land surface temperature and precipitation in prognostic simulations in all climates and also in an independent simulation with an SST offset ramped from 0K to 4K. Second, we train corrective ML using a year-long reference simulation with GFDL’s X-SHiELD global storm resolving model (dx = 3 km, 79 levels). This ML reduces annual-mean land temperature and precipitation pattern biases by up to 50% in coarse prognostic simulations and further enhances weather forecast skill.

Details

show
hide
Language(s): eng - English
 Dates: 2023
 Publication Status: Finally published
 Pages: -
 Publishing info: -
 Table of Contents: -
 Rev. Type: -
 Identifiers: DOI: 10.57757/IUGG23-2145
 Degree: -

Event

show
hide
Title: XXVIII General Assembly of the International Union of Geodesy and Geophysics (IUGG)
Place of Event: Berlin
Start-/End Date: 2023-07-11 - 2023-07-20

Legal Case

show

Project information

show

Source 1

show
hide
Title: XXVIII General Assembly of the International Union of Geodesy and Geophysics (IUGG)
Source Genre: Proceedings
 Creator(s):
Affiliations:
Publ. Info: Potsdam : GFZ German Research Centre for Geosciences
Pages: - Volume / Issue: - Sequence Number: - Start / End Page: - Identifier: -