Assessing teleconnections patterns in climate models using a combination of point correlation maps and self-organizing maps
Assessing teleconnections patterns in climate models using a combination of point correlation maps and self-organizing maps
A new method to identify and evaluate teleconnection patterns in gridded climate data is presented. A large set of point correlation maps (one for each grid point) is used to train a self-organizing map (SOM). This combines the teleconnection identification properties of point correlation maps with the ability of SOMs to group similar patterns together on a topological grid and provides a frequency of occurrence for each pattern. Once the SOM is trained it is used as a reference for comparison to other sets of correlation maps. A SOM trained using point correlation maps calculated from NCEP/NCAR sea level pressure reanalysis for the period 01.1948-12.2005 is presented and the patterns found compared to point correlation maps from several climate models. By matching each NCEP/NCAR correlation map and each model correlation map with their most similar pattern on the SOM, discrepancies between the datasets are revealed, such as differences in the frequency of occurrence or shifts in the spatial structure of teleconnections. The base points corresponding to the correlation maps for each teleconnection show the regions important for their existence. Differences in the base point locations between NCEP/NCAR and the models provide insight into the physical biases underlying the model deviations from reality. Prominent patterns are identified by the SOM, such as the NAO, ENSO and the PNA, however the flexibility of the SOM allows these patterns to be viewed as a continuum of patterns, each identifiable as a variation within a defined teleconnection. As the SOM is a non-linear method, asymmetries between patterns generated from opposite centres of action are revealed. Clustering the SOM patterns identifies the regions of the SOM corresponding to each teleconnection type by classifying similar patterns together, which retains the continuum of patterns, but allows general characterization of the teleconnections present in the data.
Hunt, F.K.
e3cb0020-9efe-4c78-8681-72cd4a5c726f
Hirschi, Joel
c8a45006-a6e3-4319-b5f5-648e8ef98906
Oliver, Kevin
588b11c6-4d0c-4c59-94e2-255688474987
Wells, Neil
4c27167c-f972-4822-9614-d6ca8d8223b5
16 April 2013
Hunt, F.K.
e3cb0020-9efe-4c78-8681-72cd4a5c726f
Hirschi, Joel
c8a45006-a6e3-4319-b5f5-648e8ef98906
Oliver, Kevin
588b11c6-4d0c-4c59-94e2-255688474987
Wells, Neil
4c27167c-f972-4822-9614-d6ca8d8223b5
Hunt, F.K., Hirschi, Joel, Oliver, Kevin and Wells, Neil
(2013)
Assessing teleconnections patterns in climate models using a combination of point correlation maps and self-organizing maps.
4th WGNE Workshop on Systematic Errors in Weather and Climate Models, Exeter, United Kingdom.
Record type:
Conference or Workshop Item
(Other)
Abstract
A new method to identify and evaluate teleconnection patterns in gridded climate data is presented. A large set of point correlation maps (one for each grid point) is used to train a self-organizing map (SOM). This combines the teleconnection identification properties of point correlation maps with the ability of SOMs to group similar patterns together on a topological grid and provides a frequency of occurrence for each pattern. Once the SOM is trained it is used as a reference for comparison to other sets of correlation maps. A SOM trained using point correlation maps calculated from NCEP/NCAR sea level pressure reanalysis for the period 01.1948-12.2005 is presented and the patterns found compared to point correlation maps from several climate models. By matching each NCEP/NCAR correlation map and each model correlation map with their most similar pattern on the SOM, discrepancies between the datasets are revealed, such as differences in the frequency of occurrence or shifts in the spatial structure of teleconnections. The base points corresponding to the correlation maps for each teleconnection show the regions important for their existence. Differences in the base point locations between NCEP/NCAR and the models provide insight into the physical biases underlying the model deviations from reality. Prominent patterns are identified by the SOM, such as the NAO, ENSO and the PNA, however the flexibility of the SOM allows these patterns to be viewed as a continuum of patterns, each identifiable as a variation within a defined teleconnection. As the SOM is a non-linear method, asymmetries between patterns generated from opposite centres of action are revealed. Clustering the SOM patterns identifies the regions of the SOM corresponding to each teleconnection type by classifying similar patterns together, which retains the continuum of patterns, but allows general characterization of the teleconnections present in the data.
Slideshow
WGNE_presentation_3.pptx
- Other
More information
Published date: 16 April 2013
Venue - Dates:
4th WGNE Workshop on Systematic Errors in Weather and Climate Models, Exeter, United Kingdom, 2013-04-16
Organisations:
Marine Systems Modelling, Physical Oceanography
Identifiers
Local EPrints ID: 354931
URI: http://eprints.soton.ac.uk/id/eprint/354931
PURE UUID: e3b3568b-e923-4215-93c6-98b09344bcf8
Catalogue record
Date deposited: 25 Jul 2013 15:47
Last modified: 14 Mar 2024 14:26
Export record
Contributors
Author:
F.K. Hunt
Author:
Joel Hirschi
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics