Reconstructing paleoclimate fields using online data assimilation with a linear inverse model
We examine the skill of a new approach to climate field reconstructions (CFRs) using an online paleoclimate data assimilation (PDA) method. Several recent studies have foregone climate model forecasts during assimilation due to the computational expense of running coupled global climate models (CGCM...
Published in: | Climate of the Past |
---|---|
Main Authors: | , |
Format: | Text |
Language: | English |
Published: |
2018
|
Subjects: | |
Online Access: | https://doi.org/10.5194/cp-13-421-2017 https://cp.copernicus.org/articles/13/421/2017/ |
id |
ftcopernicus:oai:publications.copernicus.org:cp56122 |
---|---|
record_format |
openpolar |
spelling |
ftcopernicus:oai:publications.copernicus.org:cp56122 2023-05-15T15:39:10+02:00 Reconstructing paleoclimate fields using online data assimilation with a linear inverse model Perkins, Walter A. Hakim, Gregory J. 2018-09-27 application/pdf https://doi.org/10.5194/cp-13-421-2017 https://cp.copernicus.org/articles/13/421/2017/ eng eng doi:10.5194/cp-13-421-2017 https://cp.copernicus.org/articles/13/421/2017/ eISSN: 1814-9332 Text 2018 ftcopernicus https://doi.org/10.5194/cp-13-421-2017 2020-07-20T16:23:44Z We examine the skill of a new approach to climate field reconstructions (CFRs) using an online paleoclimate data assimilation (PDA) method. Several recent studies have foregone climate model forecasts during assimilation due to the computational expense of running coupled global climate models (CGCMs) and the relatively low skill of these forecasts on longer timescales. Here we greatly diminish the computational cost by employing an empirical forecast model (linear inverse model, LIM), which has been shown to have skill comparable to CGCMs for forecasting annual-to-decadal surface temperature anomalies. We reconstruct annual-average 2 m air temperature over the instrumental period (1850–2000) using proxy records from the PAGES 2k Consortium Phase 1 database; proxy models for estimating proxy observations are calibrated on GISTEMP surface temperature analyses. We compare results for LIMs calibrated using observational (Berkeley Earth), reanalysis (20th Century Reanalysis), and CMIP5 climate model (CCSM4 and MPI) data relative to a control offline reconstruction method. Generally, we find that the usage of LIM forecasts for online PDA increases reconstruction agreement with the instrumental record for both spatial fields and global mean temperature (GMT). Specifically, the coefficient of efficiency (CE) skill metric for detrended GMT increases by an average of 57 % over the offline benchmark. LIM experiments display a common pattern of skill improvement in the spatial fields over Northern Hemisphere land areas and in the high-latitude North Atlantic–Barents Sea corridor. Experiments for non-CGCM-calibrated LIMs reveal region-specific reductions in spatial skill compared to the offline control, likely due to aspects of the LIM calibration process. Overall, the CGCM-calibrated LIMs have the best performance when considering both spatial fields and GMT. A comparison with the persistence forecast experiment suggests that improvements are associated with the linear dynamical constraints of the forecast and not simply persistence of temperature anomalies. Text Barents Sea North Atlantic Copernicus Publications: E-Journals Barents Sea Climate of the Past 13 5 421 436 |
institution |
Open Polar |
collection |
Copernicus Publications: E-Journals |
op_collection_id |
ftcopernicus |
language |
English |
description |
We examine the skill of a new approach to climate field reconstructions (CFRs) using an online paleoclimate data assimilation (PDA) method. Several recent studies have foregone climate model forecasts during assimilation due to the computational expense of running coupled global climate models (CGCMs) and the relatively low skill of these forecasts on longer timescales. Here we greatly diminish the computational cost by employing an empirical forecast model (linear inverse model, LIM), which has been shown to have skill comparable to CGCMs for forecasting annual-to-decadal surface temperature anomalies. We reconstruct annual-average 2 m air temperature over the instrumental period (1850–2000) using proxy records from the PAGES 2k Consortium Phase 1 database; proxy models for estimating proxy observations are calibrated on GISTEMP surface temperature analyses. We compare results for LIMs calibrated using observational (Berkeley Earth), reanalysis (20th Century Reanalysis), and CMIP5 climate model (CCSM4 and MPI) data relative to a control offline reconstruction method. Generally, we find that the usage of LIM forecasts for online PDA increases reconstruction agreement with the instrumental record for both spatial fields and global mean temperature (GMT). Specifically, the coefficient of efficiency (CE) skill metric for detrended GMT increases by an average of 57 % over the offline benchmark. LIM experiments display a common pattern of skill improvement in the spatial fields over Northern Hemisphere land areas and in the high-latitude North Atlantic–Barents Sea corridor. Experiments for non-CGCM-calibrated LIMs reveal region-specific reductions in spatial skill compared to the offline control, likely due to aspects of the LIM calibration process. Overall, the CGCM-calibrated LIMs have the best performance when considering both spatial fields and GMT. A comparison with the persistence forecast experiment suggests that improvements are associated with the linear dynamical constraints of the forecast and not simply persistence of temperature anomalies. |
format |
Text |
author |
Perkins, Walter A. Hakim, Gregory J. |
spellingShingle |
Perkins, Walter A. Hakim, Gregory J. Reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
author_facet |
Perkins, Walter A. Hakim, Gregory J. |
author_sort |
Perkins, Walter A. |
title |
Reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
title_short |
Reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
title_full |
Reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
title_fullStr |
Reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
title_full_unstemmed |
Reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
title_sort |
reconstructing paleoclimate fields using online data assimilation with a linear inverse model |
publishDate |
2018 |
url |
https://doi.org/10.5194/cp-13-421-2017 https://cp.copernicus.org/articles/13/421/2017/ |
geographic |
Barents Sea |
geographic_facet |
Barents Sea |
genre |
Barents Sea North Atlantic |
genre_facet |
Barents Sea North Atlantic |
op_source |
eISSN: 1814-9332 |
op_relation |
doi:10.5194/cp-13-421-2017 https://cp.copernicus.org/articles/13/421/2017/ |
op_doi |
https://doi.org/10.5194/cp-13-421-2017 |
container_title |
Climate of the Past |
container_volume |
13 |
container_issue |
5 |
container_start_page |
421 |
op_container_end_page |
436 |
_version_ |
1766370624501972992 |