Temporal- and spatial-scale dependence of three CMIP3 climate models in simulating the surface temperature trend in the twentieth century

Koichi Sakaguchi, Xubin Zeng, Michael A. Brunke

Research output: Contribution to journalArticle

11 Citations (Scopus)

Abstract

Motivated by increasing interests in regional- and decadal-scale climate predictions, this study systematically analyzed the spatial- and temporal-scale dependence of the prediction skill of global climate models in surface air temperature (SAT) change in the twentieth century. The linear trends of annual mean SAT over moving time windows (running linear trends) from two observational datasets and simulations by three global climate models [Community Climate System Model, version 3.0 (CCSM3.0), Climate Model, version 2.0 (CM2.0), and Model E-H] that participated in CMIP3 are compared over several temporal (10-, 20-, 30-, 40-, and 50-yr trends) and spatial (5° × 5°, 10° × 10°, 15° × 15°, 20° × 20°, 30° × 30°, 30° latitudinal bands, hemispheric, and global) scales. The distribution of root-mean-square error is improved with increasing spatial and temporal scales, approaching the observational uncertainty range at the largest scales. Linear correlation shows a similar tendency, but the limited observational length does not provide statistical significance over the longer temporal scales. The comparison of RMSE to climatology and a Monte Carlo test using preindustrial control simulations suggest that the multimodel ensemble mean is able to reproduce robust climate signals at 30° zonal mean or larger spatial scales, while correlation requires hemispherical or global mean for the twentieth-century simulations. Persistent lower performance is observed over the northern high latitudes and the North Atlantic southeast of Greenland. Although several caveats exist for the metrics used in this study, the analyses across scales and/or over running time windows can be taken as one of the approaches for climate system model evaluations.

Original languageEnglish (US)
Pages (from-to)2456-2470
Number of pages15
JournalJournal of Climate
Volume25
Issue number7
DOIs
StatePublished - Apr 2012

Fingerprint

twentieth century
climate modeling
surface temperature
global climate
air temperature
simulation
climate signal
climate prediction
climate
climatology
prediction
CMIP
trend
test
evaluation
distribution
comparison

Keywords

  • Climate models
  • Climate prediction
  • Decadal variability
  • General circulation models
  • Model evaluation/performance
  • Trends

ASJC Scopus subject areas

  • Atmospheric Science

Cite this

Temporal- and spatial-scale dependence of three CMIP3 climate models in simulating the surface temperature trend in the twentieth century. / Sakaguchi, Koichi; Zeng, Xubin; Brunke, Michael A.

In: Journal of Climate, Vol. 25, No. 7, 04.2012, p. 2456-2470.

Research output: Contribution to journalArticle

@article{16dc7690a8834f218d6b70ccad2caa98,
title = "Temporal- and spatial-scale dependence of three CMIP3 climate models in simulating the surface temperature trend in the twentieth century",
abstract = "Motivated by increasing interests in regional- and decadal-scale climate predictions, this study systematically analyzed the spatial- and temporal-scale dependence of the prediction skill of global climate models in surface air temperature (SAT) change in the twentieth century. The linear trends of annual mean SAT over moving time windows (running linear trends) from two observational datasets and simulations by three global climate models [Community Climate System Model, version 3.0 (CCSM3.0), Climate Model, version 2.0 (CM2.0), and Model E-H] that participated in CMIP3 are compared over several temporal (10-, 20-, 30-, 40-, and 50-yr trends) and spatial (5° × 5°, 10° × 10°, 15° × 15°, 20° × 20°, 30° × 30°, 30° latitudinal bands, hemispheric, and global) scales. The distribution of root-mean-square error is improved with increasing spatial and temporal scales, approaching the observational uncertainty range at the largest scales. Linear correlation shows a similar tendency, but the limited observational length does not provide statistical significance over the longer temporal scales. The comparison of RMSE to climatology and a Monte Carlo test using preindustrial control simulations suggest that the multimodel ensemble mean is able to reproduce robust climate signals at 30° zonal mean or larger spatial scales, while correlation requires hemispherical or global mean for the twentieth-century simulations. Persistent lower performance is observed over the northern high latitudes and the North Atlantic southeast of Greenland. Although several caveats exist for the metrics used in this study, the analyses across scales and/or over running time windows can be taken as one of the approaches for climate system model evaluations.",
keywords = "Climate models, Climate prediction, Decadal variability, General circulation models, Model evaluation/performance, Trends",
author = "Koichi Sakaguchi and Xubin Zeng and Brunke, {Michael A.}",
year = "2012",
month = "4",
doi = "10.1175/JCLI-D-11-00106.1",
language = "English (US)",
volume = "25",
pages = "2456--2470",
journal = "Journal of Climate",
issn = "0894-8755",
publisher = "American Meteorological Society",
number = "7",

}

TY - JOUR

T1 - Temporal- and spatial-scale dependence of three CMIP3 climate models in simulating the surface temperature trend in the twentieth century

AU - Sakaguchi, Koichi

AU - Zeng, Xubin

AU - Brunke, Michael A.

PY - 2012/4

Y1 - 2012/4

N2 - Motivated by increasing interests in regional- and decadal-scale climate predictions, this study systematically analyzed the spatial- and temporal-scale dependence of the prediction skill of global climate models in surface air temperature (SAT) change in the twentieth century. The linear trends of annual mean SAT over moving time windows (running linear trends) from two observational datasets and simulations by three global climate models [Community Climate System Model, version 3.0 (CCSM3.0), Climate Model, version 2.0 (CM2.0), and Model E-H] that participated in CMIP3 are compared over several temporal (10-, 20-, 30-, 40-, and 50-yr trends) and spatial (5° × 5°, 10° × 10°, 15° × 15°, 20° × 20°, 30° × 30°, 30° latitudinal bands, hemispheric, and global) scales. The distribution of root-mean-square error is improved with increasing spatial and temporal scales, approaching the observational uncertainty range at the largest scales. Linear correlation shows a similar tendency, but the limited observational length does not provide statistical significance over the longer temporal scales. The comparison of RMSE to climatology and a Monte Carlo test using preindustrial control simulations suggest that the multimodel ensemble mean is able to reproduce robust climate signals at 30° zonal mean or larger spatial scales, while correlation requires hemispherical or global mean for the twentieth-century simulations. Persistent lower performance is observed over the northern high latitudes and the North Atlantic southeast of Greenland. Although several caveats exist for the metrics used in this study, the analyses across scales and/or over running time windows can be taken as one of the approaches for climate system model evaluations.

AB - Motivated by increasing interests in regional- and decadal-scale climate predictions, this study systematically analyzed the spatial- and temporal-scale dependence of the prediction skill of global climate models in surface air temperature (SAT) change in the twentieth century. The linear trends of annual mean SAT over moving time windows (running linear trends) from two observational datasets and simulations by three global climate models [Community Climate System Model, version 3.0 (CCSM3.0), Climate Model, version 2.0 (CM2.0), and Model E-H] that participated in CMIP3 are compared over several temporal (10-, 20-, 30-, 40-, and 50-yr trends) and spatial (5° × 5°, 10° × 10°, 15° × 15°, 20° × 20°, 30° × 30°, 30° latitudinal bands, hemispheric, and global) scales. The distribution of root-mean-square error is improved with increasing spatial and temporal scales, approaching the observational uncertainty range at the largest scales. Linear correlation shows a similar tendency, but the limited observational length does not provide statistical significance over the longer temporal scales. The comparison of RMSE to climatology and a Monte Carlo test using preindustrial control simulations suggest that the multimodel ensemble mean is able to reproduce robust climate signals at 30° zonal mean or larger spatial scales, while correlation requires hemispherical or global mean for the twentieth-century simulations. Persistent lower performance is observed over the northern high latitudes and the North Atlantic southeast of Greenland. Although several caveats exist for the metrics used in this study, the analyses across scales and/or over running time windows can be taken as one of the approaches for climate system model evaluations.

KW - Climate models

KW - Climate prediction

KW - Decadal variability

KW - General circulation models

KW - Model evaluation/performance

KW - Trends

UR - http://www.scopus.com/inward/record.url?scp=84859308892&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84859308892&partnerID=8YFLogxK

U2 - 10.1175/JCLI-D-11-00106.1

DO - 10.1175/JCLI-D-11-00106.1

M3 - Article

AN - SCOPUS:84859308892

VL - 25

SP - 2456

EP - 2470

JO - Journal of Climate

JF - Journal of Climate

SN - 0894-8755

IS - 7

ER -