This Title All WIREs
How to cite this WIREs title:
WIREs Clim Change
Impact Factor: 7.385

Use of models in detection and attribution of climate change

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract Most detection and attribution studies use climate models to determine both the expected ‘fingerprint’ of climate change and the uncertainty in the estimated magnitude of this fingerprint in observations, given the climate variability. This review discusses the role of models in detection and attribution, the associated uncertainties, and the robustness of results. Studies that use observations only make substantial assumptions to separate the components of observed changes due to radiative forcing from those due to internal climate variability. Results from observation‐only studies are broadly consistent with those from fingerprint studies. Fingerprint studies evaluate the extent to which patterns of response to external forcing (fingerprints) from climate model simulations explain observed climate change in observations. Fingerprints are based on climate models of various complexities, from energy balance models to full earth system models. Statistical approaches range from simple comparisons of observations with model simulations to multi‐regression methods that estimate the contribution of several forcings to observed change using a noise‐reducing metric. Multi‐model methods can address model uncertainties to some extent and we discuss how remaining uncertainties can be overcome. The increasing focus on detecting and attributing regional climate change and impacts presents both opportunities and challenges. Challenges arise because internal variability is larger on smaller scales, and regionally important forcings, such as from aerosols or land‐use change, are often uncertain. Nevertheless, if regional climate change can be linked to external forcing, the results can be used to provide constraints on regional climate projections. WIREs Clim Change 2011 2 570–591 DOI: 10.1002/wcc.121 This article is categorized under: Climate Models and Modeling > Knowledge Generation with Models

Results of studies filtering externally driven signals based on observational data only. Top panel: a time series of global mean monthly surface temperature (topmost) is shown compared to contributions by variability from El Nino, short‐term dynamical variations in extratropics (Tdyn), and the residual that remains after removing both (each time series is offset for presentation purposes; Reprinted with permission from Ref 18. Copyright 2009 American Meteorological Society). The vertical lines indicate August 1945 and the timing of volcanic eruptions. Bottom: First discriminants of interdecadal variations in (a) January and (c) July, based on separating climate variability between long and short timescales. Changes are expressed relative to the 1916–1998 mean. Upper panels (a) and (c): discriminant pattern. Lower panels (b) and (d): canonical variates, which give the time evolution. (Reprinted with permission from Ref 20. Copyright 2001 American Meteorological Society)

[ Normal View | Magnified View ]

(a) Comparison of precipitation in zonal latitude bands between a multi‐model mean (blue time series, trend shown by red dashes) and observations (black, trend shown by black dashes), overlay colors indicate bands where both model and observations show increases (green), decreases (yellow), or neutral/disagreeing sign of change (gray), and (b) zonal pattern of variance ratio between climate models and observations. The figure shows box and whisker plots of the ratio of 5‐year 10° zonal mean precipitation variances between all‐forcing simulations and that estimated from station observations. The upper and lower ends of each box are drawn at the 75th and 25th quartiles, and the bar through each box is drawn at the median. The two bars indicate the range that would cover approximately 90% of variance ratios if the upper or lower halves of the variance ratio distribution were roughly Gaussian in shape. Individual points beyond the horizontal bars indicate outliers. (Reprinted with permission from Ref 70. Copyright 2007 Nature Publishing Group)

[ Normal View | Magnified View ]

Comparison of variability as a function of timescale for continental mean surface air temperature over the 20th century, comparing instrumental data (black) and 20th century all‐forcings simulations (colors) from 14 models. The 5–95% uncertainty ranges are given by bars. Ref 2 gives details on estimation procedure. (Reprinted with permission from Ref 2. Copyright 2007 Cambridge University Press)

[ Normal View | Magnified View ]

(a) Energy balance model simulations of the response to greenhouse gas increases, moderated by aerosols in the 20th century (red), solar forcing (green), and volcanic forcing (blue) (Reprinted with permission from Ref 58. Copyright 2003). (b) Results obtained using these simulations as fingerprints for the effect of all forcings combined in a palaeoclimatic reconstruction (black, fitted to best match the reconstruction with gray shading indicating the uncertainty in the scaling factor) compared to a fit of a coupled model (grey). The lower half of the panel shows the estimated contribution by each individual forcing time series scaled to match the reconstruction in a multiple regression, with shading again indicating the uncertainty in the scaling factor and hence in the estimated contribution by individual forcings. The uncertainty in the solar signal (green) is not shown as the effect of that forcing could not be distinguished from noise ; note that similar results are obtained using a number of other reconstructions, see paper). (Reprinted with permission from Ref 12. Copyright 2007 American Meteorological Society)

[ Normal View | Magnified View ]

Comparison between global mean temperature changes relative to the 1901–1950 average (°C) from observations (black) and simulated by climate model simulations that include (a) both human and natural influences on climate (for example, the effect of strong volcanic eruptions, marked by vertical gray bars) and (b) natural influences only. Individual model simulations are shown by thin lines, their average by a fat line (red in panel (a), blue in panel (b)). (Reprinted with permission from Ref 2. Copyright 2007 Cambridge University Press)

[ Normal View | Magnified View ]

Schematic for detection and attribution. The observed change (shown here: pattern of temperature change over the 20th century, left) is composed of a linear combination of fingerprints for all forcings combined (top, right) and for natural forcings only (center right, this combination allows rescaling of natural vs anthropogenic fingerprints in simulations of the 20th century) plus residual, unexplained variability. The resulting scaling factors and warming per fingerprint can be used to derive contributions to warming such as shown in the bottom panel, labeled panel (c), although in this instance the latter is derived from three fingerprints. It shows attributable warming estimated from a detection and attribution analysis for the 20th century, using a fingerprint of the spatial pattern and time evolution of climate change forced by greenhouse gases (red), other anthropogenic forcing (green), and solar and volcanic forcings combined (blue). The best estimate contribution of each forcing to warming in the 50‐year period 1950–1999 is given by the vertical bar and the 5–95% uncertainty in that estimate is given by the black whiskers. The observed trend over that period is shown by a black horizontal line. The different estimates are derived using fingerprints from different models. (Reprinted with permission from Ref 2. Copyright 2007 Cambridge University Press)

[ Normal View | Magnified View ]

Browse by Topic

Climate Models and Modeling > Knowledge Generation with Models

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts