A new method for assessing the performance of general circulation models based on their ability to simulate the response to observed forcing

View More View Less
  • 1 Centro de Ciencias de la Atmósfera, Universidad Nacional Autónoma de México, Circuito Exterior, 04510, CDMX, México.
  • 2 Institute for Environmental Studies, Vrije Universiteit, De Boelelaan 1087, 1081 HV Amsterdam, The Netherlands.
  • 3 Programa de Investigación en Cambio Climático, Universidad Nacional Autónoma de México, Circuito Exterior, 04510, CDMX, México.
© Get Permissions
Restricted access

Abstract

The reliability of General Circulation Models (GCMs) is commonly associated with their ability to reproduce relevant aspects of observed climate and thus, the evaluation of GCMs performance has become a standard practice for climate change studies. As such, there is an ever-growing literature that focuses on developing and evaluating metrics to assess GCMs performance. In this paper it is shown that some commonly applied metrics provide little information for discriminating GCMs based on their performance, once uncertainty is included. A new methodology is proposed that differs from common approaches in that it focuses on evaluating GCMs ability to reproduce the observed response of surface temperature to changes in external radiative forcing (RF), while controlling for observed and simulated variability. It uses formal statistical tests to evaluate two aspects of the warming trend that are central for climate change studies: 1) if the response to RF produced by a particular GCM is compatible with observations and 2) if the magnitudes of the observed and simulated rates of warming are statistically similar. We illustrate the proposed methodology by evaluating the ability of 21 GCMs to reproduce the observed warming trend at the global scale and eight sub-continental land domains. Results show that most of the GCMs provide an adequate representation of the observed warming trend for the global scale and for domains located in the southern hemisphere. However, GCMs tend to overestimate the warming rate for domains in the northern hemisphere, particularly since the mid-1990s.

Corresponding author address: Estrada F., Centro de Ciencias de la Atmósfera, Universidad Nacional Autónoma de México, Circuito Exterior, 04510, CDMX, México. E-mail: feporrua@atmosfera.unam.mx

Abstract

The reliability of General Circulation Models (GCMs) is commonly associated with their ability to reproduce relevant aspects of observed climate and thus, the evaluation of GCMs performance has become a standard practice for climate change studies. As such, there is an ever-growing literature that focuses on developing and evaluating metrics to assess GCMs performance. In this paper it is shown that some commonly applied metrics provide little information for discriminating GCMs based on their performance, once uncertainty is included. A new methodology is proposed that differs from common approaches in that it focuses on evaluating GCMs ability to reproduce the observed response of surface temperature to changes in external radiative forcing (RF), while controlling for observed and simulated variability. It uses formal statistical tests to evaluate two aspects of the warming trend that are central for climate change studies: 1) if the response to RF produced by a particular GCM is compatible with observations and 2) if the magnitudes of the observed and simulated rates of warming are statistically similar. We illustrate the proposed methodology by evaluating the ability of 21 GCMs to reproduce the observed warming trend at the global scale and eight sub-continental land domains. Results show that most of the GCMs provide an adequate representation of the observed warming trend for the global scale and for domains located in the southern hemisphere. However, GCMs tend to overestimate the warming rate for domains in the northern hemisphere, particularly since the mid-1990s.

Corresponding author address: Estrada F., Centro de Ciencias de la Atmósfera, Universidad Nacional Autónoma de México, Circuito Exterior, 04510, CDMX, México. E-mail: feporrua@atmosfera.unam.mx
Save