An “Observational Large Ensemble” to Compare Observed and Modeled Temperature Trend Uncertainty due to Internal Variability

Abstract

Estimates of the climate response to anthropogenic forcing contain irreducible uncertainty due to the presence of internal variability. Accurate quantification of this uncertainty is critical for both contextualizing historical trends and determining the spread of climate projections. The contribution of internal variability to uncertainty in trends can be estimated in models as the spread across an initial condition ensemble. However, internal variability simulated by a model may be inconsistent with observations due to model biases. Here, statistical resampling methods are applied to observations in order to quantify uncertainty in historical 50-yr (1966–2015) winter near-surface air temperature trends over North America related to incomplete sampling of internal variability. This estimate is compared with the simulated trend uncertainty in the NCAR CESM1 Large Ensemble (LENS). The comparison suggests that uncertainty in trends due to internal variability is largely overestimated in LENS, which has an average amplification of variability of 32% across North America. The amplification of variability is greatest in the western United States and Alaska. The observationally derived estimate of trend uncertainty is combined with the forced signal from LENS to produce an “Observational Large Ensemble” (OLENS). The members of OLENS indicate the range of observationally constrained, spatially consistent temperature trends that could have been observed over the past 50 years if a different sequence of internal variability had unfolded. The smaller trend uncertainty in OLENS suggests that is easier to detect the historical climate change signal in observations than in any given member of LENS.

Can an ensemble climate simulation be used to separate climate change signals from internal unforced variability?

“…This is apparent by comparing the ensemble with HadCRUT4 and JRA55 which show that practically all members of the ensemble have a higher variance. A similar result has been found for the NCAR Large Ensemble over North America (McKinnon et al 2017). The disagreement between observational data and model results can be found both globally as well as for selected regions such as the European region.…”

“…However, as with single model or multi-model studies it is important to compare results from ensemble studies with observations, where possible, to evaluate the realism of the estimates of internal variability and trends, at the same time being aware of observational uncertainty. For example, McKinnon et al (2017) found the ensemble used by Kay et al (2015) overestimated the uncertainty in trends due to internal variability for surface temperature over North America using resampling methods applied to observations to quantify uncertainty. Thompson et al (2015) point out the difficulties in trying to estimate the internal variability of the climate from models and observations.…”

The Max Planck Institute Grand Ensemble: Enabling the Exploration of Climate System Variability

“…This is a novel model evaluation method, where we transfer the methodology used on European temperature (Suarez‐Gutierrez et al, , Supplementary Figure 3) to the globe. While previous methods to investigate this have used standard deviations and often detrended quantities (Bengtsson & Hodges, ; Lehner et al, ; McKinnon et al, ), this new method allows quantification of whether the whole distribution, including the extremes, agree well with observations. Additionally, by combining the map in Figure with the method from Figure , we can investigate exactly why the model does not agree with observations in specific regions identified on the map and can differentiate between discrepancies in internal variability and the forced response.…”

“…Studies that utilize large ensembles have been extensively used to investigate the internal variability of the climate system (e.g., Dai & Bloecker, ; Fasullo & Nerem, ; Frankignoul et al, ; Smith & Jahn, ) and extreme events (e.g., Diffenbaugh et al, ; Gibson et al, ; Kirchmeier‐Young et al, ; Tebaldi & Wehner, ; Wang et al, ). They have also been used as a test bed for new methodologies such as creating an observational large ensemble (McKinnon et al, ; McKinnon & Deser, ), built by combining the forced response from CESM‐LE and the estimated internal variability from observations. These ensembles have also been used as a test bed for dynamical adjustment, which can be used to remove the internal dynamical signal and consequently bring observations closer to the forced response (Deser et al, ; Lehner et al, ).…”

The salience of nonlinearities in the boreal winter response to ENSO: North Pacific and North America

“…11, and the observed variance is indicated with a diamond. The realism of model variance has also been assessed by following the procedure used in figure 1 of McKinnon et al (2017), and results are presented in Supplemental Figures 7-9. All conclusions drawn below as to regions with biased model variance are consistent with the conclusions that can be drawn from Supplemental Figures 7-9.…”