Research brief: Accurate datasets and model weighting improve climate projections

Future North American summer maximum temperatures were studied using models from the Coupled Model Intercomparison Project phase 5 (CMIP5). The researchers tested the sensitivity of temperature projections to different observational datasets; diagnostics; ways to determine the relevant diagnostics; and the choice of spatial domain. 

When computing the multi-model average, more weight was given to the CMIP5 models that were more skillful and more closely simulated a specific process, which is potentially relevant for North American maximum temperatures. This method therefore assumed that models performing well in historical simulations also had skill in projecting future changes. 

In addition, the researchers accounted for the fact that models were not truly independent due to some shared model development history. Both performance and independence weights were assigned to the models before calculating the weighted average, following a recently proposed approach by Knutti et al. (2017).

The researchers identified a trend in shortwave radiation, mean precipitation, sea surface temperature variability and variability and a trend in maximum temperature itself to be the most promising diagnostics for constraining projected summer maximum temperature over North America. 

Compared to equally-weighted multi-model mean, the weighted mean reduced the change in summer daily temperature maxima slightly (0.05–0.45°C) over central North America. The choice of the observational dataset used to assess skill of the individual models was found to be important given that the observational and reanalysis products used for summer maximum temperatures disagree. 

The researchers concluded It was therefore essential to have reliable observations for key variables to be able to confidently constrain future projections. The more diagnostics that were included as predictors, the more similar the results across observational products became. More than one predicting diagnostic is recommended to inform the weighting.

  • Paper: Lorenz R., N. Herger, J. Sedlacek, V. Eyring, E.M. Fischer, and R. Knutti (2018), Prospects and caveats of weighting climate models for summer maximum temperature projections over North America, Journal of Geophysical Research, 123, doi:10.1029/2017JD027992.
  • Picture (top left): Beach sky, Sydney by Igor Kasalovic (Unsplash).
UNSW logo ANU logo Monash logo UMelb logo UTAS logo