Climate models predict increases in the temperature of hot extremes throughout the 21st century, but there is large uncertainty regarding the amount of warming between different models. This uncertainty makes it difficult to plan for adaptation to hotter conditions in the future.

In this study, Centre of Excellence researchers explored what controls this model uncertainty, and based on this understanding they proposed a way to reduce the uncertainty about future hot extremes.

The researchers found that the future uncertainty in hot extremes is controlled by two factors, both related to amplification of hot extremes through land-atmosphere interactions:

Firstly, future changes in seasonal precipitation prior to heat events, which affect the soil moisture in summer, determined the amount that hot extremes increased. Models with a stronger decrease in seasonal precipitation prior to heat events simulated stronger increases in hot extremes.

Secondly, the strength of the relationship between soil moisture and the temperature response found in climate models also affected the future simulated warming of hot extremes. Models with a stronger “soil moisture/temperature” relationship under current climate conditions simulated stronger future increases in hot extremes. 

The researchers then evaluated the moisture-temperature relationship in the models against observations, and found that several models tended to overestimate the strength of this relationship. Excluding those models with unrealistically strong moisture-temperature relationships led to a reduced probability of the strongest increases in hot extremes, and reduced the overall uncertainty of future simulated changes.

This study illustrates how a process-based evaluation of climate models can help to obtain more robust information about future climate changes.

  • Paper: Donat, M. G., Pitman, A. J., & Angélil, O. (2018). Understanding and reducing future uncertainty in midlatitude daily heat extremes via land surface feedback constraints. Geophysical Research Letters, 45, 10,627–10,636.