Estimating transient climate response using consistent temperature reconstruction methods in models and observations

Abstract:

Observational temperature records such as HadCRUT4 typically have incomplete geographical coverage and blend air temperature over land with sea surface temperatures over ocean, in contrast to model output which is commonly reported as global air temperature. This complicates estimation of properties such as the transient climate response (TCR). Observation-based estimates of TCR have been made using energy-budget constraints applied to time series of historical radiative forcing and surface temperature changes, while model TCR is formally derived from simulations where CO2 increases at 1% per year. We perform a like-with-like comparison using three published energy-budget methods to derive modelled TCR from historical CMIP5 temperature series sampled in a manner consistent with HadCRUT4. Observation-based TCR estimates agree to within 0.12 K of the multi-model mean in each case and for 2 of the 3 energy-budget methods the observation-based TCR is higher than the multi-model mean. For one energy-budget method, using the HadCRUT4 blending method leads to a TCR underestimate of 0.3±0.1 K, relative to that estimated using global near-surface air temperatures.