Twentieth century temperature trends in CMIP3, CMIP5, and CESM-LE climate simulations – spatial-temporal uncertainties, differences and their potential sources

Abstract

The twentieth century climate simulations from the Coupled Model Intercomparison Project Phase 3 (CMIP3) and Phase 5 (CMIP5) are compared to assess the models' ability to capture observed near-surface air temperature trends at global, continental, and regional scales. We computed trends by using a nonparametric method and considering long-term persistence in the time series. The role of internal variability is examined by using large ensemble climate simulations from the National Center for Atmospheric Research model Community Earth System Model (CESM). We computed temperature trends for three periods: (1) the twentieth century, (2) the second half of the twentieth century, and (3) the recent hiatus period to contrast the roles of external forcing and internal variability at various spatial and temporal scales. Both CMIP ensembles show statistically significant warming at global and continental scales during the twentieth century. We found a small but statistically significant difference between CMIP3 (0.57?±?0.07?°C/century) and CMIP5 (0.47?±?0.06?°C/century) twentieth century temperature trends, with the CMIP3 estimate being closer to the observations. The spatial structure of long-term temperature trends, and top-of-the atmosphere net radiation trends, suggests that differences in model parameterizations and feedback processes that lead to a smaller net radiative forcing are likely contributing to the differences between CMIP3 and CMIP5. The estimate of internal variability based on the CESM large ensemble spans 24% of the uncertainty in CMIP5 for the twentieth century temperature trends, and 76% for the recent hiatus period, both at global scales, and 43% and almost 100% during the corresponding time periods at regional scales.

More information

Identifiers

ORCID for Justin Sheffield: orcid.org/0000-0003-2400-0630

Catalogue record

Export record

Altmetrics