Process-level improvements in CMIP5 models and their impact on tropical variability, the Southern Ocean, and monsoons

• Published in: Earth System Dynamics
• Main Authors: A. Lauer, C. Jones, V. Eyring, M. Evaldsson, S. Hagemann, J. Mäkelä, G. Martin, R. Roehrig, S. Wang
• Format: Article in Journal/Newspaper
• Language: English
• Published: Copernicus Publications 2018
• Subjects: Science; Q; Geology; QE1-996.5; Dynamic and structural geology; QE500-639.5; Southern Ocean
• Online Access: https://doi.org/10.5194/esd-9-33-2018; https://doaj.org/article/2cf17fbb0b0e401a9850b3a6d2b4c0d4

The performance of updated versions of the four earth system models (ESMs) CNRM, EC-Earth, HadGEM, and MPI-ESM is assessed in comparison to their predecessor versions used in Phase 5 of the Coupled Model Intercomparison Project. The Earth System Model Evaluation Tool (ESMValTool) is applied to evaluate selected climate phenomena in the models against observations. This is the first systematic application of the ESMValTool to assess and document the progress made during an extensive model development and improvement project. This study focuses on the South Asian monsoon (SAM) and the West African monsoon (WAM), the coupled equatorial climate, and Southern Ocean clouds and radiation, which are known to exhibit systematic biases in present-day ESMs. The analysis shows that the tropical precipitation in three out of four models is clearly improved. Two of three updated coupled models show an improved representation of tropical sea surface temperatures with one coupled model not exhibiting a double Intertropical Convergence Zone (ITCZ). Simulated cloud amounts and cloud–radiation interactions are improved over the Southern Ocean. Improvements are also seen in the simulation of the SAM and WAM, although systematic biases remain in regional details and the timing of monsoon rainfall. Analysis of simulations with EC-Earth at different horizontal resolutions from T159 up to T1279 shows that the synoptic-scale variability in precipitation over the SAM and WAM regions improves with higher model resolution. The results suggest that the reasonably good agreement of modeled and observed mean WAM and SAM rainfall in lower-resolution models may be a result of unrealistic intensity distributions.