Abstract. Earth System Models (ESMs) participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5) showed large uncertainties in simulating atmospheric CO2 concentrations. By comparing the simulations with satellite observations, in this study we find slight improvements in the ESMs participating in the new Phase 6 (CMIP6) compared to CMIP5. We utilize the Earth System Model Evaluation Tool (ESMValTool) to evaluate emission driven CMIP5 and CMIP6 simulations with satellite data of column-average CO2 mole fractions (XCO2). The satellite data are a combined data product covering the period 2003–2014 based on the SCIAMACHY/ENVISAT (2003–2012) and TANSO-FTS/GOSAT (2009–2014) instruments. In this study the Observations for Model Intercomparisons Project (Obs4MIPs) format data product version 3 (O4Mv3) with a spatial resolution of 5° × 5° and monthly time resolution has been used. Comparisons of XCO2 time series show a large spread among the model ensembles both in CMIP5 and CMIP6, with differences in the absolute concentrations of up to approximately 20 ppmv relative to the satellite observations. The multi-model mean has a bias of approximately +10 and +2 ppmv in CMIP5 and CMIP6, respectively. The derived atmospheric XCO2 growth rate (GR) is typically slightly overestimated in the models, with a slightly better average and lower spread for CMIP6. The interannual variability of the growth rate is well reproduced in the multi-model mean. All models capture the expected increase of the seasonal cycle amplitude (SCA) with increasing latitude, but most models underestimate the SCA. Most models from both ensembles show a positive trend of the SCA over the period 2003–2014, i.e. an increase of the SCA with XCO2, similar to in situ ground-based measurements. In contrast, the combined satellite product shows a negative trend over this period. Any SCA derived from sampled data can only be considered an effective SCA, as sampling can remove the peaks or troughs. This negative trend can be reproduced by the models when sampling them as the observations. The average effective SCA in the models is higher when using the SCIAMACHY/ENVISAT instead of the TANSO-FTS/GOSAT mean data coverage mask, overall leading to a negative trend over the full period similar to the combined satellite product. Models with a larger difference in the average effective SCA between the two coverages also show a stronger trend reversal. Therefore, this trend reversal in the satellite data is due to sampling characteristics, more specifically the different data coverage of the two satellites that can be reproduced by the models if sampled the same way. Overall, the CMIP6 ensemble shows better agreement with the satellite data in all considered quantities (XCO2, GR, SCA and trend in SCA), with the biggest improvement in the mean XCO2 content of the atmosphere. This study shows that the availability of column-integral CO2 from satellite provides a promising new way to evaluate the performance of Earth System Models on a global scale, complementing existing studies that are based on in situ measurements from single ground-based stations.
Studying recent hydroclimatic variability in Madagascar despite deficient measurement networks: use of CHIRPS and GRACE data at the scale of the Mahajunga province
