Search

This paper provides a systematic evaluation of the ability of 12 Earth System Models (ESMs) participating in the Coupled Model Intercomparison Project Phase 6 (CMIP6) to simulate the spatial inhomogeneity of the atmospheric carbon dioxide (CO2) concentration. The multi-model ensemble mean (MME) can reasonably simulate the increasing trend of CO2concentration from 1850 to 2014, compared with the observation data from the Scripps CO2Program and CMIP6 prescribed data, and improves upon the CMIP5 MME CO2concentration (which is overestimated after 1950). The growth rate of CO2concentration in the northern hemisphere (NH) is higher than that in the southern hemisphere (SH), with the highest growth rate in the mid-latitudes of the NH. The MME can also reasonably simulate the seasonal amplitude of CO2concentration, which is larger in the NH than in the SH and grows in amplitude after the 1950s (especially in the NH). Although the results of the MME are reasonable, there is a large spread among ESMs, and the difference between the ESMs increases with time. The MME results show that regions with relatively large CO2concentrations (such as northern Russia, eastern China, Southeast Asia, the eastern United States, northern South America, and southern Africa) have greater seasonal variability and also exhibit a larger inter-model spread. Compared with CMIP5, the CMIP6 MME simulates an average spatial distribution of CO2concentration that is much closer to the site observations, but the CMIP6-inter-model spread is larger. The inter-model differences of the annual means and seasonal cycles of atmospheric CO2concentration are both attributed to the differences in natural sources and sinks of CO2between the simulations.

As one of the participants in the Subseasonal to Seasonal (S2S) Prediction Project, the China Meteorological Administration (CMA) has adopted several model versions to participate in the S2S Project. This study evaluates the models’ capability to simulate and predict the Madden-Julian Oscillation (MJO). Three versions of the Beijing Climate Center Climate System Model (BCC-CSM) are used to conduct historical simulations and re-forecast experiments (referred to as EXP1, EXP1-M, and EXP2, respectively). In simulating MJO characteristics, the newly-developed high-resolution BCC-CSM outperforms its predecessors. In terms of MJO prediction, the useful prediction skill of the MJO index is enhanced from 15 days in EXP1 to 22 days in EXP1-M, and further to 24 days in EXP2. Within the first forecast week, the better initial condition in EXP2 largely contributes to the enhancement of MJO prediction skill. However, during forecast weeks 2–3, EXP2 shows little advantage compared with EXP1-M because the increased skill at MJO initial phases 6–7 is largely offset by the degraded skill at MJO initial phases 2–3. Particularly at initial phases 2–3, EXP1-M skillfully captures the wind field and Kelvin-wave response to MJO convection, leading to the highest prediction skill of the MJO. Our results reveal that, during the participation of the CMA models in the S2S Project, both the improved model initialization and updated model physics played positive roles in improving MJO prediction. Future efforts should focus on improving the model physics to better simulate MJO convection over the Maritime Continent and further improve MJO prediction at long lead times.

This paper provides a comprehensive assessment of Asian summer monsoon prediction skill as a function of lead time and its relationship to sea surface temperature prediction using the seasonal hindcasts of the Beijing Climate Center Climate System Model, BCC CSM1.1(m). For the South and Southeast Asian summer monsoon, reasonable skill is found in the model’s forecasting of certain aspects of monsoon climatology and spatiotemporal variability. Nevertheless, deficiencies such as significant forecast errors over the tropical western North Pacific and the eastern equatorial Indian Ocean are also found. In particular, overestimation of the connections of some dynamical monsoon indices with large-scale circulation and precipitation patterns exists in most ensemble mean forecasts, even for short lead-time forecasts. Variations of SST, measured by the first mode over the tropical Pacific and Indian oceans, as well as the spatiotemporal features over the Ni˜no3.4 region, are overall well predicted. However, this does not necessarily translate into successful forecasts of the Asian summer monsoon by the model. Diagnostics of the relationships between monsoon and SST show that difficulties in predicting the South Asian monsoon can be mainly attributed to the limited regional response of monsoon in observations but the extensive and exaggerated response in predictions due partially to the application of ensemble average forecasting methods. In contrast, in spite of a similar deficiency, the Southeast Asian monsoon can still be forecasted reasonably, probably because of its closer relationship with large-scale circulation patterns and El Ni˜no–Southern Oscillation.

Using hindcasts of the Beijing Climate Center Climate System Model, the relationships between interannual variability (IAV) and intraseasonal variability (ISV) of the Asian-western Pacific summer monsoon are diagnosed. Predictions show reasonable skill with respect to some basic characteristics of the ISV and IAV of the western North Pacific summer monsoon (WNPSM) and the Indian summer monsoon (ISM). However, the links between the seasonally averaged ISV (SAISV) and seasonal mean of ISM are overestimated by the model. This deficiency may be partially attributable to the overestimated frequency of long breaks and underestimated frequency of long active spells of ISV in normal ISM years, although the model is capable of capturing the impact of ISV on the seasonal mean by its shift in the probability of phases. Furthermore, the interannual relationships of seasonal mean, SAISV, and seasonally averaged long-wave variability (SALWV; i.e., the part with periods longer than the intraseasonal scale) of the WNPSM and ISM with SST and low-level circulation are examined. The observed seasonal mean, SAISV, and SALWV show similar correlation patterns with SST and atmospheric circulation, but with different details. However, the model presents these correlation distributions with unrealistically small differences among different scales, and it somewhat overestimates the teleconnection between monsoon and tropical central-eastern Pacific SST for the ISM, but underestimates it for the WNPSM, the latter of which is partially related to the too-rapid decrease in the impact of El Niño-Southern Oscillation with forecast time in the model.

A time-lagged ensemble method is used to improve 6–15 day precipitation forecasts from the Beijing Climate Center Atmospheric General Circulation Model, version 2.0.1. The approach averages the deterministic predictions of precipitation from the most recent model run and from earlier runs, all at the same forecast valid time. This lagged average forecast (LAF) method assigns equal weight to each ensemble member and produces a forecast by taking the ensemble mean. Our analyses of the Equitable Threat Score, the Hanssen and Kuipers Score, and the frequency bias indicate that the LAF using five members at time-lagged intervals of 6 h improves 6–15 day forecasts of precipitation frequency above 1 mm d−1and 5 mm d−1in many regions of China, and is more effective than the LAF method with selection of the time-lagged interval of 12 or 24 h between ensemble members. In particular, significant improvements are seen over regions where the frequencies of rainfall days are higher than about 40%–50% in the summer season; these regions include northeastern and central to southern China, and the southeastern Tibetan Plateau.