1. Aerosol-precipitation interaction modelling.

Climate scientists rely on numerical climate models to project temperature changes in the coming decades. The credibility of the projections is argued to arise from the fact that models are able to reproduce well the temperature trend in the past century. However, success in reproducing the past does not necessarily mean credible projection for the future, because the former may be a result of compensating biases in the models, which potentially misrepresent the relationship between changes to the energy balance and aerosols/clouds. In our study, compensating biases were found to exist between rain formation and other processes in multiple models. Specifically, models tend to generate rain too frequently against satellite observations, whereas satellite-based improvements with respect to the rain process yield an unrealistically strong aerosol indirect effect that can substantially cool the surface. This means that dual efforts are required to derive trustworthy projections of future climate changes: on the one hand, the rain-formation process should be represented realistically in accordance with observations, while on the other hand, errors in the interaction between rain and aerosol/cloud should be mitigated.

2. Solar spectrum’s impact on climate.

Not only total solar irradiance (TSI) but also spectral solar irradiance (SSI) matter for our climate. Different surfaces can have different reflectivity for the visible (VIS) and near-infrared (NIR). The recent NASA Total and Spectral Solar Irradiance Sensor (TSIS-1) mission has provided more accurate SSI observations than before. The TSI observed by TSIS-1 differs from the counterpart used by climate models by no more than 1 Wm⁻². However, the SSI difference in a given VIS and NIR band can be as large as 4 Wm⁻² with opposite signs. Due to different VIS-NIR spectral reflectance contrasts between icy (or snowy) surfaces and open water, climate model simulations show that assigning more irradiance in the VIS, and less in the NIR, leads to less solar absorption by the high-latitude surfaces, ending up with colder polar surface temperature and larger sea ice coverage. The difference is more prominent over the Antarctic than over the Arctic. Our results suggest that, even for the identical TSI, the surface albedo feedback can be triggered by different SSI partition between the VIS and NIR. The results underscore the importance of continuously monitoring SSI and the use of correct SSI in climate simulations.

3. Cloud macro-structure parameterization.

Climate models usually have grid sizes much larger than a single cloud body, leaving cloud shape unresolved and having to be parameterized. The latter matters for both precipitation and radiation calculations, both of which depend on how clouds overlap verticallyand spread horizontally. It is natural to think that cloud shape is connected to the weather status in which the cloud is embedded. However, this connection is largely ignored in climate models. Our study linked cloud overlap to the strength of atmospheric convection in the tropics based on simulations of a global cloud-resolving model (NICAM), with a linear regression approach. This simple and dynamic-related representation of cloud overlap leads to a significant improvement in simulation of both cloud cover and radiation fields compared with traditional overlap treatments. This study provided a physically justifiable approach to parameterize cloud overlap in the tropics in GCMs.