Understanding cloud impacts on Amazonia carbon cycle via radiation and precipitation using convection-parameterized and resolved GCMs

The coupling of the global water, energy and carbon cycles has important impacts on multi-scale climate and better understanding of the mechanisms is key to future climate projections. In this study, we use the comprehensive Community Earth System Model (CESM) and its convection-permitting version (the "single-moment super-parameterization" model, referred to as SP) separately to quantify cloud impacts on carbon cycles via radiation and precipitation feedbacks in the Amazon region. The two experiments are both run under the present-day climate for 10 years with the same land module Community Land Model (CLM5.0) and are mainly different in the representation of convection related processes. We first decompose the Gross Primary Production (GPP) response to atmospheric conditions into the contribution of Light Use Efficiency (LUE) and that of Photosynthetically Active Radiation (PAR). The contribution from LUE variation is roughly three times of the contribution from PAR in both experiments. We also find the SP model is more sensitive to LUE when compared with CESM. Moreover, we decompose the change of LUE by two driving factors: soil moisture and the ratio of diffuse visible incident solar radiation to direct visible incident solar radiation (R). LUE is revealed to be mainly driven by the ratio R in both experiments while the constrain of soil moisture to LUE plays a more important role in SP than CESM. These findings from decompositions are correspondent with the fewer clouds and less precipitation in SP. In SP, lower cloud amount results in the lower diffusive visible radiation that can be efficiently used by plants (along with the lower ratio R) though it has higher total surface incoming radiation than CESM; and less precipitation leads to larger water stress, especially in the northeast Amazon, though generally the adequate rainfall in the rainforest would not limit the plant productivity to a large extent, further causing LUE and GPP in SP to decrease correspondingly. More detailed and quantitative results will be shown at the conference.