We investigated aerosol composition and sources and the interaction between secondary organic aerosol (SOA) and clouds at a regional mountain site in southeastern China. Clouds efficiently scavenge more oxidized SOA; however, cloud evaporation leads to the production of less oxidized SOA. The unexpectedly high presence of nitrate in aerosol particles indicates that nitrate formed in polluted areas has undergone interactions with clouds, significantly influencing the regional background site.

A global continental merged high-resolution (PBLH) dataset with good accuracy compared to radiosonde is generated via machine learning algorithms, covering the period from 2011 to 2021 with 3-hour and 0.25º resolution in space and time. The machine learning model takes parameters derived from the ERA5 reanalysis and GLDAS product as input, with PBLH biases between radiosonde and ERA5 as the learning targets. The merged PBLH is the sum of the predicted PBLH bias and the PBLH from ERA5.

The performance of the Surface Urban Energy and Water Balance Scheme (SUEWS) is evaluated against the observed surface exchanges (fluxes) of heat and carbon dioxide in a densely built neighborhood in Beijing. The heat flux modeling is noticeably improved by using the observed maximum conductance and by optimizing the vegetation phenology modeling. SUEWS also performs well in simulating carbon dioxide flux.

A CMAQ EnSRF-based regional inversion system was extended to resolve satellite retrievals into biogenic source–sink changes. The size of the assimilated biosphere sink in China inferred from GOSAT was −0.47 Pg C yr⁻¹. The biosphere flux at the provincial scale was re-estimated following the refined description in the regional inversion.

We find a new kind of non-stationarity. This new kind of non-stationarity is caused by the intrinsic randomness. Results show that the new kind of non-stationarity is widespread in small-scale variations of CO₂ turbulent fluxes. This finding reminds us that we need to handle the short-term averaged turbulent fluxes carefully, and we also need to re-screen the existing non-stationarity diagnosis methods because they could make a wrong diagnosis due to this new kind of non-stationarity.