Despite a large number of studies, out of all drivers of radiative forcing, the effect of aerosols has the largest uncertainty in global climate model radiative forcing estimates. There have been studies of aerosol optical properties in climate models, but the effects of particle number size distribution need a more thorough inspection. We investigated the trends and seasonality of particle number concentrations in nucleation, Aitken, and accumulation modes at 21 measurement sites in Europe and the Arctic. For 13 of those sites, with longer measurement time series, we compared the field observations with the results from five climate models, namely EC-Earth3, ECHAM-M7, ECHAM-SALSA, NorESM1.2, and UKESM1. This is the first extensive comparison of detailed aerosol size distribution trends between in situ observations from Europe and five earth system models (ESMs). We found that the trends of particle number concentrations were mostly consistent and decreasing in both measurements and models. However, for many sites, climate models showed weaker decreasing trends than the measurements. Seasonal variability in measured number concentrations, quantified by the ratio between maximum and minimum monthly number concentration, was typically stronger at northern measurement sites compared to other locations. Models had large differences in their seasonal representation, and they can be roughly divided into two categories: for EC-Earth and NorESM, the seasonal cycle was relatively similar for all sites, and for other models the pattern of seasonality varied between northern and southern sites. In addition, the variability in concentrations across sites varied between models, some having relatively similar concentrations for all sites, whereas others showed clear differences in concentrations between remote and urban sites. To conclude, although all of the model simulations had identical input data to describe anthropogenic mass emissions, trends in differently sized particles vary among the models due to assumptions in emission sizes and differences in how models treat size-dependent aerosol processes. The inter-model variability was largest in the accumulation mode, i.e. sizes which have implications for aerosol-cloud interactions. Our analysis also indicates that between models there is a large variation in efficiency of long-range transportation of aerosols to remote locations. The differences in model results are most likely due to the more complex effect of different processes instead of one specific feature (e.g. the representation of aerosol or emission size distributions). Hence, a more detailed characterization of microphysical processes and deposition processes affecting the long-range transport is needed to understand the model variability.

Headwater streams are important in the carbon cycle and there is a need to better parametrize and quantify exchange of carbon-relevant gases. Thus, we characterized variability in the gas exchange coefficient (k(2)) and dissolved oxygen (O-2) gas transfer velocity (k) in two lowland headwaters of the River Avon (UK). The traditional one-station open-water method was complemented by in situ quantification of riverine sources and sinks of O-2 (i.e., groundwater inflow, photosynthesis, and respiration in both the water column and benthic compartment) enabling direct hourly estimates of k(2) at the reach-scale (similar to 150 m) without relying on the nighttime regression method. Obtained k(2) values ranged from 0.001 h(-1) to 0.600 h(-1). Average daytime k(2) were a factor two higher than values at night, likely due to diel changes in water temperature and wind. Temperature contributed up to 46% of the variability in k on an hourly scale, but clustering temperature incrementally strengthened the statistical relationship. Our analysis suggested that k variability is aligned with dominant temperature trends rather than with short-term changes. Similarly, wind correlation with k increased when clustering wind speeds in increments correspondent with dominant variations (1 m s(-1)). Time scale is thus an important consideration when resolving physical drivers of gas exchange. Mean estimates of k(600) from recent parametrizations proposed for upscaling, when applied to the settings of this study, were found to be in agreement with our independent O-2 budget assessment (within <10%), adding further support to the validity of upscaling efforts aiming at quantifying large-scale riverine gas emissions.