By combining observations and numerical simulations, we investigated the responses of the surface energy budget and the convective boundary layer (CBL) dynamics to the presence of aerosols. A detailed data set containing (thermo)dynamic observations at CESAR (Cabauw Experimental Site for Atmospheric Research) and aerosol information from the European Integrated Project on Aerosol, Cloud, Climate, and Air Quality Interactions was employed to design numerical experiments reproducing two typical clear-sky days, each characterized by contrasting thermodynamic initial profiles: (i) residual layer above a strong surface inversion and (ii) well-mixed CBL connected to the free troposphere by a capping inversion, without the residual layer in between. A large-eddy simulation (LES) model and a mixed-layer (MXL) model, coupled to a broadband radiative transfer code and a land surface model, were used to study the impacts of aerosols on shortwave radiation. Both the LES model and the MXL model results reproduced satisfactorily the observations for both days. A sensitivity analysis on a wide range of aerosol properties was conducted. Our results showed that higher loads of aerosols decreased irradiance imposing an energy restriction at the surface, delaying the morning onset of the CBL and advancing its afternoon collapse. Moderately to strongly absorbing aerosols increased the heating rate contributing positively to increase the afternoon CBL height and potential temperature and to decrease Bowen ratio. In contrast, scattering aerosols were associated with smaller heating rates and cooler and shallower CBLs. Our findings advocate the need for accounting for the aerosol influence in analyzing surface and CBL dynamics.

We investigated the impact of aerosol heat absorption on convective atmospheric boundary-layer (CBL) dynamics. Numerical experiments using a large-eddy simulation model enabled us to study the changes in the structure of a dry and shearless CBL in depthequilibrium for different vertical profiles of aerosol heating rates. Our results indicated that aerosol heat absorption decreased the depth of the CBL due to a combination of factors: (i) surface shadowing, reducing the sensible heat flux at the surface and, (ii) the development of a deeper inversion layer, stabilizing the upper CBL depending on the vertical aerosol distribution. Steady-state analytical solutions for CBL depth and potential temperature jump, derived using zero-order mixed-layer theory, agreed well with the large-eddy simulations. An analysis of the entrainment zone heat budget showed that, although the entrainment flux was controlled by the reduction in surface flux, the entrainment zone became deeper and less stably stratified. Therefore, the vertical profile of the aerosol heating rate promoted changes in both the structure and evolution of the CBL. More specifically, when absorbing aerosols were present only at the top of the CBL, we found that stratification at lower levels was the mechanism responsible for a reduction in the vertical velocity and a steeper decay of the turbulent kinetic energy throughout the CBL. The increase in the depth of the inversion layer also modified the potential temperature variance. When aerosols were present we observed that the potential temperature variance became significant already around 0.7z i (where z i is the CBL height) but less intense at the entrainment zone due to the smoother potential temperature vertical gradient.

No abstract

Research has shown that personalization of health interventions can contribute to an improved effectiveness. Reinforcement learning algorithms can be used to perform such tailoring using data that is collected about users. Learning is however very fragile for health interventions as only limited time is available to learn from the user before disengagement takes place, or before the opportunity to intervene passes. In this paper, we present a cluster-based reinforcement learning approach which learns across groups of users. Such an approach can speed up the learning process while still giving a level of personalization. The clustering algorithm uses a distance metric over traces of states and rewards. We apply both online and batch learning to learn policies over the clusters and introduce a publicly available simulator which we have developed to evaluate the approach. The results show batch learning clearly outperforms online learning. Furthermore, clustering can be beneficial provided that a proper clustering is found.

No abstract