The formation of shallow cumulus cloud streets was historically attributed primarily to dynamics. Here, we focus on the interaction between radiatively induced surface heterogeneities and the resulting patterns in the flow. Our results suggest that solar radiative heating has the potential to organize clouds perpendicular to the sun's incidence angle. To quantify the extent of organization, we performed a high-resolution large-eddy simulation (LES) parameter study. We varied the horizontal wind speed, the surface heat capacity, the solar zenith and azimuth angles, and radiative transfer parameterizations (1-D and 3-D). As a quantitative measure we introduce a simple algorithm that provides a scalar quantity for the degree of organization and the alignment. We find that, even in the absence of a horizontal wind, 3-D radiative transfer produces cloud streets perpendicular to the sun's incident direction, whereas the 1-D approximation or constant surface fluxes produce randomly positioned circular clouds. Our reasoning for the enhancement or reduction of organization is the geometric position of the cloud's shadow and its corresponding surface fluxes. Furthermore, when increasing horizontal wind speeds to 5 or 10 m s⁻¹, we observe the development of dynamically induced cloud streets. If, in addition, solar radiation illuminates the surface beneath the cloud, i.e., when the sun is positioned orthogonally to the mean wind field and the solar zenith angle is larger than 20°, the cloud-radiative feedback has the potential to significantly enhance the tendency to organize in cloud streets. In contrast, in the case of the 1-D approximation (or overhead sun), the tendency to organize is weaker or even prohibited because the shadow is cast directly beneath the cloud. In a land-surface-type situation, we find the organization of convection happening on a timescale of half an hour. The radiative feedback, which creates surface heterogeneities, is generally diminished for large surface heat capacities. We therefore expect radiative feedbacks to be strongest over land surfaces and weaker over the ocean. Given the results of this study we expect that simulations including shallow cumulus convection will have difficulties producing cloud streets if they employ 1-D radiative transfer solvers or may need unrealistically high wind speeds to excite cloud street organization.

We investigate the effects of thermal radiation on cloud development in large-eddy simulations (LESs) with the UCLA-LES model. We investigate single convective clouds (driven by a warm bubble) at 50 m horizontal resolution and a large cumulus cloud field at 50 and 100 m horizontal resolutions. We compare the newly developed 3-D Neighboring Column Approximation with the independent column approximation and a simulation without radiation and their respective impact on clouds. Thermal radiation causes strong local cooling at cloud tops accompanied by a modest warming at the cloud bottom and, in the case of the 3-D scheme, also cloud side cooling. 3-D thermal radiation causes systematically larger cooling when averaged over the model domain. In order to investigate the effects of local cooling on the clouds and to separate these local effects from a systematically larger cooling effect in the modeling domain, we apply the radiative transfer solutions in different ways. The direct effect of heating and cooling at the clouds is applied (local thermal radiation) in a first simulation. Furthermore, a horizontal average of the 1-D and 3-D radiation in each layer is used to study the effect of local cloud radiation as opposed to the domain-averaged effect. These averaged radiation simulations exhibit a cooling profile with stronger cooling in the cloudy layers. In a final setup, we replace the radiation simulation by a uniform cooling of 2.6 K day⁻¹. To focus on the radiation effects themselves and to avoid possible feedbacks, we fixed surface fluxes of latent and sensible heat and omitted the formation of rain in our simulations. Local thermal radiation changes cloud circulation in the single cloud simulations, as well as in the shallow cumulus cloud field, by causing stronger updrafts and stronger subsiding shells. In our cumulus cloud field simulation, we find that local radiation enhances the circulation compared to the averaged radiation applications. In addition, we find that thermal radiation triggers the organization of clouds in two different ways. First, local interactive radiation leads to the formation of cell structures; later on, larger clouds develop. Comparing the effects of 3-D and 1-D thermal radiation, we find that organization effects of 3-D local thermal radiation are usually stronger than the 1-D counterpart. Horizontally averaged radiation causes more clouds and deeper clouds than a no radiation simulation but, in general less-organized clouds than in the local radiation simulations. Applying a constant cooling to the simulations leads to a similar development of the cloud field as in the case of averaged radiation, but less water condenses overall in the simulation. Generally, clouds contain more liquid water if radiation is accounted for. Furthermore, thermal radiation enhances turbulence and mixing as well as the size and lifetime of clouds. Local thermal radiation produces larger clouds with longer lifetimes. The cloud fields in the 100 and 50 m resolution simulations develop similarly; however, 3-D local effects are stronger in the 100 m simulations which might indicate a limit of our 3-D radiation parameterization.

The recently developed 3-D TenStream radiative transfer solver was integrated into the University of California, Los Angeles large-eddy simulation (UCLA-LES) cloud-resolving model. This work documents the overall performance of the TenStream solver as well as the technical challenges of migrating from 1-D schemes to 3-D schemes. In particular the employed Monte Carlo spectral integration needed to be reexamined in conjunction with 3-D radiative transfer. Despite the fact that the spectral sampling has to be performed uniformly over the whole domain, we find that the Monte Carlo spectral integration remains valid. To understand the performance characteristics of the coupled TenStream solver, we conducted weak as well as strong-scaling experiments. In this context, we investigate two matrix preconditioner: geometric algebraic multigrid preconditioning (GAMG) and block Jacobi incomplete LU (ILU) factorization and find that algebraic multigrid preconditioning performs well for complex scenes and highly parallelized simulations. The TenStream solver is tested for up to 4096 cores and shows a parallel scaling efficiency of 80–90 % on various supercomputers. Compared to the widely employed 1-D delta-Eddington two-stream solver, the computational costs for the radiative transfer solver alone increases by a factor of 5–10.

One‐dimensional radiative transfer solvers are computationally much more efficient than full three‐dimensional radiative transfer solvers but do not account for the horizontal propagation of radiation and thus produce unrealistic surface irradiance fields in models that resolve clouds. Here, we study the impact of using a 3‐D radiative transfer solver on the direct and diffuse solar irradiance beneath clouds and the subsequent effect on the surface fluxes. We couple a relatively fast 3‐D radiative transfer approximation (TenStream solver) to the Dutch Atmosphere Large‐Eddy Simulation (DALES) model and perform simulations of a convective boundary layer over grassland with either 1‐D or 3‐D radiative transfer. Based on a single case study, simulations with 3‐D radiative transfer develop larger and thicker clouds, which we attribute mainly to the displaced clouds shadows. With increasing cloud thickness, the surface fluxes decrease in cloud shadows with both radiation schemes but increase beneath clouds with 3‐D radiative transfer. We find that with 3‐D radiative transfer, the horizontal length scales dominating the spatial variability of the surface fluxes are over twice as large as with 1‐D radiative transfer. The liquid water path and vertical wind velocity in the boundary layer are also dominated by larger length scales, suggesting that 3‐D radiative transfer may lead to larger convective thermals. Our case study demonstrates that 3‐D radiative effects can significantly impact dynamic heterogeneities induced by cloud shading. This may change our view on the coupling between boundary‐layer clouds and the surface and should be further tested for generalizability in future studies.

Cloud radiative properties play a significant role in radiation and energy budgets and are influenced by both the cloud top height and the particle size distribution. Both cloud top heights and particle size distributions can be derived from 2-D intensity and polarization measurements by the airborne spectrometer of the Munich Aerosol Cloud Scanner (specMACS). The cloud top heights are determined using a stereographic method (Kölling et al., 2019), and the particle size distributions are derived in terms of the cloud effective radius and the effective variance from multidirectional polarized measurements of the cloudbow (Pörtge et al., 2023). In this study, the accuracy of the two methods is evaluated using realistic 3-D radiative transfer simulations of specMACS measurements of a synthetic field of shallow cumulus clouds, and possible error sources are determined. The simulations are performed with the 3-D Monte Carlo radiative transport model MYSTIC (Mayer, 2009) using cloud data from highly resolved large-eddy simulations (LESs). Both retrieval methods are applied to the simulated data and compared to the respective properties of the underlying cloud field from the LESs. Moreover, the influence of the cloud development on both methods is evaluated by applying the algorithms to idealized simulated data where the clouds did not change during the simulated overflight of 1 min over the cloud field. For the cloud top height retrieval, an absolute mean difference of less than 70 m with a standard deviation of about 130 m compared to the expected heights from the model is found. The elimination of the cloud development as a possible error source results in mean differences of (46±140) m. For the effective radius, an absolute average difference of about $(-\mathrm{0.2}\pm \mathrm{1.30})$ µm from the expected effective radius from the LES model input is derived for the realistic simulation and $(-\mathrm{0.03}\pm \mathrm{1.28})$ µm for the simulation without cloud development. The difference between the effective variance derived from the cloudbow retrieval and the expected effective variance is (0.02±0.05) for both simulations. Additional studies concerning the correlations between larger errors in the effective radius or variance and the optical thickness of the observed clouds have revealed that low values in the optical thickness do not have an impact on the accuracy of the retrieval.