Strategies to Improve Simulations of Cloud and Fog Droplet Number Concentrations in Weather and Climate Models

Clouds reflect sunlight, leading to a net cooling effect on the Earth’s surface. Aerosol particles, depending on their size, concentration, and chemical properties, can influence cloud droplet number concentrations (Nd) and cloud lifetime; hence affecting the optical properties of clouds and indirectly impacting Earth’s climate. The increase in aerosol concentrations in the last 200 years has intensified these indirect effects. However, observations of the pre-industrial atmosphere are extremely limited, and future observations will be inherently unavailable. Even today, in-situ aircraft measurements are costly and sparse, while satellite-derived data come with their own uncertainties. Hence, to complement these in-situ and remotely sensed observations, and to better understand past and future atmosphere, global and regional atmospheric models are used to simulate aerosol indirect effects and estimate how aerosol-cloud interactions have changed Earth’s radiation balance since the pre-industrial era. Unfortunately, these estimates are uncertain and have high variability among different models. In this dissertation, I address several challenges and explore potential strategies to improve the representation of aerosol-cloud interactions in weather and climate models.

Aerosol activation is the process in which droplets form on an aerosol particle and grow spontaneously in a supersaturated environment. In climate and weather models, activation parameterizations are used to simulate activation and calculate the resulting Nd based on model-simulated aerosol properties, updrafts and thermodynamic parameters such as temperature. Activation parameterizations are typically derived from cloud parcel models, which represent activation in a rising air parcel from first principles but are too expensive to include in a climate model. We recommend adjustments to three constant parameters in the equations of the popular Abdul-Razzak and Ghan (ARG) activation scheme, which improve the performance of the parameterization compared to the parcel model. Implementing the updated equations in the UK Met Office Unified Model (UM, an unified weather prediction and climate model) and using a global simulation (about 135 iv km grid spacing), we find significant changes in the estimates of cloud radiative effects.

We continue testing and developing the representation of aerosol activation in the UM by using regional model simulations. Fog is a weather hazard. By reducing visibility, fog disrupts transport, resulting in severe economic loss. However, fog is also a low-level cloud (cloud base near the surface), reflecting sunlight and impacting the present and future climate. Fog lifecycle is complex and the model microphysics code is usually not designed to simulate fog. Numerical weather prediction models, including the UM hence struggle to accurately forecast fog onset and dissipation. Given the unified nature of the UM, we use a high-resolution (500 m horizontal spacing) regional configuration around Paris in an attempt to improve our understanding of the droplet budget of several fog events during the ParisFog field campaign. We find that the default model (with default ARG) severely underestimates Nd during the fog cases we test, despite only a slight underestimate of the aerosols. However, with the updated ARG scheme, we produce Nd in better agreement with the observations. We find that the fog properties are highly sensitive to aerosol activation.

Unlike other low-level clouds, fog droplets are often formed via radiative cooling of the ground surface during winter nights. So far, we only activate droplets via adiabatic cooling generated by updrafts. However, radiative cooling is a key driving mechanism for the formation and growth of fog droplets. Therefore, we extend this analysis by incorporating radiative cooling into the aerosol activation parameterization, in addition to adiabatic cooling. This time, we also simulate a fog event in the UK during the LANFEX field campaign to use measurements of fog vertical microphysical structure. Although our modifications improve the simulated liquid water path (integrated total liquid water in columns) and the fog vertical structure, Nd is overestimated in some cases. We design multiple sensitivity study to address the bias in Nd. We also find that radiative cooling is the dominant driving mechanism for Nd, however, at times, especially after fog formation, both cooling sources can contribute significantly.

In the final part of the thesis, we investigate the updrafts used in the ARG activation v scheme (wact) to calculate Nd. In a rising air parcel, cooling is generated by updrafts, which is a main source of supersaturation, leading to activation. Updrafts resolved in a model depend on the amount of turbulence resolved, which depends on the size of the model grid. Hence, the unresolved component (subgrid variability) of the updraft is parameterized. To update the existing parameterization, we set several regional configurations of the UM, ranging from 100 m to 30 km horizontal grid spacing. We use aircraft observations and radar products to understand model performance. We derive a scal?ing factor dependent on the grid spacing and altitude and use that with the planetary boundary layer height and the standard deviation of the unresolved updrafts to calculate wact. This approach moderately improves model performance by reducing biases in simulated updrafts. Together, these studies advance the representation of aerosol activation in atmospheric models, reducing key uncertainties in cloud and fog microphysics. These improvements have the potential to enhance both climate simulations and operational weather forecasts by providing a more physically consistent treatment of aerosol-cloud interactions.