Speaker: Dan Grosvenor (Leeds)
Title: Aerosols and clouds across time and space
Abstract:
In this talk I will present work that I have done on aerosols and clouds in UKCA-based models: the UK Earth System climate model (UKESM) and the regional high resolution UKCA model. The work spans a large range of time and spatial scales ranging from a case study of volcanic event at 4km resolution through to climate modelling.
In the high resolution modelling work we use an eruption of sulphur dioxide from the Kilauea volcano in Hawaii as a natural laboratory to evaluate aerosol-cloud interaction processes against satellite observations. The model produces an increase in aerosol optical depth due to the volcano that is too large compared to observations, but the increase in cloud droplet number concentration compares well. The cloud liquid water path (a measure of cloud thickness) and cloud fraction responses are in the opposite direction what is observed, although with a large observational uncertainty. This suggests some issues with the model, but also underscores the difficulty in observing aerosol-cloud interactions.
However, Machine Learning (ML) is one approach that might help overcome such observational difficulties. I present an ML approach to isolate the causal impact of the aerosols upon marine stratocumulus cloud liquid water path (LWP) using satellite data and meteorological fields from reanalysis. The results show an increase in LWP with increasing cloud droplet number concentration (Nd, used as a proxy for aerosol loading) at low Nd followed by a smaller decrease at higher Nd for weakly precipitating clouds, which is consistent with results from high resolution large eddy simulations. The ML approach for determining causal aerosol-LWP relationships from instantaneous satellite fields is tested for accuracy using climate model simulations by training the ML model on snapshots of the model fields (akin to what would be observed by satellite instruments). The results show that the ML-derived relationship compares well to the actual relationship determined by varying aerosols in the model giving confidence in its ability to do the same thing from the observations. The climate model response to aerosol has been evaluated using the relationship derived from the observations and shows that the model performs well in high precipitation regimes but poorly in low precipitation regimes. This work opens the door to improvements in the model physics and a reduction in the uncertainty in model aerosol forcing.
I’m now working in the UKESM development team at the Met Office. Therefore I will also look forwards towards the next iterations of the UKESM model with respect to aerosols and clouds, where we hope to bring together the things learnt from the research across various time and spatial scales that has been performed by scientists across the UK and the world.