Island convection and its representation in numerical weather prediction models

Cloud Trails (CTs) are organised, thermally-forced bands of shallow convection which form downwind of small heated islands. Increased availability of high-frequency satellite imagery offers new opportunities to investigate CT occurrence. There are few examples of the environments associated with CT occurrence, with few links to the CT system in general, and none for Bermuda. Furthermore, Numerical weather prediction (NWP) systems are currently evolving toward higher resolution. As a result, CT-producing islands can now be partially represented by regional models. This motivates renewed exploration of controls on CT occurrence, the role of CTs for the evolution of the larger-scale, and how CT representation is degraded when poorly-resolved. We extended our understanding of CT and shallow convection with a two-pronged observational-numerical approach. First, a five-year climatology of CT occurrence covering the warm season at Bermuda is constructed using a novel automated detection algorithm trained on manually classified, half-hourly visible satellite imagery and surface wind observations. A total of 16,400 images are classified with more than 5,000 CT images identified. CT are most common in July and the afternoon: coincident with peak solar heating and the peak strength of the large-scale subtropical ridge and associated settled summertime weather in Bermuda. Additionally, CT form more frequently when the low-level environment is warmer and more humid than climatology. Wind speed and direction are of further importance because of their control on the total heating of the flow as it crosses the island. Next, the idealised UM is used to perform numerical experiments to examine the CT system more closely. A control experiment based on observations from Bermuda help us to extend the existing conceptual model of the CT system by expanding the role of the warm plume in stabilising the boundary layer in the wake of the island. Variants on the control experiment, are used to investigate changes in CT behaviour in response to their forcing and environment. We find that CT strength varies linearly with island heating as expected from scale analysis. Then we demonstrate that a CT circulation of equivalent strength can form without clouds given a sufficiently dry environment. Finally, we show that the CT system evolves as it intensifies with decreasing wind speed. The CT becomes precipitating at lower wind speeds in our experiments, with more impact on the larger-scale environment and potentially for the public on or downwind of such islands. The final part of this work introduces the problem of representing shallow convection at grid lengths where it is poorly resolved (i.e. the shallow convection "grey zone"). We show that shallow convection becomes too-strong and too-deep in the UM for grey zone resolutions. A crude representation of the CT system is possible for grid lengths of 1.6 km (similar to the resolution of the MetOffice's UKV). But we show that when larger grid spacing is used, the CT becomes more intense and more strongly precipitating compared to a well-resolved CT. We then echo calls for the use of a shallow convection parametrisation in kilometre-scale models, and demonstrate some of the current challenges.