This project was funded by the Natural Environment Research Council (NERC) with the grant reference - NE/I021012/1 - and was led by Dr Christopher Holloway (University of Reading).

Tropical cloud systems and rainfall help drive the global circulation of the atmosphere, transferring heat from near the Earth's surface upward for many kilometres. These convective systems can be found in groups of many different sizes, from isolated showers and thunderstorms to equatorial waves to tropical cyclones to the Madden-Julian Oscillation (MJO), an eastward-propagating weather system composed of superclusters of convection several thousand kilometres across which dominates tropical weather variability on weekly to monthly time scales. Global numerical weather forecast and climate models still do not adequately simulate these organized storm clusters and, as a result, have too little (or incorrect) variability of tropical rainfall. Improvement of the representation of organized tropical convection, and therefore the accuracy of weather forecasts, would greatly improve the lives of billions of people who rely on rainfall for agriculture in the tropics and subtropics; better forecasts of strong storms and flooding would also save countless lives and reduce property damage. Furthermore, these processes may change in the future as the climate changes due to human activities, so an improvement of the ability of global models to simulate organized convection will lead to better predictions of possible climate change scenarios over the whole globe.

Global weather and climate models divide the Earth into grid boxes about 100 km across. These boxes are too large to directly simulate the motions responsible for small-scale rainstorms, instead estimating total rainfall based on average conditions in the box. This simplified rain estimation, necessary because of limited computer resources, ignores the interaction of isolated rain showers with each other and regional weather conditions.

An exciting new research area is the study of organized convection in high-resolution idealized models. These models, with constant sea surface temperatures and constant sunlight, can now be run on domains several thousand kilometres across and with grid boxes of only a few kilometres long, allowing convection to be represented explicitly. These models are beginning to provide insight into processes that lead to spontaneous growth of convective clusters which can ultimately grow to a single large cluster accounting for all of the rainfall in the domain. These processes act over a wide range of spatial scales which are not fully resolved in global models.

However, the processes which lead to organized convection in idealized models are still not well understood, and it is not known whether they are also important for organizing tropical convection in nature. This project exploited a large archive of high-resolution model runs, forecast analyses, and observations from satellites to make more direct comparisons between idealized cases and observed phenomena. Ultimately, this endeavour had the potential to lead to improvements in the way that global models, especially the UK Met Office Unified Model, simulate tropical rainfall and with it global weather and climate.

This project benefited from collaborations with University of Reading, the Met Office and with other scientists approaching similar problems using different models.

The improvement of global model simulations of tropical rainfall, along with associated cloud fields and atmospheric heating, is key to reducing uncertainty in predictions of high-impact weather and climate change. Global models still fail to produce realistic large-scale clusters of tropical clouds and rainfall, which are fundamental to predicting variability on weekly and monthly timescales and have major effects on phenomena such as tropical cyclones, El Nino, and the Asian monsoon, as well as the entire global circulation. Recent high-resolution simulations for idealized conditions over tropical oceans have shown spontaneous self-aggregation of convection, with likely causes including feedbacks between convection, moisture and clouds, and radiation, as well as between convection and surface fluxes.

This project aimed to clearly identify processes important for self-aggregation of convection in idealized models and then to test whether these processes, or different processes, are active in convective organization in nature. The second part of this goal was an open question in the field, and this fellowship has the potential to connect a rapidly expanding theoretical research area with ongoing efforts to improve the understanding and prediction of tropical variability. The focus on the Unified Model benefited weather and climate prediction in the UK by exchanging ideas with Met Office scientists who were directly involved in testing and improving the model.

The specific objectives of this project, with increasing priority downwards, were to address the following three questions:
1. What processes are responsible for convective self-aggregation in idealized models?
2. To what extent are these idealized model processes important for convective organization in nature as represented by observations, forecast analyses, and high-resolution model case studies, particularly in conditions resembling those of idealized models? What processes lead to the simplest cases of convective organization observed?
3. How do more complex features of nature, such as equatorially-trapped waves, land-sea contrast, and extratropical features, interact with more idealized mechanisms for convective organization?

Addressing these specific objectives allowed for the improvement of numerical weather and climate prediction by contributing new ideas for the development of both high-resolution models with explicit convection and lower resolution models with parameterized convection.