Abstract
The skill of global models in predicting African rainfall on the “local” scale of a few kilometres is very poor, even at only one day lead time, whereas in the midlatitudes we have become used to those daily local forecasts being excellent. In contrast, while we struggle to make accurate seasonal predictions for Europe, parts of Africa enjoy useful seasonal predictions due to the influence of sea surface temperature patterns. By considering the model forecast skill as a function of temporal averaging and spatial aggregation, it is possible to understand these differences, which appear to be remarkably consistent across a range of different models. There is evidence from theory, going back to Lorenz in the 1960s, that these midlatitude-tropical differences are fundamental to the predictability of our chaotic, rotating atmosphere. Approaches to weather prediction need to accommodate the principle that predictability is different in the African tropics.
Extreme rain is a perennial hazard in many parts of Africa. If the 24-hour forecast is unreliable at the local scale, people need “nowcasting” (0-6 hour data-based forecasts) – more than any other continent – but nowcasting is almost unheard of in most African countries. A few recent projects have started to turn that around.
Nowcasting has always been a data-based process and it’s no surprise that the science of nowcasting is now embracing the revolution in AI-based prediction. Although purely data-based approaches show impressive skill, the AI models are also demanding a fresh look at the fluid dynamics and physics of weather systems. The next few years will see a revolution in African nowcasting capability, and the priority will be for this science to be owned and driven by the African scientific community.
About the speaker
Doug is a Meteorologist, with interests in the prediction of high-impact weather in our changing climate. He is based at the University of Leeds, between the Schools of Earth & Environment and Mathematics, and holds an adjunct position at NORCE in Bergen, where he is working mostly on weather and climate services for Africa.
Doug has led, and participated in, some large projects and field campaigns in the UK, Africa and India. He has used a range of research tools to solve practical problems, making use of field measurements, satellite remote sensing, computer models and mathematical theory.
He is particularly interested in weather forecasting, both scientifically and as an operational challenge. He was part-funded by the UK Met Office for 9 years, and he has collaborated closely with national weather services in West, East and Southern Africa. Weather prediction is a transdisciplinary challenge which requires us to understand social and economic factors in addition to the mathematics, physics and data science needed to make a prediction.
He is leading an initiative to bring short-term weather predictions (nowcasting) to Africa: the FASTA project provides real-time data to smartphones.