Evaluating Climate Models with an African Lens

Climate models are becoming ever more complex, and increasingly relied upon to inform climate change adaptation. Yet progress in model development is lagging behind in many of the regions which need the information most, including in Africa. The vast majority of climate models have been developed outside of Africa, and mid-latitude processes have historically received more attention. Tropical phenomena, such as organised convection, and strong teleconnections to local and remote sea-surface temperatures, are still a struggle for global climate models.

Targeted model development for Africa is crucial; and so too is targeted model evaluation. Assessment of model performance in specific regions often follows a “validation” approach, focusing on mean biases; but if models are to be improved, it is important to understand how they simulate regional climate dynamics: to move from validation to process-based evaluation. Process-based evaluation, focusing on the systems which matter for Africa, could provide insights for model development, and inform confidence assessments, helping discern which, if any, climate projections can be trusted for adaptation planning. Collaboration with experts in African research institutions and met services is crucial: the processes which matter may be different for every region, and “one-size fits all” could overlook important, region-specific phenomena.

Here we report on a concerted effort to better evaluate and improve the UK Met Office Unified Model over Africa, though collaboration between the UK modelling centre and African scientists. For each of Central, East, Southern, and West Africa, we review current understanding of climate models, and demonstrate an example of targeted evaluation, including investigation of moist circulations, teleconnections, and modes of variability. Our analysis is based on the Met Office Unified Model, but using diagnostics which might be applied to other models. These examples are intended to prompt further discussion among climate modellers and African scientists about how to best evaluate models with an African lens.

As part of the next generation of coupled climate models (CMIP6), there is drive to advance evaluation through routine deployment of community-based analysis tools. If diagnostics which are important for African climate are included, this new infrastructure could faciliate dramatic improvements understanding of model behavior for this important continent. We encourage the development of a model evaluation hub for Africa, to collectively identifying priorities for evaluation, share research insights and analysis methods, and move towards identifying and developing diagnostics to be incorporated into the CMIP evaluation toolkit.

We have recently published a paper in BAMS on this topic, and this conference session represents an ideal opportunity to promote further discussion between experts in African climate about how to fast-track understanding of model behavior over Africa.