Climate Prediction

Weather forecasts are based on models, which is what this part of CONFER is focused on. We want to improve the accuracy of existing models to get better forecasts for East Africa, as well as obtain a new level of seasonal forecast skill based on numerical models and high-resolution satellite data.

Objectives

Despite impressive gains in recent decades, predicting key features of the climate system (e.g. precipitation, rainy season onset, extreme weather, floods, droughts, heatwaves) at lead time of months to seasons remains a stubborn challenge. Shortcomings in modelling systems, sparse observational networks and complex interactions between components of the earth system all conspire to slow progress. The good news is that researchers have reason to believe that there is considerable predictability in the Earth system that is currently under-exploited. CONFER scientists take a multi-pronged approach to improving seasonal forecasts which will consequently impact the models used by stakeholders that provide climate services in the region.

How will we go about doing it?

 First, we want to increase the resolution of atmospheric models. This will improve the realism of the simulated climate. Second, we plan to address persistent sources of errors, such as land surface initialisation and mean state errors, in global models. Additionally, CONFER scientists are working towards smarter construction of ensembles, which take advantage of models that correctly reproduce known sources of predictability. Lastly, there are indications that combining statistical and dynamical approaches can lead to significant improvements for regions where there are strong links between ocean variability and climate on land. CONFER is adapting these methods to East Africa.

Where are we at now?

Recently, CONFER published a paper about the connection between ocean temperatures and rainfall in East Africa. The paper states that the connection between the Indian Ocean Dipole and rainfall in East Africa tends to be exaggerated by the ECMWF model. Knowing this, we can take that into account when using the model to create the weather forecast in the East African region. Long term, this knowledge can also help improve the models we use to predict the weather.  

Improvements to the models will serve as input to downstream impact models which can be used to predict hydrological and crop conditions. These models themselves are undergoing testing and improvement in CONFER and will link to existing services provided by ICPAC. Prototypes for improved streamflow (essential for predicting flood hazards and energy potential) and crop (critical for food security assessments) forecasting systems are expected by the end of the project.

While these research activities are ongoing, discussions with climate services developers and training providers are happening in parallel. In this way, the prediction research in CONFER can be targeted to the needs of ICPAC and the NMHSs, as well as the downstream users of seasonal climate information that they serve. By targeting existing operational climate services, we will minimize the gap between research and operations and ensure rapid uptake of improvements.