The outcomes! The project was organized into three primary Work Packages (WPs), each contributing to various goals. The first WP (WP1: research and understanding), tends to reach the scientific objectives of the project. Here is a simplified summary of the key outcomes of WP1 and how they are being used and shared.
WP1 has yielded significant results aimed at improving predictions of extreme events in the Angola-Benguela Upwelling System (ABUS). First, the project explored the limits of seasonal climate prediction in the ABUS, examining the existing dynamical prediction models. For this work, we evaluate the prediction skill of 2 state-of-the-art Earth System Models (ESMs), including the EC-EARTH, NorCPM and 7 dynamical predictions models from the North American Multi-Model Ensemble (NMME) project, and pre-operational models from the Copernicus Climate Change Service (C3S). The results showed that, despite significant efforts to enhance forecast quality, all dynamical systems exhibited low skill in predicting critical dynamic and thermodynamic features in the ABUS. These limitations were particularly evident in predicting the Sea Surface Temperature (SST) during the main season of Benguela Niño and Niña events.
After identifying the limitations of ESMs, the project assessed the source of error in historical Coupled Model Intercomparison Project Phase 6 (CMIP6) hindcasts model outputs. The goal was to understand why ESMs fail in simulating the variability in the ABUS and especially the extreme warm and cold Benguela events. The achievements included the identification of the key physical precursor's mechanisms leading to a wrong development of the events. This is valuable outputs for the scientific community to improve their capabilities and therefore has been shared with the climate prediction scientific community during a conference in 2022 and have been summarized in a forthcoming paper in a scientific Journal. In response to the challenges faced in predicting Benguela events, the project explored the potential of machine learning-based prediction models. A deep learning model was developed to predict Benguela events. Remarkably, the this model outperformed dynamical forecasting systems and demonstrated great capacities in predicting the peak-season of Benguela events, offering accurate forecasts several months in advance. These findings were presented to the research community in 2023 and will be showcased in a future international conference in 2024. A paper summarizing these results will be released soon! These results represent significant advancements in our ability to predict extreme events in the ABUS, offering promising avenues for improved forecasting and furthering our understanding of this critical marine ecosystem.
What's Next? The significant advancement in forecasting ABA variability and Benguela Niño-Niña events, achieved through deep learning models, has attracted considerable interest within the scientific community. Stemming from this interest, there is an initiative to create a user-friendly web-based warning system, designed to provide forecasting of extreme occurring events. Once launched, this innovative platform will grant open access to researchers, stakeholders, policymakers, and the broader public, offering the latest data and predictions. The website involves providing direct forecasts of Benguela events in the ABA, generated from the most recent satellite data available through Copernicus.
