eFATA: an efficient fata morgana algorithm for climate change forecasting

Abstract An enhanced Fata Morgana optimizer, called eFATA is presented to improve convergence speed and robustness on complex landscapes. We target FATA because, while effective, it can lose diversity and stagnate on multimodal problems; eFATA adds Opposition-Based Learning (diversified initialization) and a Local Escaping Operator (adaptive local exploration) to rebalance exploration and exploitation. On the CEC’22 benchmark suite, eFATA ranked first overall (Friedman mean rank $$=1.25$$ ), achieved the best mean fitness on 11/12 functions, and reached the known global optimum on 11/12 benchmarks, with a success rate of 91.7%. Beyond synthetic tests, we apply eFATA to tune Support Vector Regression (SVR) for maximum temperature forecasting across nine Egyptian governorates (Agricultural Research Center data), yielding consistently lower RMSE and more stable convergence than competing optimizers. These results show that eFATA advances metaheuristic optimization and provides a reliable, accurate framework for practical climate-change forecasting.