This thesis explores the application of advanced time series forecasting models, specifically N-BEATSx and N-HiTS, to understand complex dynamics within marine ecosystems in the North Sea. By integrating exogenous variables, such as sea surface temperature (SST), and employing counterfactual explanations, this research addresses the question: ”How and to what extent can counterfactual explanations enhance our understanding of deep learning models for time series forecasting with exogenous variables in marine ecosystems?” The findings reveal that incorporating SST as an exogenous variable alone does not improve the forecasting models’ performance. However, counterfactual explanations generated using ForecastCF-Exog offer valuable insights into how changes in external factors influence fish populations. This framework not only advances the interpretability of deep learning models but also provides actionable insights for sustainable policy-making and ecosystem management. By enhancing our understanding of complex environmental systems through interpretable models, this research contributes a framework that offers novel insights for effective decision-making in natural sciences and other applications.