Bottom-up discrete systematic modelling for analysis and prediction of future trends for land-sea environmental pollution systems

Abstract

Addressing the challenge of marine environmental pollution requires accurate forecasting systems, yet current methods relying on artificial intelligence and big data face limitations due to insufficient and inconsistent marine data. This paper proposes an integrated bottom-up prediction system featuring a grey modeling module and an optimization module. The grey modeling module introduces a novel nonlinear nested systematic grey model, which employs multi-structure discretization to capture intricate and variable system changes by leveraging the dual nonlinearity of cumulative generated sequences and time trends. Nested solutions minimize prediction errors through discrete recursive equations with underlying-upper structures. The optimization module utilizes an efficient bio-inspired meta-heuristic algorithm, the Coati Optimization Algorithm, which balances global and local search strategies. The accuracy and robustness of the proposed model are enhanced by optimizing its nonlinear parameters through comparative analysis with other algorithms and conducting a sensitivity analysis on the number of iterations. The proposed prediction system is validated in representative Chinese provinces (Zhejiang, Shandong, and Hainan), demonstrating high prediction accuracy for both land-based and marine environmental pollution systems, with mean absolute percentage errors below 8% in both training and test sets. Projections indicate that by 2030, land-based source pollution intensities in these provinces will decrease by approximately 5.71%, 0.16%, and 1.64%, respectively, while marine pollution indices are expected to decline by 3.28%, 4.86%, and 2.89%. This study provides a valuable predictive tool for addressing the complexity, variability, and uncertainty of marine environmental pollution, while contributing to the promotion of land-sea integration and the protection of marine ecosystems.