Plastic waste leakage in the oceans: Socioeconomic influences, predictive modeling with artificial intelligence and wavelet analysis
Abstract
Plastic waste pollution in the oceans represents the second most significant global environmental threat, surpassed only by climate change. Brazil ranks among the largest contributors to this problem, largely due to rapid population growth, intense economic activity, and ineffective solid waste management. This study aimed to: (i) examine the relationship between municipal estimates of leak-prone plastic waste in the oceans (LPW) and socioeconomic variables from the Brazilian Institute of Geography and Statistics (IBGE); (ii) evaluate the predictive modeling of LPW using artificial neural networks (ANNs). The analysis focused on the 20 Brazilian municipalities with the highest LPW potential based on 2022 data. Methodologies included simple linear regression, Pearson correlation, wavelet analysis, Taylor diagrams, and geospatial mapping. Results indicated that Sao Paulo, Rio de Janeiro, and the Southeast region exhibited the highest absolute LPW potential. Among the variables analyzed, the municipal population had the strongest influence on LPW, followed by demographic density and gross domestic product (GDP). In predictive modeling, the Nonlinear Autoregressive with Exogenous Inputs (NARX) neural network achieved the highest accuracy, followed by the Nonlinear Autoregressive (NAR) and Long ShortTerm Memory (LSTM) models. However, all ANNs showed limitations in forecasting extreme LPW values in major urban centers. These findings highlight the urgent need for targeted public policies to improve the monitoring and control of plastic waste leakage in the marine ecosystems of Brazil and similar regions worldwide.