Integrating Machine Learning and GIS for Environmental Impact Assessment and Spatial Prediction of Municipal Waste Management Practices in Port Harcourt Metropolis, Nigeria

Abstract

Municipal solid waste mismanagement is an important environmental challenge in rapidly urbanising cities, where inadequate collection and improper disposal can increase environmental and public-health risks. This study aimed to integrate statistical analysis, machine learning (ML), and Geographic Information Systems to assess environmental impacts and spatially predict municipal waste-management risk in Port Harcourt Metropolis, Nigeria, comprising Port Harcourt City and Obio/Akpor Local Government Areas. Data were collected using a well- tailored structured questionnaire administered to 388 respondents selected from the study area. The analytical framework combined descriptive statistics, chi-square goodness-of-fit tests, Spearman rank correlation, Environmental Waste Management Risk Index construction, four machine-learning regression models, and GIS-based spatial prediction. The machine-learning models comprised Random Forest (RF), Gradient Boosting (GB), Support Vector Regression with a radial basis function kernel, and Multilayer Perceptron . The dataset was divided into 80% training and 20% testing subsets, while model performance was evaluated using coefficient of determination (R2), root mean square error , and mean absolute error . GIS was used to integrate spatial predictor layers and classify predicted EWMRI into very low, low, moderate, high, and very high risk zones. Results showed that daily household waste generation was predominant (55.2%), while 42.0% of respondents rarely or never separated waste before disposal. Roadside/open-space disposal accounted for 21.1%, burning for 14.2%, and disposal into drains or waterways for 7.7%. Significant chi-square results were obtained for all waste-management, environmental-impact, and service indicators (p < 0.001). EWMRI was positively associated with illegal dumping (ρ = 0.568), blocked drainage (ρ = 0.441), and open burning (ρ = 0.458), while waste-collection frequency showed a negative association (ρ = −0.367), all at p < 0.001. Overall, 71.9% of observations were classified as high or very high risk. Among the ML models, Gradient Boosting performed best, with R2 = 0.831, RMSE = 6.107, and MAE = 5.002, while illegal dumping (0.321), open burning (0.217), and blocked drainage (0.195) were the most influential predictors. The spatial prediction framework further indicates that areas where multiple waste pressures co-occur should receive priority attention, particularly for enforcement, drainage maintenance, waste infrastructure and community-level interventions. The study concludes that poor waste-management practices constitute a substantial environmental risk in Port Harcourt Metropolis and that integrating ML IJGEM IJGEM IIARD International Journal of Geography & Environmental Management with GIS provides a useful framework for predicting and spatially prioritising environmental risk. The study recommends strengthening waste-collection coverage and reliability, improving collection points, enforcing regulations against illegal dumping and open burning, maintaining drainage systems, promoting household waste separation, enhancing environmental awareness, and using GIS-based risk maps to target monitoring and intervention resources.

Keywords

References