Research papersNetwork environmental analysis based ecological risk assessment of a naphthalene-contaminated groundwater ecosystem under varying remedial schemes

•A network environmental analysis based ecological risk assessment model is developed.•The model incorporates both direct and indirect ecosystem interactions.•The model is applied to a naphthalene-contaminated groundwater remediation site.•Naphthalene biodegradation is simulated by a two-dimensional finite difference model.•Results imply the most desirable remedial schemes with the lower integral risks.

Many of the existing ecological risk studies for groundwater ecosystems paid little attention to either small-scale regions (e.g., an industrial contamination site) or ignored anthropogenic activities (e.g., site remediation). This study presented a network environmental analysis based ecological risk assessment (ERA) framework to a naphthalene-contaminated groundwater remediation site. In the ERA, four components (vegetation, herbivore, soil micro-organism and carnivore) were selected, which are directly or indirectly exposed to the contaminated groundwater ecosystem. By incorporating both direct and indirect ecosystem interactions, the risk conditions of the whole ecosystem and its components were quantified and illustrated in the case study. Results indicate that despite there being no input risks for herbivores and carnivores, the respective integral risks increase to 0.0492 and 0.0410. For soil micro-organisms, 58.8% of the integral risk comes from the input risk, while the other 41.2% of the integral risk comes from the direct risk. Therefore, the risk flow within the components is a non-negligible risk origination for soil micro-organisms. However, the integral risk for vegetation was similar to the input risk, indicating no direct risk. The integral risk at the 5-year point after remediation was the highest for the four components. This risk then decreased at the 10-year point, and then again increased. Results from the sensitivity analysis also suggest that the proposed framework is robust enough to avoid disturbance by parameter uncertainty.