Data-driven approaches linking wastewater and source estimation hazardous waste for environmental management

Area of Science:

• Environmental Science
• Industrial Ecology
• Data Science

Background:

• Industrial enterprises generate significant contaminants, necessitating effective regulation for sustainable development.
• Estimating firm-level contaminant generation, especially hazardous waste (HW), is challenging due to enterprise heterogeneity and a lack of universal methods.
• Automatic sensors for wastewater monitoring are widespread, offering a potential data source for indirect HW estimation.

Purpose of the Study:

• To develop a data-driven methodology for predicting hazardous waste (HW) generation at the enterprise level.
• To establish a generic framework applicable across diverse industrial sectors.
• To address challenges in data distribution and computational efficiency for HW prediction.

Main Methods:

• Utilized wastewater big data as a proxy for HW generation prediction.
• Developed a generic framework incorporating representative variables from various sectors.
• Employed a data-balance algorithm to handle long-tail data distribution.
• Incorporated causal discovery for feature screening and enhanced computational efficiency.

Main Results:

• The methodology demonstrated high fidelity in predicting HW generation, achieving an R² of 0.87.
• The study successfully tested the framework on 1024 enterprises across 10 sectors in Jiangsu, China.
• The prediction model utilized a large dataset of 4,260,593 daily wastewater data points.

Conclusions:

• The developed data-driven methodology provides an accurate and efficient approach for predicting industrial hazardous waste generation.
• This approach offers a valuable tool for environmental regulation and sustainable development by enabling firm-level contaminant tracking.
• The framework's adaptability across sectors and its use of wastewater data highlight its practical applicability.