China's rapid industrialization and economic development over the past few decades have produced large amounts of solid waste, poses great risks to the environment1. Currently, waste streams from agriculture, construction, industry and post-use sources can be consolidated to more than 10 Gigatonnes (GT/A) per year2,3,4,5. Among them, China's Industrial Solid Waste (ISW) counts approximately 4 GT/A. This is more than half of the world's coal or nearly twice the world's iron ore production.5,6,7.
According to the updated solid waste code, there are 17 major categories and hundreds of subcategories in China (except for dangerous industrial waste). The top six categories are metallurgical slag, fly ash, furnace slag, coal knots, tailings and desulfurization plaster, respectively. ISW can be used as secondary materials, landfills, incineration, or temporary stockpile for future treatment. China's 13th Five-Year Plan (2016-2020) set a target for a comprehensive utilization rate of 73%, but by 2022 only 57.7% of ISW had been reused (China's Ecological Environment Status Breaking Report 2022). Cumulative ISW strains processed environmentally safely have reached about 60-70 GT8has led to various environmental decompositions ranging from soil erosion, groundwater pollution, habitat destruction, and loss of biodiversity across the country.
Despite the urgent challenges, the complete tempo spatial dataset of ISW generation in urban-level China has not been established, impeding a comprehensive understanding of waste management and recycling and further investigation. Functioning environmental monitoring and data collection systems allow statistically tracking and reporting industrial waste from factories to factories to cities by city. However, this takes time and requires expensive investments in environmental infrastructure. Before high-resolution industrial waste information is fully recorded, machine learning can provide a powerful tool to estimate regional and regional industrial waste production patterns in China and bridge gaps.
Machine learning is increasingly being applied to environmental management because of its cost-effectiveness, predictive accuracy and robustness. It is already widely used for making predictions, extracting features, detecting abnormalities, and discovering new materials and chemicals9. Machine learning is also used in a variety of solid waste research.10,11,12. Lin et al.13 We conducted an important review of the application of deep trends in solid waste management and found that this method is highly effective in predicting waste volume and composition. The amount of waste can be notified by combining one or more machine learning methods. Cannangara et al.14 Neural networks and decision trees were employed in modelling and estimating the generation and diversion of Canadian regional waste. Ma et al.15 We also used artificial neural networks to investigate the composition of urban solid waste (MSW). Chan et al.16 We predicted MSW generation with five machine learning approaches.
This current study aims to establish urban-level ISW datasets in China from 1990 to 2020 by adopting a non-intervention-based machine learning approach. It covers industrial waste from all 337 administrative regions in China from 1990 to 2022. This includes 293 provincial level cities, 7 provincial level regions, 30 autonomous prefectures, 3 provincial level leagues, and 4 municipalities (as seen in the dataset).17). For simplicity, use “city” to refer to these administrative departments in the text below. In addition to the ISW gross flow, we will also try to develop city-level inventory for 2022 for six major ISW subcategories. These include metallurgical slag, fly ash, furnace slag, coal knots, tailings, and desulfurization plaster.
ISW is one of the largest and most widely spread solid waste streams that creates environmental and ecosystem risks in many regions. However, regional and mesoscale ISW generation is often fluctuating due to local industry changes and irregular operational factors. Such dynamics can hardly be captured by traditional spatial interpolation methods. Therefore, we have developed a data-driven and non-interventional machine learning framework to improve prediction accuracy and spatial resolution. We achieve this complete and consistent estimate of ISW in Chinese cities. The methodology and practices provided in this current study could also help inform waste generation patterns in the future and in other fields.
