An inventory of industrial solid waste in 337 cities of China: Applying machine learning for data completion

Rapid industrialization of China generated a massive quantity of waste, among them industrial solid waste contributed the biggest flow to some 60 gigatonnes (Gt) in the past two decades. A complete tempo-spatial dataset of industrial waste, however, is absent in many areas in China, due to numerous waste producers and insufficient statistical coverage. To fill up the gap, we collected current available data from thousands of sources. We further developed six machine learning models to complete the dataset across all the 337 cities in China for the period 1990–2022. Bayesian optimization was employed to obtain the best estimation model for each city and to enhance its performance and resilience. In addition to the aggregate waste amount, generation of six major subcategories of industrial waste, i.e., metallurgical slags, fly ash, furnace slags, coal gangue, tailings, and desulfurization gypsum, are presented for more than half of the cities in 2022. This dataset can help researchers and policymakers recognize and address challenges brought by industrial waste.

中国快速工业化进程产生了海量废弃物，其中工业固体废物在过去二十年中贡献了最大的产生量，累计约达600亿吨（Gt）。然而，由于废弃物产生主体众多、统计覆盖范围不足，中国诸多地区仍缺乏完整的工业废弃物时空数据集。为填补这一数据缺口，本研究从数千个数据源收集了现有可获取的相关数据。在此基础上，我们开发了六种机器学习模型，以补全1990-2022年间中国全部337个城市的工业废弃物时空数据集。本研究采用贝叶斯优化（Bayesian optimization）方法，为每个城市筛选最优估算模型，以提升模型的性能与鲁棒性。除了工业废弃物总产生量外，本数据集还提供了2022年中国超半数城市的六大类主要工业废弃物子类别的产生量，具体包括冶金渣、粉煤灰、炉渣、煤矸石、尾矿及脱硫石膏。本数据集可助力科研人员与政策制定者认知并应对工业废弃物带来的各类挑战。