Norwegian Institute for Water Research (NIVA) , Gaustadalléen 21 , NO-0349 Oslo , Norway.
Department of Biostatistics, Institute of Basic Medical Sciences , University of Oslo , Oslo , Norway.
Environ Sci Technol. 2019 Feb 19;53(4):1994-2001. doi: 10.1021/acs.est.8b05389. Epub 2019 Jan 29.
Modeling and prediction of a city's (Oslo, Norway) daily dynamic population using mobile device-based population activity data and three low cost markers is presented for the first time. Such data is useful for wastewater-based epidemiology (WBE), which is an approach used to estimate the population level use of licit and illicit drugs, new psychoactive substances, human exposure to a wide range of pollutants, such as pesticides or phthalates, as well as the release of endogenous substances such as oxidative stress and allergen biomarkers. Comparing WBE results between cities often requires normalization to population size, and inaccuracy in the measured population can introduce high levels of uncertainty. In this study mobile phone data from 8-weeks in 2016 was used to train three linear models based on drinking water production, electricity consumption and online measurements of ammonium in wastewater. The ammonium model showed the best correlation with R = 0.88 while drinking water production and electricity consumption showed more discrepancies. The three models were then re-evaluated against 5-week of mobile phone data from 2017 showing mean absolute errors <10%. The ammonium-based estimated mean annual population for Oslo in 2017 was 645 000 inhabitants, 4% higher than the "de jure" population reported by the wastewater treatment plant. Due to changing conditions and seasonality, drinking water production underestimated the population by 27% and electricity consumption overestimated the population by 59%. Therefore, the results of this work showed that the ammonium mass loads can be used as an anthropogenic proxy to monitor and correct the fluctuations in population for a specific catchment area. Furthermore, this approach uses a simple, yet reliable indicator for population size that can be used also in other areas of research.
首次提出了一种使用基于移动设备的人口活动数据和三种低成本标志物对城市（挪威奥斯陆）日常动态人口进行建模和预测的方法。这种数据可用于基于污水的流行病学（WBE），这是一种用于估计人口水平使用合法和非法药物、新精神活性物质、人类接触各种污染物（如农药或邻苯二甲酸盐）以及内源性物质（如氧化应激和过敏原生物标志物）释放的方法。比较城市之间的 WBE 结果通常需要归一化为人口规模，而测量人口的不准确会引入高水平的不确定性。在这项研究中，使用 2016 年 8 周的手机数据来训练三种基于饮用水生产、电力消耗和污水中铵在线测量的线性模型。铵模型与 R = 0.88 的相关性最好，而饮用水生产和电力消耗的差异较大。然后，将这三种模型重新评估 2017 年 5 周的手机数据，平均绝对误差<10%。基于铵的 2017 年奥斯陆估计年平均人口为 645000 人，比污水处理厂报告的“法定”人口高 4%。由于条件变化和季节性，饮用水生产低估了 27%的人口，电力消耗高估了 59%的人口。因此，这项工作的结果表明，铵质量负荷可用作监测和校正特定集水区人口波动的人为代理。此外，这种方法使用一种简单、可靠的人口规模指标，也可用于其他研究领域。
