Physically-informed probabilistic modelling of air pollution in Kampala using a low cost sensor network

使用低成本传感器网络对坎帕拉空气污染进行基于物理的概率建模

• 批准号: EP/T00343X/1
• 负责人: Richard Wilkinson
• 金额: $ 52.14万
• 依托单位: University of Sheffield
• 依托单位国家: 英国
• 项目类别: Research Grant
• 财政年份: 2019
• 资助国家: 英国
• 起止时间: 2019 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=EP%2FT00343X%2F1
• 关键词: Physically; informed; probabilistic; modelling; air

Ambient air pollution is estimated to contribute to over three million premature deaths each year. Particulate matter (PM) pollution in particular is a likely contributor to this toll. Unfortunately there is only limited monitoring of air pollution in Sub-saharan Africa, in part because accurate monitoring equipment is too expensive, making it hard to develop or assess policy at national and local level. Low-cost particulate sensors are available, but their limited accuracy means that the data cannot be used reliably without correction. This project will test the hypothesis that when used in combination with a reference instrument and combined with physical insight, low-costs sensor networks can be used to produce models to accurately predict PM, gain insight, and plan policy. We focus on Kampala, where the project team have built a low-cost sensor network over the previous four years. Kampala is a rapidly growing city with persistent dangerous levels of particulate pollution, which regularly exceeds ten-times the WHO's guideline annual mean limit. Many factors contribute to this, including Kampala's geography, its partly unmetalled road network, and activities such as domestic burning of garbage and cooking on solid fuel stoves.Aims and Objectives: The project team have previously installed a low-cost sensor network, and provide predictions of pollution across the city using a mathematical model known as a Gaussian process. This type of model only uses correlations between measurements, which means that external inputs, such as wind-direction, are not properly handled. Moreover, this type of model can't be used to anticipate the effect of an intervention (for example modelling the impact of a road closure), as this involves extrapolating outside of the training data. We have previously worked with the Kampala Capital City Authority (KCCA) to install fifty sensors across the city, and in this project, we will work with them to develop possible interventions to improve air quality, model their potential impact, and then measure their effectiveness.The project's mathematical aims are specifically around the development of a new modelling paradigm for models of space and time, and the challenges these pose for training the models on observational data. The purpose is threefold. Firstly, they will allow us to include realistic approximations of physical processes, such as the movement of pollution around a city. Secondly, they will let us work out what is producing the pollution, where and when. Thirdly, they will help the KCCA answer "what if?" questions, e.g. "What if we close Luwum Street to motor traffic?" The models predictions must also report their confidence, so that the KCCA and others know if the results can be trusted.Applications and benefits: Even small improvements in air quality in Kampala would improve the health of its population. By providing policy makers and civil society with the tools for making predictions, we will enable them to plan and assess policy interventions to improve air quality. We anticipate considerable international impact, first through implementation by city authorities in neighbouring countries. Second, by supporting academic research in the field. And third, by supporting the development of practical interventions such as cleaner fuels and support active travel and other issues around 'double burden'.In summary, the project will lead to considerable high-impact improvements in quality-of-life associated with improved air quality. The Kampala Capital City Authority (KCCA), the local government and civil authority for Kampala, have the potential take action to achieve improvements in air quality. But they lack the information and evidence to make or motivate policy decisions in this domain. This project will provide the data, packaged and presented in a clear and actionable manner, in a format and context most useful to policy makers.

据估计，环境空气污染每年导致300多万人过早死亡。颗粒物(PM)污染尤其可能是造成这一损失的原因之一。不幸的是，撒哈拉以南非洲对空气污染的监测有限，部分原因是准确的监测设备过于昂贵，难以制定或评估国家和地方一级的政策。低成本的颗粒传感器是可用的，但它们的精度有限，这意味着如果不进行校正，数据就无法可靠地使用。该项目将测试这样一个假设，即当与参考仪器结合使用并与物理洞察相结合时，低成本传感器网络可以用于生成模型，以准确预测PM、获得洞察和规划政策。我们把重点放在坎帕拉，项目团队在过去四年里在那里建立了一个低成本的传感器网络。坎帕拉是一个快速发展的城市，颗粒物污染持续处于危险水平，经常超过世卫组织指导性年平均限值的十倍。造成这一现象的因素很多，包括坎帕拉的地理位置、部分非金属的道路网络，以及家庭焚烧垃圾和在固体燃料炉上做饭等活动。目的和目标：该项目团队以前安装了一个低成本的传感器网络，并使用一种称为高斯过程的数学模型来预测全市的污染情况。这种类型的模型只使用测量之间的相关性，这意味着没有适当地处理外部输入，如风向。此外，这种类型的模型不能用于预测干预的效果(例如，对道路封闭的影响进行建模)，因为这涉及到在培训数据之外进行外推。我们之前与坎帕拉首都城市管理局(KCCA)合作，在全市安装了50个传感器，在这个项目中，我们将与他们合作，开发可能的干预措施来改善空气质量，对其潜在影响进行建模，然后测量其有效性。该项目的数学目标明确地围绕着为空间和时间模型开发新的建模范式，以及这些挑战给根据观测数据训练模型带来了挑战。其目的有三个方面。首先，它们将允许我们包括对物理过程的现实近似，例如污染在城市周围的移动。其次，他们将让我们找出是什么在何时何地产生了污染。第三，他们将帮助KCCA回答“如果呢？”问题，例如“如果我们关闭卢武姆街，禁止机动车通行呢？”模型预测还必须报告其置信度，以便KCCA和其他机构知道结果是否可信。应用和好处：即使坎帕拉空气质量的微小改善也会改善人口的健康。通过为决策者和民间社会提供预测工具，我们将使他们能够规划和评估改善空气质量的政策干预措施。我们预计会产生相当大的国际影响，首先是通过邻国城市当局的执行。第二，支持该领域的学术研究。第三，通过支持实际干预措施的发展，如更清洁的燃料和支持积极出行和其他有关“双重负担”的问题。总而言之，该项目将导致与改善空气质量相关的生活质量的显著提高。坎帕拉首都城市管理局(KCCA)是坎帕拉的地方政府和民政当局，有可能采取行动改善空气质量。但他们缺乏信息和证据来制定或推动这一领域的政策决策。该项目将以政策制定者最有用的格式和背景，以清晰和可操作的方式提供打包和列报的数据。

项目成果

AI-driven environmental sensor networks and digital platforms for urban air pollution monitoring and modelling

人工智能驱动的环境传感器网络和数字平台，用于城市空气污染监测和建模

• DOI: 10.1016/j.socimp.2024.100044
• 发表时间: 2024
• 期刊: Societal Impacts
• 作者: Bainomugisha E

Using a Network of Locally Developed Low Cost Particulate Matter Sensors for Land Use Regression Modeling of PM2.5 in Urban Uganda

• DOI: 10.20944/preprints202006.0158.v1
• 发表时间: 2020-06
• 作者: E. Coker;Ssematimba Joel

Adjoint-aided inference of Gaussian process driven differential equations

高斯过程驱动微分方程的伴随辅助推理

• DOI: 10.48550/arxiv.2202.04589
• 发表时间: 2022
• 作者: Gahungu P

Air pollution and mobility patterns in two Ugandan cities during COVID-19 mobility restrictions suggest the validity of air quality data as a measure for human mobility.

• DOI: 10.1007/s11356-022-24605-1
• 发表时间: 2023-03
• 期刊: ENVIRONMENTAL SCIENCE AND POLLUTION RESEARCH
• 影响因子: 5.8
• 作者: Galiwango, Ronald;Bainomugisha, Engineer;Kivunike, Florence;Kateete, David Patrick;Jjingo, Daudi
• 通讯作者: Jjingo, Daudi

Multi-task Causal Learning with Gaussian Processes

• 发表时间: 2020-09
• 期刊: ArXiv
• 作者: Virginia Aglietti;T. Damoulas;Mauricio A Álvarez;Javier Gonz'alez