A Gap Analysis of Low-Cost Outdoor Air Quality Sensor In-Field Calibration

Applications of low-cost sensing technologies for air quality monitoring and exposure assessment: How far have they gone?

Abstract: Over the past decade, a range of sensor technologies became available on the market, enabling a revolutionary shift in air pollution monitoring and assessment. With their cost of up to three orders of magnitude lower than standard/reference instruments, many avenues for applications have opened up. In particular, broader participation in air quality discussion and utilisation of information on air pollution by communities has become possible. However, many questions have been also asked about the actual benefits of these technologies. To address this issue, we conducted a comprehensive literature search including both the scientific and grey literature. We focused upon two questions: (1) Are these technologies fit for the various purposes envisaged? and (2) How far have these technologies and their applications progressed to provide answers and solutions? Regarding the former, we concluded that there is no clear answer to the question, due to a lack of: sensor/monitor manufacturers' quantitative specifications of performance, consensus regarding recommended end-use and associated minimal performance targets of these technologies, and the ability of the prospective users to formulate the requirements for their applications, or conditions of the intended use. Numerous studies have assessed and reported sensor/monitor performance under a range of specific conditions, and in many cases the performance was concluded to be satisfactory. The specific use cases for sensors/monitors included outdoor in a stationary mode, outdoor in a mobile mode, indoor environments and personal monitoring. Under certain conditions of application, project goals, and monitoring environments, some sensors/monitors were fit for a specific purpose. Based on analysis of 17 large projects, which reached applied outcome stage, and typically conducted by consortia of organizations, we observed that a sizable fraction of them (~ 30%) were commercial and/or crowd-funded. This fact by itself signals a paradigm change in air quality monitoring, which previously had been primarily implemented by government organizations. An additional paradigm-shift indicator is the growing use of machine learning or other advanced data processing approaches to improve sensor/monitor agreement with reference monitors. There is still some way to go in enhancing application of the technologies for source apportionment, which is of particular necessity and urgency in developing countries. Also, there has been somewhat less progress in wide-scale monitoring of personal exposures. However, it can be argued that with a significant future expansion of monitoring networks, including indoor environments, there may be less need for wearable or portable sensors/monitors to assess personal exposure. Traditional personal monitoring would still be valuable where spatial variability of pollutants of interest is at a finer resolution than the monitoring network can resolve.

From air quality sensors to sensor networks: Things we need to learn

Abstract: As a potential complement to traditional regulatory instruments, low-cost air quality sensors (LCAQS) can be deployed in dense monitoring networks to provide timely and comprehensive snapshots of pollutant concentrations and their spatial and temporal variability at various scales with relatively less cost and labor. However, a lack of practical guidance and a limited understanding of sensor data quality hinder the widespread application of this emerging technology. We leveraged air quality data collected from state and local monitoring agencies in metropolitan areas of the United States to evaluate how low-cost sensors could be deployed across the U.S. We found that ozone, as a secondary pollutant, is more homogeneous than other pollutants at various scales. PM2.5, CO, and NO2 displayed homogeneities that varied by city, making it challenging to design a uniform network that was suitable across geographies. Our low-cost sensor data in New York City indicated that PM2.5 sensors track well with light-scattering reference methods, particularly at low concentrations. The same phenomenon was also found after thoroughly evaluating sensor evaluation reports from the Air Quality Sensor Performance Evaluation Center (AQ-SPEC). Furthermore, LCAQS data collected during wildfire episodes in Portland, OR show that a real-time (i.e. in situ) machine learning calibration process is a promising approach to address the data quality challenges persisting in LCAQS applications. Our research highlights the urgency and importance of practical guidance for deploying LCAQS.

Greedy function approximation: A gradient boosting machine.

Abstract: Function estimation/approximation is viewed from the perspective of numerical optimization in function space, rather than parameter space. A connection is made between stagewise additive expansions and steepest-descent minimization. A general gradient descent “boosting” paradigm is developed for additive expansions based on any fitting criterion.Specific algorithms are presented for least-squares, least absolute deviation, and Huber-M loss functions for regression, and multiclass logistic likelihood for classification. Special enhancements are derived for the particular case where the individual additive components are regression trees, and tools for interpreting such “TreeBoost” models are presented. Gradient boosting of regression trees produces competitive, highly robust, interpretable procedures for both regression and classification, especially appropriate for mining less than clean data. Connections between this approach and the boosting methods of Freund and Shapire and Friedman, Hastie and Tibshirani are discussed.

XGBoost: A Scalable Tree Boosting System

Abstract: Tree boosting is a highly effective and widely used machine learning method. In this paper, we describe a scalable end-to-end tree boosting system called XGBoost, which is used widely by data scientists to achieve state-of-the-art results on many machine learning challenges. We propose a novel sparsity-aware algorithm for sparse data and weighted quantile sketch for approximate tree learning. More importantly, we provide insights on cache access patterns, data compression and sharding to build a scalable tree boosting system. By combining these insights, XGBoost scales beyond billions of examples using far fewer resources than existing systems.

Particulate Matter Air Pollution and Cardiovascular Disease An Update to the Scientific Statement From the American Heart Association

Abstract: In 2004, the first American Heart Association scientific statement on “Air Pollution and Cardiovascular Disease” concluded that exposure to particulate matter (PM) air pollution contributes to card...

Artificial neural networks: a tutorial

Abstract: Artificial neural nets (ANNs) are massively parallel systems with large numbers of interconnected simple processors. The article discusses the motivations behind the development of ANNs and describes the basic biological neuron and the artificial computational model. It outlines network architectures and learning processes, and presents some of the most commonly used ANN models. It concludes with character recognition, a successful ANN application.