Environmental Insights: Democratizing Access to Ambient Air Pollution Data and Predictive Analytics with an Open-Source Python Package

TL;DR

Environmental Insights delivers an open-source Python package that democratizes access to ambient air pollution data and provides a lightweight ML-based forecasting workflow. It furnishes high-resolution England data at hourly 1km^2 and global hourly 0.25° resolutions, along with UK Daily Air Quality Index derivations and prediction intervals to support risk-informed decisions. The package emphasizes two-way stakeholder engagement through visualizations, scenario testing, and OpenStreetMap context, enabling soft and hard interventions in policy and planning. By lowering technical barriers and coupling data with interactive tools, it empowers a broad audience to explore air pollution futures and advocate for meaningful improvements.

Abstract

Ambient air pollution is a pervasive issue with wide-ranging effects on human health, ecosystem vitality, and economic structures. Utilizing data on ambient air pollution concentrations, researchers can perform comprehensive analyses to uncover the multifaceted impacts of air pollution across society. To this end, we introduce Environmental Insights, an open-source Python package designed to democratize access to air pollution concentration data. This tool enables users to easily retrieve historical air pollution data and employ a Machine Learning model for forecasting potential future conditions. Moreover, Environmental Insights includes a suite of tools aimed at facilitating the dissemination of analytical findings and enhancing user engagement through dynamic visualizations. This comprehensive approach ensures that the package caters to the diverse needs of individuals looking to explore and understand air pollution trends and their implications.

Figures (8)

• Figure 1: DAQI sums across all grids of the typical days of air pollution. Each of the five air pollutants NO$_2$, O$_3$, PM$_{10}$, PM$_{2.5}$, SO$_2$ each of an assignment of a given index or band for each location, with the highest of the five determining the overall DAQI for that locations. Figure \ref{['fig:drivingAirPollutants']} shows how, for rural locations, O$_3$ is driving the air quality, whereas in the urban regions, particularly London, the relationship is more complex.
• Figure 2: DAQI sums across all grids of the typical days of air pollution. Similarly to Figure \ref{['fig:aqiBandExample']}, where spatially different air pollutants drive poor air quality, the same can be true temporally. For example, the high AQI sum on Sunday is driven by O$_3$, a well-observed phenomena sicard:2020:OzoneWeekend. In contrast, the peaks in the later parts of each day are likely more complex, driven by potentially NO$_2$ concentrations accumulating from rush hour or O$_3$ production from a day of intense sunshine.
• Figure 3: Chesterfield Loundsley Green Road NO$_x$ Emissions. The AURN measurements for the Chesterfield Loundsley Green Road monitoring station are shown alongside a range of different model predictions for air pollution concentrations for NO$_x$. Of particular note is the size of the 90% interval, where the predictions for the model mean are broadly similar between 2015-03-15 and 2013-03-18; however, the prediction interval provides considerable further assurance of the prediction. The prediction interval highlights that the predictions for 2013-03-18 are considerably less certain.
• Figure 4: Prediction Interval examples for full spatial map of England. Further to individual locations, the prediction interval provides insight into the locations in which the model is less sure about predictions, with the size of the interval denoting a more extensive range of possible values a prediction can take.
• Figure 5: Comparison of future states of O$_3$ air pollution concentrations across Greater London, when decreasing wind by 20%. Decreasing wind speed highlights the relationship between O$_3$ and wind, assuming all other variables remain the same. The models allow for quick computation of future hypothetical scenarios depending on the user's desires, allowing for more complex scenarios, such as changing wind speed, rainfall, and transportation.