Eclipse: An End-to-End Platform for Low-Cost, Hyperlocal Environmental Sensing in Cities

An integrated cyberGIS and machine learning framework for fine-scale prediction of Urban Heat Island using satellite remote sensing and urban sensor network data

“…First, nodes with more powerful edge processors that can be customized with project-specific sensor packages are being deployed to replace AoT nodes in Chicago as part of a National Science Foundation Mid-Scale Research Infrastructure development project called SAGE (Beckman et al, 2019). Second, the AoT team partnered with Microsoft Research, JCDecaux, and the Environmental Law and Policy Center in 2021 to deploy 115 sensor nodes on bus shelters throughout Chicago, each measuring PM 2.5 , temperature, relative humidity, and multiple air pollutant gases (Daepp et al, 2022). Using these and other new data sources, the framework will be enhanced to pursue near real-time prediction of UHI, which is critical to help people living urban areas to be better prepared for extreme heat situations.…”

In situ drift correction for a low-cost NO2 sensor network

“…Several studies have already been working on network-specic solutions, such as reference site collocation, transfer standard collocation, remote network calibration, or a proxy model for this data quality issue, most of which recommended semi-regular adjustments to the calibrations to adjust for dri or seasonal bias. [9][10][11][12][13][14][15] This network management is an additional challenge on top of individual device calibration.…”

“…LCS calibration often involves calibration models that are trained and tested using collocated reference data. These calibration models can include linear regression, 9 multivariate linear regression, 16 support vector regression, 17 generalized additive models, 18 random forest models, 10 machine learning algorithms, 19 or hybrid models. 14 For electrochemical-based NO 2 LCSs, sensor drift, seasonal bias, ozone cross-interference, and sensor degradation are all issues that should be accounted for in the calibration.…”

“…In this work we were able to deploy 12 LCSs to locations with an O 3 EPA-approved federal equivalent method (FEM) instrument to ensure the highest accuracy in our correction for ozone interference. While other sensor network studies have used more sensors in their networks, they did not collocate them all together at a reference site, 9,10 by initially co-locating all 12 of our LCSs at a NO 2 FRM site we were able to quantify variability among sensors from the same manufacturing run. This allowed us to detect and correct outlier sensors with poor performance compared to the average sensor performance.…”

Analysis and Visualization of Hyperlocal Emissions Data in New Delhi and Mumbai

“…Moreover, the work in [4] presented a study to showcase trends in particulate air across small towns and villages in India. The work in [7] proposed Eclipse, a platform for sensing low-cost environmental trends using solar powered technologies. The work in [6] developed a machine learning framework to model air pollution using meteorological and tra c data points.…”