Discovering Localized Pollution Hotspots Using Sparse Sensor Measurements
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Keywords: Spatial Interpolation, Air Pollution, Hotspots, Sparse Data, Sensor Networks
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Abstract: Air pollution is one of the key challenges in sustainable urban development and
its management consists of two aspects, monitoring and governance policy. Most
cities only deploy a sparse coarse-resolution sensor network for air pollution mon-
itoring due to the high costs of industrial-grade gold standard air quality sensors.
Similarly, most city governments tend to focus on more heavily polluted areas of
the city, the so-called “hotspots”, due to a lack of budget. In this paper, we first
show that the space-time resolution of the public sensor network is insufficient to
detect and analyze hotspots effectively. Taking New Delhi as our study area, we
augmented the government network with 28 low-cost sensors that monitor PM 2.5
concentrations and collected pollution data over a 30-month period, from May 1,
2018, to Nov 1, 2020. Applying established definitions of hotspots to this data, we
show the existence of additional hotspots that are missed by the public network.
Going further, we show that using machine-learning-based spatial interpolation
methods, like Gaussian Process regression, on the sparse sensor data allows us
to find these hotspots in a manner that is robust to sensor failures and is extensi-
ble to new spatial locations that don’t have sensors deployed. We also compare
this approach with traditional dispersion modeling and provide the pros and cons
of both approaches. Finally, we present the findings of our analysis and provide
recommendations for changes to the New Delhi government’s current policy.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 3869
Loading