Adaptive spatial sampling design for urban air quality mapping

Authors: Eun-hye Yoo*, University At Buffalo (SUNY)
Topics: Geography and Urban Health
Keywords: adaptive geostatistical designs, air quality, low-cost portable sensor
Session Type: Paper
Day: 4/10/2018
Start / End Time: 12:40 PM / 2:20 PM
Room: Balcony N, Marriott, River Tower Elevators, 4th Floor
Presentation File: No File Uploaded


Recent advancements in sensing technology enable investigators and communities to collect air quality measurements at a time and location of interest. The low-cost and real time sensing capability of portable sensors increase spatial and temporal resolution and data availability, although the reliability and quality of measurements often need improvement. Similarly, computer simulation models, such as community multiscale air quality models (CMAQ), and remote sensing provide information on air quality with exhaustive spatial coverage, whereas these data are not directly compatible to ground air quality measurements.
In this paper, we investigated the optimal sampling locations for low-cost sensor measurements using adaptive geostatistical designs (AGD). The AGD enables the collection of data over time to depend on information obtained from previous information to optimize data collection toward the analysis objective. To illustrate our point, we assessed the contribution of portable sensor data on the quality of reconstructed the PM2.5 concentration surface at a fine scale using a simulation example. The three data sets --- 24-hr PM2.5 measurements from four fixed monitoring stations, CMAQ modelled values of PM2.5 at a grid at 12 km resolution, and portable sensor measurements collected from two sampling scheme (AGD and a randomly selected points and time instants)---were generated from realistically simulated daily PM2.5 concentration surfaces in Erie and Niagara counties in New York States, US, for 30 days. We assessed the effects of AGD by comparing the prediction accuracy of urban scale air quality prediction to the results of random sampling techniques.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login