Data Overview

This data set contains data from two Captor nodes. These are prototype nodes that were developed at the Universitat Politècnica de Catalunya (UPC) to study the effects of the sensor data-gathering process on the use of low-cost sensors for air quality monitoring. Specifically, the captor nodes have tropospheric ozone, nitrogen dioxide, and nitrogen monoxide electrochemical sensors. They also have a temperature and relative humidity sensor inside the box. Two Captor nodes were placed in a reference station in Barcelona (Spain) for four months (January 2021 to May 2021), at a sampling frequency of 0.5 Hz. The data set contains a "readme" file with a brief description of the five files that make up the data set.

The dataset contains: Raw: NO2, O3, and NO in quantum steps of a 10-bit A/D converter (values between 0 and 1024, with 0 equivalent to 0 Volts and 1024 equivalent to 1.1 Volts), temperature in degree C, relative humidity in %.

The data set corresponds to samples from January 16 at 00:00 to May 15 at 00:00. There is a data gap of 3 days, from 2021-03-13 00:00:00 to 2021-03-15 23:59:59.

The data description can be found at https://www.sciencedirect.com/science/article/pii/S2352340922007934#fig0001 and downloaded at https://zenodo.org/records/5770589.

The ground truth air quality data are from an authorized air reference station in Palau Reial, Barcelona. The air quality data at the same time of sensor measurement are downloaded from: https://mediambient.gencat.cat/es/05_ambits_dactuacio/atmosfera/qualitat_de_laire/vols-saber-que-respires/descarrega-de-dades/descarrega-dades-automatiques/

Value of the Data

High-frequency measurements of low-cost sensors located in situ at a reference station, deployed by authorities, are useful for conducting sensor calibration studies. In addition, having the raw sensor data available at such a high frequency allows investigation of different techniques related to sampling, filtering, and further analysis of the raw sensor signals.

• The data from these electrochemical sensors comprise measurements of three different pollutants (O3 , NO2 and NO), with the actual concentrations measured by the reference station being available. Hence, the main beneficiaries are researchers specialized in the use of low-cost sensors for the measurement of air pollution, those who focus on the calibration of these sensors, and their possible use in regulated air pollution monitoring networks.

• Most air pollution data published in repositories are post-processed data with a sampling resolution of half an hour or one hour. Using the data presented in this paper, researchers can perform studies that require a high sensor sampling frequency. These data, allow studying the impact of sampling frequency on calibration, the impact of the type of aggregation used for calibration, and even the use of signal filtering techniques to improve sensor calibration. Therefore, as they are not pre-processed or already aggregated data, they can be of great interest for future studies that require a high sampling frequency.

• Reference stations give very accurate data, but they are expensive and are used with low spatial resolution. The use of low-cost air pollution sensors, although they cannot be used to generate alarms, can be used to generate social awareness, and as an indication that there is more pollution than desired in certain locations. The presented data set will help researchers and engineers, who want to deploy their low-cost sensor nodes, to test various calibration techniques, with varying sampling frequencies, thus allowing them to better learn how to adjust the parameters of the nodes they deploy.