Sensor and Analysis
The PAM (Personal Air Quality Monitor) from Atmospheric Sensors is being used for our 2021 and 2023 measurement campaign. This measures fine dust particles (PM1, PM 2.5, PM 10), gas concentrations (NO2, NO, CO, O3), noise levels, temperature and humidity. The GPS position of the measurement and acceleration values are also saved. At the end of a measurement day, the sensor is placed in the charging station. The collected data is then sent to UFZ server.
Analysis
Calibration
The calibration phase involved all 12 sensors, which were meticulously prepared for a pollution measurement campaign. These sensors were placed in an UFZ room for a period of two days, exposing them to consistent levels of ambient environmental conditions and pollution. This step ensured that each sensor was aligned to accurately reflect similar pollution levels. To enhance the precision of each sensor's readings, a correction factor was applied to adjust their outputs, thereby reducing the deviation from the true value.
Modelling
Following the calibration, the focus shifted to the modelling stage. Utilizing the corrected sensor data as a foundation, individual predictive models were developed for each sensor using the XGBoost algorithm. These models were specifically trained to predict pollution levels throughout the entire measurement campaign, based on the initial calibration data. This approach allowed for a refined and accurate assessment of exposure levels, utilizing the groundwork laid during the calibration process
