(7b) Long-Term Calibration Models to Predict Ozone Levels with a Metal Oxide Sensor | AIChE

(7b) Long-Term Calibration Models to Predict Ozone Levels with a Metal Oxide Sensor


Sayahi, T. - Presenter, University of Utah
Garff, A., University of Utah
Le, K., University of Utah
Becnel, T., University of Utah
Powell, K., The University of Utah
Gaillardon, P. E., University of Utah
Butterfield, A., University of Utah
Kelly, K., University of Utah, Assistant Professor
Ozone (O3) is a strong oxidant that is associated with adverse health effects. Low-cost o O3 sensors, such as metal oxide (MO) sensors, can complement regulatory O3 measurements and improve the spatiotemporal resolution of measurements. However, the quality of MO sensor data remains a challenge. The University of Utah has a network of low-cost air quality sensors (called AirU) that primarily measures PM2.5 levels around the Salt Lake City Valley (Utah, U.S.). The AirU package also has a low-cost MO sensor ($8) that measures oxidizing/reducing species. These MO sensors exhibited excellent laboratory response to O3 although they showed some intra-sensor variability. Field performance was assessed by placing eight AirUs at two Division of Air Quality (DAQ) monitoring stations with O3 federal equivalence methods for one year to develop long-term multiple linear regression (MLR) and artificial neural network (ANN) calibration models to predict O3 concentrations. Six sensors served as train/test sets. The remaining two sensors served as a holdout set to evaluate the applicability of the new calibration models in predicting O3 concentrations for other sensors of the same type. A rigorous variable selection method was also implemented by least absolute shrinkage and selection operator (LASSO), MLR and ANN models. The variable selection indicated that the MO oxidizing species and temperature from the AirU and solar radiation from DAQ were the most important variables. The MLR calibration model exhibited moderate performance (R2 = 0.491), and the ANN exhibited good performance (R2 = 0.767) to estimate the O3 concentrations of the holdout set. The ANN model was able to help address the intra-sensor variability challenge. These low-cost MO sensors combined with a long-term ANN calibration model can raise awareness about the utility in low-cost air quality sensors in understanding geospatial and temporal differences in O3.