Performance of artificial neural networks and linear models to quantify 4 trace gas species in an oil and gas production region with low-cost sensors

Testing the performance of field calibration techniques for low-cost gas sensors in new deployment locations: across a county line and across Colorado

“…We found that the best-performing model inputs and model type depended on circumstances associated with individual case studies, such as differing characteristics of local dominant emissions sources, relative timing of model training and application, and the extent of extrapolation outside of parameter space encompassed by model training. In agreement with findings from our previous study that was focused on data from a single location (Casey et al, 2018), ANNs remained more effective than LMs for a number of these case studies but there were some exceptions. For CO 2 models, exceptions included case studies in which training data collection took place more than several months subsequent to the test data period.…”

“…When the error Table 3. The best-performing models, as determined for each gas species, in the previous study (Casey et al, 2018). associated with the validation dataset is no longer being reduced, training stops early.…”

“…SJ Basin sampling sites spanned a range of elevations, including some that were higher and some that were lower than the BAO tower, coinciding with a wide range of atmospheric pressure at the distributed sampling sites. We began by testing the best-performing CO 2 model, as determined in our previous work (Casey et al, 2018), on data from this case study, during the summer of 2015. ANNs were trained for each U-Pod using data from the BAO tower with the following inputs from each U-Pod: eltCO2 (ELT S-300 CO 2 sensor), temp (temperature), and absHum (absolute humidity); the ANNs were then tested on data collected at the BAO tower and at sampling sites in the SJ Basin.…”

“…Public health concerns have arisen about the increasing number of people living alongside these industrial activities and emissions McKenzie et al, 2014;McKenzie et al, 2012McKenzie et al, , 2017. We previously developed methods to quantify ozone (O 3 ), carbon dioxide (CO 2 ), methane (CH 4 ), and carbon monoxide (CO) using low-cost gas sensors in an area where the ambient mole fractions of these species are influenced by oil and gas production activities (Casey et al, 2018). Such low-cost sensor measure-ments could enable greater understanding of air quality in oil and gas production basins, informing the spatial and temporal scales on which people live and work in a way that current technologies used by regulatory agencies cannot feasibly accomplish.…”

“…

We assessed the performance of ambient ozone (O 3 ) and carbon dioxide (CO 2 ) sensor field calibration techniques when they were generated using data from one location and then applied to data collected at a new location. This was motivated by a previous study (Casey et al, 2018), which highlighted the importance of determining the extent to which field calibration regression models could be aided by relationships among atmospheric trace gases at a given training location, which may not hold if a model is applied to data collected in a new location. We also explored the sensitivity of these methods in response to the timing of field calibrations relative to deployment periods.

…”