Volatile Organic Compound Prediction Bias in a Regulatory Air Quality Model in Houston, Texas

High Resolution Image Download MS PowerPoint Slide This study evaluates the performance of a regulatory air quality model in simulating volatile organic compounds (VOCs) by comparing observed concentrations with public and private surface monitoring data. Significant negative biases were found, with normalized mean bias (NMB) values ranging from −33.18% for benzene to −71.62% for ethane, particularly during nighttime hours (10 p.m. to 6 a.m.). Median observed concentrations at monitors ranged from 0 to 6.84 ppb for ethane, 1.48–2.87 ppb for propane, 0–0.37 ppb for benzene, and 0.39–1.33 ppb for ethene, while median modeled concentrations ranged from 1.29 to 2.21 ppb, 0.56–1.47 ppb, 0.08–0.48 ppb, and 0.34–1.23 ppb for those same species. Trajectory analyses on high ozone days revealed air masses moving from industrial regions east of downtown Houston to downwind monitors registering large VOC discrepancies, suggesting potential errors in the modeled emissions inventory. These errors may stem from the inventories’ reliance on annual emission totals, average ozone season daily emission rates, or averaged temporal profiles that do not capture short-term variability. Additional errors may also stem from incomplete VOC speciation, and the reliance of averaged VOC speciation profiles in modeling inventory development. Ethane provides a unique speciation and quantification challenge because it is classified as an exempt VOC by the EPA and therefore is not required to be reported by industry. To assess ozone formation potential (OFP), the study incorporated both VOC abundance and reactivity. Despite incomplete VOC measurements, observed OFP consistently exceeded modeled OFP, indicating underrepresentation of reactive VOCs in the model. This discrepancy underscores limitations in the emissions inventory and model representation of VOCs. A sensitivity analysis doubling VOC emissions showed improved model performance, particularly for benzene and ethene, though alkane biases remained relatively unchanged. Overall, the findings highlight the need for refined anthropogenic VOC emissions data from industrial sources found in Houston, with emphasis on temporal resolution and speciation of alkanes and aromatics. Improving the modeled emission inventory is critical for enhancing model accuracy and understanding VOC-driven ozone formation.