Evaluation of the Houston ozone modeling emissions inventory via inverse modeling with sensitivity coefficients from the decoupled direct method (DDM)

Effective management of urban photochemical oxidant pollution is heavily dependent on the availability of accurate inventories of volatile organic compounds (VOCs) and nitrogen oxide (NOx) emissions. Application of a variety of techniques for evaluating the accuracy of emission estimates is therefore an essential element of the air quality planning process. Recent advances in photochemical modeling of ozone episodes provide an opportunity to evaluate emission inventories via an inverse modeling technique in which the sensitivity of ozone formation to small changes in emissions (calculated via the Direct Decoupled Method) are used to reconcile emissions estimates with ambient air quality measurements. We have applied inverse modeling to the evaluation of an emissions inventory for Houston, Texas. The Texas Air Quality Study (TexAQS) in year 2000 developed a rich database of aircraft measurements for ozone and precursors (NOy, formaldehyde) during high ozone episodes. Modeling the TexAQS episode periods suggests that there are important differences between the emission inventory and actual emissions. We developed emission inventory adjustments for specific emissions source categories and geographic sub-regions in Houston under a fixed realization of meteorological fields. Results were checked by comparing revised model results based on adjusted inventories against both aircraft and surface based air quality measurements. Major findings from the inverse modeling were (1) an adjustment to industrial olefin emissions suggested by other analyses (under which olefins are scaled to NOx) appears to result in a near optimal adjustment of point source VOCs; (2) Model performance can be further improved by a downward adjustment in biogenic VOCs (the episode period was a drought, and the GloBEIS3 biogenic emissions model estimates lower biogenic VOC emissions if drought is accounted for); and (3) model performance can also be improved by increasing area/mobile source VOC, especially in combination with lowering biogenic VOC. These results demonstrate the value of inverse modeling as an emission inventory evaluation tool.