Incremental Reactivity Uncertainty in So Cal Philip Martien

Incremental Reactivity & Uncertainty in So. Cal Philip Martien, Robert Harley, et al (2003) Presented by Nasim Mullen & Zack Subin CE 218 C, 4/1/08

Motivation / Contribution n n Focus on VOCs with greatest impact on ozone Extends previous studies by including n n More species 3 D model Updated mechanisms Extensive uncertainty analysis

Incremental Reactivity n n Sensitivity of ozone to emissions of a particular VOC Absolute Incremental Reactivity (AIR) Relative Incremental Reactivity (RIR) Maximum Incremental Reactivity (MIR)

Approach n Mechanism: 104 species & 246 rxns. n n n 3 D Decoupled Direct Method Several trajectory runs estimate 2 nd order sensitivities / uncertainties n n 8 lumped species: 3 alkane, 2 aromatic, 2 alkene, and 1 terpene. Monte Carlo w/ Latin Hypercube Sampling June 1987 Pollution Episode

3 D Decoupled Direct Method n 3 D version of simultaneous, time-dependent sensitivity method presented in class ci = concentration (of species i), u = wind, K = diffusivity, R = net chemical production, E = emissions, J = ? , s*i, j = semi-norm. sensitivity of species i concentration with respect to species j, εj = scaling factor for emissions of j

Spatial Dependence n 8 sites n n n 2 coastal (Hawthorne & Long Beach) 2 central (downtown LA & Anaheim) 4 inland (Burbank, Azusa, Claremont and Rubidoux) Ozone increases going inland Anthropogenic emissions peak in central locations Biogenic emissions peak downwind of urban sites

Modeling Domain

Uncertainty Analysis n n Sensitivities of base case ozone concentrations and incremental reactivities 33 uncertain input parameters n n n Rates, e. g. NO 2 photolysis, NOx deposition Emissions, e. g. NOx & NMVOCs from vehicles Incorporated as multiplicative factors n Lognormal distribution with mean 1

Uncertainty Analysis (cont’d) n n Multivariate linear regression analysis applied to Monte Carlo results to identify influence of input variables 1 -D trajectory version of the air quality model, with trajectories ending at several sites at time of maximum observed ozone

Results

(Selections from) Results n AIR much higher inland than on the coast n n Alkanes not very reactive; alkenes more reactive than previous 1 D model n n Inland sites downwind of most emissions Low level of radicals inland alkenes that produce radicals had big effect Biogenic alkenes less influential n Mostly emitted downwind of sites

More Results HCHO RIR varied widely across sites Decreased moving inland n n Higher at sites where low radical concentrations limit O 3 production, because HCHO can initiate radical formation Benzaldehyde RIR < 0, because of NOx sink

More Results n n Normalizing reactivities reduced variability across locations AIR sensitivity to NOx > 0 in NOx-limited locations, e. g. Claremont n n Meanwhile, sensitivity to radical-increasing parameters is negative AIR sensitivity to NOx < 0 in NOxsaturated locations, e. g. Azusa n But positive for radical-increasing parameters

More Results n 2 sources of error n n n Variability due to environmental conditions Uncertainty due to model inputs In this case, comparable in magnitude for most cases the box model results were about as useful for evaluating IR as the more complex models One of the exceptions: coastal regions have higher RIR than inland regions for C 4 and C 5 alkanes and carbonyls

More Results n “This study supports previous studies that suggest that using relative rather than absolute reactivity metrics can reduce both variability and uncertainty. ”