What do correlations tell us about anthropogenic-biogenic interactions in the Sacramento plume during CARES?

 

Authors

Larry Kleinman — Brookhaven National Laboratory
Chongai Kuang — Brookhaven National Laboratory
Arthur J Sedlacek — Brookhaven National Laboratory
Gunnar I. Senum — Brookhaven National Laboratory
Stephen R. Springston — Brookhaven National Laboratory
Jian Wang — Washington University in St. Louis
Qi Zhang — University of California, Davis
John E Shilling — Pacific Northwest National Laboratory
Rahul Zaveri — Pacific Northwest National Laboratory

Category

Aerosol Properties

Description

During the Carbonaceous Aerosols and Radiative Effects Study (CARES) field campaign the G-1 aircraft was used to sample aerosol and gas phase compounds upwind, over, and downwind of Sacramento. We present data from 13 flights in which the wind direction was from the southwest. Our data set is further restricted to be from the boundary layer on transects perpendicular to the wind direction. There were a total of 66 transects. Our objective is to empirically determine the fraction of organic aerosol (OA) that can be attributed to anthropogenic and biogenic sources. Of particular interest is the question of whether the simultaneous presence of anthropogenic and biogenic precursors leads to enhanced concentrations of OA as has been found by Setyan et al. (2012) and Shilling et al. (2013).

CO and MVK+MACR were used as tracers of anthropogenic and biogenic emissions. MVK+MACR has a short atmospheric lifetime, so that its presence only explicitly addresses biogenic inputs to an air mass over a few-hour time span. It is, however possible, that elevated concentrations of MVK+MACR are indicative of meteorological conditions such as temperature, sunlight, ventilation, and wind direction that are favorable for the occurrence and accumulation of biogenic volatile organic compounds (VOCs).

Three sets of calculations were performed. First is a correlation/regression analysis testing the relation between OA and various combinations of CO and MVK+MACR, treating each of the 66 transects as a separate data set. The average R^2 for OA vs. CO is 0.63. In the bi-linear regression [OA]s = b1[CO]s +b2[MVK+MACR]s, where "s" indicates a standardized variable, R^2 = 0.69 and the average ratio b1/b2 shows that CO has approximately 10 times the weight of MVK+MACR in explaining OA. Adding an A-B interaction term, b3[CO]s[MVK+MACR]s to the bi-linear regression yields R^2= 0.72, b1/b2 still >10, and on average a negative sign for b3. Second, perturbations (Deltas) were defined for each transect as the 90th percentile of concentration minus the 10th percentile. Correlations were done amongst transects, thereby bringing in flight to flight concentration variations. In regressions with a single independent variable Delta CO, MVK+MACR or ozone could explain, respectively, 69%, 14%, or 85% of the variance of Delta OA. Third, the transect Deltas were split into nine subsets. The subset having high Delta CO and high DeltaMVK+MACR had a greater DeltaOA than would be predicted based on trend lines from the other data sets. Our third calculation differs only in detail from those published.