Comment on “The 2003 North American electrical blackout: An accidental experiment in atmospheric chemistry” by L. T. Marufu et al.

[1] In order to assess the air-quality impact of the 14 Aug. 2003 blackout in the United States and Canada, Marufu et al. [2004] compare air-quality measurements over Selinsgrove, Pennsylvania on 15 Aug. 2003 to concentrations measured over the same site during reportedly similar synoptic weather conditions on 4 Aug. 2002. From these comparisons, Marufu et al. found that the SO2 and O3 mixing ratios and particle light scattering near the surface at Selinsgrove were lower in 2003 than in 2002 by >90%, ∼50% and ∼70%, respectively. The authors report measurements of CO and particle light absorption, provide results of back trajectory models, and estimate the influence of reductions in power plants emissions on air quality at Selinsgrove. Ultimately, the authors conclude that the decreased concentrations are “predominantly due to reductions in power-plant emissions hundreds of km upwind of the study area” and, through forward-trajectory analysis, conclude that “these improvements in air quality benefited much of the eastern U.S.” [2] Clearly, the decrease in power-plant emissions lowered concentrations of directly emitted species, i.e., SO2 and NOx, from levels that would have been experienced otherwise. It may be reasonable to expect regional improvements in O3 and light extinction from levels observed during seemingly similar meteorological regimes in the past as a result. However, a quantitative estimate of the effect of the 2003 blackout – including the impact on SO2, NOx, secondary species such as O3 and sulfate, and visibility – cannot be estimated solely from the data presented in the manuscript. Even in a qualitative sense, a review of the manuscript indicates that there are problems with the representativeness of the aircraft data that should be taken into account before interpreting the Selinsgrove data and inferring the impact of the blackout on air quality. [3] The most important limitation of the manuscript is the failure to consider the variability associated with concentrations of atmospheric species. Marufu et al. [2004] state that synoptic weather patterns were fairly similar in the Selinsgrove region during the measurement flight in 2002 and the blackout measurement flight of 2003. However, this does not necessarily mean that concentrations of atmospheric species should be similar. For example, a comparison of observed O3 maps from EPA's AirNow archives shows that O3 levels during 4 Aug. 2002 were markedly higher over Pennsylvania, Maryland and New Jersey than during 14 Aug. 2003, i.e., the day of the onset of the blackout – a day that was also synoptically similar to 4 Aug. 2002 and whose peak O3 concentrations would have been unaffected by the blackout. Moreover, local meteorology could have influenced the differences in air quality at Selinsgrove between 4 Aug. 2002 and 15 Aug. 2003. The dissimilarity in the backward trajectories shown for Selinsgrove, which differ by 30° to 80° depending on the start height of the trajectory, demonstrates that local dissimilarities in atmospheric conditions can be large enough to result in a large variability in SO2 and O3 mixing ratios. [4] Another aspect of the variability associated with the concentrations of atmospheric species is the regional representativeness of a single location: a relative change in concentration at one location may not be representative of relative changes over a broad geographical region. This can be checked by comparing 24-hour PM2.5 data from EPA's Air Quality System (AQS) (http://www.epa.gov/ttn/airs/airsaqs) for the day of the onset of the blackout to similar measurements on the day after; since 14 Aug. 2003 and 15 Aug. 2003 had synoptically similar meteorology (a slow-moving high pressure system), “persistence” would have dominated PM2.5 levels in the region. (Although some power generation units tripped in Pennsylvania during the blackout's cascade, the transmission grid isolated these outages and electricity distribution within the state was fairly unaffected, except for Erie, PA. Therefore, measurement data from EPA's AQS monitors are available.) In addition to addressing the issue of the representativeness of a single location, this analysis also provides insight into the potential impact of the blackout on PM2.5. For 15 stations with valid data for both days within the state of Pennsylvania, PM2.5 concentrations are found to decrease by an average of 8%, ranging from a decrease of 31% to an increase of 37% (Table S1). As shown in Figure 1, sites with moderate PM2.5 levels exhibited little change in particle concentrations, whereas those sites with higher PM2.5 levels exhibited greater variability in the day-to-day comparisons. Considering the 10 sites within 200 km of Selinsgrove, an average decrease in PM2.5 concentration of 13% is observed. [5] These relatively small average decreases (8–13%) in PM2.5 observed over Pennsylvania from the day before to the day after the blackout can be contrasted with the 70% decrease in light scattering (a common surrogate for PM2.5) inferred by Marufu et al. [2004] from vertical profiles measured a year apart. In addition, the large variance in the surface data shows how the absolute values and relative changes in PM2.5 observed at a single site can be unrepresentative of the region as a whole. Clearly, a more complete analysis of individual gas and particle components over several stations, in addition to an analysis of the prevailing meteorological conditions is necessary to assess the role of the blackout in these observations. [6] The SO2 measurements made during the vertical spiral in 2002 served as the basis for calculating the effect of the blackout on SO2 concentrations at Selinsgrove. Figure 4 of the manuscript depicts vertical profiles taken during 2002 and 2003 for SO2, O3, particle light scattering, and particle light absorption. All 2002 profiles, save that for SO2, indicate a well-mixed convective boundary layer below 2 km. The SO2 profile indicates a strong vertical gradient with a maximum value near the surface, decreasing monotonically to 2.5 km above the surface. Such a vertical profile is typical of highly photochemically reactive compounds with a surface source, such as isoprene, but not of much less-reactive gases such as SO2 assumed to be from elevated point sources. Although we do not question the accuracy of the 2002 measurements of Marufu et al. [2004], they may not be characteristic of SO2 profiles typically encountered in the atmosphere. To explore this issue, the 2002 SO2 profile was compared with a review of 146 vertical profiles documented as part of the Sulfate Regional Experiment [Keifer et al., 1981a, 1981b]; a similar vertical gradient in SO2 is observed only twice when O3 and light scattering are relatively uniform through the convective boundary layer. Thus, the comparison shows that this profile may be considered atypical and underscores the importance of understanding local atmospheric dynamics in order to ascertain the reasons for the observed differences in SO2 levels. Indeed, simply being within or outside of an individual plume could yield profoundly different SO2 measurements on a given day when emissions remain constant. [7] Another limitation of the analysis is the method for estimating upwind emissions in 2002 and 2003. Hourly data collected by continuous emissions monitors (CEMs) for NOx and SO2 were integrated for 24-hr periods within 100-km wide swaths straddling backward trajectories (possibly from three starting heights) from the vertical profile location. This oversimplifies the temporal evolution of the emissions and their atmospheric dispersion and leads to some questionable choices of what emissions sources to include. For example, two power plants included in the 2003 swath [Marufu et al., 2004, Figure 3] are actually to the ENE of Selinsgrove and not likely to have influenced air quality there since airflow was from the northwest. From this perspective, some of the observed differences attributed to power-plant shutdowns during the blackout could be artifacts of their estimation methodology. [8] Marufu et al. [2004] contend that motor vehicle emissions were unaffected by the blackout based on 1) traffic counts in Pennsylvania, 2) observation that 2003 CO levels remained within of the standard deviation of 2002 measurements and consistent with CO levels over the Maryland/Virginia sites, and 3) observation that particle light absorption was almost twice as high in 2003 as in 2002 (for which they offered no explanation). Consequently, changes in air quality are attributed predominantly to reductions in power-plant emissions during the blackout. Yet, power plant emissions in Pennsylvania were virtually unaffected during the blackout. Outside of Pennsylvania, the story was different. Namely, in other regions, such as eastern Michigan, northern Ohio, New York and southern Ontario, where electricity distribution was disrupted significantly during the blackout, it is reasonable to assume that the overall composition of emissions – including vehicles, industry, emergency backup generation and power plants – differed considerably from typical conditions. The levels of particle light absorption and CO over Selinsgrove need to be reevaluated with consideration of the full suite of emissions disruptions in these upwind areas, lack of emissions disruptions in most of Pennsylvania, and atmospheric variability. [9] A final limitation of Marufu et al. [2004] is the method used to conclude that the improvements in air-quality attributed to the blackout benefited much of the eastern United States. Through forward trajectory analysis the authors indicate that air parcels passing through Selinsgrove during their measurements reached an unspecified location in Maryland where the O3 forecast was 125 ppb, whereas the actual observed value was only 90 ppb. Since the difference in O3 predictions is greater than the RMS (root mean square) forecast error of 10 ppb, the authors provide this as evidence of air quality impacts from the blackout. A RMS error of 10 ppb does not preclude the model forecast error from being as high as 35 ppb on any particular day since the RMS error is an ensemble statistic. In addition, considering the PM2.5 spatial variability demonstrated above, the use of forward trajectories from Selinsgrove to extrapolate inferred changes there to regional air-quality impacts is insupportable. [10] In summary, Marufu et al. [2004] leave unresolved important issues that preclude a quantitative attribution of the changes in pollutant concentrations and light scattering to changes in emissions between the two days in separate years, much less to changes only in power-plant emissions during the blackout. Analyses of observed concentrations of atmospheric constituents across a broader range of locations performed over several meteorologically similar time periods, in addition to the data of Marufu et al., are warranted when attempting to understand the influence of the blackout on pollutant levels. Auxiliary material for this article contains a table of PM2.5 levels at EPA AQS sites in Pennsylvania before (14-Aug-2003) and after (15-Aug-2003) the 2003 blackout. Additional information is provided in the README.txt. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.