This event primarily took place during January 2014. After visual inspection, a key indicator was a very low negative value reported by beta-attenuation monitor (BAM) component of the instrument. For periods when the BAM reported a value below −70 μg m−3, two hours prior to this data point and 24 hours of data afterward were set to missing. As a result, a total of 1.9% of the data were flagged and removed from analysis. After this initial processing, data were imported and analyzed using R version 3.5.1, with packages including gdata, openair, data.table, and plyr.
NTA analysis was conducted via MATLAB, using the one-minute monitoring data at the R2PIER site and meteorological data at a Newark Airport (KEWR) as inputs. NTA calculated local back-trajectories per minute using inverse distance-squared weighting interpolation of the wind data at the R2PIER site and KEWR Automated Surface Observation System (ASOS) 1-minute wind data. The ASOS wind data are two-minute averages reported every minute from a cup anemometer while the R2PIER data were produced by an ultrasonic 3-D anemometer. The wind speed and direction at a point was determined as the inverse-distance-squared weighted average of the wind velocity components from these two sites. The equations for the trajectory and other NTA calculations are detailed in Henry (2008) and most recently in Henry et al. (2019).
3. RESULTS AND DISCUSSION
3.1. Study Sampling Conditions and Overall Averages
Shipping activity data fluctuated throughout the study, with TEUs and cargo ship number not consistently aligned (Fig. 2). The spring to summer of 2015 indicates an overall increase in TEUs relative to previous years, however the number of vessels did not increase in proportion to the TEUs. In general, vessel activity near the Port was dominated by hoteling operations, while in-transit operations comprised a small percentage of activity. The wind conditions throughout the sampling study are an important consideration to the connection of the high time resolution monitoring data to nearby source areas. Prevailing winds were typically from the west, ranging from the SW to NW, with wind speed between 2–4 m s−1. However, over each calendar year sampled, monitoring data were collected over all wind angles, with the high time resolution measurements equating to thousands of observations for any given 10-degree wind angle. An example showing data collection frequency by wind speed and direction for the four complete summer quarters is provided in Fig. S2. Summary data on pollutant concentrations and meteorology, as well as by season, are provided in Table 2. Seasonality is evident for all parameters measured, with maximum concentrations occurring during fall and winter months (October through March) and lower concentrations in the summer.
Comparing this location with other areas in the United States for the years 2012–2015, via the U.S. EPA's trends reports (https://www.epa.gov/air-trends), the CO data measured in this study was below the 10th percentile of 44 Trend Sites reported nationally, NO2 values were similar to the 10th percentile value, SO2 values were well below the 10th percentile, and PM2.5 levels were slightly below the 90th percentile.
The nearest regulatory monitoring station to the R2PIER site was New Jersey Department of Environmental Protection's Elizabeth Lab Chemical Speciation Network (CSN) site, located approximately 2 km to the southwest (Latitude: 40.641440; Longitude: −74.208365). Whereas the R2PIER site measured PM2.5 continuously with the SHARP monitor that combined nephelometry and beta-attenuation, the Elizabeth Lab site measured PM2.5 using an integrated filter collection approach with subsequent laboratory analysis. The integrated PM2.5 filters were collected on a 1-in-3 day sampling schedule. Comparison of the particulate matter measurements were made between the two sites, with the R2PIER data reduced and averaged to match the Elizabeth Lab CSN site sampling schedule (Fig. 3). Despite the differences in the detection methods and distance, the two PM2.5 daily average values had moderate correlation (R2 = 0.72). Although the R2PIER daily PM2.5 concentrations were within 6% of that measured at the Elizabeth Lab site, the difference in the daily concentration between the two sites ranged fairly widely (relative standard deviation of 37%). The R2PIER BC concentration was compared against elemental carbon (EC) measured at Elizabeth Lab on the 1-in-3 day schedule. Both BC and EC are operationally defined methods, with BC measured optically and converted to a mass concentration, while EC is measured through a thermal-optical method that separates organic and elemental carbonaceous mass fractions. Both methods are commonly used and a past study intercomparing the two methods in the New York City area observed moderate correlation but with slopes ranging from 0.86 to 1.23 (Venkatachari et al., 2006). When comparing BC at R2PIER and EC at Elizabeth Lab sites, the two methods had moderate correlation (R2 = 0.43), with BC concentrations lower than EC.
3.2. Temporal Analysis
In order to visualize the coupled effects of wind direction, day of week and time of day, two pollutants (BC and SO2) affected primarily by direct emissions were visualized in a polar annulus graphs oriented by wind direction and time (Fig. S3). For SO2, maximum concentrations occur during winds from the NE and SW, with highest values during morning periods. Lowest concentrations appear to occur on Sundays, during afternoon periods and with wind from the NW. BC has similar temporal trends, however appears to have higher concentrations associated with winds from the N and some periods of higher concentration under winds from the E/SE compared to SO2. Shipping activity data indicate a repeatable diurnal pattern with higher number of pings in the evening hours and lowest mid-day, and with weekends having lower activity than weekdays (Fig. S4). The general range of variation, however, is modest with less than 10% variation above/below the average.
Throughout the study period, the monitoring station exhibited variation in hourly concentrations (Fig. 4). The coefficient of variation (CV) was calculated by computing the standard deviation of the hourly averages divided by the mean of the averages. The CV was highest for NO (2.65), followed by SO2 (1.45), BC (1.21), PM2.5 (0.79), NO2 (0.65), and CO (0.60). In a near-source environment, the high concentration excursions of directly emitted pollutants would likely be related to proximate emissions. To explore how the overall concentrations varied over time, as well as the occurrence of high concentration episodes, monthly percentiles (50th, 75th, and 95th) were compared (Fig. 5), normalized to their respective January-2013 concentration for ease of visualization. BC stands out in having a slightly different seasonal signal in 2013, with higher concentrations in summer 2013 while other parameters had a local minima. For other periods, the general seasonal trends are similar across the pollutant types. SO2 stands out in both its decreased frequency of outliers (Fig. 4) and an associated decrease in the 95th percentile, by approximately half, in spring/summer 2015 compared to the same time periods in 2013 and 2014. This shift is likely related to the introduction of lower sulfur fuel requirements which were enforceable in January 2015.
Pollution wind roses and wind-directional averages of SO2 (Fig. 6) for successive summer quarters (July through September) of the four monitored years reveal higher concentrations associated with winds from the southwest and northeast. Compared to earlier summers, the mean decreases from these two wind sectors in summer 2015 and shows little wind directionality. Comparing 2014 and 2015 and selecting for winds between 15 and 75 degrees (winds from the NE), mean concentrations from that sector were significantly lower by a factor of 0.55 with a Kruskal-Wallis test p-value < 0.01. For the SW wind sector, winds between 195–255 degrees, there was a significant decrease (p-value < 0.01) in mean SO2 by a factor of 0.60.
3.3. Nonparametric Trajectory Analysis
A major problem in assessing the impact of emissions changes in the Port is the presence of large nearby sources. Nonparametric Trajectory Analysis (NTA) was used to separate out the source areas for the purpose of sharpening the trends in the air quality data affected by mandated control strategies. The study period had a total of 1,623,765 one-hour back-trajectories computed for each minute and ending at the R2PIER site. Of these, 100,521 trajectories passed directly over the Port to the NE of the site (Fig. 7) while passing over no other known nearby source areas. Periods low wind speeds (calms) posed little difficulty for the trajectory calculations. Wind speeds at Newark airport (KEWR) are reported in whole knots and only 0.52% of the observations were less than 1 knot (0.514 m s−1). Wind speeds at the R2PIER site were reported to an accuracy of 0.1 m s−1 and only 0.16% of the observations were less than this.
This technique allowed the impact of the Port and other source areas to be isolated for trend analysis. This high volume of trajectories revealed expected SO2 concentrations at the monitoring site on a quarterly basis as air masses passed over surrounding geographic areas (Fig. 8). The map visualizations show higher concentrations associated with areas to the southwest and northeast of the site, with several quarters showing higher source area contributions from the bay area due east of the monitoring site. BC and NOx results (Figs. S5–S6) also indicate higher concentrations associated with areas to the northeast of the site, but relatively less from the southwest compared to the SO2 results. Specifically, the NTA maps show that the highest average concentrations at the R2PIER site were associated primarily with transport from the Port to the northeast and the container terminal to the southwest, and occasionally with transport from Newark Bay. The third quarter (July through September) maps show little year-to-year variation in geographic pattern, but a great reduction in maximum mixing ratio from 6 to 1 ppb.
Focusing on the subset of trajectories that pass solely over nearby geographic areas associated with ship traffic and cargo handling, Table 3 provides the change in concentration at the monitoring site for on-source trajectories comparing the last 5 quarters of the study (July 1, 2014–October 1, 2015) to the first 5 quarters of the study (July 1, 2012–October 1, 2013). These longer time spans were selected to provide a significant volume of trajectories to assess temporal shifts in source emission areas over the study period and use identical spans of months to minimize bias due to seasonality. This analysis indicates a decrease of 34–41%, 11–17%, and 28–41% for SO2, NOx, and BC, respectively, for air masses passing over three specific nearby source regions. An important caveat to these findings is that the concentrations input to the NTA analysis were absolute values and not background-subtracted. While the influence of background seasonality should be lessened by using the same 15 months for the "before" and "after" time windows and focusing on pollutants expected to be highly influenced by nearby emissions, additional research to separate the background component from the time series would further refine these estimates.
4. CONCLUSIONS
This study utilized research-grade and trace-level instrumentation to maximize the potential air quality insights from the data through careful siting and sustained collection of high time-resolution data. Comparison of the R2PIER site against a nearby regulatory monitoring site revealed correlation in measured PM2.5 and BC. Although TEUs increased and ship activity (vessel pings) did not markedly change, NTA analysis indicated decreases of SO2, NOx, and BC associated with port-related geographic areas when comparing the last 5 quarters measured to the first 5 quarters. The trends indicated by NTA analysis were directionally consistent with "port-wide" emission inventory changes over 2012–2015 for SO2 (PANYNJ, 2019b) and both analyses found relatively less change in NOx although in differing directions. This comparison is limited, as this study's NTA analysis emphasizes geographic areas surrounding the monitoring site position and the in situ monitoring data may be affected by sources outside the scope of the Port Authority's inventory (e.g., emissions related to activities from privately-owned terminals).
As air monitoring rates increase with the evolution of air monitoring technology (e.g., Snyder et al., 2013), retention and analysis of high time resolution air monitoring data may enable the isolation of local emission signals. Coupled measurements of directly emitted pollutants and wind, along with monitor siting to minimize collinearity in sources, increases the likelihood of resolving geographic source emission areas influencing the measurements and gaining insight into the impacts of emission reduction efforts.