Ammonia Exchange between the Atmosphere and ... - ACS Publications

Estimation of ammonia exchange at the land–ocean boundary condition of Sundarban mangrove, northeast coast of Bay of Bengal, India. Atmospheric ...
0 downloads 0 Views 116KB Size
Environ. Sci. Technol. 2001, 35, 4731-4738

Ammonia Exchange between the Atmosphere and the Surface Waters at Two Locations in the Chesapeake Bay RANDOLPH K. LARSEN, III, JOSEPH C. STEINBACHER, AND JOEL E. BAKER* Chesapeake Biological Laboratory, University of Maryland Center for Environmental Science, 1 Williams Street, Solomons, Maryland 20688

Excess phytoplankton production, which contributes to hypoxic conditions, is nitrogen limited in the Chesapeake Bay during the summer months. Therefore, understanding the flux of ammonia by direct deposition to the biologically active surface layer is critical to understanding the nutrient dynamics of the bay. This paper presents the results of a 2-yr study measuring gaseous ammonia (NH3) and aerosol ammonium (NH4+) in Baltimore and Solomons, MD, from which direct atmospheric loading of total ammonia (Nt ) NH3 + NH4+) to the Chesapeake Bay is estimated. Mean atmospheric concentrations of total ammonia for Baltimore and Solomons were 2.7 ( 1.7 and 1.0 ( 0.8 µg of N m-3, respectively. Monte Carlo estimates of gross dry deposition ranged from 18.2 mΩ deionized water was added and hand agitated as described in the manufacturer’s directions. Extractions were placed in a -20 °C freezer until analyzed. Samples were analyzed on a Technicon TrAAcs-800 nutrient analyzer, which determines ammonia concentrations by colorimetric technique using the Berthelot reaction. Filter packs and denuder/filter systems were prepared and extracted in a nitrogen-filled glovebag or in a laminar flow clean bench where two inlet glass fiber prefilters were placed before the high-efficiency particulate air (HEPA) filter. One prefilter was coated with 0.01 M oxalic acid, while the second was coated with 0.01 M sodium carbonate solution to minimize ambient ammonia or nitrates from contaminating the samples during preparation or extraction. The sodium carbonate solution consisted of 1 g of sodium carbonate, 50 mL of >18.2 mΩ deionized water, 50 mL of methanol, and 1 mL of glycerol. QA/QC. Field blanks were collected at each sampling site. Field blanks consisted of filter packs and denuders that were taken to the sample site, unpacked, and attached to the sampling train; endcaps were removed, then immediately repackaged, and returned to the lab for extraction as described above. Blank correction consisted of determining whether the sample concentration was greater than 3× the site-specific mean blank concentration. For the filter packs, this meant comparing the concentration of the composited front-middle filters against 3× the mean front-middle blank filter as well as comparing the back filter against 3× the mean blank back filter. Composited front-middle filters exceeded 4732

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001

3× the mean blank 75% and 88% of the time at Solomons and Baltimore, respectively. Back filters exceeded 3× the mean blank 24% and 31% for Solomons and Baltimore, respectively. If the sample concentration exceeded 3× the mean blank, the sample concentration was reported unchanged. If the sample concentration did not exceed 3× the mean blank, a value of two-thirds the mean blank concentration was reported. Assuming a typical air sample volume of 3.1 m-3, the limit of detection (LOD) is equal to 0.8 and 1.1 µg of N m-3 for Solomons and Baltimore, respectively. The minimum detection level (MDL) of the Technicon TrAAcs-800 nutrient analyzer was 3 µg of N L-1, which corresponds to an ambient sample concentration of 0.01 µg of N m-3. The MDL was calculated as 3× the standard deviation of seven replicates of a 7 µg of N L-1 ammonia standard. Estimating Dry Flux of Ammonia/Ammonium. The dry flux of ammonia/ammonium is expressed as the difference of the gross volatilization and the gross dry deposition:

gross volatilization ) Ve[NH3]eq

(1)

gross dry deposition ) -(Vefg[NHx] + Vd(1 - fg)[NHx]) (2) net air-water exchange flux ) gross volatilization + gross dry deposition (3) where fg is the fraction of the total ammonia, [NHx] present as ammonia; Ve is the exchange velocity of ammonia across the air-water interface; [NH3]eq is the atmospheric concentration of ammonia at equilibrium with the dissolved surface water ammonia; and Vd is the deposition velocity of particulate ammonium. A negative flux indicates a net transfer into the bay; if positive, the flux will be out of the water (i.e., net volatilization). To calculate [NH3]eq requires knowledge of the surface water pH, temperature, and ionic strength as well as the dissolved ammonium concentrations. These variables or their surrogates are measured monthly at numerous locations throughout the bay by the Chesapeake Bay Program. Through equilibrium theory, the [NH3]eq is calculated by first determining the ratio of dissolved ammonium to ammonia (NH3(w)):

NH4+ ) NH3(w) + H+

(4)

The ratio of ammonia to ammonium is expressed as the acid dissociation constant (Ka). The log Ka of eq 4 is 9.23 for an ideal solution (19). However, Ka is dependent on temperature as expressed:

Ka ) 5.67e-10 exp(-6286(1/T - 1/298.15))

(20)

(5)

The effects of ionic strength on ammonia and ammonium must be incorporated to accurately predict the ammonia/ ammonium ratio. Adjustment for ionic strength is expressed by the Davies modification of the Debye-Huckel limiting law:

-log γNH4+ ) Aiz(I 0.5/(1 + I 0.5)) - 0.2I

(21)

(6)

where γ NH4+ is the activity coefficient for ammonium, z is the charge of ammonium, I is the ionic strength of the water, and Ai is the temperature correction factor expressed as

Ai ) 0.4896844 - 0.0007477T + 2.729 × 10-6T 2

(21) (7)

TABLE 1. Literature Review for Over-Water Exchange Velocities of Hydrophilic Gases compd

Ve (mm s-1)

location

method

ref

NH3 NH3 NH3 NH3 NH3 NH3/HNO3 NH3 HNO3 HNO3 HNO3 HNO3

8-20 8.33 8.33 8 7.6 x¯ ) 7 2-15, x¯ ) 5 6 1-27, x¯ ) 6.5 2-7.2 3-10

Atlantic Basin Australian sector, Southern Ocean NE Pacific North & Baltic Seas North Sea Tampa Bay North Sea Chesapeake Bay Chesapeake Bay Long Island Sound Great Lakes

Moguntia/HAMOCC3 by ref to 37 similarity to H2O not reported two-resistance modela iterative bulk exchange model three-resistance modelb RAMS model iterative bulk exchange model two resistance modela ASTRAP model

21 38 39 40 41 36 42 43 17 35 44

a The two-resistance model calculates V ) 1/(aerodynamic resistance + laminar boundary layer resistance). This model assumes that the water e phase resistance is negligible. b The three-resistance model calculates Ve ) 1/(aerodynamic resistance + laminar boundary layer resistance + water phase resistance).

The activity coefficient of ammonia can be expressed as

γNH3 ) 1 + 0.085I

(22)

(8)

The ratio of the major ions in seawater is relatively stable. Therefore, ionic strength can be extrapolated from the salinity of the water:

I ) 0.00147 + 0.01988S + 2.08357 × 10-5S2

(23)

(9)

where S is the salinity in ppt. The waters of the Chesapeake are a mixture of both freshwater and seawater. Therefore, a systematic error exists using this assumption, but it provides a first estimate of the ionic strength of the bay. Equation 4 includes the term H+; therefore, the pH of the bay, which varies over 1 pH unit, will also affect the partitioning. Once the concentration of NH3 in the surface water is calculated, it can be combined with the temperature-corrected Henry’s law coefficient (KH) to determine [NH3]eq:

log KH ) 4092/T - 9.7

(24)

(10)

Thus, the gas-phase concentration of ammonia in equilibrium, [NH3]eq, with the ammonia in the bay [NH3(w)] can be expressed from the combination of eqs 5-10:

[NH3]eq (µg of N m-3) )

C × MW[NHx]/RTKH(1/γNH3 + 10-pH/γNH4+Ka) (11)

where C is a unit conversion constant, MW is molecular weight of nitrogen, R is the ideal gas constant, and [NHx] is the combined molar concentration of ammonia and ammonium in the bay. Equations 1-11 revel that gross volatilization is linear with respect to the dissolved ammonia concentration (due to Henry’s law) and exponentially related to pH (equilibrium partitioning) and temperature (Ka and Henry’s law equations). The percent of dissolved ammonia complexed with sulfate and chlorine was investigated using Minequal+ software. The available dissolved ammonia decreased by less than 5% when complexation was considered; therefore, complexation was not included in the [NH3]eq calculation. Table 1 illustrates ammonia and other hydrophilic gas exchange velocities (Ve). The bulk exchange method employed by Valigura (17) iteratively solves the turbulent flux equations of heat, moisture, and momentum. From these equations, the dimensionless heat transfer coefficient is found. This coefficient is proportional to the aerodynamic resistance of sensible heat, which is assumed to equal the aerodynamic resistance of HNO3. The inverse of which is the exchange velocity of HNO3. Combinations of temperature

and humidity gradients from the water surface to the point of measurement as well as wind speed contribute to the variations in Ve. The frequency histogram of Ve presented in Valigura (17) was modeled and incorporated into the Monte Carlo flux simulations. This data set was chosen for the flux equation in this study since it represents a long-term study (∼18 months) of velocities found over the Chesapeake Bay (17). This deposition velocity represents a conservative estimate of the mean annual velocity since data were not collected during winter months due to the potential for ice damage to field equipment. Winter deposition velocities are likely higher than annual mean velocities due to a greater magnitude and frequency of unstable atmospheric conditions. Several factors influence particle deposition velocities, including the particle size and wind speed. There exists an abundance of literature concerning ammonium ion particle size distribution that show median ammonium ion size distributions ranging between 0.1 and 1.0 µm (25, 26). From this size distribution, the particle deposition velocity can be calculated from the William’s model (27). Size-dependent deposition velocities calculated by the William’s model are functions of wind speed, atmospheric stability, and sea conditions. Such calculations have recently been performed using particle size distributions and weather conditions found in Solomons, MD, from December 1992 to July 1993 (28). The deposition velocities for particles with aerodynamic diameters of 0.125-1.34 µm ranged from 0.2 to 0.8 mm s-1 using the average of all applicable weather conditions. Furthermore, element specific deposition velocities were calculated, which found the fine sulfur fraction had a mean Vd of 0.23 mm s-1. Therefore, assuming that the fine ammonium fraction was a form of ammonium sulfate, the Monte Carlo flux simulation assumed a log-normal distribution of deposition velocities with a mean of 0.23 mm s-1 and 2σ of 0.8 mm s-1. Error Analysis. An error analysis of each individual flux estimate was performed via a two-step process. Initially, propagation of error in each measured variable (ammonium concentration, pH, temperature, salinity, and air volume) using the standard δ technique was performed to determine the error of each calculated variable ([NH3]eq, Ka, KH, ionic strength, ammonia and ammonium activity coefficients, and total atmospheric [NHx]) (29). Since values of Ve and Vd were taken from the literature, the error associated with either is dependent on their distribution. By definition, both Ve and Vd must be positive values; therefore, they appear lognormally distributed. The δ technique is an inappropriate method for error propagation of Ve and Vd because it assumes that the error of each variable is normally distributed. Additionally, the δ technique cannot be used for error propagation of the gas fraction (fg), which is a β distribution VOL. 35, NO. 24, 2001 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

4733

FIGURE 1. Total ammonia concentrations in atmosphere at (a) Baltimore and (b) Solomons, MD. with an absolute range of 0-1. Therefore, a Monte Carlo analysis was performed to determine the error associated with the gross deposition, gross volatilization, and net flux. For each deposition, volatilization, and flux estimate, 10 000 simulations were made with a Monte Carlo program written with SAS software, using normal distributions of error for the calculated variables ([NH3]eq and total atmospheric [NHx]), log-normal distributions of error for Ve and Vd, and β-distributed error for fg. After the results were sorted in ascending order, the 5000th, 250th, 500th, 9500th, and 9750th datum points represent the median, 2.5, 5, 95, and 97.5 percentiles. The range between the 2.5 and 97.5 percentiles is the 95% confidence interval, likewise between the 5th and the 95th percentiles is the 90% confidence interval.

Results and Discussion Arithmetic mean atmospheric concentrations of total ammonia (ammonia plus ammonium) for Baltimore and Solomons, MD, were 2.7 ( 1.7 and 1.0 ( 0.8 µg of N m-3, respectively (Figure 1a,b). The Baltimore data were normally distributed while the Solomons data were neither normally or log-normally distributed (Kolmogorov-Smirnov normality test). The median Baltimore ammonia concentrations were statistically higher than the rural Solomons location using the Mann-Whitney rank sum test (P e 0.001), indicating that those local sources such as industry, fossil fuel combustion, and sewage treatment plants impact the local atmosphere, even though such nonagricultural sources account for a small fraction of the total emissions to the bay air shed (9). To further investigate this finding, a review of the ammonia releases from industrial sources, as listed by the 1997 Environmental Protection Agency’s Toxic Release Inventory, found a total of 150 000 kg of ammonia reportedly released in Maryland in 1997 from industrial stacks and fugitive 4734

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 35, NO. 24, 2001

emissions. Nine facilities within Baltimore city were responsible for 73% of the total industrial releases of ammonia in Maryland (30). Emission rate estimates of ammonia from California automobile fleets range from 49 to 61 mg of NH3 km-1 (10, 11). Combining these rates with the Maryland Department of Transportation estimate of vehicle miles driven provides an estimate of ∼3 × 105 kg of NH3 yr-1 from mobile sources in Baltimore. While releases from automobiles and those reported in the Toxic Release Inventory do not account for all sources of ammonia, they do indicate that industrial urban areas are responsible for a large quantity of the ammonia found in the local urban atmosphere. Table 2 lists the ambient atmospheric ammonia concentrations reported in other parts of the United States. Maryland does not have the highest or least atmospheric ammonia/ ammonium concentrations. However, this table indicates the degree to which ammonia levels may increase or decrease. At both the Baltimore and Solomons sites, mean atmospheric concentrations were not significantly different between 1997 and 1998. Therefore, the influences of interannual variability in atmospheric conditions, such as annual rainfall, were not observed during this study. The monthly mean concentrations of ammonia for July, January, and December at the Solomons location were significantly different from the annual mean (P e 0.001), indicating the presence of a seasonal variation with highest ammonia concentrations present during the summer and lowest during the winter. No seasonal signal was detected at the Baltimore location. This may be due to a smaller sample size, or it may be due to constant local emission that overpowers the seasonal background signal. Atmospheric ammonia concentrations are expected to undergo seasonal fluctuation with peak concentrations in the spring and summer due to increased fertilizer usage during the spring and summer, faster volatilization rates due to warmer temperatures, and enhanced ammonia production from increased microbial activity. In addition, vertical mixing of the atmosphere may also play a role controlling ground-level ammonia concentrations. During the winter, vertical mixing is maximized due to the large difference in air and water temperatures. This means that the volume in which the ammonia emissions mixed is greater, resulting in decreased ground-level concentration. From January to May 1999 plus one sample from August 1998, denuder/filter systems were deployed to determine the partitioning of ammonia and ammonium. Eight samples were taken in Baltimore, and 15 were taken in both Solomons sites. The ammonia gas fractions (fg) were normally distributed based on the Kolmogorov-Smirnov normality test, P ) 0.184 and 0.247 for Baltimore and Solomons, respectively. The mean ammonia fraction of the combined results was 50% ( 26% (N ) 23). A t-test comparing the ammonia fractions in Baltimore to those in Solomons found no significant differences (P ) 0.107). These data were used to determine the parameters of the β distribution of fg, which was then used in the Monte Carlo error analysis. The gross dry deposition of ammonia is presented in Figure 2a,b. Mean estimates of gross deposition are on the order of 1000 µg of N m-2 d-1 near Baltimore and about 50% less near Solomons. The gross dry deposition is greater in Baltimore due to the increased atmospheric ammonia concentrations. The majority of the gross dry deposition is attributable to the gasphase ammonia, even though the ratio of ammonia: ammonium is nearly 1. This is due to the gas exchange velocity being an order of magnitude greater than the particle deposition velocity. The EPA’s Chesapeake Bay Program monitors the water quality of the bay by collecting monthly samples at various locations throughout the Chesapeake (31). The gross volatilization was calculated using eq 1 and their surface water

TABLE 2. Literature Review of Ambient Atmospheric Ammonia/Ammonium Concentrations in the United Statesa location FLb

Tampa Bay, (36) NE Pacificc (45) Cheeka Peak, WAc (45) Niwot Ridge, COc,d (46) Sherwood Island State Park, CTc (35) Hammonasset State Park, CTc (35) Harvard Forest, MAe (47) Los Angeles, CAc (10) Sampson Co., NCb,f (48) Mt. Mitchell, NCb (49) Baltimore, MD,c this study Solomons, MD,c this study

ammonia (µg of N m-3) 1.4 0.015 (0.08) 0.009 (0.004)

0.132 (0.154) 1.6 7.52 (4.58) 2.25 (1.54) 0.7

NH3 + NH4+ (µg of N m-3)

0.7 0.07 (0.03) 0.07 (0.04) 0.134 ( 0.018 0.069 ( 0.019 0.87 0.78

time frame 8/96-6/99 5/87 5/87 May-Sept 1998 & 1999 Sept 1998-May 1999 spring 1991-winter 1993 spring 1991-winter 1993 summers 1991-1995 9/21/93 5-7/98 & 10-12/98

0.16 2.7 (1.7) 1.0 (0.8)

5-9/88 & 5-9/89 3/97-3/99 3/97-5/99

a Mean values in parentheses indicate standard deviation if reported. b Samples collected by denuders/filter and analyzed by colorimetry. Samples collected by filter packs and analyzed by colorimetry. d Values after ( represent 90% confidence intervals. e Samples collected with Teflon prefiltered mist chambers and analyzed by ion chromatography. f Maximum and minimum mean concentrations of four sample sets. c

FIGURE 2. Gross dry deposition of NH3 and NH4+ at (a) Baltimore and (b) Solomons, MD. Vertical bars represent the 95% confidence intervals; b represents the mean of the estimate. quality data at the two sampling sites nearest to this study’s air sampling sites, LE 1.4 (38°18′ N, 76°25′ W) and WT 5.1 (39°12′ N, 76°31′ W) for Solomons and Baltimore, respectively. Figure 3a,b shows the mean estimated gross volatilization ranging from