Nitrogen attenuation in the Connecticut River, northeastern USA; a comparison of mass balance and N₂ production modeling approaches

Abstract

Two methods were used to measure in-stream nitrogen loss in the Connecticut River during studies conducted in April and August 2005. A mass balance on nitrogen inputs and output for two study reaches (55 and 66 km), at spring high flow and at summer low flow, was computed on the basis of total nitrogen concentrations and measured river discharges in the Connecticut River and its tributaries. In a 10.3 km subreach of the northern 66 km reach, concentrations of dissolved N₂ were also measured during summer low flow and compared to modeled N₂ concentrations (based on temperature and atmospheric gas exchange rates) to determine the measured “excess” N₂ that indicates denitrification. Mass balance results showed no in-stream nitrogen loss in either reach during April 2005, and no nitrogen loss in the southern 55 km study reach during August 2005. In the northern 66 km reach during August 2005, however, nitrogen output was 18% less than the total nitrogen inputs to the reach. N₂ sampling results gave an estimated rate of N₂ production that would remove 3.3% of the nitrogen load in the river over the 10.3 km northern sub-reach. The nitrogen losses measured in the northern reach in August 2005 may represent an approximate upper limit for nitrogen attenuation in the Connecticut River because denitrification processes are most active during warm summer temperatures and because the study was performed during the annual low-flow period when total nitrogen loads are small.

Appendices

Appendix A—Uncertainty analysis computations

Mass balance

Uncertainties in the mass balance measurements at each sampling site consisted of potential errors in the flow measurements, and analytical errors in determining the concentrations of nitrate and total ammonia plus organic nitrogen (which are summed to compute the total nitrogen concentrations). Because the rivers were well mixed, and samples were collected using depth- and width-integrated sampling (discharge weighted), sampling error was assumed to be negligible relative to the discharge and analytical uncertainties. For the ground-water component of the nitrogen input to each study reach, however, the standard deviations of nitrate and total ammonia plus organic nitrogen among the seven ground-water seep samples were used to determine the uncertainty in the ground-water nitrogen concentration.

Discharge records at the USGS streamflow gaging stations used in the study were rated “good,” which the USGS defines as within 10% of the true value at 95% confidence (Kiah et al. 2005). Manual discharge measurements were rated “fair” (or better), which the USGS defines as within 15% of the true value at 95% confidence (Sauer and Meyer 1992). Analytical uncertainty was approximately 5% at 95% confidence for nitrate (Antweiler et al. 1993), and approximately 10% at 95% confidence for total ammonia plus organic nitrogen (Patton and Truitt 2000).

The uncertainties in the source terms (such as analytical error) were combined as the square root of the sum of their squares for each computation step, where the relative (percent) errors were squared and summed when terms were multiplied, and the absolute errors were squared and summed when terms were added (Taylor 1982).

N2 analysis and modeling

Uncertainty in gas transfer rates is derived from the scatter in SF₆:Br data and the uncertainty in the Schmidt number coefficient used to calculate N₂ transfer rates from SF₆ transfer rates (see Laursen and Seitzinger 2002). To compute the upper and lower confidence limits of N₂ transfer rates, the maximum and minimum slopes of the regression line of SF₆:Br versus time were calculated using ±1 standard deviation (Fig. 4). N₂ gas transfer rates were then re-calculated using the maximum and minimum SF₆ transfer rates derived from these slopes, with Schmidt number coefficients of $\raise0.7ex\hbox{$2$} \!\mathord{\left/ {\vphantom {2 3}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$3$}$ (for flat surfaces, no waves) and $\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}$ (wavy surfaces not broken by white caps) (Jähne et al. 1987). A Schmidt number coefficient of $\raise0.7ex\hbox{$2$} \!\mathord{\left/ {\vphantom {2 3}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$3$}$ was used for the baseline model, whereas a Schmidt number coefficient of $\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}$ represented maximum error.

N₂ concentrations at locations C-3 and C-9 (used as model inputs) were calculated for all possible scenarios among the two sources of error—uncertainty in Ar and uncertainty in N₂:Ar. Argon was modeled using measured Ar (at location C-3) ±1 standard deviation (0.04 for replicate samples). The standard deviation associated with replicate N₂:Ar measurement was ±0.019. N₂ concentrations were calculated at locations C-3 and C-9 as ±0.27 standard deviates. The probability of actual N₂:Ar being within +0.27 standard deviate of measured N₂:Ar at location C-3 while, simultaneously, actual N₂:Ar being within −0.27 standard deviate of measured N₂:Ar at location C-9, was approximately 70% (probability associated with 1 standard deviation). Channel depth is well constrained by the many transect measurements made along the study reach; for modeling purposes, however, uncertainty in average channel depth was assumed to be ±5%.

Appendix B—Sample calculations for gas transfer rates

• 1. K_SF6 = 0.120 h⁻¹ (regression of ln-transformed, bromide-corrected SF₆ concentrations versus time (Fig. 4)).

• 2. Average temperature over reach (locations C-1 to C-6) = 23.58°C

• 3. Schmidt number for SF₆ @ 23.58°C = 808*

• 4. K₆₀₀ = K_SF6 × (600/808)⁻ⁿ

• 5. K₆₀₀ = 0.146 h⁻¹ ${\left( {{\text{n}}\, = \,\raise0.7ex\hbox{$2$} \!\mathord{\left/ {\vphantom {2 3}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$3$}} \right)}$, = 0.139 h⁻¹ ${\left( {{\text{n}}\, = \,\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}} \right)}$. K₆₀₀ values are used to calculate N₂ and Ar transfer rates at each time step given in situ temperature. For example, @ 20°C:

• 6. K_N2 = K₆₀₀ × (580/600)⁻ⁿ (Schmidt number for N₂ @ 20°C = 580)*

• 7. K_N2 = 0.134 h⁻¹ ${\left( {{\text{n}}\, = \,\raise0.7ex\hbox{$2$} \!\mathord{\left/ {\vphantom {2 3}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$3$}} \right)}$, = 0.126 h⁻¹ ${\left( {{\text{n}}\, = \,\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}} \right)}$

• 8. K_Ar = K₆₀₀ × (518/600)⁻ⁿ (Schmidt number for Ar @ 20°C = 518)*

• 9. K_Ar = 0.144 h⁻¹ ${\left( {{\text{n}}\, = \,\raise0.7ex\hbox{$2$} \!\mathord{\left/ {\vphantom {2 3}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$3$}} \right)}$, = 0.133 h⁻¹ ${\left( {{\text{n}}\, = \,\raise0.7ex\hbox{$1$} \!\mathord{\left/ {\vphantom {1 2}}\right.\kern-\nulldelimiterspace} \!\lower0.7ex\hbox{$2$}} \right)}$

* Calculated based on Wanninkhof (1992).