Simulations of transient pesticide concentrations in Long Island Sound for late summer 1999 with a high resolution coastal circulation model.

ABSTRACT Results for hindcast simulations are presented for the transient concentration distributions in Long Island Sound for the pesticides malathion, resmethrin and sumithrin for the period August 1, 1999 through October 19, 1999. The high resolution of the simulations provide information concerning the detailed spatial patterns in time varying concentration distributions that are dependent on the distribution of pesticide loadings around the perimeter of the basin. Plume concentrations attenuate very rapidly with distance from the coast to below 0.001 [micro]g/L; high concentrations are detected only in confined coastal embayments and regions immediately adjacent to the coast. The decay rates used in the simulations have a strong influence on the concentration fields. Concentrations in plumes for pesticides dispersing into the western Sound from the East River also attenuate rapidly with distance from the East River.

KEY WORDS: pesticides, concentrations, model, lobster, Homarus americanus

INTRODUCTION

During the late summer and early fall of 1999, counties in New York State were subjected to pesticide spraying to combat mosquitoes potentially carrying the West Nile Virus. This spraying included coastal counties bordering western and central Long Island Sound. Approximately coincident with this spraying, a major mortality of the American lobster (Homarus americanus H. Milne Edwards, 1837) occurred during summer and fall of 1999 throughout much of the central and western Sound. The purpose of this study is to characterize the time varying spatial patterns in near-bottom pesticide concentrations within Long Island Sound associated with the pesticide spraying. This information is intended to contribute to an assessment of the possible role of pesticide spraying in the lobster mortality event. This late summer period was characterized by distinctive meteorologic events. On August 30th a strong cold front associated with a Canadian high pressure system passed through the region. This system interacted with Hurricane Dennis to the south leading to strong winds, which contributed to rapid and complete mixing of the water column. In addition, Hurricane Floyd passed through the region (as a tropical storm) on September 16th. It was associated with strong wind stirring and high precipitation leading to high fresh water inflows to the basin.

METHODS

The method used here involves hindcast simulations for pesticide concentrations for the period of interest using an existing coastal circulation model for the Sound. We used pesticide loadings estimated from application data provided by New York State Department of Environmental Conservation (NYDEC). We also used available meteorologic data to define surface heat and momentum flux time series, stream flow data to define inflow time series to the basin, and available hydrographic data to define initial temperature and salinity fields as well as open boundary values for temperature and salinity.

Crowley (2005) applied the 3-dimensional primitive equation model ROMS Version 2.1 (Haidvogel et al. 2000) to Long Island Sound for the purpose of describing the seasonal cycle in thermohaline circulation, and we used her model set-up with contemporary forcing and boundary conditions. The model domain (Fig. 1) extends from Throgs Neck in the west to Block Island in the east; it does not include the East River tidal strait, which connects the western Sound with the Upper Bay of New York Harbor. The model uses orthogonal curvilinear coordinates in the horizontal and sigma-coordinates in the vertical. The grid is 40 by 350 cells with a mean horizontal resolution of approximately 500 m; resolution in the western Sound is considerably higher. The grid has 10 evenly spaced levels in the vertical. ROMS uses a split-explicit time-stepping scheme; the time step used for this study was 6 s for the external mode and 120 s for the internal mode. A third-order upstream-biased advection scheme was used for both the momentum and tracer equations. The horizontal eddy viscosity coefficient was scaled by the grid spacing with an average value of approximately 10 [m.sup.2] [s.sup.-1]; no explicit horizontal tracer diffusivity was defined. The vertical eddy coefficients were calculated using a k-[omega] turbulence closure. Crowley (2005) has shown that the model has very good skill for the prediction of temperature and salinity, both for magnitude and vertical structure. Figures 2, 3 and 4 show examples of comparisons between observed and modeled surface and bottom temperature and salinity for the months of May, July and August; these are taken from Crowley (2005).

[FIGURES 1-4 OMITTED]

Because the model domain shown in Figure 1 does not include the East River tidal strait, we elected to carry two pesticide tracer equations during the simulations. One provides tracer concentrations forced by pesticide loadings applied locally or directly to the model domain. A second provides tracer concentrations forced by pesticide loadings applied remotely, outside of the model domain within the East River. These tracer concentrations were then forced through tracer boundary conditions specified at the western end of the basin.

Open Boundary Conditions

Surface elevation was prescribed at the open boundaries using available observations. Elevation time series were derived from hourly observations at Willets Point in the far western Sound, and Montauk Point, Vails Beach, Sandy Point and Point Judith in Block Island Sound. The data were linearly interpolated along the southern and eastern boundaries, but the Willets Point data were applied uniformly across the East River. Periods of missing data were filled by constructing tidal predictions using harmonics obtained from the existing data and adding the nontidal component extracted from a nearby station. In the external mode, the depth-averaged velocities normal to the boundaries were calculated from a linearized momentum equation. A modified Orlanski radiation condition was applied to the internal mode velocities, as well as the tangential velocities in the external mode. Temperature and salinity at the open boundaries were nudged toward observed values on inflow.

Pesticide tracer fields were nudged to zero on the open boundaries at the eastern end of the basin. At the western end of the domain, a radiating condition was applied to the locally forced tracer field; this allowed transient disturbances to propagate through the boundary. On the western boundary the remotely forced tracer fields were nudged strongly (6-hour relaxation time during periods of inflow to the Sound) to a specified time varying concentration obtained by dividing the total mass discharged in a specified time into the East River by the volume of the river.

Initial Conditions and Physical Forcing

The model was started from rest using initial temperature and salinity fields derived from available Connecticut Department of Environmental Protection hydrographic data. To create an initial field, each of the temperature and salinity profiles from a survey was first interpolated vertically to standard depths. The data at each level were then interpolated to the grid using optimal interpolation with a value of 12 km for the radius of influence. Finally, the values at each cell were vertically interpolated to the depths of the local s-coordinate levels. The initial values for temperature for surface and bottom sigma-levels are shown in Figure 5. The distribution of bottom temperature emphasizes the dependence on water column depth. The model fields adjust quickly, thereby limiting the influence of inaccuracies in initial conditions to only a few days. Initial pesticide concentrations were set to zero.

[FIGURE 5 OMITTED]

The discharge from 4 rivers: the Housatonic, the Quinnipiac, the Connecticut and the Thames, are applied to the model as point sources. Daily-mean streamflow data for seven stations within the drainage basins of these rivers were obtained from United States Geological Survey. Figure 6 shows that river discharge was low until Hurricane Floyd on September 16th. The surface heat and momentum fluxes were calculated by the model using the bulk parameterizations of Fairall et al. (1996). These calculations require meteorologic data consisting of wind velocity (Fig. 6), air temperature, relative humidity and air pressure (Fig. 7); as well as rainfall, cloud cover and incoming shortwave radiation. Hourly observations collected at LaGuardia International Airport, along with pyranometer data measured at Brookhaven National Laboratory in Upton, New York were used for these simulations. Crowley (2005) has discussed the correction of LaGuardia meteorological data for over-water application. Figures 6 and 7 show the fluctuations in wind, pressure and air temperature associated with the cold front passage on August 30th and Hurricane Floyd on September 16th.

[FIGURES 6-7 OMITTED]

Pesticide Loadings

The 3 pesticides that are considered here are malathion, resmethrin and sumithrin. They were applied in counties or boroughs of Queens, Bronx, Westchester, Manhattan, Nassau and Suffolk during August through October. The specific period for the hindcast simulations described here was August 1, 1999 through October 19, 1999. Pesticide application data were supplied by NYDEC. These data were itemized by zip code; application data from land-locked areas more than two zipcodes from the coast were excluded from the model simulations. These exclusions resulted in slightly lower and more discriminatory loadings for each of the three pesticides, as compared with the loadings used by Miller et al. (2005). Factoring in those exclusions, the total loads of active ingredient applied in 1999 to the Long Island Sound basin were 2819 kg of malathion, 61 kg of resmethrin and 102 kg of sumithrin. For purposes of running model simulations, weekly loadings distributed by zip codes were applied to coastal grid cells as surface fluxes (kg [m.sup.-2] [s.sup.-1]). Figures 8, 9 and 10 show the spatial distribution of loadings (kg) in each coastal cell of the model summed over the entire simulation period for malathion, resmethrin, and sumithrin, respectively. Malathion applications were concentrated in Queens, Bronx, eastern Nassau and eastern Suffolk. Resmithrin applications were similarly concentrated in Queens and Nassau-Suffolk. Sumithrin loadings were high in Westchester and in Bronx. The total loads of active ingredient applied in 1999 to the East River reservoir were 1402 kg of malathion, 37 kg of resmethrin and 24 kg of sumithrin. For purposes of running model simulations, weekly loadings were applied to the East River reservoir with an assumed volume of 2.0745 x [10.sup.8] [m.sup.3] to obtain concentrations used to nudge the western boundary.

[FIGURES 8-10 OMITTED]

In the tracer equations, pesticides are lost through first order decay. The half-lives assumed in the calculations were as follows: malathion--48 h, resmethrin--2 h in the surface sigma level, 120 h elsewhere, sumithrin--10 h in the surface sigma, 120 h elsewhere. The half-life for malathion is consistent with base catalyzed hydrolysis at a seawater pH of 8 in the temperature range of 20[degrees]C to 25[degrees]C (Wolfe et al. 1977, Freed et al. 1979, Wang & Hoffmann 1991). It is assumed that the decay of the two pyrethroids is controlled by more rapid photolysis in the surface layer of the model and by slower abiotic or biotic transformations in deeper zone. Photolytic half-lives are assumed to be somewhat longer than the 12-47 min (Rand 2002) reported for clean waters because of the diurnal cycle, shading by suspended particles, and sorption of these hydrophobic chemicals. There is a paucity of data of peer-reviewed degradation rates for sumithrin, and half-lives of 9-14 h are given on the EXTOXNET website. There is also a paucity of data on decay rates for either pyrethroid that would apply to the dark water column, and the value of 5 days that is used here is based on observations of rapid decay in the environment after spraying and on the 36-day half-life for aerobic metabolism of resmethrin determined for pond sediment (Rand 2002); the shorter half life used (5 days) reflects the expectation that sorption to pond sediment would have retarded the rate of microbial degradation, and that structurally related sumithrin would have a similar rate of loss in the water column. Note for loadings applied in the East River, only the longer half-lives were used for resmethrin and sumithrin.

RESULTS

Analysis of the model hydrodynamics should be interesting during this period of unique meteorological forcing, but that is beyond the scope of this project. Results presented here are limited to a characterization of only the time varying pesticide concentrations over the period of simulation. As discussed earlier, the concentration distributions for each of the three pesticides are partitioned into a locally forced and a remotely forced concentration field. Note that the concentration fields presented in the following sections are for the lowest or bottom sigma level, they represent daily averages, and only concentrations >0.001 [micro]g/L are presented. The concentration fields provide a wealth of information concerning the detailed patterns in time varying concentration distributions.

Malathion

Model-simulated concentrations fields for malathion associated with loadings directly into the basin showed significant bottom concentrations within the period August 30th to October 15th. These concentration fields are characterized by two distinct plumes. One plume develops in the far western sound from applications in Queens and Bronx; concentrations >0.001 [micro]g/L, are confined to an area west of 73.6[degrees]W, concentrations >0.5 [micro]g/L are confined to coastal embayments. Figure 11 shows the detailed structure for maximum bottom concentrations that occurred on September 9th. A second distinct plume develops off Long Island in response to applications in Nassau and Suffolk. Details of the along-shore and across-shore extent of this plume are shown in Figure 12 for maximum bottom concentrations that occurred on October 3rd; bottom concentrations above 0.5 [micro]g/L are confined to coastal cells.

[FIGURES 11-12 OMITTED]

Figure 13 shows the spatial patterns for bottom concentrations in the plume associated with malathion dispersing into the western Sound from the East River. Maximum concentrations occurred on September 10th. Maximum concentrations are at the East River boundary and attenuate eastward; concentrations above 0.001 [micro]g/L are again confined to an area west of 73.65[degrees]W.

[FIGURE 13 OMITTED]

Resmethrin

Model simulations showed detectable bottom concentrations for resmethrin associated with local loadings into the basin over the basin period September 19th to October 15th. The plumes in the far westernsound from applications in Queens and Bronx are confined to the immediate vicinity of cells that were loaded. A significantly larger plume develops off Long Island in response to applications in Nassau and Suffolk; maximum spatial extent is attained on October 3rd (Fig. 14).

[FIGURE 14 OMITTED]

Figure 15 shows the spatial pattern for bottom concentrations in the plume associated with resmethrin dispersing into the western Sound from the East River. As with the malathion plume, concentrations are maximum at the East River boundary and attenuate eastward; maximum spatial extent is attained on September 22nd.

[FIGURE 15 OMITTED]

Sumithrin

Time varying concentrations fields for sumithrin associated with loadings directly into the basin showed detectable bottom concentrations over the period September 10th to October 15th. These concentration fields emphasize an interaction of transient plumes from applications in Queens, Bronx and Westchester. The detailed structure of this composite plume for maximum bottom concentrations on September 25th shown in Figure 16; maximum concentrations are confined to coastal cells with high concentrations off Westchester and concentrations decrease rapidly offshore. The distinct plume produced by applications in Nassau and Suffolk attains maximum extent on September 30th. Figure 17 shows the spatial extent of the plume associated with sumithrin dispersing into the western Sound from the East River. Maximum eastward extent is attained on September 16th.

[FIGURES 16-17 OMITTED]

DISCUSSION

The concentration distributions shown earlier emphasize that plumes associated with pesticide application at the perimeter of the basin are confined to near-shore regions. There is little evidence for bottom concentrations >0.001 [micro]g/L within the interior of the basin for any of the three pesticides considered. These concentration distributions are the result of the combined effects of cross-shore and along-shore horizontal dispersion and first order decay. Concentration distributions were strongly influenced by the imposed loading distributions around the perimeter of the basin.

Concentrations in plumes associated with pesticides dispersing into the western Sound from the East River attenuated rapidly with distance from the western boundary. The eastward extent of these plumes should be influenced by the combined effects of longitudinal dispersion, longitudinal transport and first order decay. Crowley (2005) has shown that the net nontidal longitudinal volume transport in the western Sound is correlated with the longitudinal sea surface slope (Fig. 6). The total concentration could be considered as the sum of concentrations in the plumes produced by local and remote loading.

The concentration distribution for malathion is characterized by several localized and distinct plumes. A persistent plume off the coast of Nassau and Suffolk counties was confined near the coast and dispersed along the coast; the dispersion pattern was presumably influenced by circulation off the coastal promontories. A second localized plume developed in the far western Sound in response to applications in Queens and the Bronx. A third plume developed in the western Sound in response to malathion dispersing from the East River. This plume would interact with the plume produced by loading in Queens and Bronx.

The most prominent feature in the concentration distribution for resmethrin was the persistent plume off the coast of Nassau and Suffolk counties. Plumes that developed in the far western Sound in response to applications in Queens and Bronx were of very limited spatial extent. The plume that developed in the western Sound in response to pesticide dispersing from the East River showed maximum spatial extent during the third week of September, possibly in response to eastward directed nontidal volume transport through the East River.

Sumithrin exhibited a very narrow coastally confined plume off the coast of Nassau and Suffolk counties. The most prominent feature was the plume that developed along the coast of Westchester county in the far western Sound. This plume interacted with plumes produced by applications in Queens and Bronx. The plume that developed in the western Sound in response to sumithrin dispersing from the East River showed maximum spatial extent during the beginning of the third week of September.

It should be noted that we consider these concentration distributions conservative because we included landlocked zip codes into our coastal cell loadings and assumed that all of the pesticide sprayed on land reached Long Island Sound. This includes concentrations predicted for direct loadings into the basin and those predicted for remote loadings into the East River.

ACKNOWLEDGMENTS

This project received support from the National Sea Grant College Program and the National Marine Fisheries Service of the United States Department of Commerce's National Oceanic and Atmospheric Administration under award # NA16RG1354 to the Research Foundation of State University of New York for New York Sea Grant. The views expressed herein do not necessarily reflect the views of any of those organizations.

LITERATURE CITED

Crowley, H. 2005. The seasonal evaluation of thermohaline circulation in Long Island Sound. PhD thesis, Marine Sciences Research Center, Stony Brook University, Stony Brook, NY. 134 pp.

EXTOXNET. Pesticide Information Profiles, Available at: http:// ace.orst.edu/info/extoxnet/pips/ghindex.html

Fairall, C. W., E. F. Bradley, D. P. Rogers. J. B. Edsom & G. S. Yang 1996. Bulk parameterization of air-sea fluxes for Tropical Ocean-Global Atmosphere Coupled-Ocean Atmosphere Response Experiment. J. Geophys. Res. 101:3747-3764.

Freed, V. H., C. T. Chiou & D. W. Schmedding. 1979. Degradation of selected organophosphate pesticides in water and soil. J. Agric. Food Chem. 27:4.

Haidvogel, D. B., H. G. Arango, K. Hedstrom, A. Beckman, P. Malanotte Rizzoli & A. F. Shehepetkin. 2000. Model evaluation experiments in the North Atlantic Basin: simulations in nonlinear terrain-following coordinates. Dynamics of Atmospheres and Oceans 32:239-281.

Miller, R. L., J. R. Wands, K. N. Chytalo & R. A. D'Amico. 2005. Application of water quality modeling technology to investigate the mortality of lobsters (Homarus americanus) in western Long Island Sound during the summer of 1999. J. Shellfish Res. 24(3):859-864.

Rand, G. M. 2002. Hazard assessment of Resmethrin: I. effects and fate in aquatic systems. Ecotoxicol. 11:101-111.

Wang, T. C. & M. E. Hoffmann. 1991. Degradation of organophosphorous pesticides in coastal water. J. Assoc. Off. Anal. Chem. 74:883-886.

Wolfe, N. L., R. G. Zepp, J. A. Gordon, et al. 1977. Kinetics of chemical degradation of malathion in water. Environ. Sci. Technol. 11:88-93.

ROBERT E. WILSON, * HEATHER A. CROWLEY, BRUCE J. BROWNAWELL AND ROBERT L. SWANSON

Marine Sciences Research Center, Stony Brook University, Endeavour Hall, Stony Brook, New York 11794

* Corresponding author. [e-mail]