Western Long Island Sound (wLIS) has experienced a long-term decline in the July/August summer minima bottom dissolved oxygen concentrations. This decline continues despite New York City having eliminated routine raw discharges of sewage, upgraded sewage treatment to nearly complete secondary, and introduced nitrogen control. It is our conclusion that long-term changes in physical oceanographic processes are having an impact on the hypoxia problem in wLIS. Specifically, we show that interannual variations in summertime thermal and haline stratification contribute to variations in vertical mixing which controls the ventilation of bottom waters. Analyses of bottom dissolved oxygen and density stratification point directly to the importance of wind-induced current shear in controlling stratification and vertical mixing; numerical simulations support this result. Interannual variations in both the direction and directional constancy of summertime winds over wLIS are shown to control the ventilation of bottom waters and thereby the seasonal development of hypoxia.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Since the mid-1980s, western Long Island Sound (wLIS) has experienced a number of summers when hypoxia (bottom dissolved oxygen concentrations less than 3 mg L−1) occurred. The summers of 1987, 1988, 1990, 1998, and 2002–2006 are notable. Hypoxic conditions are routine, the impacted geographic area is extensive (often on the order of 400 km2), and the minimum concentrations declining.
 In the 1970s, Koppelman et al.  and D. F. Squires (Marine Sciences Research Center, State University of New York at Stony Brook, unpublished data, 1971) noted that wLIS was prone to low-dissolved oxygen (DO) concentrations in bottom waters. Parker and O'Reilly  reviewed the historic DO data for wLIS and indicated that the severity and extent of low-DO conditions appeared to be occurring with greater frequency. They speculated that changes in sewage treatment practices by New York City (NYC) might be a contributing factor. Specifically, that a higher fraction of the total nitrogen load is dissolved and retained in the surface layer, where estuarine transport carries it eastward into the sound. Welsh  also suggested that the increasing tendency toward hypoxia in wLIS might be related to upgrading of NYC Water Pollution Control Plants (WPCPs) to full secondary treatment. She speculated on the continued importance of dissolved organic nitrogen which is difficult to remove with secondary treatment. She also speculated that reduced turbidity and enhanced production could compensate for reductions in inorganic nitrogen. More recently, O'Shea and Brosnan  discussed general trends of several parameters associated with eutrophic conditions in wLIS. They concluded that over the last decade or so, there has been a decline in bottom DO that may be linked to increasing temperature stratification.
 There are six WPCPs along the East River (Figure 1), the tidal strait that connects LIS with the Atlantic Ocean through New York Harbor; four of these are on the upper East River. Statistics regarding these plants are provided in Table 1. All facilities are now operating at full secondary treatment with the exception of Newtown Creek. Sewage now treated by the Red Hook WPCP had previously entered the East River untreated. Full secondary treatment at the other four WPCPs had been attained prior to 1970. It should also be noted that a small screening plant located on Hart Island was shut down in the 1970s and its influent rerouted to the Hunts Point WPCP. The plants were upgraded in the mid-1970s. Thus, by the mid-1980s, sewage treatment along the East River was almost totally compliant with the requirements of the Federal Water Pollution Control Act of 1972 (P.L. 92–500).
 Secondary treatment converts organic nitrogen to inorganic nitrogen and it is generally believed that these inorganic forms are more biologically available [Parker and O'Reilly, 1991]. This coupled with the rapid advection of the relatively fresh surface waters containing this inorganic nitrogen from the East River into the more sluggish wLIS can result in increased primary production in wLIS. Water typically has a residence time of several days in the upper East River compared to several months in wLIS [Welsh, 1995]. Ultimately, the interaction of these processes could lead to supersaturation of DO in surface waters [O'Shea and Brosnan, 2000] and DO depletion in bottom waters of wLIS. Welsh and Eller  emphasized the importance of physical oceanographic processes in controlling and maintaining hypoxia in wLIS. Swanson et al.  suggested that while the above processes may be operative, oceanographic climatic processes may also be contributing to increased likelihood of hypoxia.
 The hypoxia issue in wLIS became even more complex in the early 1990s when NYC complied with the requirements of the Ocean Dumping Ban Act (P.L. 100–688). In order to reuse or otherwise dispose of sewage sludge on land, the city commenced dewatering sewage sludge using centrifuges at selected WPCPs. The centrate from this process was rerouted through the WPCPs. Swanson  estimated that the dewatering process would increase the nitrogen load entering the East River WPCPs by about 24 t d−1, a 38% increase relative to 1990. In the hope of reducing the likelihood of hypoxia in wLIS, the Long Island Sound Study (LISS)  established a goal that WPCPs should reduce nitrogen effluent discharges to pre-1990 levels.
 In the late summer and fall of 1999, concern about hypoxia was once again heightened as wLIS experienced mass mortalities of the American lobster (Homarus americanus), the single largest commercial fishery remaining in LIS. The pattern of mortalities is consistent with the outbreak of a communicable disease whose cause is currently unknown. However, lobstermen and others raised the possibility that the infection was a consequence of stress, possibly brought on by hypoxia [Coastlines, 2000]. This latest crisis has thus reignited concerns about hypoxia in wLIS.
 It is our contention that the hypoxia problem in wLIS is more complex than an ecosystem response to nutrient stimulation, what Cloern  has identified as the phase I conceptual model of coastal eutrophication. Rather, wLIS falls into his phase III model based around five concepts to guide coastal science. These concepts include coastal system attributes (including but not limited to physical attributes) that modulate the response to nutrient enrichment as well as other environmental stressors; the forcing functions including nutrient enrichment and climate variability; linkages among ecosystem stressors and associated responses; and management strategies to ameliorate environmental impacts.
 We examine the influences of physical water column processes on hypoxia in wLIS using the historic environmental water quality data of the NYC Department of Environmental Protection (NYC DEP) as well as the nitrogen discharge data from the four WPCPs along the upper East River. Our objective is to show that physical water column processes have a pronounced effect upon the hypoxia problem of wLIS and that simply reducing nitrogen loadings may not achieve the desired improvements.
 Analyses of the relationship between bottom dissolved oxygen and water column structure in wLIS were motivated by analyses presented by HydroQual . They used linear regression to establish a relationship between summertime thermal stratification and bottom DO in wLIS for the period 1963–1993. For this period they found, for example, that 0.47 of the variance associated with interannual variations in bottom DO averaged from 15 July to 15 August for each year, was explained by variations in averaged surface minus bottom temperature. They found also that when they limited their analysis to 1981–1993, the variance fraction explained increased to 0.74. It is potentially important to note that HydroQual  had to subject their data to a 3-year moving average to achieve these relatively high values for R2, thereby removing variance associated with fluctuations with periods shorter than 3 years. This paper complements the HydroQual  analyses in several respects. First, we have extended the data set in time to include the period 1994–2006 in order to update the description of interannual variations in bottom DO. Second, we have examined the relationship between temporal variations in summertime bottom DO and total density stratification for the period 1988–2006 for which good quality salinity data were available with the objective of evaluating the contribution of haline stratification. Third, we have examined the relationship between short-period fluctuations in density stratification and external forcing to define the mechanisms responsible for producing variations in both stratification and bottom DO in wLIS. In this connection we have considered recent three dimensional modeling results for wLIS by R. E. Wilson et al. (Circulation and mixing in western Long Island Sound during a major summertime hypoxia event, submitted to Estuaries, 2008) (hereinafter referred to as Wilson et al., submitted manuscript, 2008) and interesting results concerning of the ventilation of bottom waters of wLIS by wind [O'Donnell et al., 2006].
2. Data and Methods
 We have based our analyses primarily on the extensive database of the NYC DEP. The NYC DEP routinely conducts a water quality survey throughout the harbor, usually in surface and bottom waters, including 11 stations in the upper East River and wLIS. Beginning in 1946, the sampling period is weekly to biweekly from June through September. Sampling was expanded seasonally beginning with the Long Island Sound Study in 1988. We have also considered time series of total nitrogen discharged from the four upper East River WPCPs as an indicator of anthropogenic forcing of wLIS.
 To examine recent trends in bottom DO (DOb) over the past two decades we used simple regression techniques applied to monthly averaged data from NYC DEP Stations E4 (Hell Gate) in the East River and E10 (Hart Island) in wLIS (Figure 1). NYC DEP stations E4 and E10 are the most relevant to this analysis as they are at the extremities of the area of interest and in the deep part of the main waterway; many of the other NYC DEP stations are located in adjoining embayments. Station E4 is the closest sampling location to the Wards Island WPCP and has a depth of 21 m. Station E10 is located in 27 m of water and is the deepest along the East River transect. Station E4 is well flushed and E10 is in the more sluggish wLIS.
 Temperature and DO data are available for E10 from 1946 to 2006. The salinity record is less complete than that for temperature and DO. Surface and bottom salinity data became available in 1966, but temperature-corrected specific gravity hydrometers were used until 1986. A calibrated YSI instrument was used from 1986 to 1991. Beginning in 1992, NYC DEP has used a SEABIRD SBE 25-03 CTD to make temperature and salinity measurements at approximately 1 m below the water surface and 1 m above the bottom. The data cover a considerable length of time, and sampling protocols and analysis techniques have changed. Discussion of contemporary analysis techniques is covered in Appendix 2 of the New York City 1999 Regional Harbor Survey [Swanson et al., 2000]. From 1920 to 1984, the azide modification of the Winkler method was used to measure dissolved oxygen. From 1985 through 1987, YSI meters were used in addition. These were calibrated each sampling day using the azide modification of the Winkler method [American Public Health Association, 1985]. The Winkler method was used almost exclusively after 1988.
 The availability of salinity data from 1986 to 2006 allows us to evaluate the relative contributions of thermal and haline stratification to the total density stratification. These data also allow us to describe the extent to which fluctuations in the haline and thermal contributions covary, and the relationship between fluctuations in density stratification and DOb. As part of our analysis of the response of DOb to variations in stratification, time series data for DO and density stratification at E10 are used to evaluate terms in the two-layer model proposed by HydroQual  relating DOb to vertical exchange and water column and benthic consumption at E10. Additionally, we examined the relationship between short-period fluctuations in density stratification and wind direction. This analysis was facilitated by results from a numerical study (Wilson et al., submitted manuscript, 2008) of wind-forced circulation and mixing in wLIS during the 1988 hypoxia event which define the directional response of water column to winds. Finally, time series for total nitrogen effluent loadings from the four NYC Upper East River WPCPs (Figure 1) were developed and considered to be representative of local anthropogenic forcing.
3.1. Long-Term Trends in Dissolved Oxygen and Thermal Stratification
 Time series for DOb at E10 averaged over July and August (Figure 2) show a long-term decline with marked interannual variability. We have partitioned the records into two segments: 1946–1985 and 1986–2006 because of changes in instrumentation and sampling protocols, as discussed in section 2. The rate of decrease in DOb for the early segment is 0.031 mg L−1 a−1 (Figure 2); it increases to 0.061 mg L−1 a−1 for the second segment. The variance for DOb is very broad banded but it has a peak in a band centered on approximately 15 years.
 The time series for bottom temperature (Tb) at E10 averaged over July and August (Figure 3) also shows a long-term decline with significant interannual variability. The rate of decrease in Tb for the early segment is 0.028°C a−1 and 0.039°C a−1 for the later segment. The time series for Tb and to some extent for DOb suggest that for 1986–2006 shorter-period fluctuations make an increased contribution to the variance.
 Surface temperature (Ts) at E10 averaged over July and August shows long-term variability, but the decline in Tb is the primary factor contributing to a long-term increase surface minus bottom water temperature (ΔT) (Figure 4) with an associated increase in density stratification αΔT where α is the thermal expansion coefficient. The rate of increase in ΔT at E10 averaged over July and August is 0.013°C a−1 for the early segment, and 0.018°C a−1 for the later segment. This long-term increase in ΔT is associated with the higher rate of decrease in Tb.
 Relationships between DOb and ΔT (Figures 2 and 4) were examined through simple regression analysis applied to the two segments of the records. We did find a very weak relationship between July/August DOb and ΔT with a regression coefficient of approximately −0.04 mg L−1 °C−1 for both record segments, but the variance explained was extremely low and the p value was high. Furthermore, for the later record segment, regression of DOb against Δρ rather than ΔT, where Δρ is the bottom minus surface density, did not result in an improved relationship. The weakness of these relationships is in contrast to the findings of HydroQual  which, however, were based on data subject to a 3-year moving average thereby removing short-period variance. This emphasizes that a significant fraction of the interannual variance in summertime DOb does not have a simple linear dependence on thermal stratification averaged over July and August.
 Comparisons between surface and bottom dissolved oxygen data from stations E10 and E9 confirm that E10 is indeed representative of the deep stations in wLIS in the vicinity of Hart Island (Figure 1). The correlation coefficient for DOb concentrations at stations E9 and E10 during June through September for 1999 and 2000, for example, was 0.9. In direct contrast to E10, the mean July/August surface DO (DOs) and DOb have improved considerably at station E4 over the period 1985–2000. The improvement for both surface and bottom waters is significant at the 5 percent level and R2 = 0.78 and 0.75, respectively (Figure 5).
3.2. Influence of Stratification on Vertical Mixing and Dissolved Oxygen
 The availability of good quality salinity data beginning in 1986 allowed us to examine the relationship between short-period fluctuations in total density stratification Δρ (thermal plus haline) and DOb for the period 1986–2006. HydroQual  and O'Shea and Brosnan  considered only the relationship between DOb and ΔT. For April through September, the thermal and haline contributions to Δρ are comparable. The haline contribution dominates in spring and late summer while the thermal contribution during midsummer. Density stratification is variable and characterized by fluctuations with periods ranging from two weeks (the approximate Nyquist period for these observations) to one month, and the haline and thermal and contributions tend to covary positively. 1988 and 2004 are examples of severe hypoxia years, while 1992 and 2000 exhibited relatively high July/August DOb (Figure 2). A comparison of summertime density stratification in 1988 with that in 1992 (Figure 6) shows that one significant difference was the persistence of the stratification; 1992 was characterized by frequent destratifying events whereas 1988 was characterized by more stable stratification. Similarly, in 2000, stratification was characterized by repeated destratifying events beginning in August, and in 2004 a moderately low but stable stratification persisted throughout the summer. Another difference between the stratification in 1988 and 1992 was that total stratification in 1988 tended to be greater than that in 1992 and with a greater thermal contribution.
 Time series for DOs and DOb shown in Figure 6 indicate that fluctuations in DOb covary negatively with fluctuations in Δρ, and that the rate of decline in DOb increases during increasing stratification and decreases or even reverses during destratifying events. Figure 6 emphasizes the influence destratifying events have on the seasonal decline in DOb. The degree of summertime hypoxia is controlled by the frequency and intensity of destratifying events, and hypoxia years are characterized by prolonged periods of relatively stable density stratification during July and August.
3.2.1. Mechanisms Controlling Stratification and Vertical Mixing
 Results from a model study of circulation and mixing in wLIS during the 1988 hypoxia event (Wilson et al., submitted manuscript, 2008) emphasize that vertical mixing in wLIS is episodic, and that time variations in vertical mixing in the interior of the water column are controlled by variations in vertical density stratification which in turn are driven by wind-induced fluctuations in the vertical shear in along-channel current (Figure 7). Results shown in Figure 7 are for a station in wLIS at 40.874°N, 73.733°W and are from hindcast numerical simulations for LIS during 1988 performed by Crowley  using ROMS with a k-ω closure. These simulations involved assimilation of temperature and salinity observations within the basin, forcing with observed sea level, temperature and salinity on the open boundaries, and forcing with surface wind stress and heat flux derived from observed meteorological variables. Wilson et al. (submitted manuscript, 2008) showed that vertical current shear and stratification are highly dependent on wind direction. They found a maximum positive correlation with surface wind stress directed along 23°T, while wind stress along 203°T was most effective in causing destratification. These wind directions which give maximum response are not exactly along the local channel axis which is 42°T, but approximately 25° to the left (counterclockwise). Wilson et al. (submitted manuscript, 2008) interpreted this as an effect of rotation. This directional dependence indicates that the straining mechanism described by Scully et al.  in the York River and as seen by O'Donnell et al.  in wLIS is operative. This mechanism is distinct from the wind stirring mechanism which is proportional to wind speed cubed and independent of wind direction.
 The time history of modeled vertical eddy diffusivity in Figure 7. indicates that there are basically two mixing states. During brief periods of reduced or negative current shear, density stratification decreases and vertical mixing increases abruptly and remains active until the shear changes sign. After restratification, vertical mixing becomes extremely weak in the interior of the water column. Figure 7 shows also that wind-induced current shear and stratification are both surface intensified.
3.2.2. Bottom Dissolved Oxygen Response
HydroQual  introduced a quasi-steady two-layer model to relate changes in DOb to vertical mixing and to the combined effects of water column respiration and benthic demand; horizontal advection is unaccounted for. They applied it to observations at E10 for the summer of 1988; in their notation, the model is
where DOb and DOs are bottom layer and surface layer DO concentrations, V is the volume of the lower layer, Kz is vertical eddy diffusivity, Ai and Ab are the areas of the interface and bottom, Δz is the thickness of the pycnocline, R is community respiration rate for the lower layer, and S is benthic demand. Using a simplified formulation for Kz = βΔz/Δσ which varies inversely with the vertical density gradient, where Δσ is the density difference between the upper and lower layers, they obtained the relationship:
and used it to estimate the total water column and benthic consumption rate of 0.28 mg L−1 d−1 at E10 for August 1988. In (2)β is not to be confused with ∂ρ/∂S. Using time series data DOb, DOs and Δσ we regressed ΔDO against Δσ to obtain estimates for [R+S/h]h/β for 1988–2006. For the values of β and h used by HydroQual , the mean value for [R+S/h] estimated for the 19 years of data is 0.22 mg L−1 d−1 with a standard deviation of 0.068 mg L−1 d−1. Results from this steady state model imply that in 10 days, DOb would decrease by approximately 2.2 mg L−1 without significant ventilation.
 In light of observations shown in Figure 6 emphasizing the response of DOb to transient mixing events, we considered the time-dependent version of (1)
and defined two mixing states for the interior of the water column. During periods of increasing density stratification (very weak mixing), we estimated [R + S/h] simply as ΔDOb/Δt. During periods of decreasing density stratification (active mixing), we used (3) to calculate vertical exchange as the difference between ΔDOb/Δt and the time interpolated of estimates for [R + S/h].
 Results (Figure 8) provide estimates of terms in (3) for 1988 and 1992 as examples of hypoxic and nonhypoxic years, respectively. These emphasize that the balance between consumption and vertical exchange controls the seasonal evolution of DOb. For 1988 between the first week in May and the second week in August, only one mixing event produced vertical exchange adequate to exceed consumption which had a median value of −0.18 mg L−1 d−1. For 1992, at least four mixing events produced vertical exchange which exceeded or approached the consumption rate of −0.16 mg L−1 d−1 between the first week in May and the second week in August. The hypoxic years are characterized by prolonged periods of weak vertical exchange, and more frequent and more intense vertical exchange characterizes the nonhypoxia years. As discussed below, the vertical exchange was dependent on wind direction. It is also noteworthy that the estimates for the median values of [R+S/h] for hypoxia years 1988 and 2004 of −0.18 and −0.22 mg L−1 d−1, were consistently higher than for the nonhypoxia years 1992 and 2000 with estimates of −0.16 and −0.09 mg L−1 d−1, respectively.
 Almost 1.9 × 106 m3 d−1 of secondary sewage effluent are discharged by four NYC-operated WPCPs to the upper East River. The Wards Island WPCP discharges at Hell Gate; the three others discharge to the northeast of Hell Gate (Figure 1). Hell Gate is considered by oceanographers to be the boundary between LIS and the East River because it represents a demarcation in vertical stratification; waters to the southwest are typically well mixed whereas the waters to the northeast are vertically stratified in summer. Water parcels have a residence time of 2 to 3 days in the East River [Caplow et al., 2004]. The residence time in the western LIS is on the order of several months [Welsh, 1995].
 Nitrogen released as effluent on an annual basis from all East River WPCPs and raw discharges remained relatively unchanged over the period 1960–1990 [Swanson, 1993]. This occurred despite a 118% increase in flow from WPCPs relative to 1960, a nearly 100% reduction in raw discharges, significant upgrades to existing facilities, and construction of the Red Hook WPCP.
 In the early 1990s, annual average DIN loading from the four upper East River WPCPs increased from 35.3 t d−1 to 44.9 t d−1, a consequence of sewage sludge dewatering associated with the requirement to end ocean dumping (Figure 9). By 2002, this discharge dropped to 28.1 t d−1 as part of a removal program implemented by the city. However, by 2005 the nitrogen load once again was on the increase, reaching 38.7 t d−1. Still, this was nearly a 14% decrease from the peak in 1993. The increase from 2002 to 2005 was apparently associated with upgrading construction occurring at WPCPs.
 The mass load of inorganic nitrogen delivered to the waters of the wLIS has been targeted as a major factor contributing to the area's hypoxic conditions [LISS, 1994]. Improvement in the quality of sewage effluent from East River WPCPs since the 1970s and 1980s has led to better water quality in the East River (NYC DEP Station E4) through reductions in organic carbon and total suspended solids [Swanson et al., 2000]. However, Welsh  reports that within NYC's WPCPs, organic forms of nitrogen are converted to more biologically available inorganic forms. This nitrogen load is quickly advected to wLIS where it stimulates production of phytoplankton (as evident in the high surface and at times supersaturated DO concentrations particularly in the mid-1990s) that eventually dies and settles or is exported in zooplankton feces to the relatively quiescent bottom waters as highly labile organic matter. The increased load of organic matter consumes much of the limited amount of DO in bottom water of wLIS in summer, thus often creating hypoxic conditions. The implication is that now there is possibly a more direct linkage between nitrogen loadings from WPCPs and DOb concentrations in the western Sound than there was some three or more decades ago when sewage was untreated or less than fully secondary treated. Of course, the East River has benefited by the reduced biological oxygen demand (BOD) resulting from secondary treatment. From the 1960s to the 1990s, BOD declined some 60% while the summer DOb minima increased by 1.3 to 1.8 mg L−1 [O'Shea and Brosnan, 2000].
 However, considering the recent NYC harbor water quality surveys and WPCP discharge data, this latter explanation seems too simplistic. Assuming that conventional wisdom is correct, DOb concentrations would have responded accordingly to the DIN loadings. That is, July/August DOb concentrations would have declined dramatically in the early 1990s, remained low in the mid-1990s, and then rebounded as DIN loadings declined beginning in 1997. The July/August DOb concentrations reached a minimum in the late 1980s. Following that, there was a period of limited interannual variation prior leveling off in the late 1990s. The first half decade of the new millennium experienced a significant decline in DOb concentrations.
 There is no obvious relationship between DIN from NYC WPCPs and DOb concentrations in wLIS. However, other sources of nitrogen such as from the atmosphere, the Connecticut and Housatonic Rivers, nonpoint source runoff, inputs from the Atlantic Ocean through eastern LIS, and numerous other but quite small WPCPs, may be important. Anderson and Taylor , for example, established that DOb concentrations are correlated to allochthonous inputs of NH4+ associated with precipitation. Thus storm water and combined sewer overflow (CSO) events may be more important than previously thought.
 We have shown that the time series for NYC DEP Station E10 extended to 2006 exhibit the long-term trends for July/August averaged DOb, Tb and ΔT. We partitioned our time series into two segments because of changes in instruments and sampling protocol including sampling depths in 1986. In contrast to the findings of HydroQual  we found only a very weak relationship between July/August averaged DOb and ΔT. We found that July/August DOb depended instead on the history of density stratification and vertical mixing.
 Model results by Crowley  and Wilson et al. (submitted manuscript, 2008) discussed in section 3.2.1 indicate that wind-induced variations in stratification have a direct effect on the evolution of summertime DOb. Direction histograms for July/August winds (Figure 10) calculated from winds at LaGuardia afford a comparison between a hypoxia  and nonhypoxia year . Wind direction during summer months for the hypoxic year had a modal value close to that which would produce maximum stratification with little directional variability. During the nonhypoxic year winds showed increased directional variability with a significant percentage of winds from the northeast.
5. Regional Winds
 Insight into the long-term variations in directionality of regional summertime winds is afforded by National Centers for Environmental Prediction/National Center for Atmospheric Research reanalysis surface monthly averaged surface wind data from 1948 to 2007 at grid point 40°N, 72.5°W (Figure 11). These data suggest a long-term change in July/August resultant wind direction with the direction approaching the preferred direction of 203°T which produces maximum stratification. There is also some evidence that the rate of change in wind direction has increased since 1980.
Figure 12 shows that the directional frequency of July/August National Climatic Data Center daily averaged winds at LaGuardia Airport for 1984–2007 is bimodal with significant interannual variability. The mean histogram for 1984–2007 has one distinct peak in a bin centered on approximately 70°T, and another at approximately 195°T; these two wind directions are associated with the U.S. East Coast summer ridge (70°T) and trough (195°T) system. The red lines on Figure 12 define the band of wind directions from 113°T to 293°T which produce water column stratification in wLIS (Wilson et al., submitted manuscript, 2008). A stationary trough system will, therefore, contribute to the maintenance of water column stratification. The percentage of summer wind from 113°T to 293°T calculated from results shown in Figure 11 is a useful discriminate between hypoxia (1988, 2003, 2004) and nonhypoxia (1992, 1995, 2000) years. However, another important index can be obtained by examining directional constancy defined as the ratio of vector mean wind speed to scalar mean wind speed times 100 [Swanson and Zimmer, 1990]. As discussed in section 4, not only was wind direction important but also variability. The interannual variations in the percentage of August winds which are directed from 113°T to 293°T and which have a constancy of greater than 50 are shown in Figure 13. Comparison with Figure 2 shows that this percentage index can discriminate between hypoxia and nonhypoxia years, and that a sufficient condition for August hypoxia is a percentage greater than approximately 40.
 NYC has reduced the quantity of nitrogen in the effluent of the four upper East River WPCPs over the period of 1993 through 2002. However, there was not a corresponding increase in the concentration of DO in bottom waters of wLIS, a result that had been anticipated in the Long Island Sound CCMP. In fact, the summertime DOb concentrations have declined over the past six decades. The correlation between interannual variations in minimum summer DOb concentrations (Figure 2) and nitrogen loads from NYC WPCPs (Figure 11) is low, and so the cause of hypoxia in wLIS appears to be more complex than nitrogen loadings from WPCPs. It is possible that benthic processes buffer DOb conditions. The effects of changes in nitrogen loadings may not register until some time in the future because sediments are a reservoir of BOD and organic nitrogen. However, it appears that climatic processes and specifically wind-induced mixing operating through the straining mechanism described by Scully et al.  and as seen by O'Donnell et al. , distinct from wind stirring proportional to wind speed cubed, play a dominant role in controlling the evolution of summertime hypoxia in wLIS. It should be noted in addition, that our two-layer model provided estimates for the consumption term which was lower in the nonhypoxic years than in hypoxic years. If the degree of super saturation in surface waters is considered as a proxy for relative production, we estimate using available temperature, salinity and dissolved oxygen that for 1991–2006 the average percent saturation of surface waters for February through June is 158. For nonhypoxic years 1995 and 2000 the average percent saturation for February through June was estimated to be 134 and 140, respectively. This could indicate that interannual variation in production, possibly also influenced by stratification, could contribute to interannual variations in DOb.
 Thus, despite concerns that low DO may be impacting the largest commercial fishery in LIS, there may be only limited local management actions that can reverse the apparent long-term trend toward decreasing DOb. Clearly, Cloern's  suggestion that coastal eutrophication is more complex than the traditional nutrient input and response model is applicable in wLIS.
 We want to thank NaJi Yao and Beau Ranheim of the NYC DEP for making the city's water quality data accessible. We appreciate the assistance of Heather Crowley, Ruta Rugabandana, and Nickitas Georgas in manipulating the data files. We are especially grateful to Christine O'Connell for reviewing historical NYC DEP data reports to evaluate changes in sampling protocol. We acknowledge very helpful comments and suggestions from two anonymous reviewers. This work was supported by U.S. EPA grants LI972862040 and LI 972862050 and NY State Sea Grant Institute grants NA46RG0090 and NA16RG1645.