Water Column Stratification and Mixing Events in Narragansett Bay During Winter and Spring

We deployed a buoy in the West Passage of Narragansett Bay, approximately one km east of Hope Island, within a bathymetric depression of approximately twice the surrounding water depth. The buoy contained an onboard logging computer and had three YSI-600 instruments, each capable of measuring temperature, salinity, dissolved oxygen, pH and pressure, logging at three different depths in the water column at a 20 minute sampling interval . The first deployment started on 6 January 1997 and lasted 18 days before problems required retrieving the buoy. Additionally, problems with the salinity sensors in this deployment made it impossible to use most of the salinity data. We fixed the salinity and electronic problems, modified the mooring system, and redeployed the buoy on 10 March 1997. We collected data for two months in this second deployment, until 8 May 1997. Using the temperature, salinity (when available), dissolved oxygen and pressure data from these deployments, we investigated the relationship between physical stratification as determined by density, and oxygen variations. We also used the predicted tides, observed tides from the moored, bottom instrument, wind data from the GSO Pier and from T. F. Green Airport, USGS stream flow data and NOAA-NWS precipitation data from T.F. Green Airport to investigate forcing relationships for the physical stratification and destratification cycles. Using the difference between the bottom and surface parameters as an index for stratification, we determined that the density related stratification results in increased vertical oxygen gradients, and that a 1 kg m-3 difference between surface and bottom in density is accompanied by approximately a l O % saturation difference over the same depth. The stratification cycles are related to the spring-neap tidal cycles for extended periods, and to strong wind forcing for shorter time scales. Large fresh water inputs can affect stratification as a result of estuarine entrainment whereby the vertical salinity gradient increases. This relationship suggests a linkage between the deep estuarine waters at the buoy station and the surface offshore waters, from Rhode Island Sound.

Calibration and profile-buoy offsets for salinity Page 74 sensors during March deployment Table 3 Calibration and profile-buoy offsets for dissolved Page 87 oxygen sensors during March deployment Table 4 ADCP transects and profiles indexed by filename Page 13 1 and figure number Table 5 Stratification and destratification periods in the Page 169 to 171 March deployment, and possible factors involved in causing stratification and destratification  Chart depiction of study area centered on Hope Page 8 Island Figure 3 Zoomed view ofbathymetry around Hope Island Page 12 Figure 4 Drawing of January deployment mooring system Page 17 Figure 5 Drawing of March deployment mooring system Page 20 Figure 6 Chart depiction of 5 April profiles and ADCP Page 24 transects Figure 7 Comparison of wind data between Green Airport Page 29 and GSO pier Figures 8a and b Profiles from 24 January and 31 January at buoy Page 33 and 34 station Figure 9 January Meteorological Data from GSO Pier Page 37 Figure 10 Temperature data from buoy for January Page 40 deployment Figure 11 Uncorrected data from salinity sensors for January Page 42 deployment Figure 12 Salinity sensor response over range of temperatures Page 45 Figure 13 Corrected buoy salinity data Page 47 Figure 14 Uncorrected dissolved oxygen buoy data from Page 50 January deployment Figure 15 Corrected buoy dissolved oxygen data for January Page 55 deployment Figure 16 Pressure readings from the three buoy YSI-600's Page 57 during January deployment  Raw pressure from three buoy YSI's against ~Tin Page 99 days Figure 29 Corrected pressure data for three buoy YSI's during Page 101 March deployment Figures 30a and b Observed tidal signal, bottom pressure minus Page 103 and surface pressure, and predicted tides for Newport, 104 RI from 10 March to 8 May. Figure 31 Profile from 24 March at 13: 40 Page 108 Figure 32 Profile 2 from 5 April at noon Page 110 Figure 33 Profile 5  USGS stream flow data for the Blackstone River Page 128 from Woonsocket, RI and precipitation data from NOAA from Green Airport Figure 40 Temperature from January 1997, s~owing well-Page 137 stirred and isothermal periods Figure 41 Thermal stratification and tide height versus time in Page 139 January Figure 42 Temperature, predicted tide and wind speed and Page 141 direction, 16 to 1 7 January 1997 Figure 43 Bottom temperature and tide height between 6 Page 144 January and 19 January Figure 44 Oxygen gradient and tide height versus time in Page 148 January Figure 45 Stream flow for the Blackstone River and daily Page 151 precipitation averages for T.F. Green Airport, January deployment Salinity at three depths, observed tide height, and Page 174 24-hr moving average of tide height from 11 April to 20 April 1997 Figure 51 Wind speed and direction from 11 April to 20 April Page 177 1997 Figure 52 T-S Plots for surface, mid-depth and bottom Page 179 instruments from 11 April to 20 April 1997 Figure 53 Dissolved Oxygen at three depths, and observed Page 182 tide height from 11 April to 20 April 1997 Figures 54a Extracted ADCP sections based on hours after slack Page 185 to through 54f high water, zonally across the hole. 190 Figure 55 Predicted and observed tide height at the buoy Page 192 station on 5 April 1997 Figure 56 Sketch of surface and bottom current structure in Page 195 the vicinity of Hope Island approximately one to two hours before low water Figures A-1

Introduction
In the water column, stratification due to a vertical density gradient may be established through modification of the surface salinity due to rainfall and riverine outflow (Hansen and Rattray 1966), or through modification of the surface temperature due to solar heating (Simpson and Bowers 1984). A result of this increased stratification is that vertical mixing is inhibited (Grasshoff 1975;Waldichuk 1984). Further, water trapped in bathymetric depressions may be likewise inhibited from horizontal mixing and advection (Kuo and Park 1992;Largier, et al. 1992). Such isolated water may be subject to notable changes in the concentration of dissolved gases such as oxygen and carbon dioxide due to respiration, the ongoing downward flux and decompositon of organic matter, and interaction with the sediments (Skirrow 1975). Water column stratification can also be a factor in primary productivity (Frenette, et al. 1994;Haas, et al. 1980;Peeters, et al. 1995;Rasmussen 1995) and it limits the amount of air-sea exchange (Downing and Truesdale 1955;Liss 1973;Waldichuk 1984). Three mechanisms which serve to reduce the level of stratification are tidal stirring (Haas 1977), wind-induced surface mixing (Blumberg and Goodrich 1990;Goodrich, et al. 1987), and convective overturning (Stanley and Nixon 1992;Lund-Hansen, et al. 1996;Simpson and Rippeth 1993;Simpson and Bowers 1984).
A great deal of effort has been devoted to studying chemical effects of stratification in large anoxic basins such as the Black Sea (Brewer, et al. 1970;Brewer and Murray 1973;Spencer and Brewer 1971), in fjord-type systems such as the Baltic Sea (Lund-Hansen, eta/. 1996;Rasmussen 1995;Skyum, eta/. 1994), at abyssal depths (Craig 1971 ;Fanning and Pilson 1972), within rivers and in estuaries dominated by freshwater-runoff (Anderson and Devol 1973;Peeters, et al. 1995;Prestidge and Taylor 1995;Simpson and Rippeth 1993), and in lakes (Frenette, et al. 1994). Most of these studies have been on seasonal time scales. Figure 1. A bathymetric chart of Narragansett Bay, designed from 50 m-grid soundings, formatted and plotted in GMT by Pockalny (1997). The mid-and up-Bay yellow linear Jb.thyml!try (m ) bathymetry of the Bay reveals the presence of deeper areas in both the East and West Passages ( Figure 1). Some of these areas, produced during pauses in the most recent glacial retreat, are two to three times the depth of the surrounding water (McMaster, 1984). Considering the importance of the benthos in contributing to the remineralization of nutrients (Olsen, et al. 1981) and in overall carbon utilization (Frithsen 1989), these deeper depressions may play a special role in the ecology of Narragansett Bay, in the development and persistence of the stratified condition, and in the modification of dissolved chemical constituents such as oxygen, carbon dioxide, and redox potential of these waters. Do these deeper areas act as nutrient traps, where debris falling from above is accumulated and oxidized? Do the bottom features modify the currents and vertical mixing such that material from the bottom of the depression is brought to the surface? Does stratification in these areas produce a physical setting for the "seeding" of a phytoplankton bloom? Do the physical bounds of the depression prevent the circulation of the deep waters such that oxygen is removed and C02 is continually added through the ongoing decomposition and respiration processes?
An analysis of waters in the viCinity of one of these depressions may indicate the importance of mixing, advection and diffusion processes, and may display short term changes between stratified layers as a result of physical, chemical and biological variability.
The variability associated with these depressions may be seen to influence biological productivity, especially in colder and less productive winter conditions and in the transition to spring production. The area chosen for this investigation is in the mid-Bay region between Hope and Prudence Island (Figure 2). Though not the deepest of these Narragansett Bay depressions, the Hope Island feature was selected for a number of Figure 2. A chart depiction of the study area, centered on Hope Island. The blue-shaded depths are less than 2 meters, the white-shaded areas are from 2 to 7 meters depth, the gray shaded areas are from 7 to 15 meters depth, and the black shaded area is greater than 15 meters depth. The area of study is between Hope and Prudence Islands, and the position of the buoy is denoted by the filled circle.
,. '-. . : "-" scientist with considerable experience in sampling and scuba diving in the Bay, he mentioned that the deep areas of the Bay are of two types: erosional and depositional. The erosional deeps have strong tidal currents that prevent sediment accumulation, whereas the depositional deeps have fine-grain mud sediments (Klos, personal communication, 1995). The Hope Island trough is a depositional depression. It is readily accessible from our boat launch facilities, in Wickford and Allen's Harbor. It is not located in one of the maritime traffic areas within Narragansett Bay, so that deployment of a surface buoy would not pose a hazard to navigation, or to the scientific instruments. Also, it is situated at the juncture between the East and West Passage, and between the upper and lower Bay.
Thus, its location in the transition zone between differing areas of the Bay may be important.
Using high resolution time series measurements at the surface, mid-depth, and near the bottom in the area of the Hope Island depression, a time-series of meteorological data from the GSO Dock and T.F. Green Airport, discrete v~rtical profiles over several days and ADCP data over a tidal cycle in ·and around this study area, we will determine the level and periodicity of stratification in the late winter/early spring, link these episodes of stratification to forcing mechanisms such as tides and winds, assess the current structure in the vicinity of the study area, and describe how these changes in the physical and current structure can affect the chemistry of Narragansett Bay.

Study Area
A depression down to about 15.3 m (50 ft) depth is located 1 km east of Hope Island in the West Passage ofNarragansett Bay (Figure 3). This hole is approximately twice the depth of the surrounding areas and is relatively limited in horizontal extent.
Along a north-south axis, the hole is approximately 660-m long, and from east-to-west, across the center of the feature, the hole is about 500-m. The south and west slopes, closest to Hope Island, are steeper than those on the north and east. The western slope drops about 7 meters over 160 meters laterally, while the eastern run is about 420 meters for the same drop. The southern slope drops 6 meters over 180 meters laterally, while the northern slope drops 7 meters over about 760 meters laterally. The absolute center of the hole, in the deepest area, is about 80-m long by 50-m wide, with a maximum depth of about 15 meters.
An oceanographic buoy, capable of simultaneously measuring temperature, salinity via conductivity, dissolved oxygen concentration, pH, and in-situ pressure at three depths, was deployed to determine the chemical and physical characteristics of the water column, although we chose not to discuss pH due to calibration and processing complexities. We chose to deploy the buoy within the hole, close to the western slope of the feature. In the initial deployment, we regularly observed depths from the deepest instrument of up to 15 m (48 to 50 feet) at mean lower low water. However, due to the modification in the instrument mooring in the second deployment, the bottom instrument resided between 12 and 14 m. A twenty-minute time interval was chosen for measurement as a compromise between resolving time-scales of less than one hour and providing a reasonable schedule

January Deployment
The initial deployment began 6 January 1997 and lasted for 25 days, but the data file spanned only 18 days. Problems with servicing the instruments and an electronic problem within the buoy' s computer, coupled with errors in calibration, caused this first attempt to end sooner than originally planned. In addition, an apparent fouling of the bottom sensor' s conductivity cell has made analysis of the data set problematic. The January data provide some useful information on winter-time processes in the Bay. Based on this initial deployment, modifications were made in the buoy computer electrical wiring and in the mooring configuration.

March Deployment
On I 0 March 1997, we redeployed the buoy with a modified instrument mooring, so that we could more easily access and service the instruments on station. The instruments were all recalibrated, the electronic problem was fixed, and the buoy was reset to log at 20 minute intervals. At 06:40 on 5 April 1997, we downloaded the accumulated data in memory and set the logging interval to 5 minutes. Through the following tidal cycle we conducted ADCP transects and stationary time-series in the vicinity of the study area with GSO's 20' Maritime Skiff At about 11 :00, the HIY Melanitta began taking discrete YSI-6000 profiles within the study area and remained on scene conducting profiles and towed surface transects until approximately 16:00. At 18:00, the buoy' s memory was accessed and data were once again downloaded. The buoy was then set to log at a 20 minute interval once again. On 22 April, we again visited the buoy, downloaded the data, conducted a profile and calibrated all three instruments. The final visit to the buoy for this study was on 8 May, and we again downloaded buoy data and conducted a profile.

14
The buoy consists of a PVC-constructed spar which is foam-filled, and upon which is mounted a solar panel, a flashing beacon, and within which is housed a battery and datalogging computer. This buoy's computer records and stores the water-quality measurements of three SDI-12 compliant instruments. In this study, we outfitted the buoy with three YSI model 600XL instruments. Each YSl-600XL is capable of measuring temperature, conductivity (with which we determine salinity), dissolved oxygen, pH and depth. Each is essentially a mini-CTD, mounted in a serial array and capable of acquiring marine environmental data as rapidly as once every three minutes. For our study, we chose a 20 minute interval, to observe hourly-to-daily changes in the water column.

Instrument Calibration
Each 600XL required a pre-deployment calibration for all sensors except temperature. The salinity was calibrated using a secondary salinity standard which was determined relative to IAPSO Standard Seawater using a Guildline Salinometer.
Dissolved oxygen and depth were calibrated in a moist atmosphere of known barometric pressure, and pH was calibrated using a three-buffer system, allowing independent determination of an empirical Nemst slope, from which we determined actual pH from a millivolt reading. After calibration the instrument was deployed immediately, to reduce sensor drift . More information regarding the YSI instruments and their calibration may be found in Appendix 1.

Rigging and Mooring for First Deployment
We used a two point mooring system (Figure 4 ), with legs of three-eighths inch galvanized steel chain, each 24 meters in length, and each moored by a 91 kg mushroom anchor. Each leg terminated at surface flotation which were then joined on each side by 4. 6 meter lengths of three-eighths inch chain, with the buoy between them. The axis of the mooring was north-south, and an instrument mooring extended perpendicularly to the east. This leg was also 24 meters in length, with a YSI-600XL attached within 3.7 meters of the bottom mooring, a 34-kg grapnel anchor. This attachment location placed the YSI-600 approximately 1 m from the seafloor, considering the chain's catenary when the mooring was deployed. Another YSI-600XL was attached within 3.7 meters of the top of this leg. With catenary, the middle instrument was situated at about 4 meters depth, and the deep instrument at about 14.8 meters. The deep instrument had a 9 kg flotation buoy tied into the chain at its location to provide buoyancy, and keep it from impacting the bottom. The problem associated with this arrangement was that the middle and deep instruments were difficult to retrieve and service from a small boat, due to the weight of the grapnel and chain. The surface YSI-600XL was mounted in a PVC pipe mounted on the side of the spar buoy with holes for water flow to the sensors. This instrument was about 0.5 m beneath the sea surface. Flotation bodies are gray-filled, and the position of the YSI-600XL's are denoted by "Y" characters. The mooring leg, a 50 ft, 3/8 in chain mooring with the 75 lb grapnel, is oriented such that this leg is coming out of the page, along a third axis.

Mooring Design for the Second Deployment
In the subsequent deployment, the chain instrument mooring perpendicular to the two-point buoy mooring was eliminated ( Figure 5). Instead, the middle instrument was also fed through the PVC pipe on the side of the spar buoy and was freely suspended by 18 its data cable and a light nylon line to hang at about 4.5 m depth. The instrument closest to the bottom was rigged away from the buoy itself. Its data cable ran from the buoy, and was affixed by cable ties along the length of the southward-running mooring chain to the southern surface float . The excess data cable was made up in a coil and secured to this surface float. A polypropylene line, with a 15-lb diamond anchor, moored the instrument to the bottom. This instrument was made fast approximately 1 meter above the bottom and an 8-lb float was attached to the line just above the instrument to provide buoyancy, keeping the instrument away from the seafloor. As mentioned earlier, this arrangement allowed us to more easily access the middle and bottom instruments, without the use of heavy lifting equipment. The removal of the instrument leg, however, moved the positioning of the bottom instrument approximately 15 meters to the west, such that it was moored partway up the western slope of the depression at a depth of 11. 5 to 14 m depending on the tide. This depth is still 4 to 5 meters greater than the seafloor that surrounds the depression.

Buoy Deployment
The initial buoy deployment was performed by Rhode Island Mooring Company (RIMS) on 6 January 1997. Downloading operations were undertaken on 15 January and 24 January, and the buoy was hauled for repairs on 3 I January. The subsequent buoy deployment was performed by RIMS on 10 March 1997. Downloading operations were Flotation bodies are gray-filled, and the position of the YSl-600XL's are denoted by "Y'' characters. In this layout, the perpendicular mooring leg has been removed and the entire scheme is oriented in only two dimensions. RD Instruments gives other specification and performance data in its Technical Manual (RD Instruments, 1995). It should be noted that the times in the ADCP data files were one hour and two minutes greater than those from the buoy for the 5 April data. The ADCP data were all gathered with this offset and corrected during extraction and analysis to the correct buoy time.

ADCP Sampling
In order to evaluate the circulation at all depths entering the study area, we conducted three transects in a single circuit ( Figure 6). Transect 1 was followed from west-to-east, south of the depression, for about 0.75 nautical miles (nm). Transect 2 was followed from east-to-west, across the center of the hole, for about 0.5 nm. Transect 3 extended from southwest-to-northeast, at the north end of the hole and was followed for about 0. 75 nm. The first circuit was begun at approximately 07:00 and ended just after 08:00. We attempted to conduct all circuits within an hour, and when extra time was available, we conducted time-series measurements of over 50 to 100 ensembles at the buoy site. We conducted six circuits between 07:00 and 12:50, three of which included time-series data, and we continued this sampling regime.until the southerly winds began to increase and waves increased to about 0.5 m wave heights. Between 13 :23 and 13 :32, we conducted a transect from the south end of the hole to the north end of the hole, however, data quality during our final circuit and in this north-south transect was lower than the earlier circuits due to increased sea-state. We conducted time series measurements at the buoy location from 14:00 until 18:00, for the remainder of our sampling day. For this entire period, the buoy was logging temperature, salinity, dissolved oxygen, pH and pressure at the three instrument depths at a five-minute sampling interval. Problems with the ADCP data extraction were encountered with one of the files, Transect   profiles were taken throughout the study area, along both the longitudinal and zonal axes, and at the buoy station. We used these profiles to verify the data from the buoy, to determine any calibration drift of the buoy instruments' sensors over time since deployment, and to gain a sense of the characteristics of the water masses in the study area.

YSI Profiles
We attempted to follow a consistent procedure with regards to all vertical profiles taken during this study using the YSI-6000. These prqfiles were taken at the buoy station on 24 January, 31 January, 24 March, 5 April, 22 April and 8 May. Profiles north and south of the buoy station were taken on 5 April. We used the profiles to resolve the water quality parameters from less than a meter off the bottom, to the surface in one to two meter intervals. Parameters measured by the YSI-6000 included depth via a stainless steel pressure strain gauge, temperature via a thermistor, conductivity via a four electrode cell, dissolved oxygen via a patented Rapid Pulse Polarographic sensor, pH via a glass combination electrode, oxidation and reduction potential (ORP) via a platinum tip, two electrode system, and turbidity by 90° optical scattering. More detailed instrumentation data is contained in Appendix 1. 26 The method for profile measurement was the result of observations concerning the responsiveness of some of the sensors, notably that for dissolved oxygen (Kester and Magnuson, 1994). The YSI-6000 was lowered to the deepest sampling depth, usually within one half to one meter of the bottom, and the sensors were allowed to equilibrate for approximately five minutes. After this period, the instrument was raised to the next sampling depth. Successively sampled depths required only minor equilibration periods of approximately two minutes. We sampled in one to two meter depth intervals toward the smface and continuously monitored the instrument by a laptop computer interface via a 30.5-m (100-ft) data cable. The shortest sampling interval, four seconds, was set for the vertical profiles so that a number of readings were recorded at each depth. The final five readings at a particular depth were then averaged to arrive at a smoothed value for each depth. This smoothing procedure was performed to eliminate microscale variability at a particular depth after the probes had equilibrated, and was performed for all parameters except dissolved oxygen. The longer equilibration time for the teflon membrane on the oxygen sensor required us to use only the last value before the change in depth, without averaging. The membrane is not particularly susceptible to the microscale variability as found in the other probes, but the delay for equilibration increases the accuracy of the measurements.

Meteorological Time-Series
In addition, we accessed the GSO Pier Meteorological time series, to help ascertain the effects of changes in wind forcing. The meteorological data were taken via Sensormetrics ENV-50 modules (air temperature, wind speed, and wind direction). These data were recorded in ten minute intervals, and for comparison were reduced to a 20 minute interval by averaging across three readings, centered at the 20 minute node. This record is continuous from 10 March 1997 until 24 March 1997. There are gaps in the data after this date, between 24 March and 28 March, and again between 1 April and 7 April. During these gaps in the GSO meteorological data, hourly wind data taken from Theodore F. Green State Airport in Warwick, RI, were obtained by Andrea Magnuson and Wendy Woods as part of a long-term time-series monitoring project. These data were interpolated to a 20 minute period. Some overlapping portions of the T. F. Green (PVD) data and the GSO Pier (GSO) data were plotted together and showed good agreement ( Figure 7). Therefore, we have produced a data set combining measurements from both sources, to be used for wind-related analyses, which provides data coverage for most of the gaps in each data set. . . ··-1.

January Deployment
From the first deployment, we have a record of the water temperature, salinity, and dissolved oxygen at three different depths as reported by pressure readings from each of the three instruments from 6 January until 24 January 1997. We will not discuss the pH data from the buoy in this thesis, however, we hope to include them in a later publication.
We also acquired profile data at the buoy station on 24 January and 31 January, via a YSI model 6000 instrument. Each profile reported temperature, salinity, dissolved oxygen, pH, redox potential, and turbidity at depths indicated by a pressure sensor, however, we will show only the data relating to the present work, of temperature, salinity and dissolved oxygen. Problems associated with the initial YSl-600 calibration made interpretation of the salinity records for the middle and bottom instruments difficult. The calibration problem consisted of only partially immersing the conductivity sensor in the secondary seawater standard, so that the instrument was calibrated with an erroneous cell constant.
In order to correct this error, we attempted to raise the .instrument mooring leg, to remove the instruments, recalibrate them on station and redeploy them on the mooring. Our attempts to raise the mooring, were seen in the buoy data on 15 January, as a broad increase in the oxygen and salinity of the bottom instrument. We were unable to lift the heavy chain and anchor onto the boat, and therefore had to reset the mooring without calibrating either the middle or bottom instrument.
We used temperature as the primary tracer for water stratification, and, using a post-deployment calibration temperature test, were able to derive corrections for the middle and bottom salinities. However, the bottom salinity exhibited an additional drift, starting on about 10 January, which made use of the later part of the salinity record impossible. Comparing with the predicted tides for Newport RI and data from a timeseries of meteorological data from the GSO dock, we used temperature and dissolved oxygen data to qualitatively discuss stratification/destratification events for the first deployment, especially within the depression.

31
The general structure of the water column for this time of year may be represented from vertical profiles performed at the buoy location on 24 January and 3 1 January 1997 ( Figures 8a and 8b). We measured temperature, salinity, and dissolved oxygen in% saturation, at six discrete depths from 0.36 m to 12.92 m for 24 January, and eight discrete depths from 0. 07 m to 13 . 02 m for 31 January. The 3 1 January profile showed much larger gradients from the surface to the bottom in practically every parameter, and much lower dissolved oxygen saturation levels, especially at depth. The profile was taken about two hours before high tide, during the neap phase of the lunar period. The 24 January profile was taken between two and three hours after high tide, during a spring phase, and weaker gradients were noted.
The 31

GSO Pier Meteorological Data
The parameters measured in the time-series of meteorological data being collected at the GSO Pier include air temperature, wind speed and direction, solar radiation, and barometric pressure. We plotted these data over the time span of our buoy deployment ( Figure 9). Unfortunately, we experienced a loss of data from midday on 11 January until the evening of 14 January, due to low battery voltage on the meteorological data logger.
The recorded data were averaged to match the sampling interval of the buoy data. For the first part of the meteorological record, data were recorded every ten minutes so the averaging was over three values centered on the twenty-minute node. For the second part of the data, ten minute sampling was commenced on five minutes after the hour. In this case, we averaged over four values, two to either side of the twenty-minute node. These data were then plotted for comparison to the buoy data.
The wind data showed several high wind events, greater than 10 meters per second. The directions were variable during the first 4 days of the deployment, were coherent with a passing front on 10 January, and showe~ alternating bands of variability and invariability from 15-22 January. · There were only two days of very low solar radiation. These occurred on 9 January and 16 January. Solar radiation exceeded 550 W m-2 only once, on 17 January. Barometric pressure showed the passing of several atmospheric fronts during the deployment. There was a broad increase at the end of the record, to values higher than 770 mmHg.

Temperature
The time-series of temperature at three depths shows an overall cooling of the waters during this period in January ( Figure 10). We observed periods when the bottom water was 1. 0 to 1. 5 °C warmer than the surface water. Periods of thermal stratification terminated with periods of isothermal conditions through the water column. The temperature variations indicate episodic stratification and destratification events over time scales of 2-5 days.

Salinity and Salinity Correction
It became apparent from a review of the salinity data that calibration errors occurred in two of the sensors; the mid-depth and bottom instruments gave salinity values of between 42 and 46 o/oo, and between 41 and 53 o/oo, respectively ( Figure 11). Upon retrieval of the instruments on 31 January, we determined that the erroneous values were due to incorrect conductivity cell constants in the middle and bottom instruments, the determination of which is performed during calibration. We surmised that ifthe conductivity cell was not fully immersed in the seawater standard during the initial calibration, this would account for an abnormally high cell constant and the observation of abnormally high salinity data. In an attempt to salvage the data from the time-series, an experiment was conducted, after the instrument retrieval, to determine the salinity readings for all three instruments in a seawater medium, 15°N in the Pacific, with a salinity at 36.1 F . Figure 10. Temperatures at three depths as recorded by the buoy. The red trace denotes the surface instrument, at approximately 0.5 meters, the green trace denotes the mid-depth instrument, at about 4. 5 meters, and the blue trace denotes bottom temperature, at between 13 and 15 meters. This color scheme will be maintained throughout the paper, in which red, green and blue represent surface, middle and bottom data, respectively.   conductivity does, and since salinity is calculated from conductivity by the YSI-600, this experiment provides a method of arriving at a correction factor for salinity, and accounting for an observed, temperature-dependent error in the salinity calculation.
Equations 1 and 2 give the corrections for the middle and bottom salinity data, respectively.
The corrected salinity data for the middle instrument from the buoy followed the variations seen at the surface extremely well, however the bottom instrument showed features which indicate that other factors affected its values ( Figure 13). We suspect that the bottom instrument may have been too close to the bottom, so that the conductivity cell

Dissolved Oxygen
The dissolved oxygen record at the three depths showed some variation between instruments, even during mixing events ( Figure 14). These inter-instrument variations during these periods were probably not real, as we would expect fairly close agreement among the concentrations of dissolved oxygen at all depths during mixing events over the entire water column (Table 1). First, we compared the oxygen data from the profile on 24 January to the readings obtained at the end of the buoy's operation. At the surface, the buoy was reading values between 108.3 % and 109.6%, and the profile's dissolved oxygen value at the depth of0.5 mat 10:53 was 106.9 % . Because the buoy stopped logging at 10:20, it was necessary to estimate what the buoy might have been reading at the surface at 10:53 . High water occurred at 08:20 and coincided with the local minimum in surface      24-Jan 17.88 3.6 n/a n/a I 7.88 5 n/a n/a ..  January gave an average difference between surface oxygen and middle oxygen of 4.5 ± 0.12 % saturation. The average difference between surface and bottom was 5.8 ± 0.23 % saturation. Averaging 27 values between 05:00 and 14:00 on 17 January gave average differences from surface oxygen levels of 4. 7 ± 0.15 % and 6.2 ± 0.40 %, for middle and bottom, respectively. The estimate for the buoy' s middle instrument dissolved oxygen reading for 10:50, the time of the profile datum from the middle instrument's depth of 3.96 m, was approximately 103.3%. The profile value for the mid-depth was 106.9%, giving a difference of 3.6%. A value for the bottom instrument, buoy minus profile, was calculated as 8.3, however this value gives rise to over-correction at the end of the deployment. A value of 5 was used instead, and was determined by observing the difference between raw bottom oxygen data and middle and surface corrected data for 9 readings on 22 January, the longest stretch of consecutive readings between gaps in the buoy data. The over correction was due to comparing bottom data which was affected by the interaction with the seafloor. Figure 15 shows the corrected dissolved oxygen data for the three instruments.
One sees a broad stratification period to begin the deployment, which is reflected in the salinity data. This condition is broken down on 8 January, and remains destratified until approximately 13 January. This later stratification eventually increases, but is again broken down over 16 and 17 January, after which, the water column once again stratifies with respect to oxygen. The deep spikes down below 90 % are believed to be due to the instrument contacting the seafloor. Additionally, the large distribution of downward peaks and the low bottom 02 values on the afternoon of 15 January can be attributed to our tying the boat to the buoy. When we did this, the drag of the boat pulled slack in the instrument mooring, causing the catenary, and thus the bottom instrument to move lower.
The instrument was already close to the seafloor, and these activities certainly brought the instrument into contact with the seafloor.

Depth or Pressure
The pressure readings from the three instruments ( Figure Figure 16. A plot of the pressure readings from each instrument in meters depth of water.
On 10 January, a surge of water entered the Bay and pulled slack into the instrument mooring. The spike on 15 January was our attempted retrieval of the instrument mooring. to the surface. Also obvious in the depth record was the apparently wind-and barometric pressure-induced event on I 0 January in which unusually high water levels and a surge of water entered the Bay. This event resulted in coastal flooding in the northern portion of the Bay. The dynamics of the mooring system can be deduced by observing the middle and bottom pressure records. Between 6 January and 9 January, the instrument mooring is extremely taut, and takes up some of the buoyant force of the buoy on high tides. The truncated signal in the bottom instrument and the full signal by the middle instrument indicate this. As the surge on 10 January impacts the buoy, the instrument mooring is pulled towards the buoy, and when normal water levels resume, this mooring now carries more slack. The more fully recorded tidal signal in the bottom instrument, and the fact that the mid-depth instrument begins recording the atmospheric changes, similarly to the surface instrument, suggest the increased slack in the instrument mooring. The downward spike shows when we lifted the mooring, and after this point, no tidal variations are noted in the middle instrument, suggesting much more slack in the mooring.

March Deployment
On 10 March 1997, the buoy was redeployed on the same two anchor-float systems as before but the instrument mooring leg was removed and changes made to facilitate retrieving the YSI-600 instruments (see page 15 and Figure    events which may be recognized as partial mixing in which the parameters in question converge, but the water column does not quite become homogeneous before stratification once again occurs. These partial events occurred just after noon on 19 April, prior to noon on 24 April, as a partial stratification on 2 May, and as a partial destratification on 3 May.

Raw Salinity Data and Development of Correction Factors
A number of factors must be considered to apply post-collection calibration to the observed data. To determine finally calibrated values, we employed the field calibration offsets from calibrations on 24 March and 22 April, and used the profiles from 24 March, 5 April, 22 April and 8 May to constrain the possible offsets among the sensors.
Unfortunately, the deployment operations made it impossible to conduct a profile of the waters at the site, so we have no additional data about the initial deployment conditions, other than those from the buoy. The tidal lunar phase was just past the spring tide conditions of 8 March (Figure 20a and b ). Thus, there was a relatively large amount of tidal energy available for mixing. The weather was still cold, and water temperatures were between 3.5 and 4.5°C, and close to isothermal (see Figure 17a  The 5 April values were all disregarded due to the large, negative offsets observed for all profiles on this day. We suspect that the problem resided in the YSI-6000 salinity calibration, and thus, the profile-buoy comparison could not be used, in order to avoid applying a correction unsupported by the other observations. In the 22 April profiles, the offsets for the surface and middle instruments were of similar magnitude and opposite direction as the offsets for the calibration on the same day. The profile values were These values were then plotted against time and a curve was fit to the data ( Figure 21 ).
Additional data points were added at 17 March, and 30 April, to constrain the data. These

81
From the raw oxygen data, a few needed corrections are apparent (Figure 19a (Kester and Magnuson, 1994). We determined the overall drift by comparing the last value at the 5 minute interval, taken at 18: 00 with that from the first succeeding 20-minute sample, a 5.6 % difference for the surface instrument, and a 6.7 % difference for the middle instrument, and a 3.2 % difference for the bottom instrument.
This change over the 11 hour period was incrementally applied from the start of the 5minute sampling until the return to 20-minute sampling ( Figure 24).  the middle and bottom averages, respectively. As in the salinity correction, the offsets were plotted versus time since deployment of the buoy and a curve in 20 minute intervals was fitted to the offset data. Points as midpoints of the line described by adjacent offset data were added on 17 March and 31 April, to constrain the curves ( Figure 25). These correction factors were then added to the preliminarily corrected raw data to obtain corrected oxygen time-series for the surface, middle and bottom instruments (Figure 26a and b).
In the corrected oxygen data, we observed trends in stratification which mirror those in temperature and salinity. There was a period of stratification from about 16-26 March, brief and understated stratification on 29 March and 3 April, and a broad and persistent period of stratification from 13 April until 3 May, which is interrupted on 19 April, 24 April and 28 April by partial mixing events. In general, the surface instrument recorded the highest % saturation readings of the three instruments. The middle \: The average for this value was computed using only the 5 and 7 profile, as the offset from the 2 profile was greatly different from the other two offsets.
b The offsets for all three profiles for the middle instrument were of similar, and minor magnitude, and mixed sign, therefore no offset was applied. c This offset was greatly different from both the profile offset, and from the other offsets for this instrument, therefore, the 2.1 offset from the profile comparison was applied, vice the calibration offset.
d The offsets for the middle instrument for 22 April were of similar magnitude and opposite direction for the profile and calibration, so no offset was applied. However, an examination of the raw data suggested that the -8.7 % calibration offset was applied in error. We therefore removed the effects of this calibration offset from the middle instrument's oxygen data from 22 April until the end of the record.
e The raw data offset was 11. 2 %, however, after adding in the 8. 7 % from the erroneous offset on 22 April, the difference becomes 2. 5%, and this value was used for the 8 May offset. The open black squares indicate the positions of the data used from Table 3 in generating the curves. The filled black squares indicate mid-point, tie-points added to constrain the curves to a form supported by the data.  ; 120 There are two factors which must be accounted for in correcting the pressure data from the buoy. First, the sensors are prone to drift from their calibrated values over time.
Second, the instruments measure absolute pressure via a strain gauge, meaning that in addition to the pressure of the water column above the instrument, the instruments also indicate the atmospheric pressure above the instrument as well. Thus, when atmospheric pressure is low, the pressure reading is lower for an equal height of water (of equal density), as when the atmospheric pressure is higher. In order to obtain observed tide   information from the bottom instrument, one must first remove the effects of these two factors from the data set. To remove the drift from calibration, the data set was separated into two sections, before 22 April and after 22 April, because after the 22 April calibration, the drift rate may have changed for each instrument. Figure 28 shows the linearly-estimated drifts for each instrument both before and after the 22 April calibrations,  Of seven profiles taken on 5 April, in conjunction with the ADCP fieldwork, three were obtained in the vicinity of the buoy. Profiles 2, 5 and 7 were taken at 11:57, 14: 11 and 15 :38, respectively, and were obtained less than 100-m northeast of the buoy station.
Profiles 1 and 4 were taken to the south, and profiles 3 and 6 were taken to the north.
These other profiles do not add to this analysis, and therefore will not be discussed . At the time of profiling at the buoy station, the Bay was approaching low tide, with slack water calculated from the Newport, RI prediction at 12:09 pm. The following high tide was at 18:00.   layer between this middle mixing layer and the deep layer from about 9 meters down to almost 13 meters. The structure was a result of the increasing tidal currents with the recent maximum in the spring-neap cycle on 6 May, and the gradients in temperature, <2°C over 13 m, and salinity, <20/00 over 13m, and <7% oxygen saturation over 13 meters reflected the increased mixing of the period.

120
As mentioned in the previous section, a record of meteorological data was compiled by filling gaps in the GSO Pier meteorological time-series with data from T.F.
Green Airport. These data span the period of the buoy deployment (Figure 37a and b). In the wind record, we observed maximum velocities of approximately 20 to 22 mis. These major wind events occurred twice during the buoy deployment, once on 31 March-I April, and again on 19 April. The modal velocity was approximately 4 mis, and over 90% of the wind velocities during the period were less than 10 mis (Figure 38a). Wind directions were much more variable, and displayed somewhat of a bimodal distribution, with over 500 frequency events of directions coming from both 210° and 340° (Figure 38b). The majority of wind directions were from the northwest, between 270° and 360°. In order to facilitate plotting and quantitative consideration, all directions less than 90° were plotted as 360° plus the observed direction, so that directions between 360° and 450° actually represent those winds between 0° and 90°.
Two sources of data pertaining to fresh water delivery were accessed ( Figure 39).
We obtained a record of daily precipitation from the NOAA, National Climatic Data Center compilations for T .F. Green Airport in Providence, RI, for March -May of 1997.
Data are reported in inches of water equivalent precipitation per day. These data are   the Blackstone River, Woonsocket Station, station number 01112500. These data are reported in cubic feet per second. Since the Blackstone River is the source of the largest riverine input of freshwater to Narragansett Bay (Olsen, et al., 1992;Pilson, 1985), this stream flow data can represent the overall riverine freshwater input to the Bay.
The precipitation delivered during the 1 April Blizzard is immediately obvious, and .
the resulting increase in river flow is also readily observed, the fresh water from the rain and snow melt from this storm caused the stream flow to almost triple. One also notes that the largest stream flow volume resulted from a weather event occurring from 17 April to 19 April, in which a total of about 1. 5 inches of precipitation was delivered over three days. The longer lag between precipitation and stream flow volume associated with the 1 April storm compared to the 17-19 April storm was likely a result of the type of precipitation, snow versus rain.

ADCP Images
To explain the filename convention, the files are named HOL21XXA, where HOL refers to the location of our data collection, 2 indicates that it is our second time performing ADCP work in the area, 1 represents the fact that two steps were required in image processing and the filename changed from 0 to 1. The XX represent a sequential number referring to the chronology on scene, starting with 1 and ending with 28. The A indicates that the raw data was translated into ASCil format. There are two basic types of ADCP image files, and they are based upon the type of sampling operation (Table 3). The first type of sampling consisted of transects, in which the instrument was pulled through the water over some lateral distance, generally over 150 meters. Almost all of these transects were oriented roughly east-west. Only one of them, filename HOL2122A, consisted of a south-to-north transect across the middle of the hole. The rest were across the hole in an east-west direction, either north or south of the feature . We have seventeen transect files, not including HOL2122A, covering from 07:00 until 13:00. This sampling was conducted during the morning, and the ebbing tide? when surface conditions were much more conducive towards transect operations. Each of the transect files in a triplet (HOL2101A through HOL2103A) is part of a circuit, starting south of the hole, running back across the feature and finally back across at the north end of the hole. For ease in discussion, transects which ran south of the island were numbered 1, transects abeam of the island were numbered 2 and those north of the island were numbered 3. Files HOL2101A through HOL2108A, HOL21l1A through HOL2113A, HOL2115A through HOL21l7A, and HOL2119A through HOL2121A are transects. HOL2009's file was corrupted and unreadable.
The other type of file in this study is a time-series, in which the ADCP was held at or near the buoy and was allowed to measure the currents in the same area over some period of time. These time-series formats were predominately conducted later in the day, when surface conditions worsened and transect data quality deteriorated. These data were taken while the skiff was fast to the buoy, allowing the instrument to sample over a long period, starting from file HOL2123A at 13:50, until file HOL2028_2A, which ended at 17:58, or through most of the flood tide. HOL2010A, HOL2014A and HOL2018A were brieftimeseries operations inserted between transect circuits earlier in the day, during the ebb tide.
Filenames HOL2124A and HOL2126A were aborted files. The display images were created with a C-shell written by Dr. Robert Pockalny from GSO, and convert all transects into planar images oriented from west to east, looking from left to right, with the viewer's perspective looking northward. The color scheme is such that red contours are northward currents, into the page and blue contours are southward currents moving out of the page.
The time-series display images show an integrated flow magnitude and direction at the top of the page, and raw and processed data below. In all images, absolute horizontal magnitudes and directions are shown· in the bottom two panels, and vertical velocities can be observed in the middle panel, labeled Transect Parallel Flow Magnitude.

Morning (Falling) Tide
The early period of this study showed mostly a uniform picture for the area. The ebb tide in general was directed between about 165° and about 190°. Flow magnitude was between 40 and 45 cm s· 1 in the main axis of the current. This main axis was found close to the island at the north end of the circuit (Figs. A-3, A-6, A-12) and further to the Table 4. A chart of ADCP transects and profiling, indexed by figure number. The ADCP images are contained in an appendix, and thus, the images are labeled A-1 through A-26.
The transects are broken into three categories, representing transects performed south of the hole, across the hole, or north of the hole, respectively. Time-series profiles are predominantly found in the latter portion of the day, and are denoted by TS . Transect 23, figure A-21 , was the last transect performed, and was taken across the hole from south to north. show the development of a back eddy close to the island, on the shelf between the island and the depression. As time progressed, the ebb current weakened in the deep layer, and the cross-section of ebbing water shoaled and moved eastward. This can be seen in the first time series (Fig. A-9), where current flow was on the order of 5 to 10 cm s· 1 below 5 m depth. Between the time 9:56 and 10:50, a significant shear layer was set up at the buoy station, between about four and eight meters depth (Figs. A-9, A-13). This appears to be a result of the developing condition, especially close to the island and within the deep layer, in which water was moving into the Bay from below and from the west, while the surface water was still ebbing. Velocities in the deep northward-flowing region were much slower than those in the surface ebb jet (Fig. A-11), however, this counter-current region continued to expand in cross-section through the end of the ebb period (Figs. A-13 and A-15) until the tide reversed, when current magnitudes became very low, and current direction was extremely variable (Figs. A-18, A-19, and A-20).

Evening (Rising) Tide
After the tide turned, the conditions in the small boat worsened. Through the afternoon and evening seas built from negligible conditions to approximately 0.5 m waves.
The southerly wind picked up in intensity as well. We determined that the boat time might be better and more safely spent conducting time-series data acquisition. Therefore, files HOL2123 to HOL2128 (Figs. A-22, A-23, A-24, A-25 and A-26) are completely time-series data. The quality of data is lower than that from the morning, however, the prevailing situation during the flood-tide is easily discemable. Over the course of the four hours from 14:00 to 18:00, the incoming surface flow, above approximately 5 m depth intensified from about 28 cm s· 1 , to about 37 cm s· 1 . Again a counter-flowing current developed at the bottom and, though it was less well defined than that observed earlier, by the final time-series, it had become firmly established. The surface layer, initially broad and uniform, began to develop shear until there was a significant shear layer between 4 and 8 meters depth by the end of the last time-series. In general, the deep layer flow appeared broad and of uniform velocity, averaging between 15 and 25 cm s· 1 , but its direction appeared to vary more with depth. In the surface, the opposite prevailed, in that the velocity magnitude increased with decreasing depth, and direction was more uniform within the layer.

Vertical Velocities
Considering that the morning period, ebb tide condition was sampled during a time of negligible wind effects, we considered vertical current data from only this period. The flood period showed higher magnitude vertical structure, however, due to the pitching of the boat and the higher waves, we felt that some of these data may have been biased by the boat movement. We are more confident that the vertical velocities represented in the earlier transect data were unaffected by wind forcing and boat movement. As expected, vertical velocities were much lower than the horizontal velocities. It appeared that, early in the ebb, water was moving downward at the north end of the hole, especially over the deepest area (Figs. A-3 and A-6). Later in the ebb, slight but upward currents could be noted in some of the transects (Figs. A-12, A-16, and A-20). Across the center of the hole (Figs. A-2, A-5, A-8, A-11, and A-19) water was moving upward on the western slope of the hole, near where the buoy was placed, and was moving downward near the eastern slope. As the ebb ended, alternating bands of upwelling and downwelling were observed across the diameter of the hole (Fig. A-19).

January Deployment
The January deployment contains some very valuable information, because it shows events occurring at the buoy site during mid-winter conditions. This initial deployment suffers from a relatively short duration and it required substantial postdeployment calibration corrections for salinity and oxygen as described in the Results section.

135
The temperature time-series in January did not require corrections to the measured values. Over the twelve-day period of 6-17 January, the waters showed a general cooling trend at all depths, with periods of an isothermal water column, a well-stirred and nearly isothermal water column, and a significant thermal gradient in the water column with warmer waters near the bottom and cooler waters at the surface ( Figure 40). The difference in bottom temperature minus surface temperature provides an index of thermal stratification ( Figure 41). The water column was ''well-stirred" from 8 to 10 January, and briefly on 12 January, and on 16 January. This condition was represented by a narrow difference between the surface and bottom temperatures, but the difference is significant (~T ~ 0.25°C) and it appears to vary with the tidal state. At other times the water column is isothermal. This situation is represented in the time-series on 10 January and on 17 January. This more dramatic destratification was indicated by insignificant differences between surface and bottom conditions (~T ~ 0.05°C) and it appeared to be related to meteorological forcing from high velocity wind events such as observed on 10 January, or perhaps persistent winds of moderate velocity such as on 17 January, which may result in the release of potential from a "wind set-up" situation ( Figure 42). During this period, the These warm water pulses occur during high tides and are quickly mixed to lower temperatures during the outgoing tides. The rapidity of change of the bottom temperature increase during the flood is suggestive of an advective process rather than a mixing event.
It is also interesting that the local low temperature at the bottom occurs at the end of the ebb, but prior to, slack low water.
Given that we had no meteorological data from 11 January to 14 January, and that most of the salinity record was unusable, we decided to investigate a few large scale events, discussing only the usable buoy data and meteorological and tidal data. The first event investigated was the destratification early in the record, on 7 January. The second event is the stratification event between 14 and 16 January, and the third event is the stratification on 17 January, which lasted, at various strengths, until the end of the data set.
The destratification on 7 January appeared to be produced through the contribution of several forces. First, the spring-neap tide phase was extremely important, as a new moon, and thus a maximum "in the lunar phase, occurred on 9 January. From the first low tide in which water heights were below mean lower low water (MLL W), early on 8 January, until the first low tide in which water levels were above MLLW, the thermal stratification was held to very low levels, LiT averaged about 0.43°C between low tide on 8 January and low tide on 14 January. This was a period of high tidal amplitude, exceeding 1. 6 m difference between high and low tide and this source of energy in the water column could be a factor in breaking down the stratification observed prior to 8 January. January. The local incidences of warmer water are coincident with high tides. These warm waters are rapidly mixed on the outgoing tide between 8 January and 14 January.
Between 14 January and 18 January, during the deployment's lower tidal range, the warmer waters associated with high tide are more gradually mixed during the outgoing tide. Another factor important in this event was the wind. From wind data taken from the GSO Pier, during the time of the January deployment, we observed that wind directions from 6 January until late on 9 January were consistently from the north and northwest and that there was an increase in wind speed starting on 7 January and peaking at almost 10 mis, midday on 8 January. After this peak, winds gradually decreased to about 1 to 2 mis late on 9 January, before a frontal system moved through, in the early hours of 10 January.
This system was responsible for wind speeds in excess of 10 mis over most of 10 January.
The direction of these winds began from the southeast and south, and gradually clocked to the west and eventually the north and northwest, as the system passed. It is thought that the combination of high wind speeds and wind setup produced through an extended wind fetch, in which winds which persistently blow at moderate to high velocities from one particular direction for long periods of time, up to 24 hours or more. This set-up may cause Narragansett Bay's waters to be forced and held in the leeward section of the Bay, until the winds subside, when the waters are free to move back to lower elevations. In the case of this destratification event, the northerly winds may have caused a set-up in which estuarine waters were forced out into Rhode Island Sound, until late on 9 January, when the change in wind direction came about and speeds increased. In fact, the entrance to the West Passage exhibited anomalously large waves, approximately 2-3 meters in height, directed into the Bay, on 10 January. Release ofthis set-up condition, the contribution of spring condition tidal currents and stirring effects of> 12 ml s winds all contributed to the isothermal event on 10 January and the destratification which was observed from 8 January until 13 January. The effects of the destratification on the oxygen data was to mix oxygen down into the deep layer, and saturation levels rose from a low near 92% during the early stratified period, to a sustained level near 100% saturation during the 4 days of the mixing event, with the localized peaks associated with the tidal cycle superimposed on this constant oxygen saturation level. The Li02 was calculated in a manner similar to that for temperature. Values for bottom oxygen which appeared to be affected by instrument contact with the seafloor were removed by averaging across adjacent, reasonable values.
The difference between surface 0 2 and bottom 0 2 were plotted along with tidal height ( Figure 44). Periods of significant 02 gradients are observed between 6 and 8 January, from 13 January to 16 January, and intermittently from 17 January to 22 January.
During the period from 8 January to 13 January, there was only a small difference between the surface and deep oxygen saturation levels, mostly less than 2 % . In addition, the features of oxygen variability are very similar to those for temperature. These results indicate the sensitivity of oxygen variations to stratification events, even in mid-winter.
Following the extended period of destratification between 8 and 13 January, the surface and deep conditions began to diverge, as surface temperature continued to decrease from 4°C on 13 January to approximately 2°C .on 14 January, while deep temperatures leveled off at about 4°C to 5°C between 12 January and 16 January. The lunar tide component continued to weaken as the neap phase approached, and tidal stirring was insufficient to break the developing stratification, although the tidal excursion made its impression on the time-series at all three depths during this brief period of stratification.
It appeared that, in the absence of the mixing event from 16-17 January, the stratification might have had the opportunity to persist and strengthen as surface temperatures dropped considerably from the mixed conditions during 17 January, into the much more strongly -""" 00 January. The input of fresh water decreased over the period 11 January to 15 January as measured by USGS stream gauge data from the Blackstone River, at Woonsocket, RI, to very low levels, and the lack of fresh water input likely reduced the magnitude and stirring energy of the non-tidal circulation as well ( Figure 45). Unfortunately, there is a lack of wind data between 11 January and 14 January, however observing the pressure record from the YSl-600 from the surface, and recognizing that the variations observed in this record give an indication of atmospheric pressure changes, we note that the surface pressure changes, and thus, atmospheric pressure changes are minimal between 11 January and 14 January, suggesting that there were no major storm or wind events during that time. Though the stratification was small, it amounted to a difference between surface and bottom of between 2°C and 3°C and led to an increase in surface oxygen of about 6 %, to 106% saturation, late on 15 January. After this, a wind event with 14 mis winds from the southwest caused another isothermal event, and the stratification was extremely weak until late on 17 January. On 18 January, wind speed dropped briefly near zero, but gradually climbed to about 5 mis over the day and by 19 January~ the wind had been blowing at almost 5 mis, from the north and northwest continuously for almost 72 hours when the stratification eroded on 19 January. This occurred at low tide, and well mixed conditions persisted only until the following high tide, when winds were decreasing to below 2 mis.  stratification, as the bottom water was much warmer than that at the surface or middle. It appears, for the period of the January buoy deployment that the winds were responsible for short term, transient mixing events, on the order of hours, and that the tides were more closely related to longer term mixing events, on the order of days. The direction of the wind and its persistence may affect the bottom conditions by allowing a set-up of higher water levels to the leeward side of the Bay's basin, and possibly a fjord-like venting through pycnocline tilting (Kuo and Park, 1992). The lack of salinity data prevented a more quantitative test of this possibility. One must also consider that the sense of thermal stratification for this deployment, that is, for more strongly, thermally-stratified conditions in January, the surface is much colder than the deep, and the greater this sense of thermal stratification, the lesser the density stratification will be. We can see the relative effects of the tides and winds in these vertical mixing processes. We also note the sensitivity of the dissolved oxygen data to the presence of stratification in the water column. When stratification is permitted to persist for a few consecutive days, dissolved oxygen is able to respond with supersaturated levels in the surface waters,. even in the middle of winter.
Even though benthic oxygen utilization is low for the cold winter season, bottom oxygen levels are able to drop to and below 90 % saturation, further increasing the gradient in dissolved oxygen.

March Deployment:
Though the position of the buoy did not change between deployments, it is important to note that the positions of the middle and deep sensors changed due to the change in the mooring configuration. Previously, the bottom instrument was deeper, and further towards the center of the hole. In the second deployment, the bottom instrument was shallower because of the setup of the mooring, with about 1 meter of line between the anchor and the instrument with no catenary because of the float above the instrument, and because the mooring was situated further to the west, up on the western slope of the hole.
However, in the re-rigging of the instrument mooring, we gained the opportunity to more accurately measure the observed tides, since the bottom instrument was a fixed distance above the bottom. The float and slack in the line above the instrument isolated it from the movements of the buoy.
On about 29 March, surface temperatures began to consistently exceed bottom water temperatures, interrupted only by the 1 April blizzard (Figures 17a and b ). On 3 April, surface temperatures warmed following the storm event, and except for during destratification events, remained warmer than the bottom waters. There were three "modes" of variability in the water column temperature record for the surface and middepth waters. The first mode was of smaller scale variability in which the middle instrument closely followed the features in the surface instrument. This mode was especially evident during longer duration mixing events, where there was not a large gradient between surface and bottom.· This mode also occurred during stratified periods coincident with weak wind speeds, such as on 14 March, between 23 March and 26 March, and during most of 17 April. The second mode was the condition wherein the middle temperature was somewhat following the surface temperature, and then made a sharp and rapid change to conditions similar to the bottom instrument, and back again to surface-like conditions. This mode can be observed on 31 March, 14 April, 19 April and 29 April. This mode seems to follow or precede mixing events, and may signify water column conditions in the midst of a transition between two stable states, a mixed and a stratified water column. The third mode was that of rapid oscillations of larger magnitude, over extended periods between 24 and 96 hours in length. This mode can be observed on 17 March, on 3-4 April, and from 21 to 24 April. Over the 21 to 24 April event, wind speeds are very low, there was no precipitation reported and stream flow for the Blackstone River is falling rapidly. The oscillations in the middle instrument appear to be on a half-tidal cycle, but the bottom water shows no oscillations, only a steady increase over several tidal cycles. This suggests that the oscillations of this mode are related to a response of off-shore waters following the outflow of a large volume of fresh water. The deep layer appeared to be "re-filling" and the oscillations in the middle instrument, apparent in both temperature and salinity, but not oxygen, may be related to the up-Bay movement of the deep water which has been in contact with the surface, perhaps out in Rhode Island Sound.
The mean salinity for the water column appeared to follow the river flow data (Figures 22a and b, and Figure 45). The destratified-condition salinity before 1 April was approximately 28.30/00. It began to decrease after 1 A.pril but rose again between 7-9 April. It then began a long-term decrease to approximately 28. Periods of destratification based on salinity correspond well with those based on the temperature record. They are also of varying duration, from approximately 1 hour to about 2 to 4 days. In the bottom layer, both stratification and destratification appear to occur in stages, mostly as seen on 18 through 20 March, 30 March, 6 April, and 1 through 3 May. In these incremental events, the changes in water column structure are based on tidal cycle events, and especially in a destratification, it is the outgoing tide that primarily modifies the conditions of the bottom water ( Figure 46). Examination of the timing of changes in the water column at the surface and in the deep layer indicates that, for some of these destratification events, the bottom salinity begins its decrease an hour or more prior to the increase in surface salinity.
There are a few occasions where stratification is a more precipitous event, such as The bottom oxygen values are mostly between 90 % and 100 % before 13 April. In fact, the oxygen variability is relatively subdued in the first half of the data set, and even surface oxygen, though supersaturated, only briefly approaches 120% saturation three times. The 1 April storm disrupted the Spring phytoplankton bloom. Four days following this brief but thorough mixing event was the maximum in tidal amplitude, which encompassed the peak period of fresh water runoff from the Blackstone River, and high levels of precipitation, which could provide a significant nutrient input from further up in the Bay (Pilson, 1985).
The combination of these conditions and warming of the local waters, with the seasonal increase in solar radiation, resulted in a large-scale oxygen stratification, with surface saturations greater than 130% during the last half of April. However, this bloom was interrupted by three apparent mixing events, on 19 April, 24 April, and 28 April.
After both the 19 April and 24 April destratifications, surface dissolved oxygen grew exponentially over the first day, but surface oxygen concentration growth slowed as time progressed. The surface dissolved oxygen progression flattened out in the case of the 24 April event, and began to drop in the case of the 19 April event, just prior to the mixing.
From 27 April to 29 April, surface density gradually increased despite the weakening of the spring-neap tidal cycle. The winds, however, increased in magnitude to over 10 mis and direction varied between approximately 350° and between 180° and 200°. Freshwater river flow decreased from an absolute peak of over 3600 ft 3 s-1 on 20 April to around 1500 ft 3 s-1 on 27 April. Eventually, the deep waters were affected, late on 28 April, as density dipped about 0.75 kg m-3 over a tidal cycle, but it rebounded to previous levels. In the following surface oxygen bloom, the surface values were lower than those previously, and the surface oxygen decreased steadily from about 30 April until the end of the data set on 8May.
With respect to bottom oxygen, we also noted a recurring feature . The bottom oxygen data showed occasions, all associated with high tide, wherein the oxygen level would rapidly decrease, on the order of 10 to 30 %, and then return to its previous level.
These spikes appeared to be coincident with rapidly changing wind directions, and frequently appeared immediately before a destratification event, although a strong spike appeared in the middle of a strongly stratified period on 16 April. The relationship between the bottom oxygen and bottom salinity variations do not indicate a hydrographic or water-mass cause for the low oxygen spikes (Figure 48). The oxygen variability is probably a result of processes occurring locally within the hole, and are not due to horizontal exchange with the East Passage of Narragansett Bay. Observing that they occur after a period of stratification or during strong stratification, they may indicate the normal fate of the isolated, deep water, before the outg<;>ing tide is able to mix the oxygendepleted water out of the hole.
The difference between bottom density and surface density and one-tenth of the difference between surface oxygen and deep oxygen, can be used as indices for stratification in the physical structure, and the gradient in oxygen, respectively. The oxygen saturation values were scaled by a factor of 0. 1 to allow plotting of both indices on the same graph (Figures 49a and b). We plotted the two indices along with the Newport Tide Height predictions for 10 March through 8 May, 1997. The tide data show periods of greater and lesser tidal amplitudes. Higher tidal amplitude and thus higher    II II II II II II II II II II II II II  °' ~ water column energy for mixing are found between 6 and 9 April and between 5 and 8 May. The two stratification indices coincide remarkably closely. The fact that the peaks and the mixing periods plot so closely suggests that dissolved oxygen stratification and hydrological stratification are related and that a 1 kg m-3 difference in the surface to bottom density can be accompanied by an approximately 10 % surface to bottom dissolved oxygen difference. The data from this deployment were summarized in a table of conditions observed in these two months, to indicate the forcings responsible for stratification and destratification (Table 5). A destratification event was defined as a difference between the bottom and surface densities of about 0. 5 kg m-3 or less for the 12 m deep water column. The event on 19 April was an exception to this, in that the difference between the two depths was about 1 kg m-3 . However, considering the nature of the event, in which winds were over 18 m s-1 , and the previous stratification was between 2.5 and 4.0 kg m-3 , we classified this event as a destratification. Events of stratification and destratification seem to correspond to the departure from predicted tides, which can indicate either localized winds or large scale weather patterns which can drive subtidal flushing and filling of the estuary (Goodrich, 1988). It appears that the critical wind speed for inducing destratification is between 7 and 9 m s-1 , although lower winds are required when the tidal amplitude is high, as on 6-7 April. Conversely, after the delivery of a large volume of fresh-water, which may strengthen stratification, this stratification may resist higher velocity winds, as on 9-12 April.
Observing the predicted high tide data for Newport, RI and the temperature, salinity.and dissolved oxygen records, one sees that early in March and April, periods of thorough vertical mixing coincided with periods of greater tidal range. The periods in Table 5. A compilation of parameters over periods of stratification and destratification in the March to May 1997 buoy deployment. The code in column 2 indicates the condition, S for stratified and D for destratified, column 3 is the bottom instrument density minus the surface instrument density in kg m-3 , the observed tidal range from high tide to low tide in meters is in column 4, column 5 is the difference between the observed tidal signal as represented by the corrected bottom YSl-600 pressure record minus the Newport predicted tides corrected in time and amplitude to the southern tip of Prudence Island. The next two columns are the wind speed in meters per second, and direction represented by the cardinal direction description from the wind time series. The dash separating individual wind directions indicates the progression of wind direction over time (i.e. N-E indicates that the wind switched from north to east, while NE indicates winds out of the northeast). The final column is the fresh water input in both river outflow from the Blackstone River, as gauged by the USGS at Woonsocket, RI and the rainfall in inches of precipitation as measured by NOAA National Weather Service at T.F. Green Airport in Warwick, RI. which stratification is sustained for days are coincident with lower tidal ranges. We observed an asymmetry in the tide curve which was most notable in spring phase tidal events ( Figure 30). The flood portion of the tidal cycle was extended in time, up to 8 hours, and the ebb portion was as short as 4 hours. This difference may also account for the much stronger mixing observed during the ebb tide, since the change in height was the same for a much shorter period on the ebb, than on the flood.
The data set can be broken naturally into two sections, that before 9 April and that after 9 April. Conditions prior to 9 April are more indicative of winter conditions, in that water temperatures are lower, bottom temperatures are warmer, stratification is more limited and storms, as indicated by higher winds, are more frequent. After 9 April, the water column is warming, surface temperatures begin regularly exceeding bottom temperatures, stratification is stronger and more persistent, and high wind events are fewer, although the diel southwesterly sea-breeze becomes prevalent and strong after 30 April.
On 13 April, the most dramatic stratification ev~nt of the time-series occurred.
Slightly after mid-day on 13 April, surface density began to drop precipitously with the outgoing tide and bottom density started a steep climb which persisted over the next tidal cycle, and finally steadied at high tide, mid-day on 14 April. An extremely well-stratified condition persisted until the occurrence of a storm event on 19 April. This stratification was remarkable in that it represented a classic evolution of estuarine entrainment in the deeper layer as a result of increased fresh water flow in the surface layer. The overall period from 14 April until 16 April was marked by damped tidal excursion and a general decrease in the day-to-day water level. Figure 50 shows a broad decrease in the surface April. Meanwhile, a change in the wind orientation, from the clocking wind system between 11 April and 13 April, to steady winds from the north between 13 April and 15 April occurred (Figure 51). The combination of winds from the north and the increased river flow represented by the decreasing surface salinity, damped tidal range, and general emptying of the estuary, resulted in deep entrainment of cold, high salinity waters from offshore, and the stratification between the warm, fresh surface outflow and the cold, saline deep inflow. This period was followed by a re-filling of the estuary as the tidalheight moving average indicated rising water levels between late on 16 April and a peak late on 18 April. This peak in the estuarine re-filling occurred just prior to a storm on 19 April as marked in Figure   Alternately, the surface waters originating in the Providence River and in Mount Hope Bay can be assumed to support greater photosynthesis (Martin, 1957), and these waters may eventually be advected to the West Passage and thus, to the buoy station. This second scenario is more likely, given the northerly winds, and the high level of supersaturation in the surface waters, a condition which would be more readily mixed to saturation offshore. The deep waters, though exhibiting undersaturation, did not show as low a level of oxygen saturation as one might expect given a several day isolation from the atmosphere. Two possibilities are suggested. First, it may be that there was photosynthesis occurring in the deep layer over the period. However, the low levels of surface photosynthesis, and the generally high turbidity of the waters we observed during buoy servicing, suggest another mechanism. If one assumes that the deep waters were from offshore, in the Rhode Island Sound vicinity, and that this offshore surface water was subducted under the less-dense estuarine outflow, then perhaps the deep estuarine waters had been in recent contact with the atmosphere while still offshore, as surface waters.
This association between the deep estuarine waters and the surface offshore waters may help explain the lag in warming between the deep waters and the surface waters as observed in the temperature time-series ( Figures 17 a and b).
The results from our ADCP profiling were not as useful for correlating with the buoy data as we at first hoped. We were unable to compare conditions over both the ebb and the flood, as sea conditions worsened to the point that the data we acquired on the flood was too poor to use. We were able to construct several current sections that  '..c N water level curve leads the predicted data as the winds and water begin to move in the same direction. One can observe in the sections that the water column was initially ebbing at 07: 40 and this condition was evident through the entire water column. A small area towards the island was characterized by weak currents, and the surface current maximum was centered over the deepest part of the hole. As time advanced, a progressively stronger counter-current started on the shelf between the island and the hole, and the higher surface currents were displaced away from the island towards the middle of the passage between Hope Island and Prudence Island. About 2 hours prior to low tide, this counter-current regime had begun to dominate the entire western portion of the hole, and had worked its way across the deepest part of the hole, to the eastern rise. At about the time of low tide, the western "shelf' waters and the deep center of the hole were flooding at about 20 cm s-1 , while the surface to the east was still ebbing, also at about 20 cm s-1 .
By about 12:30, almost the entire water column was flooding, in a direction parallel to the Hope Island eastern shoreline, except for a few pockets of ebbing waters. The ADCP data obtained during the afternoon of 5 April was noisy due to choppy sea state caused by increasing wind speeds. From the ebb ADCP Data and from our observations of the surface patterns while at the buoy station, we were able to create a diagrammatic sketch of the surface currents and the deep currents in the vicinity of the hole during the switch from ebbing conditions to slack low water ( Figure 56). The size of the arrows are roughly scaled, using our assumptions of the relative flows, and assume that the East and West Passage are slightly out of phase. The main branch of the East Passage begins to flood while the West Passage is still ebbing and vice versa. The pattern is extremely variable near buoy, and the bottom current floods about two hours before the surface goes slack.

Conclusions
This thesis represents the first multi-depth time-series study of water quality parameters in Narragansett Bay, and it is remarkable for both its length of study and temporal resolution, of two month duration and twenty minute sampling interval, respectively. The most important results of this thesis pertain to the relationship between the physical stratification and the oxygen variations. The physical stratification was a function of three main effects: tides, winds and fresh-water flux into the area. The stratification and destratification cycles and their effects on the oxygen saturation demonstrate the interdependence of chemical, biological and physical processes in coastal waters. We observed periods of stratification throughout the month of January, which indicates that wind speed and temperature are not the only determinants of stratification; during the winter the injection of saline offshore waters leads to stratification at the mid-Bay location. In both the January and March deployments, we observed occasions when the mid-Bay study area was affected by waters of much different characteristics than those commonly recognized as waters of estuarine origin. The results suggest the importance of exchanges between the Bay and the offshore waters, especially in the setup and destruction of stratification.
There appeared to be several factors which are related to the beginning of stratification and the beginning of destratification. Increased tidal mixing, represented by larger tidal amplitudes in the spring-neap cycle, caused a weakening of stratification, and usually occurred incrementally over several tidal cycles. The destratification caused by this factor usually commenced with a decrease in bottom density and was eventually followed by an increase in density at the surface. High velocity wind events caused rapid destruction of stratification. However, these events were of a shorter duration, and stratification, if observed before the event, was usually able to reoccur soon after the wind speed decreased. Increased freshwater flow, as represented by higher stream flow volumes and greater precipitation usually helped to strengthen stratification, but were subject to tidal and wind mixing. Thus, the fresh water effect was more difficult to quantify. However, the classic estuarine circulation in which fresher and less dense surface outflow causes entrainment and inflow of deeper, denser and more saline water was observed for a period of high river outflow. The relatively high oxygen saturation values in the deep layer and the warming lag between bottom and surface waters seemed to suggest that the surface offshore waters are entrained into the deep estuarine layer. The extremely high surface oxygen saturation and the lag and timing between peak sunlight and the local maximum in surface oxygen suggest that the surface waters at the buoy station originate in the East Passage from upper Bay locations.
The dissolved oxygen gradient appeared to co-occur with temperature stratification, however the gradient was greater in the March-May deployment because of higher dissolved oxygen concentrations in the deep water during winter. A main advantage in analyzing and discussing the March deployment was that we were able to calculate density at the surface, at mid-depth and at the bottom, and use density to indicate the stratified and destratified periods. The need to represent stratification in terms of density, rather than temperature alone as in the January deployment, was acute since seasonal wanning of the surface waters reduced the vertical temperature gradient over the period between 15 March and 1 April. With respect to winds, it is difficult to measure the effect of wind fetch, on the water column. We were able to observe from our time-series that wind plays a role in destratifying the water column by adding stirring energy at the surface, which can work its way through the entire water column, depending on wind speed and persistence. In terms of changing winds, it seemed that between 7 and 9 mis wind speed was in the critical range for causing destratification, in that winds in this range were able to begin causing destratification. This critical wind speed may be lower for persistent wind from a singular direction, where fetch can build up or for periods of higher tidal stirring. The two wind directions from which Narragansett Bay's coastline configuration might allow a maximum effect of fetch to occur are from the south and from the northeast. These two orientations would maximize the area of open water over which the winds might have effects, and the effects would be observed at the buoy.
For the tidal range, it appeared that changes in elevation of about 1. 5 to 1. 8 m were the critical levels for causing destratification through tidal mixing. This range of tidal height appeared to be sufficiently energetic to cause even the bottom water to participate in mixing, and these mixing events appeared to be marked by bottom-up destratification, as seen on 5-7 April. Although there are exceptions to this observation of a critical tidal range within the data set, one must consider that it is difficult to separate the effects of tides from that of winds. Also, for periods where stratification persists, even with high tidal ranges, higher freshwater outflow or warmer surface temperatures may serve to maintain stratification in the face of forceful mixing.

Appendix 1 YSI Performance Data and Operational Information
In investigating the relationship between physical stratification and dissolved gas concentration using in-situ time-series measurements, we tested a prototype oceanographic buoy. A 3-m long by 46-cm diameter spar buoy employed three YSI-600XL (EndecoNSI, Inc.) multi-parameter water sensors, situated at the three depths, to help measure the differences in parameters along the vertical axis. A YSI-6000 instrument was used for the profiles at the buoy station on various occasions. The YSI-600XL's, as outfitted for our study, are capable of determining water temperature, conductivity, salinity, pH, depth and dissolved oxygen concentration. They are designed from PVC and stainless steel, measure 4.06-cm in diameter, 35 .56-cm in length and weigh 0.8-kg. Unlike the YSI-6000, the YSI-600XL does not have internal battery capability or internal datalogging capability. The YSl-600XL's data were logged in a control module on board the buoy, and were be downloaded onto a microcomputer approximately weekly. The YSI-600' s were externally powered by a 12 volt battery which is also located within the buoy.
Also unlike the YSI-6000, it is not capable of recording turbidity.
The YSl-6000, in addition to the parameters measured by the YSI-600, measures turbidity, and includes an internal data logging system and internal power supply.
Parameters measured by the 6000 are listed in Table 1. The instrument communicates with an ASCII terminal or a computer with a terminal emulation program and data is easily exported to any spreadsheet program for data processing. It operates on 12 VDC provided by 8 C-sized alkaline batteries, and adequate logging and communications 200 capabilities are maintained when the charge remains above approximately 8.5 volts. The 6000 is made from PVC and stainless steel, has dimensions of 8.9-cm diameter, 49.5-cm length, and a 3. 0-kg weight with batteries.
Periodic calibration of the instruments is required. This calibration process is relatively simple, is performed in the laboratory, requires about one hour' s time, and is very similar for both the YSI-600 and YSI-6000. The dissolved oxygen sensor is calibrated in a water-saturated atmosphere with atmospheric pressure in mmHg as an entering argument. The conductivity sensor is calibrated with a secondary standard seawater consisting of 2-micron-filtered MERL water. The salinity ofthis seawater is measured by a Guildline salinometer housed at the National Marine Fisheries Building in Narragansett, RI, which is itself standardized using IAPSO Standard Seawater. The pH sensor is calibrated with a three-buffer calibration consisting of standard buffers: 0. 05 molal KH phthalate (4.008), 0.008695 molal KH2P04 with 0.03043 molal Na2HP04 (7.413), and 0.01 molal Borate (9.180); and with each buffer's pH corrected for temperature (Bates, 1954). An upgrade to the pH elect.rode allows the user to calculate a more accurate actual pH, by returning only the pH in millivolts. This allows the operator to calculate actual pH, using a regression of the calibration millivolt values and the operational form of the Nemst Equation as follows : PH.neasured PHbutrer + slope where pHbuffer is the temperature-adjusted pH of a buffer from the calibration. The Na2HP04, or 7 buffer was used since its pH is closest to what we would expect to measure in the field. The Ebutrer is the temperature adjusted millivolt reading from the the calibration for the 7 buffer. The Emeasured is the actual millivolt reading from the field . The slope is the linear regression of the calibration curve, millivolts versus pH of the three buffers. Turbidity is calibrated with a one-point calibration in distilled water with a turbidity of zero. Depth is calibrated in air, for a depth of zero. Upon completion of the calibration procedure, the instrument is stored with its sensors in a moist atmosphere.

Appendix 2
Figures A-1 through A-26. ADCP images from the 5 April field work. Images are referenced by Table 3  .. ..