An integrated measurement and modeling methodology for estuarinewater quality management

M ichael Hartnett*，Stephen NashDepartmentof Civil Engineering，National University of Ireland，Galway，Ireland Received 30 July 2014；accepted 23 October 2014 Available online 10 February 2015

?

An integrated measurement and modeling methodology for estuarinewater quality management

M ichael Hartnett*，Stephen Nash
Departmentof Civil Engineering，National University of Ireland，Galway，Ireland Received 30 July 2014；accepted 23 October 2014 Available online 10 February 2015

Abstract

This paper describes research undertaken by theauthors to develop an integratedmeasurementandmodelingmethodology forwater quality managementof estuaries.The approach developed utilizesmodeling and measurement results in a synergisticmanner.Modeling resultswere initially used to inform the field campaign of appropriate sampling locations and times，and field data were used to develop accuratemodels. Remote sensing techniqueswere used to capture data for bothmodel development and model validation.Field surveyswere undertaken to providemodel initial conditions through data assim ilation and determ ine nutrient fluxes into themodel domain.From field data，salinity relationshipswere developed w ith variouswater quality parameters，and relationships between chlorophyll a concentrations，transparency，and light attenuation were also developed.These relationships proved to be invaluable inmodel development，particularly inmodeling the grow th and decay of chlorophylla.Cork Harbour，an estuary that regularly experiencessummeralgalbloomsdue to anthropogenic sourcesof nutrients，was used as a case study to develop themethodology.The integration of remote sensing，conventional fieldwork，and modeling is one of the novel aspects of this research and the approach developed hasw idespread applicability.

?2015 Hohai University.Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license（http:// creativecommons.org/licenses/by-nc-nd/4.0/）.

Estuarinemodeling；Hydrodynam ics；Water qualitymanagement；Nutrients；Phytop lankton；Field measurements；Remote sensing

1.Introduction

Brackish waters are commonly characterized by high productivity due to frequent inputs of nutrients，notably nitrogen and phosphorous，from both freshwater and marine sources（Correll，1978；Nixon，1995）.Significant settlement of particulatematter，sometim es rich in nutrients including organic carbon，often occurs in estuaries.These nutrients prom ote the growth of phytoplankton，and algal bloomsmay occur；high productivity，combined w ith alternating salinity and temperature conditions，can result in fluctuating oxygen levels.These disturbancesand adverseenvironmental conditionsoften result in estuaries being characterized by a low biodiversity.

It is difficult to model the complex interaction of water quality processes in estuarine systems because of the large number of variables thatcan be critical to the onsetof polluted conditions（Hines et al.，2012）.These variables include（1）nutrient inputs from rivers，oceans，sediments，outfalls，and the atmosphere；（2）hydrodynam ics of river flow s，tidal dynam ics，w ind direction，and velocity；（3）shape and bathymetry of the estuary；（4）light available considering day-length，water transparency，temperature，and depth；and（5）distribution，composition，and abundance of natural fauna and flora.Water qualitymanagementissuesin estuariesare receiving increasing regulatory attention from European Union（EU）directives（e.g. Birdsand Habitats directives），and other initiatives such as the Oslo and Paris Commission（OSPAR）.International research programs，notably the EU ELOISE（Estuarine Land-Ocean Interactions Studies）and its global counterpart LOICZ（Land-Ocean Interactions in the Coastal Zone）specialize in modeling nutrientdynam ics in coastal and estuarinewaters.In 1996，aworkshop organized by the Environmental Assessment and Monitoring Committee（ASMO）of the OSPAR reviewed the available models and modeling activities used to assess eutrophication concerns in the North Seaand otherConvention waters（Ourseletal.，2014）.Theonly estuarinemodel included in the workshop was a one-dimensional（1-D）model called EcoWin.The report on the findings of theworkshop（Oursel etal.，2014）listsanumberofdisadvantagesassociatedw ith the model.For example，the model represented one spatial dimension；the m odel could only resolve parameters at a temporal scale of days and months；tides were not adequately resolved；and phytoplankton dynam ics were not well represented.In the intervening period，modelshave been improved considerably:two-dimensional models are now commonly used and models tend to provide high temporal and spatial resolutions.Also，considerable detail can now be included regarding biochemical processes and interactions between variouswater quality parameters.

The present research was undertaken in order to develop a framework，using water qualitymodeling tools，formanaging estuarine water quality in response to general issues arising pertaining to water quality managem ent.Phytoplankton dynamic models require the inclusion ofmany parameters and processes such as oxygen and nutrient cycles，dissolved and particulate organicmatter，sedimentexchanges，algal species，light attenuation，temperature，hydrodynamics，water-atmosphere exchange processes，and respiration by fish，zooplankton，and other invertebrates（Allen etal.，1980；Runca et al.，1996）.The number of water quality variables that numericalmodels require to fully describe the dynamicsof algal blooms is often in the order of 50-60（Hipsey et al.，2007）. This poses severe constraints on computational requirements when high spatialand temporal resolutionsare required.W hen large numbers of parametersare included in amodel thisalso poses difficulties for data collection.To develop a consistent model，data should be provided for initial and boundary conditions for each parameter，along w ith values for kinetic rates and constants thatgovern physicaland biochemical processes. If these data are not available then themodelw illnot be fully prescribed.

This research focuses on the integration of a conventional field survey，remote sensing，andmodeling，w ith one component being used to inform the other.Early-stage scenario modelingwasundertaken to predicthydrodynam icsand solute transport pathways.Modeling results informed the fieldwork program indicating the most appropriate locations for collecting water quality samples.Aircraft remote sensing has rarely been used to provide data for estuarine water quality models；this is a highly novel component of this research. Using specially developed sensors，water surface chlorophylla levels were remotely sensed and integrated into the project. The rem ote sensing provided high-quality spatial data for model intercomparisons；this type of data is rare，but it is highly valuable for synoptic model validation.A further significant integration of modeling and measurements was the developmentof initial fieldsofwater quality parameters；field data provided relationships between salinity and individual water quality parameters and model predictions of salinity distributions throughout the harbor allowed specification of initial fields ofwater quality parameters.

In the follow ing sectionswater quality management issues are discussed and a methodology is proposed for assessing estuarine water quality using modeling and measurements. The modeling aspects are described in Nash et al.（2010）. Details are presented here of themeasurement aspects of the methodology，which was applied to an Irish estuary.Finally，conclusions are drawn from this research.

Fig.1.Schematic diagram of phytoplanktonmodel.

2.Methodology

In this research the number of variables included in the phytoplankton model was kept to a relatively small number. As can be seen from Fig.1，themodelwas limited to the relationships between phytoplankton and the nitrogen，phosphorous，and dissolved oxygen cycles.The arrows represent processes which either increase or decrease the constituent concentrations，and arrows to shaded blocks indicate settlement to seabed.The philosophy underpinning themodel is toinclude param eters for which data could be collected to provide an accurate and efficient predictivemethodology.In this system，chlorophylla（CHL）ismodeled asa gross indicator of phytoplankton activity in thewater.Phosphorousand nitrogen in both organic and inorganic forms are included；these are generally considered essential nutrients in phytoplankton models as one or the other is usually a limiting nutrient（Harper，1992；Howarth and Marino，2006）.

Previous water quality models of estuarine and coastal systems have taken similar hydrographic and nutrient modeling approaches to the one proposed here.However，the approach adopted in this research concentrates on developing a highly integrated measurement and m odeling system.Specifically，the authors integrate numericalmodeling w ith field measurements，remote sensing，and data analyzed w ithin geographic information systems（GIS）.Through field measurements and remote sensing，an extensive data set for an estuary and its catchmentmay be developed.From these data and their analysis，site-specific relationships can be developed between salinity and key water quality parametersand kinetic coefficients used in the numericalmodel.The postulation of these relationships is fundamental to the success of the research and isa key linkage between fieldwork andm odeling.

Fig.2.Schematic diagram of integratedmeasurementand modelingmethodology.

Fig.2 presents a schematic diagram of the m ethodology proposed to develop an algal bloom predictive system.The figure shows the interdisciplinary nature of the project and how the interdisciplinary activities are integrated.All data pertaining to a hydrodynamicmodel，such asbathymetry，tidal data，and freshwater flows，are integrated in a GISsystem and transferred into a hydrodynam ic model via a specially developed interface.The hydrodynam ic model is then used to simulate water circulation patterns over complete spring-neap tidal cycles；themodel includes the effects of freshwater inflows，surfacew ind stress，and theCoriolis force.Themodel is then used to perform a number of initial solute transport simulations to determine travel pathsof solutes throughout the domain.The results from the solute transport simulations are used to inform the field campaign of the most appropriate sampling locationsand times；this isa key integrating element of the system.The field campaign consists of sampling water quality param eters spatially and temporally throughout the estuary.The results of the field campaign are input to the GIS and are used in a number of ways:to validate the numerical models；through data assimilation，to provide initial and boundary conditions to develop thewater qualitymodels；and to assist in overall trophic assessment of the estuary.Mapsof the catchments drained by the estuary are imported to the GIS and，based on catchment characteristics，discharge rates and fluxes of water quality parameters determ ined.Point sources from domestic and industrial discharges are defined in the GIS.Aircraft remote sensing of surfacewater CHL is carried out and ground-truthed against field data.The water quality model is now developed and validated against the field data and remote sensing data as appropriate.Finally，all data and modeloutputsare analyzed through the GISand trophic status assessment ismade.Details of the hydrodynam ic and solute transport model can be obtained from Falconer and Lin（2003）.

2.1.Field samp ling

Field sam pling is obviously a necessary and inherent componentof anymarinewater quality assessment.However，field campaignsare also heavily influenced by the needsof the water models；data are required to provide boundary and initial conditions and to parameterize various processes，such as the effectsof lighton CHL production.Sampling typically extends from the freshwater rivers entering an estuary throughout theestuary properand into the fully salineadjacent sea.During the field campaign，in situmeasurementsaremade of tidal dynam ics，water depths，current regime，temperature，nutrients，oxygen，CHL，salinity，and transparency.Sampling should be undertaken in autumn，w inter，spring，and summ er so that the data provide a comprehensiveoverview of seasonal variation in water quality parameters.Within this regime，samples should be collected during both spring and neap tides and at high and low waters to include affects of tidal dynamics.As mentioned above，the particular locations and times of sampling are informed by initial runs of the hydrodynamic andwaterqualitymodels.A typical field campaign is outlined in the case study below.Sampling regimes change from site to site depending on local conditions and issues.

2.2.Remote sensing

During the course of the research，two aspects of remote sensingwereutilized to analyze theestuarineenvironment.The geometric aspectrefers to theability of remotesensing imaging systems to picture effectson，ornear，the sea surface in such a way that they can bemapped and hence used to identify the positionofvarious frontsand features.The radiometricaspectof remote sensing refers to the ability to quantify light or reflectance levels.This is particularly important if it can be simultaneously achieved in different spectral or color bands.Ocean color sensorsaredesigned to pick up and quantify the different color components of light reflected from the ocean.This light containsamixture of light reflected from the surface and from w ithin thewaterbody.The lightreturning from w ithin thewater body contained the information required for this research.

An analysis of the remote sensing techniques available to the project and the research modeling requirements led to the definition of four functions for remote sensing:

（1）CHL m onitoring to validate model predictions；

（2）Dye plum emonitoring to validate performance of hydrodynam ic and transportmodels；

（3）High and low water boundary mapping to validate bathymetric model；

（4）Shoreline imaging asa reference database for land use and change detection.

In-water reflectance（R）was used to determine CHL concentrations.R is dependent on the absorption and back-scattering coefficients of water.In any small region of the reflectance spectrum（e.g.，over about 40 nm），any substantial change in R can be solely attributed to the absorption characteristics of CHL，as opposed to back-scattering，and as per Walsh（1998）it can therefore be shown that where a is the absorption coefficientof thewater，andλis the wavelength.Cork Harbour，which was used as a case in this study，was overflown in predeterm ined flight lines（identified from preliminary model results）and，using Eq.（1），the collected reflectance data were used to compute spatial distributionsof CHL in the upper 2m of thewater column.This isa unique approach to the almost simultaneousquantification of CHL throughout a large complex water body and distributions in a typical estuary can be captured in a 1-h duration flight.Sim ultaneously，water samples were collected from a large number of sites throughout the estuary using conventional field sampling and tested for CHL levels.

The spatial mapping requirements of the research were addressed using both colorandmonochromatic video imaging. The latter system was configured w ith suitable filters to enhance the dye plume imaging and thewater-land interface. The remote sensing techniques to support the various research aspects of this project are presented in Table 1.

Table 1 Remote sensing capabilities available to this project.

2.3.Description of Cork Harbour

Cork Harbour（see Fig.3）is relatively deep and long（17.72 km），w ith a large surface area（85.85 km2），draining a large freshwater catchment（1 860 km2）.It is a macro-tidal harborw ith a typicalspring tide rangeof 4.2m at theentrance to the harbor.At low water，extensive areas ofmud-flats and sand-flats are exposed in the harbor.These mudflats and the saltmarsh areas in the harbor are important ecosystems for birds.Cork Harbourwas designated a Special Protection Area（SPA）under the 1979 W ildbirds Directive（79/409/EEC）.

The level of industrialization in the harbor is amongst the highestin estuarineareas in Ireland.In addition to the industrial effluents from these industries，the harbor receivesmunicipal sewage discharges and a variety ofwaste transported seaward by the Lee River.The bulk of themunicipal and licensed industrialwaste discharged into the estuary and harborwaters is non-toxic，biodegradable organicmatter.However，at the time the research was conducted，thewater quality in the Lee Estuary immediately below Cork City was poor due to the discharge of untreated sewage.High levelsof algae have been recorded in the North Channel.This is due to high levels of both domestic and agriculturally derived nutrients dischargedthrough the Owenacurra River and poor ambient flushing characteristics.The quality recovers somewhat across Lough Mahon so that，while the average quality in Passage West is reasonable，some unsatisfactory levels can be found，principally at low tide.Thewater quality continues to improve from Black Point towards themouth of the harbor.

Of considerable im portance to theunderstandingof thewater quality in the harbor are the riverine discharges（41m3/sannually）.These dischargeshave significanteffectson salinity distributionsand flushing timeswithin theharborand theirnutrient loads，resulting from theintensiveagriculturein thehinterlands，havesignificantconsequences for primary production.

Many studieshave been undertaken on thewater quality of Cork Harbour，for example，O'Sullivan（1977）.Generally，the areas in the harbor that suffered most from low DO concentration and high BOD，phosphorus，ammonia，and nitrate concentrations were the inner estuary and the Lough M ahon area（see Fig.3）.Toxic phytoplankton species have also been recorded in Cork Harbour，e.g.Petersen et al.（1999）and O'Boyle and McDermott（2014）.

Although considerablewater quality datawere available for Cork Harbour，from a modeling perspective there were significantspatialand temporalgaps in the data.Forexample，the majority of data had been collected during the summer and autumn months，w ith gaps in w inter data.Winter data，when biological activity is at itsm inimum，give a good estimate of baseline nutrient levelsw ithin the area and are thus important. It was therefore necessary during this research to conduct additional field surveys to supplem ent existing datasets.

Fig.3.Locationmap of Cork Harbour and rivers.

Fig.4.Field samp ling sites.

Fig.5.Delineation of catchmentsand quality of flow datawithin them.

3.Cork Harbor case study

3.1.Field measurement

A setof sampling siteswereestablished for theproject；their locationswerebased on initialmodel runsand on the locations of pre-existing Irish Environmental Protection Agency（EPA）sites so that new data could be compared w ith previous data. Additionalsiteswere added in inner Cork Harbour，where high levels of nutrients occurred and more detailed data were required.Fig.4 presents the locations of the sampling sites.

Sampling was undertaken over a tidal cycle on all dates. Sampling was more frequent during the summer months，although samples were collected throughout the year to take account of seasonal variation.In total，338 nutrient samples were taken to assistw ithmodeldevelopment.Datawereused to provideboundary conditionsofboth riverand sea fluxesand to provide initial conditions throughout thedomain of themodel. Data were also collated from outfall operators to determine point source loads from industrial and dom estic discharges.

3.2.Estimation of freshwater runoff

The hinterland draining into Cork Harbour consists primarily of agricultural lands.Thus，the determination of flows and associated nutrient loads from the riverine inputs into Cork Harbourwas significant.The principal riversdischarging into Cork Harbour are the Lee River，Ownacurra River，Owenboy River，and Glashaboy River，w ith associated annual average flow s of 27，1.8，3.8，and 3 m3/s，respectively.Fig.5 delineates the catchments of these rivers.

Daily flow s were recorded by the EPA for the Lee and Owenboy rivers.However，runoff from other sections of the freshwater catchmentareawas not recorded.Fig.5 delineates the areas thathave high-quality flow data and the areaswhere flow measurements are sparse.The areas w ith poor-qualityflow data are com prised of the sm aller river catchments and the areas adjacent to each estuary，where runoff pathways are notapparent from mapping.In the first instance，flow rates for these ungauged areaswere estimated using long-term rainfall and evaporation data（from the period of 1960-1990）from Met Eireann（the Irish meteorological service，www.met.ie）. Monthlymean rainfall and evaporation values from a network of gauging sites were converted into area-based estimates utilizing point data interpolation techniques w ithin the research GIS facility.Values produced by thismethod were compared againstmeasured flow in a series of catchments. W hereas the gross annual estimates of flow swerew ithin 10% of m easured annual flow s，the distribution of runoff on a monthly basis was less satisfactory.Therefore，estimates for the ungauged catchmentswere derived from the computation of a unit area-based runoff rate from adjacent catchments where flow was gauged and applied to the ungauged catchments based on relative areas.Fig.6 presents the average monthly flows for each of the fourmain rivers.

Fig.6.Meanmonthly flows for fourmain rivers.

Fig.7.Analysis of catchment nutrient load data.

3.3.Nutrientboundary and initial conditions

From two particular aspects the field data were invaluable in developing thewaterqualitymodel.Firstly，models required boundary conditions of fluxes of water quality parameters；these were estimated from the field data.Secondly，during model development it was decided for computational efficiency not to begin model simulations w ith unpolluted or clean water，but rather to specify a concentration for each water quality parameter at eachmodel cell through the use of water quality parameter initial grids.This is a data assimilation approach（direct insertion）used inmany atmospheric and oceanographic m odeling activities.

3.3.1.Water quality flux boundary conditions

Boundary fluxesare seasonal.Hence，sampling rates varied between months and years.Boundary conditions for marine nutrient fluxes ofmaterial into Cork Harbour from the Celtic Seaweremeasured at the entrance to the harbor during flood tides.Freshwater boundary conditions of nutrient fluxeswere obtained by sampling at each of the four main rivers discharging into the harbor（Costello et al.，2001）.It is seen in Fig.6 that the Lee River is the dominant freshwater discharge into the harbor.Fig.7 presents a plan view of the catchments ofeach of the four riversand show s detailsof the quality of the nutrient input data thatwere collected during the project.The quality of the data of the nutrient loads varied throughout the river catchment；for some parts of a catchment therewere no data，for some parts data was considered reasonable，and for some parts the data were considered good.The classification of reasonable and good wasbased on the spatialand temporal data resolutionsas defined by Costello etal.（2001）.It is seen in Fig.7 that therewere good data available for themain river，the Lee River.Asmost of the hinterland drained by all rivers had similar land uses，namely dairy farming，one data set of nutrient loadswas devised based on the data for the Lee River catchm ent and applied to all rivers.

From the measured data，average monthly values were derived formarineand freshwaterwaterquality parameters；this provided the model w ith monthly averaged fluxes at the boundaries.Data collected during this projectwere added to existing EPA dataand analyzed toprovideestimatesofmonthly nutrient fluxes into Cork Harbour from the Celtic Sea and the rivers（see Table 2）.In some cases，mainly because of the limited w inter data，no seasonal trends could be fitted for a variable；in these situationsa singlemean value for thevariable wasdeterm ined.Mean valuesarepresented in Table3.In Tables 2 and 3，TN，TON，TP，SRP，and TAN represent total nitrogen，total organic nitrogen，total phosphorus，soluble reactive phosphorus，and totalammoniacialnitrogen，respectively.

Themean monthly flows for the fourmain rivers are presented in Fig.6；themeanmonthly loadsof TN and TP to the harbor during the year 2000 are presented in Fig.8.Table 4 presents themonthly inputsof TN and TP into theharbor from the four riversand themonthly averaged variations in TN and TP concentrations in the seawater at the boundary of the domain during the year 2000.Itwas estimated that the ratio of the annual average concentrations of TN in seawater to TN in freshwater was approximately 10%；seawater provides baseline levelsof nutrients in the harbor.Freshwater valuesof TN and TP are much larger in w inter months than summer months.However，algal blooms are not problematic at this time of year due to the lim iting conditions of light and temperature.

3.3.2.Initial grids

By specifying initial values of water quality parameters at each grid point w ithin a model dom ain，run times were reduced by approximately 50%.A lso，more accurate resultswere obtained.Initial gridswere developed betweenmeasured water quality parameters and salinity.Biological uptake and release of nutrientsarem inimal inw inter.Hence，January was taken as the start time of themodel；the initial nutrient grids were developed from the water quality datasets collected during January.The model was developed on a grid of 30 m×30 m cells.However，initial water quality data were only available for selected samp ling stations at a few model grid points.This presented a problem，as the model requires initial values of water quality variables for each grid cell. Many of these variables are strongly correlated w ith salinity. Hence，relationships between water quality parameters and salinity were used to extrapolate values from field measurements to the entire harbor.Table 5 presents the relationships that were developed between all measured variables（except CHL）and salinity for January 1999.All regressions were significant at the 5%level.No persistent trends in outliers were observed.The determination coefficients（r2）for these relationships are also presented.The mean concentration of CHL was 2.311，and no relationships were found between CHL and salinity.

Table 2 Seasonal cycles derived from freshwater and marine samp les taken from Cork Harbour and from existing EPA data.

Table 3 Overall average values for freshwater and marine variables in Cork Harbour where no seasonal cycle could be derived.

Table 4 Monthly averaged freshwater loads and seawater concentrations of TN and TP for 2000.

Table 5 Relationships between variables and salinity.

Fig.8.Meanmonthly inputsof TN and TP to Cork Harbour in 2000.

The solute transportmodule of thewater qualitymodelwas used to predict salinity distributions throughout the harbor.An extensive calibration/validation of the salinity m odel was carried outduring the project.Detailsof salinitymodeling and comparisons between modelled and measured salinity levels are presented in Nash et al.（2010）.Initial salinity values for each grid point in the estuary were predicted using the validated solute transport model.Then，initial values of water quality parametersateach grid pointwere computed using the salinity at the grid point and the water quality-salinity relationships given in Table 5.The average value for CHL was used as an initial value throughout the model；this appeared reasonable for a w inter value.As an examp le，a p lot of the initial grid for orthophosphates is presented in Fig.9.The use of field data to specify initialwater quality parameters in the model demonstrates the strong synergy betweenmodeling and field data.

Fig.9.Initial grid for orthophosphate concentration.

3.4.Industrial and domestic nutrient inputs to Cork Harbour

Discharge data were available for 26 companies that discharge directly into the harbor.The com position of these industrial discharges and associated flow rates are available in Costello etal.（2001）.Fig.8 shows themonthly contributions from industrial and domestic sources of TN and TP.Many smaller industries discharge to the public sewer system and were thus included w ithin the sewage discharges.Treated domestic waste is discharged to the harbor through nine wastewater treatment plants.In some cases，multiple sources discharged through the same outfall，and in such instances the inputs were added together.Thus，the number of outfalls is less than the sum of the industrial and domestic discharges.

3.5.Light attenuation coefficients

A key factor in the grow th of phytoplankton is light availability（Hartnett and Nash，2004）.When the authors initially developed thewater quality model of Cork Harbour， light attenuation coefficients were chosen from literature sources.However，it has been shown in other research carried out by Hartnett and Nash（2004）thatmodel predictions of CHL using these values were not close to measured data. Thus，itwas necessary to develop a new site-specific expression for the lightattenuation coefficient thatwould be used in thewater quality model.

Due to high levels of suspended solids，estuaries typically have relatively low lightavailability compared to open sea or lake conditions.Secchi disk measurements were taken to determine transparency for thewatersof Cork Harbour.These m easurements were converted from transparency readings to light attenuation coefficients using Walker（1980）:

where KA（m-1）is the lightattenuation coefficientand Dsd（m）is the Secchi disk depth.

During the fieldwork，simultaneousmeasurementsof CHL，sedimentconcentrations，and Secchidisk readingswere taken. These data and Eq.（2）allowed the follow ing relationship of lightattenuation as a function of CHL concentration（CCHL）to be derived:

The above coefficient includes the constant effect of sediment and the variation w ith respect to CCHL.It represents a feedback system based on the current prediction of CCHLbeing used to revise the light attenuation coefficient value for the prediction at the next computational time step.This approach provides the model w ith a light attenuation coefficient that varies both spatially and temporally w ith CHL；this is highly significant as CHL varies highly w ith seasons and also w ith locations throughout Cork Harbour.In Nash et al.（2010）results are presented for predictionsof CHL using the above light attenuation coefficient and values from literature. It is shown that the above formulation greatly improvesmodel predictions.

Light intensity，photoperiod，and temperature are all significant factors affecting phytoplankton grow th.The model includes the variations in these throughout the annual cycle.

3.6.Remote sensing results

Themostnovel aspectof the remote sensing was itsuse to detectCHL distributionsin thesurfacewatersofCork Harbour. Fig.10 shows examples of reflectance spectra from Cork Harbour，Wexford Harbour，and three typical Irish lakes for comparison.The gross notable feature is the brightnessof the reflectance over Wexford Harbour.Here，the additional sediment in the water causes a high value of backscatter and therefore relatively higher reflectance at all wavelengths. However，the shape of the reflectance versus wavelength is quite sim ilar for both Cork Harbour and Wex ford Harbour，indicating thatthe remotesensing unitisdetecting sim ilar light.

From fieldmeasurements themain features of the spectral functions are（1）absorption due to dissolved organic matterbelow 500 nm；（2）the shoulder in the absorption spectrum of purewater causing a decrease from 580 to 620 nm；（3）chlorophy ll absorption feature around 674 nm；and（4）main absorption by water itself centered around 740 nm.A ll of the above featuresare observed in each of the records presented in Fig.10.

The results of the shoreline imaging were used in the construction of the model to develop a more accurate geometric model of Cork Harbour than was possible from using thepublished Admiralty Charts.These photographic datawere also used to assess adjacent land uses.The imaging datawere collected simultaneously with the CHL data，leading to a considerable saving in both time and costs.Fig.11 presents the resultsof the remotely sensed CHL in the surfacewatersof Cork Harbour.The results present a highly spatially resolved image of the almost instantaneous distributions of CHL throughout the domain.

Fig.10.Typical reflectance spectra for estuaries and lakes.

Fig.11.Remotely sensed CHL values for surfacewaters.

Fig.12.Measured salinity athighand low tides in July and November.

4.Discussion

Toxic dinoflagellates prevail in Cork Harbour during summer，indicating that the estuary is eutrophic（O'Boyle and M cDermott，2014）.Thebloom isgenerally limited to the inner estuary，leading to deoxygenation in deeper waters.Also，the bloom moves up and down the estuary w ith the tide.

Analysis of salinity measurements illustrates some of the important mechanisms that prevail and lead to water quality issues in Cork Harbour.Fig.12 shows comparisons of averagedmeasured salinity valuesathigh and low tides taken during November and July.Themeasurements were taken at the locations shown in Fig.4；as expected，the salinity values in July are higher than the corresponding values in November. It can also be seen that the influences of the seawater，and hence its constituents，decrease w ith the distance from the open sea boundary of the Celtic Sea.Thus，water quality issuesare dominated bymajor discharges located in the vicinity of Cork City and Lough Mahon，and by poor ambient flushing（Nash et al.，2010）.It is observed from Fig.12 that there is a relatively steep salinity gradient，particularly for low tide in July，between points C4 and C7，which indicates that Lough Mahon behavesas a distinct sub-region w ithin Cork Harbour；flushing studies support thisargument（see Nash etal.，2010）. It is shown that，during November，there is relatively little difference between the salinity at low and high tidesat points C4 and C7；this implies that there is little tidal flushing in Lough Mahon during thew inter.However，during November material is flushed through Lough Mahon primarily due to the large freshwater inputs from the Lee River.During July，salinity values at low tide show that freshwater inputs are significant.However，the salinity gradient at low tide shows that the freshwater influence prevails only as far as point C7. Beyond this point，the salinity gradient is quite shallow.This illustrates that during the summer there is very little transport ofmaterialoutof LoughMahon into the Cork Harbour region. This contrasts sharply w ith the situation in November when the salinity gradient is relatively uniform between all points during low tide.From the salinity measurements it is clear that，during the critical summer period for water quality，flushing is lim ited，suggesting that the Lough Mahon regionsuffers from water quality problems.Detailed results of data collection and all loads entering the harbor can be obtained from Costello et al.（2001）.

Fundamental to the development of a water quality management model is the specification of nutrient loads from varioussources.It isworth analyzing nitrogen and phosphorus loads into Cork Harbour over-and-above the baselinemarine inputs.Fig.8 presents TN and TP loads from freshwater，industrial，and domestic sources.TN loads derived from freshwater during January to April and during November to December are significantly higher than other sources；these high riverine nitrogen loads occur during times of relatively low biological activity.It is seen that riverine contributions to phosphorous loads are relatively small and that there ismuch less inter-annual variability relative to nitrogen inputs.It is highly likely that thewater quality problemsalluded to above are primarily due to industrial and domestic sources.From a waterqualitymanagementperspective the locationsof nutrient sources are very important；considering that during highly productive times of a year considerable amounts of nutrients are derived from point-source outfalls rather than from agricultural sources，water quality issues may be addressed by providing nutrient rem oval processes in treatment works.If industrial and domestic phosphorous levels were reduced，primary production w ithin the water body would then be significantly phosphorous limited.

Current and tidalmeasurements，airborne remote sensing，and other field data were collected to develop an accurate model.The field campaign was strongly influenced bymodel data requirements，both spatially and temporally.Water sampleswere collected over different years，months，and states of the tide.The field data were analyzed and a number of extrem ely im portant results ensued.Firstly，relationships between various water quality parameters and salinity were derived.From these relationships initial conditions for the model were postulated.Initial salinity fields were computed throughout the domain using themodel.From salinity values，spatially varied initial conditions for the water quality parameterswere developed.This enabled themodel to spin up quicker and to nudge forward to more correct solutions through data assim ilation.Secondly，the simultaneous Secchi disk readingsandmeasurementsof CHLwere used to develop a relationship between the twowhichwassubsequently used in the model to construct a feedback mechanism for improved light conditions w ithin the sub-m odule controlling the grow th of CHL.The rem ote sensing data were extremely useful in efficiently detecting semi-synoptic CHL values throughout the domain.Data from remote sensing were used to construct modelgeometry andwerevery usefulwhenmodel resultswere being analyzed（Nash etal.，2010）.Thisaspectof the research wasvery successfuland the resultssuggestthatthis technology should bew idely used in futurewater quality studies.

5.Conclusions

This research presents a methodology for water quality management where the integration of conventional field surveying，remote sensing，andm odeling is key，w ith one stage being used to inform the other.Nash et al.（2010）show that models based on thismethodology arewell behaved and can be used for practical estuarinemanagement.Here，themethodology has been applied to algal blooms in estuaries.However，it can be also used in modeling other aqueous systems such as rivers and lakes，and in studying other water quality parameters such as sediments，heavymetals，toxic chem icals，and pathogens.The approach hasw idespread applicability in implementing the EU Water Framework Directive，w ith consequential benefits for society in general.

Manym odeling studies lack adequate data to provide initial and boundary conditions for hydrodynam ic and water quality parameters，resulting in inaccurate and computationally expensive models.The integration of measurement and modeling，particularly at the design stage，helps address these data issues.Initial scenario modeling was instrumental in identifying suitable monitoring locations and the most significant data gaps to be targeted in the field campaign.The development of initialwater quality grids using relationships between salinity and water quality parameters derived from field data resulted in a 50%reduction in computational cost due to the elim ination of spin up time.Detailed data assim ilation ofwater quality parameters intom odels as performed in this research is innovative and hadmajor effectsonmodel run times andmodel accuracy.

The Cork Harbour case study has shown that a 1-h flight time in a light aircraft and relatively inexpensive image capture equipment can enable production of a map of nearlyinstantaneous CHL spatial distributions across a complex estuary.Theauthorshave notseen thisapproach towaterquality monitoring before；themaps developed from thisapproach are extremely valuable data for m odel validation.The authors recomm end that remote sensing of this nature should be used more widely in the development and validation of models. Based on this research，the authors recommend that the number of transect linesbe increased to provide higher spatial resolution and that，ideally，data be captured every hour over a full tidal cycle.Further，elements of the remote sensing campaign，such as flight pathsand timing，should be informed by output from early-stage model simulation results.In particular，flushing studies should be used to inform remote sensing campaigns since poorly flushed watersare likely to be m ore prone to eutrophication due to build-up of nutrients.

Acknow ledgements

The authors would like to thank the Irish Environmental Protection Agency，Wexford County Council，Cork County Council，the Irish Marine Institute，and Cork City Council for making data available to this research project.

References

Allen，G.P.，Salomon，J.C.，Bassoullet，P.，Du Penhoat，Y.，de Grandpr，C.，1980.Effects of tides on m ixing and suspended sediment transport in macrotidal estuaries.Sediment.Geol.26（1-3），69-90.http://dx.doi.org/ 10.1016/0037-0738（80）90006-8.

Correll，D.，1978.Estuarine productivity.BioScience 28（10），646-650.

Costello，M.J.，Hartnett，M.，M ills，P.，O'M ongain，E.，Collier，L.，Johnson，M.，Nash，S.，Leslie，R.，Berry，A.，Emblow，C.，etal.，2001.NutrientDynam ics of Two Estuaries in Ireland:Wexford and Cork Harbours.Irish EPA，Wexford.

Falconer，R.A.，Lin，B.，2003.Hydro-environmental modelling of riverine basins using dynam ic rate and partitioning coefficients.River Basin Manag.1（1），81-89.http://dx.doi.org/10.1080/15715124.2003.9635194.

Harper，D.，1992.Eutrophication of Freshwaters:Principles，Problems and Restoration.Chapman&Hall，London.

Hartnett，M.，Nash，S.，2004.M odelling nutrient and chlorophy ll_a dynam ics in an Irish brackish waterbody.Environ.Model Softw.19（1），47-56. http://dx.doi.org/10.1016/S1364-8152（03）00109-9.

Hines，D.E.，Lisa，J.A.，Song，B.，Tobias，C.R.，Borrett，S.R.，2012.A network model show s the importance of coupled processes in them icrobial N cycle in the Cape Fear River Estuary.Estuar.CoastShelf Sci.106（20），45-57. http://dx.doi.org/10.1016/j.ecss.2012.04.018.

Hipsey，M.R.，Antenucci，J.P.，Romero，J.R.，Ham ilton，D.，2007.Dynam ics M odel:CAEDYM v3.1 Science Manual.Centre for Water Research，University ofWestern Australia.

Howarth，R.W.，Marino，R.，2006.Nitrogen as the limiting nutrient for eutrophication in coastal marine ecosystems:evolving view s over three decades.Limnol.Oceanogr.51（1/2），364-376.http://dx.doi.org/10.4319/ lo.2006.51.1_part_2.0364.

Nash，S.，Hartnett，M.，Dabrowski，T.，2010.Modelling phytoplankton dynamics in a complex estuarine system.Proc.ICE:Water Manag.164（1），35-54.http://dx.doi.org/10.1680/wama.800087.

Nixon，S.W.，1995.Coastalmarine eutrophication:a definition，social causes，and future concerns.Ophelia 41（1），199-219.http://dx.doi.org/10.1080/ 00785236.1995.10422044.

O'Boyle，S.，M cDermott，G.，2014.Observationsofa thin near surface layer in an estuarine environment:an exceptional bloom of the dinoflagellate Akashiwo sanguinea in the Lee estuary（Lough M ahon），Co.Cork，in September 2010.Deep-Sea Res.Part II:Trop.Stud.Oceanogr.101，244-248.http://dx.doi.org/10.1016/j.dsr2.2012.12.008.

O'Sullivan，M.C.，1977.Cork Harbour Pollution Report（Volumes 1-17），a Report to Cork County Council.Cork Corporation and Cork Harbour Comm issioners.

Oursel，B.，Garnier，C.，Pairaud，I.，Omanovi，D.，Durrieu，G.，Syakti，A.D.，Le Poupon，C.，Thouvenin，B.，Lucas，Y.，2014.Behaviour and fate of urban particles in coastalwaters:settling rate，size distribution andmetals contam ination characterization.Estuar.Coast.Shelf Sci.138，14-26. http://dx.doi.org/10.1016/j.ecss.2013.12.002.

Petersen，A.C.，Hunter，P.R.，Keevil，C.W.，Merrett，H.，Reynolds，C.S.，1999. Investigation of Toxins Produced by Cyanobacteria:Final Report.Environmental Protection Agency，Ireland.

Runca，E.，Bernstein，A.，Postma，L.，Di Silvio，G.，1996.Control ofmacroalgae bloom in the Lagoon of Venice.Ocean Coast.Manag.30（2-3），235-257.http://dx.doi.org/10.1016/0964-5691（95）00065-8.

Walker，T.A.，1980.A correction to the Poole and A tkins Secchi disc/light attenuation formula.J.Mar.Biological Assoc.U.K.60（3），769-771. http://dx.doi.org/10.1017/S0025315400040431.

Walsh，J.E.，1998.Development of a Low Altitude Remote Sensing Spectrometer.M.S.Dissertation.University College Dublin，Dublin.

This work was supported by the Irish Environmental Protection Agency under the Environmental Monitoring，R&D Sub-Programme，Operational Programme for Environmental Sciences（Grant No.EPA_97_0151）.

*Corresponding author.

E-mail address:michael.hartnett@nuigalway.ie（Michael Hartnett）.

Peer review under responsibility of HohaiUniversity.

http://dx.doi.org/10.1016/j.w se.2014.10.001

1674-2370/?2015 Hohai University.Production and hosting by Elsevier B.V.This is an open access article under the CC BY-NC-ND license（http:// creativecommons.org/licenses/by-nc-nd/4.0/）.