Science of the Total Environment 412-413 (2011) 315–323

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Modelling phosphorus dynamics in multi-branch river systems: A study of the BlackRiver, Lake Simcoe, Ontario, Canada

P.G. Whitehead a, L. Jin b,⁎, H.M. Baulch c,1, D.A. Butterfield d, S.K. Oni c, P.J. Dillon c, M. Futter e, A.J. Wade d,R. North c, E.M. O'Connor f, H.P. Jarvie g

a School of Geography and the Environment, University of Oxford, Oxford OX1 3QY, UKb Department of Geology, State University of New York College at Cortland, Cortland, NY 13045, USAc Environmental and Resource Studies, Trent University, Peterborough ON, Canada K9J 7B8d School of Human and Environmental Science, University of Reading, Reading, RG6 6AB, UKe Department of Aquatic Sciences and Assessment, Swedish University of Agricultural Sciences, SE-75007 Uppsala, Swedenf Lake Simcoe Regional Conservation Authority, Bayview Parkway, Newmarket ON, Canada L3Y 4X1g Centre for Hydrology and Ecology, Wallingford, Oxfordshire, OX10 8BB, UK

⁎ Corresponding author. Tel.: +1 607 753 2188.E-mail address: li.jin@cortland.edu (L. Jin).

1 Current address: Center for Limnology, Universit53706, USA.

0048-9697/$ – see front matter © 2011 Elsevier B.V. Alldoi:10.1016/j.scitotenv.2011.09.073

a b s t r a c t

a r t i c l e i n f o

Article history:Received 8 April 2011Received in revised form 25 September 2011Accepted 26 September 2011Available online 3 November 2011

Keywords:INCA-PLake SimcoeBranched riverPhosphorus loadingManagement practisesWater quality modelling

High rates of nutrient loading from agricultural and urban development have resulted in surface water eutro-phication and groundwater contamination in regions of Ontario. In Lake Simcoe (Ontario, Canada), anthropo-genic nutrient contributions have contributed to increased algal growth, low hypolimnetic oxygenconcentrations, and impaired fish reproduction. An ambitious programme has been initiated to reduce phos-phorus loads to the lake, aiming to achieve at least a 40% reduction in phosphorus loads by 2045. Achieve-ment of this target necessitates effective remediation strategies, which will rely upon an improvedunderstanding of controls on nutrient export from tributaries of Lake Simcoe as well as improved under-standing of the importance of phosphorus cycling within the lake. In this paper, we describe a new modelstructure for the integrated dynamic and process-based model INCA-P, which allows fully-distributed appli-cations, suited to branched river networks. We demonstrate application of this model to the Black River, atributary of Lake Simcoe, and use INCA-P to simulate the fluxes of P entering the lake system, apportion phos-phorus among different sources in the catchment, and explore future scenarios of land-use change and nutri-ent management to identify high priority sites for implementation of watershed best management practises.

y of Wisconsin, Madison, WI

rights reserved.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Eutrophication is a well-recognised threat to water quality in riverand lake systems around the world. High rates of nutrient loadingfrom agricultural and urban sources have contributed to serious prob-lems of surface water eutrophication and groundwater contaminationin many countries. Phosphorus (P) enrichment is the primary driverof eutrophication in most fresh water ecosystems (Schindler, 1971).In Lake Simcoe (Ontario, Canada; Fig. 1), P concentrations have in-creased three-fold since pre-settlement times (Evans et al., 1996;LSEMS, 1995). Associated increases in algal productivity and declin-ing hypolimnetic oxygen concentrations have led to degradation offish habitat for cold water species and have had negative impacts onan economically important recreational fishery (LSRCA, 2009).

Recently, some improvements in lake water quality have been ob-served and been attributed to a decrease in P loading (Winter et al.,

2007). However, further reductions in P loading are required toallow natural reproduction of cold water fish species including laketrout and Lake whitefish. As a result, aggressive P reduction targetshave been set. The Lake Simcoe Protection Strategy, a remediationplan mandated by legislation under the Lake Simcoe Act, requires a40% decrease in P loads to the lake by 2045 (Lake Simcoe PhosphorusReduction Strategy, 2010).

Currently, the largest P source to the lake is its tributaries (LSRCA,2009; Winter et al., 2007). Although recent data indicate that total P(TP) loads from some tributaries are declining (Winter et al., 2007),much larger load reductions are required in the future (Lake SimcoePhosphorus Reduction Strategy, 2010). These reductions will haveto be achieved in spite of continued urban growth, climatic change,and maintenance of a high level of agricultural productivity.

A strong understanding of catchment-scale nutrient dynamics isrequired to understand the relative importance of different P sourcesin catchments and the relationship between source strength and landuse, to describe source-sink behaviour in soils and sediments, and topredict P loads under future conditions. In Lake Simcoe and otherhuman-dominated watersheds, characterisation of the best meansof reducing P loads is a critical research and management goal. As

316 P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

0 5,000 10,000 m

BR15

BR14

BR13

BR12

BR06-1

BR04-1 BR03-1

BR11

BR10

BR09

BR08

BR07

BR06 BR05

BR04

BR03

BR02

BR01

N

S

Reach Boundary

River Network

Elevation (m)

High : 345.1

Low : 218.5

STW INCA Land uses

Forest

Intensive Ag

Non-intensive Ag

Urban

Wetland

0 5,000 10,000 m N

S

1

2

a b

Fig. 2. a) Land use in the Black River catchment, b) digital terrain map including INCA-P reach boundaries of Black River and locations of two sewage treatment works (STW) dis-charging effluents to reach 14 (BR14) and the Simcoe Lake, respectively.

317P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

there is often a considerable lag time between the implementation ofnutrient control measures and lake recovery, management action isoften difficult to assess, a priori.

Computer-based mathematical modelling offers an aid to under-standing current catchment dynamics and investigating the potentialeffectiveness of potential remedial actions. In this paper we considerthe application of a process-based dynamic model, INCA-P, of flow,suspended sediments and P to the Black River (Fig. 2a and b), amajor tributary of Lake Simcoe. The model is used to estimate thefluxes of P entering the lake system which then paves the way foran assessment of alternative management strategies to control Pfluxes and reverse the eutrophication process. An INCA model appli-cation is also helpful in identifying critical data needs and can beused to assess the potential consequences of future land use and cli-mate change. Although INCA-P has been applied widely in a rangeof landscapes and land uses, this paper represents the first applicationof a new version of INCA-P (Wade et al., 2009), which supportsmulti-branched systems including the main streams and their sepa-rate tributaries. This modified model structure is well-suited to den-dritic or branched catchments such as the tributaries of Lake Simcoe.

2. Integrated catchment (INCA) model

2.1. The overview of INCA model

INCA has been developed over the past 12 years and is a dynamicprocess-basedmodel that predictswater quantity and quality in riversand catchments. The philosophy of the INCA model has been to pro-vide a process-based representation of the factors controlling flow

Fig. 1. The location of Lake Simcoe, Ontario, Canada and the Lake Simcoe sub-catchments, flodeposition collectors.

and water quality dynamics in both the land and in-stream compo-nents of river catchments, while minimising data requirements andmodel structural complexity (Whitehead et al., 1998a,b). As such,the INCA model produces daily estimates of discharge, and streamwater quality concentrations and fluxes from both diffuse sourcesacross a catchment and at discrete points along river channels. TheINCA model has been applied widely in over 20 catchments in theUK, 21 catchments elsewhere across the EU, as well as in catchmentsin Brazil, Australia and Nepal. The major applications of INCA havebeen published to date in two special volumes of International Jour-nals, namely, Hydrology and Earth System Sciences, 2002, 6, (3) andScience of the Total Environment, 2006, 365, (1–3).

Over the past 12 years versions for nitrogen, phosphorus, carbon,sediments, ecological parameters (macrophytes and epiphytes), chlo-ride, mercury and a range of other metals have been created (Futteret al., 2007; Jarritt and Lawrence, 2007; Jin et al., 2011; Lazar et al.,2010; Wade et al., 2002, 2004; Whitehead et al., 1998a,b, 2009). Themodels simulate flow pathways and track fluxes of pollutants in theterrestrial and aquatic portions of catchments. The model system al-lows the user to specify the spatial nature of a river basin or catch-ment (using semi, or fully-distributed representation), to alter reachlengths, rate coefficients, land use, velocity-flow relationships and tovary input pollutant deposition loads.

2.2. Recent INCA developments

INCA is constantly evolving and, over the past two years, therehave been two major new developments to the P and sediment ver-sions of INCA. There has been a new version of the sediment model

w gauging stations, river and lake sampling stations, weather stations and atmospheric

318 P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

to integrate the sediment equations into a new version of INCA-Sed(Lazar et al., 2010; Whitehead et al., 2010). In addition, there hasbeen extensive modification of the underlying P process equationsof the INCA-P model so that the delivery and in-stream processingof particulate and soluble P plus the total dissolved P can be undertak-en, instead of simulating the delivery of TP only. These modificationsmean that the INCA-P model is better able to both simulate P trans-port and retention at the catchment-scale, and to also allow for theinteractions between the hydrology, water chemistry and the aquaticecology. Wade et al. (2009) gives a full description of all the equationsand processes in the new version of the model. Finally, the INCA-Pmodel has been modified so that we can simulate fully branchedriver networks. The original version of INCA P is a single stem of themain river that treats the sub-catchments/tributaries in the modelas aggregated inputs, whereas with the multi-branch version wecan simulate branched tributaries in a fully-distributed manner.

The software has been built on top of the latest version of ‘stan-dard’ main-stem INCA-P and can work in either mode (i.e., as a mainstem-only model or fully branched model). Multi-branch INCA-P cannow represent extremely complicated river networks; there may beany number of streams, there are no restrictions on the number oftributaries a stream may have (and indeed, each reach may have anynumber of tributaries as inputs) and no restrictions on stream order.

Fig. 3 shows the flow of information in INCA from a cell containingthe process equations, to the land use scale, with up to 6 land uses,

Fig. 3. A simplified INCA-P model structure and the actual m

then to the sub-catchment level with multiple reaches and finally tothe multi-branch set up. A schematic of the Black River is shown inFig. 3 indicating a main river branch and 3 sub-catchments. In addi-tion, point sources such as STWs can be fed directly into the mainriver system or into any of the sub-catchments. The mass balanceoperates at basic cell level for each land use, the sub-catchmentlevel to sum the land use inputs and also transport fluxes along thesub-catchment reaches and then at the full river scale to sum theloads from all the sub-catchments The catchment sub-reach structurecan be set up to be as complex and detailed as is required. However,in the case of the Black River, we have opted for an intermediate com-plexity approach with the 3 sub-catchments and the main river sys-tem. As in the original version of INCA-P, the set of differentialequations used to model the hydrology and water quality are solvedusing numerical integration routines. Mass balance is enforced at allsub-catchment reach boundaries as well as all main river reachboundaries. This ensures that a full mass balance is obtained alongthe entire river system.

3. Application of INCA-P to the Black River catchment

3.1. Study area (the Black River catchment)

Lake Simcoe (44° 25′N, 79° 20′W) is a large lake (729 km2) situat-ed at an elevation of about 216 metres above sea level (m.a.s.l.) in

ulti-branch structure used for Black River application.

319P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

southern Ontario, Canada (Fig. 1). There are more than a dozen catch-ments which drain into the lake; these vary in complexity in terms ofstream order, land use, slope, P flux and rainfall. The Black River flowsinto the southeast portion of the lake and drains a catchment of322 km2 in area. Although the catchment has substantial forest(25.6% of catchment area) and wetland (23.3%) coverage (Winter etal., 2007), agriculture is the dominant land use in the catchment(24.3% of catchment is in intensive agriculture, 16.6% in non-intensiveagriculture). Major crops in the catchment are: alfalfa, corn (90%of corn is planted for grain), soybeans, hay fodder crops, and wheat(Statistics Canada, 2006). Urban development is relatively limited(10.1% of land area). There are eight towns located in the BlackRiver catchment but Sutton and Mount Albert are the only townswith municipal sewage treatment works (STW) discharging into theLake Simcoe catchment.

The catchment is characterised by fertile soils— largely grey-brownluvisols, and areas of organic soils (Johnson, 1997). Textural qualityranges from water-logged to free-draining sand plain (Johnson, 1997).The growing season ranges from approximately 120–140 frost freedays, andmean annual temperature is approximately 6–7 °C. Precipita-tion occurs year-round, and a significant proportion (20–25%) of precip-itation falls as snow (Johnson, 1997). The peak period of P export is inthe springtime as a result of high flows during snowmelt. Phosphorusloads from this tributary are lower than from many other inflows, andthere have been no significant changes in rates of P export or P concen-trations in the Black River in recent years (1990/91–2003/04) (Winteret al., 2007).

We applied the INCA-P model to the Black River for the period2007–2009. Firstly, the multi-branch reach structure was established,with aid of digital terrain maps for the area. The selection of reachesdepends on many factors including terrain, land use, location ofwater quality and flow monitoring sites, point discharges and anyother factors that might affect flow and water quality. Also it is neces-sary to have reach boundaries where output information is requiredsuch as downstream discharge points into the lake or key sites of eco-logical significance. Fig. 2a and b shows the land use maps for theBlack River catchment together with a digital terrain map indicatingreach boundary structure set up in INCA-P. Although fully-distributedmodelling can be supported, this represents a semi-distributed appli-cation of the model.

3.2. Modelling catchment hydrology

INCA requires time-series of hydrologically effective rainfall (HER)and soil moisture deficit (SMD) to drive themodel. One approach to ob-tain these data is to utilise a model such as the HBV rainfall–runoffmodel. HBV is a conceptual rainfall–runoff model originally developedby SwedishMeteorological andHydrological Institute (SMHI). It has un-dergone several modifications over time to meet different hydrologicalobjectives (Kobold and Brilly, 2006; Lindström et al., 1997; Menzel andBurger, 2002). In HBV, time series of precipitation, temperature andphysical characteristics of the catchment are used to simulate daily run-off, HER and SMD. Parameters and simplifiedmathematical expressionswhich represent the underlying physical processes, such as snowmelt,water residence time, and evapotranspiration rates, are adjustedwithinthe recommended range in order to minimise the difference betweenmodelled and observed flow.

The model has four main storage components: snow, soil mois-ture, an upper runoff zone and a lower runoff zone. The processeswithin and between the four zones are governed by simplified ex-pressions which represent the underlying physical processes andare controlled by parameters which are adjusted during the calibra-tion process (Lawrence et al., 2009). This was achieved by manuallyadjusting the main parameters within the recommended range inorder to achieve a best fit relative to observed daily stream flows.

We applied HBV to the Black River catchment, using meteorologi-cal data, and flow data from the gauging station near the river mouth(BR04 in Fig. 2b). The version of HBV utilised here was implementedin Structural Thinking Experimental Learning Laboratory with Anima-tion (STELLA) modelling software (ISEE, 2009). HBV has been used onprevious occasions to estimate daily SMD and HER to provide inputdata for INCA (Futter et al., 2009; Oni et al., 2011). The HBV modelwas run on the Black River from 1992 to 2009 (Oni et al., 2011). Fur-ther details of the HBV model applications to the other catchments ofLake Simcoe are given by Oni et al. (2011).

3.3. Phosphorus inputs

Two small sewage treatment plants are located in the Black Rivercatchment. The Mount Albert sewage treatment serves a population of~3200 (2008) and discharges into reach BR14 (Fig. 2b). This plant hasa mean flow of 0.0077 m3 s−1 (2006–2008 data) (XCG ConsultantsLtd., 2010) and a mean P concentration of 0.063 mg L−1 (XCG Consul-tants Ltd., 2010), resulting in a load to the river of approximately15.3 kg P per year. The second sewage treatment plant is in Sutton, On-tario, serving a population of ~4100 (Ontario Center for Municipal BestPractices, 2008). This is a direct-to-lake dischargewith a load of approx-imately 88.5 kg P year−1 (2008 data; XCG Consultants Ltd., 2010). Thisplant is not considered an input to the Black River, hence is excludedfrom our model inputs.

In addition to the population served by these sewage treatmentplants, there is a large population (15,108 people) serviced by septictanks in this catchment, which can be an important additional Pinput. Because septic inputs enter soil and groundwater, the phospho-rus is expected to be largely retained by soil-processes. We assumed Prelease rates (excretion plus detergent P inputs) of 2.0 g per capitaday−1 and 57% efficiency of septic tanks in P removal (Stephens,2007). Septic tank loads (0.9 kg P ha−1 year−1) are included in theterm accounting for manure application on non-intensive agriculture.Atmospheric deposition is 0.2 kg P ha−1 year−1 (estimated based onthe nearest monitoring site; Ramkellawan et al., 2009).

Agricultural P inputs were estimated for both intensive and non-intensive agricultural land. To determine manure production in thecatchment, we used Census of Agriculture data to determine thenumber of livestock present in the catchment (Statistics Canada,2006). We assumed 14% of cows were grazed (Statistics Canada2008 in MacDonald and Bennett, 2009), and based on this, we as-sumed 14% of cattle excrement (an estimated 4424 kg P, basedon excretion data from ASAE, 2003) was deposited by grazing cattleon non-intensive agricultural land, resulting in an input of0.8 kg P ha−1 year−1. Beyond these diffuse inputs to non-intensive ag-ricultural land, we assumed a low-level of fertilisation using manureto supplement forage crops (5.1 kg P ha−1 year−1). The sumof fertiliser(manure P+grazing cattle excretion) inputs and septic inputs results inan estimate of P addition to non-intensive land of 6.8 kg P ha−1 year−1.On intensive agricultural land, we assume an intermediate level of fer-tilisation for cropland of 33.3 kg P ha−1 year−1 (Hilborn and Stone,2011). We assume this fertilisation was performed during a 60-dayperiod in spring using both manure, and inorganic fertilisers. There isa large amount of information required to set up the INCA-P modeland a full description of this information is given elsewhere (Wadeet al., 2002, 2009). A summary of key input parameters is given inTable 1.

3.4. Water quality monitoring data

Routine monitoring of TP, SRP, and SS is performed (2–4 samplesper month plus limited event-based sampling) at two sites (seeFig. 2, BR02 and BR14) in the Black River. Specifically, P was ana-lysed colourimetrically either following a digestion of the wholewater sample (TP) or following filtration without digestion (MOE,

Table 1Data sources for INCA P.

Data Data description Data source

Observed water chemistryTP, SRP, SS Routine sampling (2–4 times per month)

and event samplingTwo sampling programmes: Lake Simcoe EnvironmentalManagement Strategy, Provincial Water QualityMonitoring Network

Hydrological inputs, observed dataPrecipitation Daily time series Environment CanadaDischarge Daily time series Environment CanadaFlow-velocity relationship Measurements for flow rating Measured at site BR07, BR09SMD and HER Daily time series HBV derived

Land use data, and P inputsLand use data Ecological land classification and land

use classifications GIS layerGIS

P inputs to nonintensive agriculture Septic tanks Septic population estimates (GIS), excretion and detergentrelease estimates, septic efficiency estimates (Stephens, 2007)

Grazing inputs Cattle populations (GIS; Census of Agriculture Statistics Canada, 2006)Manure fertiliser inputs Low fertilisation estimate for forage crops (OMAFRA 2011 —

Agronomy guide for field crops; http://www.omafra.gov.on.ca/english/crops/pub811/3fertility.htm#table3-7)

P inputs to intensive agriculture Manure+fertiliser inputs Intermediate estimate (Hilborn and Stone, 2011)P inputs to whole catchment Atmospheric deposition Nearest monitoring site (Ramkellawan et al., 2009)P inputs to stream Direct sewage inflow XCG Consultants Ltd. (2010)

320 P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

1983). Suspended solids were analysed gravimetrically after filter-ing a known volume of water through a glass fibre filter (MOE,1983).

INCA-P does not directly simulate SRP but the model can be usedto estimate SRP concentrations using simulated TDP values, basedon the relationship between these two parameters in the studystream. TDP is SRP plus a dissolved (b0.45 μm) hydrolysable fraction(i.e. organic/polymeric P) which is released under acid/persulphatedigestion and autoclaving. However, TDP is not routinely analysedin the Lake Simcoe monitoring programme. In order to assess the re-lationship between SRP and TDP in the Lake Simcoe watershed, wesampled 13 tributaries of the lake between July 2008 and November2009. Sampling was performed 6–7 times per site at 1–2 sites ineach tributary (19 sites in total). Samples were filtered (0.2 μm)and analysed using standard methods (MOE, 2007; Stainton et al.,1977). The SRP–TDP relationship was found to be approximately lin-ear, described by Eq. (1) and bounded by a lower limit of zero:

SRP ¼ 0:6614 × TDP−0:0008

r2 ¼ 0:65;p b 0:001� � : ð1Þ

However, while this might be a useful relationship to use for awhole Simcoe catchment study, here we are concerned with justthe Black River catchment, and on analysing the Black River data, nostrong relationship was found between SRP and TDP. Data fromother regions (e.g. Jarvie et al., 1998; Neal et al., 2010) suggest theSRP–TDP relationship may be flow-dependent, as SRP can be dilutedby storm flow, while TDP may not be significantly flow-affected(Neal et al., 2010). Therefore, measured SRP were not used for cali-bration in this study.

3.5. Scenarios of future change

The Lake Simcoe watershed is undergoing continued urbanisation.To address these changes, and the need for concurrent P load reduc-tions despite the likelihood of increased urban P loading, we utiliseda series of scenarios to reflect likely future conditions, and maximumpossible mitigation measures. We have considered a set of scenarios

for P and sediment control for the Black River system with thesebeing:

Scenario 1a. Increased sewage loads are projected based on ap-proved growth, assuming future population growth is treated by theMount Albert water pollution control plant, using the best possibletechnology to limit P loads to 35 kg year−1 (Lake Simcoe PhosphorusReduction Strategy, 2010).

Scenario 1b. Future population growth with maximum allowable Pdischarge from the sewage treatment plant resulting in loading of75 kg year−1 (Lake Simcoe Phosphorus Reduction Strategy, 2010).

Scenario 2. Controls on fertiliser and manure addition on the inten-sive agricultural area such that the P loading is within the lowerrange of recommended rates at 16.7 kg P ha−1 year−1 (Hilborn andStone, 2011).

Scenario 3. A set of measure to control sediments in the catchmentusing buffer strips and bank erosion controls.

Scenario 4. 1a combined with 2 and 3.

Scenario 5. Assuming all P additions of either fertilisers or manuresare stopped.

Scenarios were run for a three-year period, assuming that envi-ronmental conditions were otherwise unchanged from the conditionsin 2006–2009.

4. Results and discussion

4.1. Hydrological modelling

The INCA-P model was calibrated using the observed flow andwater quality data for the Black River for a 3 year period from1/1/2007 to 31/12/2009. Flows are generally less than 10 m3 s−1 ex-cept for the spring snowmelt period (Fig. 4). Hydrological modellingof the Simcoe catchments is discussed in detail elsewhere (Oniet al., 2011). Importantly, in some cases high flow events (particularlyduring spring runoff) are overestimated or underestimated, which

01020304050

1/1/2007 1/1/2008 12/31/2008 12/31/2009

Flow

(m

3 /s) Measured ModeledBR04

0

0.05

0.1

0.15

0.2

1/1/2007 1/1/2008 12/31/2008 12/31/2009

TP

(mg/

L) Modeled MeasuredBR02

Fig. 4.Measured and simulated flow at BR04 flowmonitoring site and examples of sim-ulated and observed TP concentrations at BR02 on the Black River from 1/1/2007 to31/12/2009.

321P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

will impact modelling, particularly of SS exports. However, in general,a good fit to the observed flows are obtained, as shown in Fig. 4, withan r2 of 0.67 and a Nash–Sutcliffe statistic of 0.58.

4.2. Water quality

In general the model simulates water quality well (Fig. 4). Themodel is being used in the current study to estimate loads of phos-phorus being transported into Lake Simcoe and Fig. 5 shows themean simulated and observed monthly load estimates for the 3 yearperiod, which gives an r2 of 0.91. However, it is difficult to model riv-ers with extreme climatic conditions and in some instances, seasonalice cover on the Black River may also be contributing to differencesbetween modelled and observed water chemistry data. Ice abrasionand altered flow dynamics during periods of ice cover can contributeto very high SS concentrations during periods of ice cover and breakup (Milburn and Prowse, 2000, 2002; Prowse, 1993). The INCA-Pmodelling approach does not explicitly address ice-related sedimententrainment. In addition, the complexity of physical processes associ-ated with snowmelt, combined with high variability in snowpack canaffect the accuracy of flow simulations.

Simulations indicated that water quality in the Black River duringthe model period was fair. Median TP concentrations at BR02 (Fig. 6)were below provisional chemical targets of 0.032 mg P L−1 for streamsin this ecozone (Chambers et al., 2009), and indeed modelled medianconcentrations are near to a more protective “ideal performance stan-dard”whichhas been suggested to prevent adverse effects on ecologicalcondition (Chambers et al., 2009). However at the upstream site BR14,the median TP concentrations were 0.036 mg P L−1, which is abovethe TP target. Approximately 34% of reaches on the Black River wereabove the target, which mainly occurred in the central part of the

Fig. 5. Comparison of observed and simulated TP mean monthly loads.

catchment where intensive and non-intensive agriculture activitiesare high.We also note macroinvertebrate indicators are typically indic-ative of poor to fairly poor conditions, and fish sampling is indicative ofconditions ranging between “good” and “fair” (Louis Berger Group andGreenland International I, 2006). The range and median value of simu-lated TDP concentrations were similar at two sites shown in Fig. 6. Me-dian total SS concentrations were also relatively low but simulatedconcentrations of SS tend to be higher in the upstream reaches(Fig. 6), which are characterised by steep slopes, but settle out in shal-lower portions of the downstream catchments.

Low concentrations of P also make simulation difficult, and arepushing the lower boundaries of model behaviour. At low concentra-tions, field data are frequently near the detection limit, further chal-lenging modelling activities, as the accuracy of field data is alsoexpected to decline. There is also an issue of sample times when con-sidering the differences between the modelled and observed phos-phorus concentrations (Dean et al., 2009). INCA-P models flow andconcentrations at a daily time-step and generate daily mean values.However, grab samples are collected at a single point in time. It isknown that phosphorus concentration can vary significantly sub-daily (Johnes, 2007). Therefore, the grab samples may not be repre-sentative of daily mean concentrations.

Much of the research and management effort on Lake Simcoe istargeted to understanding, and reducing phosphorus loads from therivers to the lake. Our modelling results estimate P export from theBlack River at 4034 kg P/year (average across 3 years). This compareswell with other loading estimates which range between observed es-timates of 3709 kg P/year (LSRCA, 2009) for the period 2004–2007and 2907 kg P/year (Louis Berger Group and Greenland InternationalI, 2006) based on an export coefficient approach. Estimated catch-ment P export rates vary between 0.12 kg/ha/year (this study) and0.11 kg/ha/year (2004–2007 period; LSRCA, 2009). These catchmentexport rates are quite low compared to values for Europe, whichrange from 0.02 kg/ha/year for woodlands up to 0.65 kg/ha/year forcereal crops (Johnes, 1996). Given that the Black River catchment ismixed forest, pasture and agriculture, the export rate of 0.12 kg/ha/year seems reasonable.

A further issue to consider when applying models, is that of uncer-tainty analysis. Uncertainty is always a major problem with waterquality models and techniques such as generalised sensitivity analysishave been used byWade et al. (2001)) to evaluate both parametic un-certainty in the INCA Model and uncertainty in the data used to drivethe model. In general, it is found that there are key parameters thatcontrol the behaviour and in the case of the P model these are oftenrelated to the sediment–water column exchange and the suspendedSRP to particulate P exchange. Research by Jarvie et al. (2006) utilisedan improved sediment–water exchange equation which has been in-corporated into INCA by Wade et al. (2009). INCA has progressed sig-nificantly over the years by taking advantage of new ideas onmechanisms and improved understanding of catchment phosphorusfluxes to enhance the model. In the case of this Black River applicationthere are considerable uncertainties because of the extensive icecover in winter, difficulty in simulation of snowmelt and the low ob-served concentrations, which are sometimes close to the limits of de-tection, and create significant challenges for modelling.

4.3. Scenarios of future change in the Black River

Table 2 summarises the scenario impacts on P changes in terms ofconcentrations of TP at the bottom of the river system and in terms ofTP loads to the lake. Simulation results suggest that P load reductionsfor the Black catchment may be quite difficult to achieve. The STWinput at Mount Albert is relatively small (35 to 75 kg year−) com-pared to the overall load of P moving down the river system(~4034 kg year−1). Projected population increases suggest this loadwill increase at the catchment outlet to the lake, and the predicted

SS c

once

ntra

tions

(m

g/L

)

0

5

10

15

20

BR02 BR14

TP

conc

entr

atio

ns (

mg/

L)

0.00

0.02

0.04

0.06

0.08

0.10

TD

P co

ncen

trat

ions

(m

g/L

)

0.00

0.01

0.02

0.03

0.04

0.05

BR02 BR14 BR02 BR14

Fig. 6. Box plots summarising water chemistry modelling results (TP, TDP and SS) for two sites (BR02 and BR14) on the Black River from 1/1/2007 to 12/31/2009. The box shows75th percentile, median and 25th percentile; the top and bottom whiskers show 90th and 10th percentile, respectively.

322 P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

increases are 1.4% and 3.4% for TP concentrations assuming the bestand worst technologies, respectively. In terms of load, the TP loadswill increase by 1% and 2.5% respectively for the best and worsttechnology.

The second scenario assumes that the farmers act to reduce theirfertiliser and manure use to the lower level of the recommended ap-plication rates. This does lower TP concentration and TP loads mar-ginally by 0.1% but the effect is most significant in springtime andearly summer when a 9.4% reduction is achieved. So, relative topoint-source controls, this does appear to be an effective method oflowering TP loads across the catchment during the spring and earlysummer months.

The third scenario explores the use of buffer strips and bank ero-sion control measures to reduce SS in the river system. Riparian bufferzones have many advantages (Rabeni and Smale, 1995). They mini-mise overland transport of sediment and reduce the direct runoffflow velocity and filter the direct runoff. In a literature review, Liuet al. (2008) suggest a 45–100% sediment removal efficiency depend-ing on the width of the buffer zone and physical characteristics of thesite. Therefore, a 15% decrease of the transport capacity parameterof INCA-P model produces a realistic potential buffer strip effect(Whitehead et al., 2010). In addition, it has been assumed that asmall bank erosion control policy has been introduced to control thebanks as a source of sediments. The effects of these measures in themodel scenario run are significant in terms of SS with the measuresproducing a 6.6% reduction in SS concentrations in the river system.As shown in Table 2, this translates into an improvement in TP con-centrations and load of 6.5% and 5.6% respectively. A factor to consid-er is the source of P in the bed sediment and the subsequentremobilization back into the water column which can delay any P im-provements (Jarvie et al., 2005, 2006).

Scenario 4 is the combined strategy of Scenarios 1a, 2 and 3. It isassumed that population increases but improved technology is intro-duced at the sewage treatment plants. In addition we assume thatthere is a reduced use of fertilisers and manures and also the sedi-ment controls are introduced. Collectively these effects generate TP

Table 2Summary of the scenario simulations.

TP conc.mg L−1

TP conc.% Change

TP loadkg year−1

TP load% Change

Baseline 0.0317 4034Scenario 1a 0.0322 1.4 4073 1.0Scenario 1b 0.0328 3.4 4132 2.5Scenario 2 0.0316 −0.6 4029 −0.1Scenario 3 0.0297 −6.5 3806 −5.6Scenario 4 0.0297 −6.6 3804 −5.7Scenario 5 0.0310 −2.3 3989 −1.1

concentration and load reductions of 6.6% and 5.7% respectively.This is a quite disappointing result, especially if such small achievablereductions are true of other catchments draining into Lake Simcoe.

Finally a further scenario was run to assess the total effects ofstopping all P additions in the agriculture parts of the catchment.This is an exploratory scenario, rather than a realistic one, as muchof this land is designated “greenbelt”, based on its important to re-gional food supplies. The simulated cessation of fertilisation is notconsistent with goals of continued high agricultural productivity.Even under this extreme scenario, the TP concentrations and TPloads would only fall by 2.3% and 1.1% respectively over the 3 yearsof the simulation. Again there is a legacy effect here as there is astore of P available in the river sediments and the catchment and itwill deplete over time, although this can be relatively fast accordingto Jarvie et al. (2006).

Clearly, P reductions to Lake Simcoe will require long-term man-agement changes. Results of these management changes will mani-fest both in relatively short-term changes and longer-term loadreductions as stored P begins to decline. Importantly, Scenario 4 –

which reflects growth commitments and ambitious managementprogrammes – leads to only a ~230 kg reduction in P loads over the3-year simulation. Although greater reductions may be achievedover the long-term, this is a very small proportion of the P reductiontarget which ultimately requires annual loads to be reduced by al-most 30 tonnes (OMOE, Lake Simcoe Protection Plan, 2009). Givenload reductions for the entire Lake Simcoe catchment will have tobe attained in a changed climate, more work is required to under-stand how hydrology will change in the Black River and how alteredhydrology will affect P loads. The INCA scenarios show the difficultyin achieving large-scale reductions from the watershed, given its al-ready low rates of export. Although our model simulations are subjectto uncertainty, none of our scenarios suggested significant reductionsin P loads would be easily attainable. Reducing non-point source pol-lution is a difficult problem in many catchments, and regions of theworld. In Lake Simcoe, most point sources are already under stringentcontrol, near the minimum feasible P discharges. As a result, reducingnon-point source inputs is critical to recovery of the lake. Althoughsome degree of load reduction can be achieved from low P-exportcatchments such as the Black River, management efforts may bebest targeted towards areas which represent hotspots of P export inthe landscape, and catchments where P export is markedly higher.

Acknowledgements

Long-term monitoring data were collected by the Lake Simcoe Re-gional Conservation Authority and Environment Canada. We thankthe Lake Simcoe Conservation Authority and Statistics Canada for accessto Census of Agriculture Data. Research was funded by the Ontario

323P.G. Whitehead et al. / Science of the Total Environment 412-413 (2011) 315–323

Ministry of the Environment (Best in Science Program), and Environ-ment Canada (Lake Simcoe Clean-up Fund). We also extend thanks tonumerous people who provided prompt access to background data,and information on agricultural and landuse practises in the catchment.

References

ASAE. Manure production and characteristics. ASAE; 2003. D384.1 FEB03.Chambers P, Guy M, Dixit S, Benoy G, Brua R, Culp J. Nitrogen and phosphorus stan-

dards to protect the ecological condition of Canadian streams, rivers and coastalwaters. In: Canada Environment, editor. National Agri-Environmental StandardsInitiative Synthesis Report No. 11. Gatineau, Quebec: Government of Canada;2009. p. 79.

Dean S, Freer J, Beven K, Wade AJ, Butterfield D. Uncertainty assessment of a process-based integrated catchment model of phosphorus. Stochastic Environ Res Risk As-sess 2009;23:991-1010.

Evans DO, Nicholls KH, Allen YC, McMurtry MJ. Historical land use, phosphorus loading,and loss of fish habitat in Lake Simcoe, Canada. Can J Fish Aquat Sci 1996;53:194–218.

Futter MN, Butterfield D, Cosby BJ, Dillon PJ, Wade AJ, Whitehead PG. Modeling themechanisms that control in-stream dissolved organic carbon dynamics in upland andforested catchments.Water Resour Res 2007;43:W02424. doi:10.1029/2006WR004960.

Futter MN, Helliwell RC, Hutchins M, Aherne J. Modelling the effects of changing cli-mate and nitrogen deposition on nitrate dynamics in a Scottish mountain catch-ment. Hydrol Res 2009;40:153–66.

Hilborn D, Stone RP. Determining the phosphorus index for a field. Ontario Ministry ofAgriculture FaRA; 2011. editor.

ISEE. Software reference guide. Lebanon, U.S.A.: STELLA software technical documenta-tion; 2009

Jarritt NP, Lawrence DSL. Fine sediment delivery and transfer in lowland catchments:modelling suspended sediment concentrations in response to hydrological forcing.Hydrol Processes 2007;21:2729–44.

Jarvie HP, Whitton BA, Neal C. Nitrogen and phosphorus in east coast British rivers:speciation, sources and biological significance. Sci Total Environ 1998;210:79-109.

Jarvie HP, Jurgens MD, Williams RJ, Neal C, Davies JJL, Barrett C, et al. Role of river bedsediments as sources and sinks of phosphorus across two major eutrophic UK riverbasins: the Hampshire Avon and Herefordshire Wye. J Hydrol 2005;304:51–74.

Jarvie HP, Neal C, Juergens MD, Sutton EJ, Neal M, Wickham HD, et al. Within-river nu-trient processing in Chalk streams: the Pang and Lambourn, UK. J Hydrol 2006;330:101–25.

Jin L, whitehead PG, Siegel DI, Findlay S. Salting our landscape: an integrated catch-ment model using readily accessible data to assess emerging road salt contamina-tion to streams. Environ Pollut 2011. doi:10.1016/j.envpol.2011.01.029.

Johnes PJ. Evaluation and management of the impact of land use change on the nitro-gen and phosphorus load delivered to surface waters: the export coefficientmodelling approach. J Hydrol 1996;183:323–49.

Johnes PJ. Uncertainties in annual riverine phosphorus load estimation: impact of loadestimation methodology, sampling frequency, baseflow index and catchment pop-ulation density. J Hydrol 2007;332:241–58.

Johnson FM. The landscape ecology of Lake Simcoe basin. Lake Reservoir Manage1997;13:226–39.

Kobold M, Brilly M. The use of HBV model for flash flood forecasting. Nat Hazards EarthSyst Sci 2006;6:407–17.

Lake Simcoe Phosphorus Reduction Strategy. Draft Lake Simcoe phosphorus reductionstrategy; 2010. February.

Lawrence D, Haddeland I, Langsholt E. Calibration of HBV hydrological models usingPEST parameter estimation. Oslo, Norway: Norwegian Water Resources and Ener-gy Directorate; 2009. 44 pp.

Lazar AN, Butterfield D, Futter MN, Rankinen K, Thouvenot-Korppoo M, Jarritt N, et al.An assessment of the fine sediment dynamics in an upland river system: INCA-Sedmodifications and implications for fisheries. Sci Total Environ 2010;408:2555–66.

Lindström G, Johansson B, Persson M, Gardelin M, Bergström S. Development and testof the distributed HBV-96 hydrological model. J Hydrol 1997;201:272–88.

Liu XM, Mang XY, Zhang MH. Major factors influencing the efficacy of vegetated bufferson sediment trapping: a review and analysis. J Environ Qual 2008;37:1667–74.

Louis Berger Group and Greenland International I. Pollutant Target Loads: Lake Simcoeand Nottawasaga River Basins Report; 2006.

LSEMS. Lake Simcoe: our waters, our heritage. Lake Simcoe environmental manage-ment strategy implementation program summary of phase I progress and recom-mendations for phase II. In: Heathcote Isobel, editor. Jointly published by LakeSimcoe Region Conservation Authority, Ministry of Natural Resources, Ministry ofEnvironment and Energy and Ministry of Agriculture, Food and Rural Affairs; 1995.

LSRCA. Report on phosphorus loads to Lake Simcoe 2004–2007. Ontario: Lake SimcoeRegion Conservation Authority; 2009.

MacDonald GK, Bennett EM. Phosphorus accumulation in Saint Lawrence River water-shed soils: a century-long perspective. Ecosystems 2009;12:621–35.

Menzel L, Burger G. Climate change scenario and runoff response in the Mulde catch-ment (Southern Elbe, Germany). J Hydrol 2002;267:53–64.

Milburn D, Prowse TD. Observations on some physical–chemical characteristics ofriver-ice breakup. J Cold Reg Eng 2000;14:214–23.

Milburn D, Prowse TD. Under-ice movement of cohesive sediments before river-icebreakup. Hydrol Processes 2002;16:823–34.

MOE (Ministry of Environment). Handbook of analytical methods for environmentalsamples. Toronto, ON: Laboratory Services Branch; 1983.

MOE (Ministry of Environment). The determination of total phosphorus in water bycolourimetry, Ministry of Environment Laboratory services branch, quality man-agement unit; 2007. 26 pp.

Neal C, Jarvie HP, Withers PJA, Whitton BA, Neal M. The strategic significance of waste-water sources to pollutant phosphorus levels in English rivers and to environmen-tal management for rural, agricultural and urban catchments. Sci Total Environ2010;408:1485–500.

Oni SK, Futter MN, Dillon PJ. Landscape-scale control of carbon budget of Lake Simcoe:a process-based modelling approach. J Great Lakes Res 2011;37:160–5.

Ontario Center for Municipal Best Practices. Best practice summary report: inflow andinfiltration-increasing system knowledge through flow monitoring; 2008. Toronto,Canada.

Ontario Ministry of the Environment. Lake Simcoe Protection Plan. Queens printer forOntario; 2009. p. 92.

Prowse TD. Suspended sediment concentration during river ice breakup. Can J Civ Eng1993;20:872–5.

Rabeni CF, Smale MA. Effects of siltation on stream fishes and the potential mitigatingrole of the buffering riparian zone. Hydrobiologia 1995;303:211–9.

Ramkellawan J, Gharabaghi B, Winter JG. Application of weather radar in estimation ofbulk atmospheric deposition of total phosphorus over Lake Simcoe. Can WaterResour J 2009;34:37–60.

Schindler DW. Carbon, nitrogen, and phoshosphorus and the eutrophication of fresh-water lakes. J Phycol 1971;7:321–9.

Stainton MP, Capel MJ, Armstrong FAJ. The chemical analysis of freshwater. Can FishMar Serv Misc Spec Publ 1977;25.

Statistics Canada. Census of agriculture: farm data and farm operator data; 2006.Stephens SLS. Optimizing agricultural and urban pollution remediation measures using

watershed modeling: Review, calibration, validation and applications of the CAN-WET model in the Lake Simcoe watershed, Ontario, Canada. Trent University MSthesis 2007.

Wade AJ, Hornberger GM, Whitehead PG, Jarvie HP, Flynn N. On modeling the mecha-nisms that control in-stream phosphorus, macrophyte, and epiphyte dynamics: anassessment of a new model using general sensitivity analysis. Water Resour Res2001;37:2777–92.

Wade AJ, Whitehead PG, O'Shea LCM. The prediction and management of aquatic nitro-gen pollution across Europe: an introduction to the Integrated Nitrogen in Europe-an Catchments project (INCA). Hydrol Earth Syst Sci 2002;6:299–313.

Wade AJ, Whitehead PG, Butterfield D. The integrated catchments model of phospho-rus dynamics (INCA-P), a new approach for multiple source assessment in heteor-ogenous river systems: model structure and equations. (Vol 6, pg 583, 2002)Hydrol Earth Syst Sci 2004;8:859-859.

Wade AJ, Butterfield D, Lawrence DS, Bärlund I, Ekholm P, Lepistö A, et al. The integrat-ed catchment model of phosphorus (INCA-P), a new structure to simulate particu-late and soluble phosphorus transport in European catchments. Deliverable 185 tothe EU Euro-limpacs project, UCL, London; 2009. p. 67.

Whitehead PG, Wilson EJ, Butterfield D. A semi-distributed integrated nitrogen modelfor multiple source assessment in catchments (INCA): Part I—model structure andprocess equations. Sci Total Environ 1998a;210:547–58.

Whitehead PG, Wilson EJ, Butterfield D, Seed K. A semi-distributed integrated flow andnitrogen model for multiple source assessment in catchments (INCA): Part II— ap-plication to large river basins in southWales and eastern England. Sci Total Environ1998b;210:559–83.

Whitehead PG, Butterfield D, Wade AJ. Simulating metals and mine discharges in riverbasins using a new integrated catchment model for metals: pollution impacts andrestoration strategies in the Aries-Mures river system in Transylvania, Romania.Hydrol Res 2009;40:323–46.

Whitehead PG, Lázár AN, Futter M, Pope L, Wade AJ, Willows R, et al. Modelling sedi-ment supply and transport in the River Lugg: strategies for controlling sedimentloads. British Hydrological Society (BHS) Conference Proceedings, University ofNewcastle; 2010.

Winter JG, Eimer MC, Dillon PJ, Scott LD, Scheider WA, Wilcox CC. Phosphorus inputs toLake Simcoe 1990 to 2003: declines in tributary load and observations on lakewater quality. J Great Lakes Res 2007;33:381–96.

XCG Consultants Ltd. Review of phosphorus removal at municipal sewage treatmentplants discharging to the Lake Simcoe Watershed. Water Environment Associationof Ontario; 2010.