Modeling water and nutrients fluxes in the Buyuk

Menderes drainage basin, Turkey

O. F. Durdu and V. Cvetkovic

ABSTRACT

O. F. Durdu

Water Resources Research Center,

Adnan Menderes University,

Aydin 09100,

Turkey

E-mail: odurdu@adu.edu.tr

V. Cvetkovic

Department of Land and Water Resources

Engineering,

Royal Institute of Technology,

Brinellvagen 32,

Stockholm SE-100 44,

Sweden

E-mail: vdc@kth.se

Buyuk Menderes catchment, located in the southwestern part of Turkey, is one of the most

populated river basins in Turkey with 2.5 millions inhabitants. Due to increasing activities

in agriculture and industrial sectors, water resources management in the basin is one of the

biggest matters for the future. During the past decade, it has been observed a basinwide

shift to larger monocultural, intensively operated farm units. Therefore, there is land use

conversion from native lands to agriculture. The threat of nutrients pollution, nitrogen and

phosphorus, has become a preoccupation since many lands and rivers undergo a eutrophication

process. The discharge of nutrients from Buyuk Menderes basin to the Aegean Sea through

Buyuk Menderes river also needs to be reduced in order to bring the eutrophication problems

under lasting control. In this paper, the PolFlow model embedded in PCraster is applied to

the catchment for quantifying water and substances fluxes for the five-year period, 1999–2004.

The implementation of the model in the catchment allows describing the water balance and

thus nutrient transport on the landscape surface but also through the soil and aquifer’s layers.

Modeling process is complicated by the transfer of nutrients from diffuse and point-source

emissions, managed by retention and periodic release from storages within the catchment.

Modeling diffuse and point-source nutrient emissions contribution to river loads can be

improved by better knowledge about spatial and temporal distribution of this retention and

release in the basin.

Key words | basin scale modeling, GIS, nutrient loads, water discharge

INTRODUCTION

Excess nutrient loads through drainage basins lead to

eutropication of both coastal and marine waters of the

Aegean Sea region. The EU Water Framework Directive

(WFD), the legislative framework for water management

in Europe, sets clear objectives that a good water quality

status must be achieved by 2015 and that sustainable water

use is ensured throughout Europe. Convention for the

Protection of the Mediterranean Sea against Pollution

(1976) and protocols (1980, 1982) recommends that all

joint parties should take appropriate measures to prevent,

abate and combat pollution of the Mediterranean Sea area

caused by discharges from rivers, coastal establishments or

outfalls, or emanating from any other land-based sources

within their territories. Over the past three decades, the

Aegean Sea has experienced significant water quality

problems due to eutrophication caused by extensive flows

of nutrients from point and non-point sources. Due to high

level of farm practices in Western Turkey, nutrient loads

into Aegean Sea from agricultural lands seem a major

challenge for the future. The Nutrient Reduction Action

Plan for Turkey prepared with Global Environment Facility

(GEF) suggests environmentally friendly farming practices,

such as crop rotation, integrated pest management, early

warning system and improved livestock management,

doi: 10.2166/wst.2009.013

531 Q IWA Publishing 2009 Water Science & Technology—WST | 59.3 | 2009

a good management of manure storage areas, optimum

application of organic and inorganic fertilizers, and moni-

toring and evaluation of soil and water quality.

Buyuk Menderes catchment (Figure 1), a watershed

area of 24,976km2 and 3.2% of the total area of the country,

is located in the southwestern part of the Turkey. The length

of the Buyuk Menderes river is 584km. Mean annual

precipitation in the river basin is 635mm and total mean

annual evaporation is 2,122mm (Class A pan). Precipi-

tation occurs mainly in the winters while during the

summer irrigation period there is very little rain. Buyuk

Menderes basin is a graben area containing Paleozoic

metamorphic formations consisting of gneiss, schist, cryis-

talline limestone. Most of the agricultural land in the Buyuk

Menderes basin is arable land that is drained by artificial

ditches, and irrigated during summer. The land use in the

Buyuk Menderes river basin is as follows: 40% agriculture,

45% forest and scrubland, 10% meadow and pasture, 3%

empty, 1% settlement, 1% surface water. The agricultural

economy of the basin depends on the irrigated cotton

cultivation, corn, fig and olives. Total population of the

basin is 2.5million. 37% of this population is involved in

agricultural activities. Large part of the basin depends on

the Buyuk Menderes river, draining into the Aegean Sea, for

its irrigation water supply. The coastal waters of the Aegean

Sea have been plagued by algal blooms, leading to fish kills

and unpleasant conditions for tourism. These problems

indicate that the eutrophication in the Aegean Sea have

increased over the last century as a result of the increased

inputs from the catchment. Therefore, the improvement of

the water quality in the Aegean Sea requires a reduction in

nutrient pollution in the Buyuk Menderes basin.

Several studies have addressed different aspects of

nutrient pollution in the Buyuk Menderes River basin.

Altınbas et al. (1999) analyzed nutrient inputs from

fertilizers and agricultural chemicals to the surface waters

of Buyuk Menderes basin and investigated water quality of

the Buyuk Menderes river during the irrigation period.

Considering in particular nutrient fluxes, small amount of

work has been done in the basin. For the implementation

of the Water Framework Directive, Netherlands and

Turkey made an agreement (MATRA) to investigate the

status of water quality in the basin.

The study presented in this paper differs from the above

mentioned studies, because it takes point and diffuse sources

into account, and the nutrient transport from pollution

sources to the river outlet. It presents a large scale analysis

and a first attempt to spatially model nutrient transport

within the Buyuk Menders drainage basin. The theory of

the water fluxes and nutrients transport modeling applied

in the study is essentially based on the approach of

De Wit (1999, 2001), that proposes two models to describe

both water flow and nutrients transport through the soil and

surface water (PolFlow model) as well as comprehensive

method to evaluate diffuse and punctual pollution emis-

sions at the river basin scale. This modeling approach was

successfully applied for the Rhine, Elbe, Po and Lake Peipsi-

Chudskoe basins (DeWit 2001;DeWit&Bendoricchio 2001;

Mourad & van der Perk 2004). The objectives of this study is

to: (i) construct and implement a first spatial information

database for the modeling of water fluxes and Nitrogen (N)

and Phosphorus (P) surface and subsurface transport in

the Buyuk Menderes catchment; (ii) give an overview of

the average N and P fluxes in the Buyuk Menderes basin;

and (iii) identify key parameters/processes that are most

influential for determining accurate nutrient loading.

METHODS

Modeling water flow

The approach used for modeling is based on the GIS-based

PolFlow model (De Wit 1999). The PolFlow model isFigure 1 | Buyuk Menderes Drainage Basin.

532 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

embedded in PCRaster, a raster GIS suitable for both static

and dynamic modeling. All spatial inputs have been

converted to a generic map format using a 1km2 grid cell

size (Mourad & van der Perk 2004). The water flow module

of PolFlow simulates long-term water fluxes within the

drainage basin. The module is based on three determinant

factors: (i) the long-term average total runoff or precipi-

tation surplus Q (mm/yr): Q ¼ (P 2 Ea) where P is the long-

term average annual runoff (mm/yr) and Ea is the long-term

average actual evapotranspiration (mm/yr). The denomina-

tion “long term” always refers to an annual average made

over several years (here ten years); (ii) the groundwater

recharge index Qgw/Q, where Qgw is the long-term average

total groundwater recharge (mm/yr). In this study, deep

groundwater recharge is considered, i.e. the groundwater

recharge that flow in the deeper part of the aquifer and has

a longer average residence time that the shallow ground-

water; (iii) the average groundwater residence time RTgw

(year) (Darracq et al. (2005)). The groundwater residence

times are based on Wendland (1992) and Meinardi et al.

(1994), and is described in detail in De Wit et al. (2000).

These hydrological characteristics are estimated for each

km2 in the Buyuk Menderes basin as a function of average

annual precipitation, average annual temperature, eleva-

tion, land cover, soil data, and lithology.

Nutrients transport through the soil/groundwater

system

Nutrients are assumed to follow water flow paths and the

results from the water model, therefore, build the basis for

the nutrient transport module of PolFlow (Darracq et al.

2005). The fraction of the five-year average surplus at the

soil surface (SSS) that is leached, immobilized, volatilized,

or discharged to the surface water is calculated as a func-

tion of average total runoff, groundwater recharge indices,

groundwater residence time, soil, lithology, slope, land

cover, and the type of pollutant (De Wit & Bendoricchio

2001). PolFlow model works on a five years time step.

Different parameters or factors are used to describe and

evaluate different paths of circulation for nutrients at the

surface of the soil and through the soil and groundwater

(Darracq et al. (2005)). The data used to model water fluxes

in the Buyuk Menderes basin is shown at Table 1.

Nutrients transport through the river network

The local drainage direction map (ldd map) is used to route

the nutrients through the river system. A ldd map is a

network of connected cells and developed using a digital

elevation map (DEM by USGS) by connecting each cell

(1 km2) to the lowest neighboring cell all the way down to

the outlet of the basin. Nutrients transport is governed by

a transport fraction and in each cell, there is a certain

fraction of the nutrient loss. The fraction that is transported

from a cell to its lowest neighboring cell is described as

a function of the average annual discharge (accumulated

average annual total runoff), slope, and the presence of

lakes (De Wit & Bendoricchio 2001). In case of increasing

discharge and slope, the relative loss in a cell decreases.

As shown in Figure 2, Lx21 and Lx represent the nutrient

load respectively in the cells x 2 1 and x.

Table 1 | Data used to model water fluxes in the Buyuk Menderes River Basin

Data Resolution Sources

Average annual precipitation – GLOBALSOD NOAA http://www1.ncdc.noaa.gov/pub/data/globalsod

Average annual temperature – http://www1.ncdc.noaa.gov/pub/data/globalsod

Hydrogeological map 1/250,000 DSI and MTA resources

Digital Elevation Model (DEM) 1 degree, DEM 30 arcsecond or 1/250,000

GTOPO30, US Geological Survey

Slope 1 km2 Derived from digital elevation model

Land cover 1/250,000 http://glcf.umiacs.umd.edu/data/landcover/

Soil Map 1/5,000,000 http://www.lib.berkeley.edu/EART/fao.html

Discharge data for rivers www.dsi.gov.tr

533 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

The factor tf represents the fraction of nutrients that is

transported from one cell to the downstream one. (1 2 tf)

represents the retention, loss and decay in the river. tf is a

function of the average discharge in the cell and the terrain

slope. A tf value of 1 means that there is absolutely no

retention, loss or decay in the cell (Darracq et al. (2005)).

The factor tf can be expressed as:

tf ¼ 121

1þ ðrn1:ð1000:slopeÞ þ 1ÞÞ:qrn2

d

h i

in which rn1 is the loss in the river network (s/m3); rn2 is

the loss in the river network; qd is average discharge in

the cell (m3/s); slope is derived from digital elevation model.

The data used for the analysis of nutrients fluxes in the

Buyuk Menderes basin is shown at Table 2.

RESULTS

Water flux model

The water flux model was only validated by measured

discharges on rivers in the Buyuk Menderes catchment.

For water flux, no calibration was done in the model since

none of parameter had to be calibrated. Concerning the

discharge measurements, the monthly data were available

from the EIE and DSI resources which measures dis-

charges at 15 stations (Figure 3). Those data was used to

have a value of measured annual average discharge (m3/s)

so that it could be compared with modeled values of

discharges. Total groundwater recharge index (Qgw/Q)

should have been compared to a ratio between average

annual discharge and average weekly lowest discharge.

Since weekly measurements were not available, this

comparison was not implemented. As demonstrated in

Figure 4a, the comparison between modeled and measured

discharge values indicates that the results are globally in

good agreement.

Figure 2 | Nutrient transportation through the river network.

Table 2 | Data used for the analysis of nutrient fluxes in the Buyuk Menderes River Basin

Data Resolution Reference

Population numbers Basin sub-catchments Turkish Statistical Institute

Connection rate to sewage system and WWTP Basin sub-catchments Basin municipalities

Livestock numbers Basin sub-catchments Department of Agriculture

Agricultural land use Basin cities Department of Agriculture

Crop yields Basin cities Department of Agriculture

Fertilizer use Basin sub-catchments Department of Agriculture

Industrial emissions Basin cities Local Chamber of Industry

Long term average total runoff (Q) 1 km2 Modelling the water flow

Shallow groundwater recharge index (Qgsw) 1 km2 Modelling the water flow

Deep groundwater recharge index (Qgwd) 1 km2 Modelling the water flow

Average annual discharge (qd) 1 km2 Modelling the water flow

Average residence time shallow groundwater (RTsgw) 1 km2 Modelling the water flow

Average residence time deep groundwater (RTdgw) 1 km2 Modelling the water flow

Local drainage direction map 1km2 Derived from elevation map,GTOPO30/HYDRO1K, USGS

534 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

Water flux model uses two different approaches (Wend-

land 1992; Meinardi et al. 1994) for calculating average

annual discharges (De Wit 2001). It appears that both

methods give results in the same range of values. However,

the results from Meinardi et al. (1994) approach are closer

to measured values. The results from Wendland (1992) are

slightly over estimating discharge values. Therefore, annual

average discharge values calculated byMeinardi et al. (1994)

approach were chosen to use in the following nutrients

fluxes modeling. Analysis on modeled and measured

discharge values indicates that absolute error or difference

increases with the discharge value (Figure 4). This is

reasonable because it is acceptable to make an error of

5m3/s on a 50m3/s value rather than on a 0.1m3/s value.

Analysis on relative error, ratio between discharge absolute

error and measured discharge, is demonstrated in Figure 5a.

It appears that the error on discharge lies under 25% for

most of them. The results obtained for the deep ground-

water recharge index were compared with the reference

values (Figure 5b). The comparison results indicated that

some of the values were overestimated, some of them were

underestimated, and only the half of the results were close

to the supposed real indexes values. These variations can be

explained in several ways: 1) the reference recharge indexes,

the model results are compared with, are calculated since

no direct measurements can be done to evaluate them. An

empirical ratio was used to compute them, which is rather

rough and could probably be improved; 2) the results

concerning deep groundwater recharge indexes are very

sensitive and differ a lot with a short distance, since they

depend on many parameters.

The eastern part of the catchment and the corridor

along the river bed have higher recharges in comparison to

the other part of the drainage basin. Model results point out

that the total groundwater recharge index is ranged from 0.5

and 1, except for the lakes where it is considered as zero

(Figure 6a). The deep groundwater recharge is much more

dependent on the aquifer quality. Thus, the soil and rock

characteristics are assign the distribution of the deep

groundwater recharge in the region. Figure 6b demonstrates

the deep groundwater recharge index and it is mostly

between the values of 0.15 and 0.2. The results for residence

times in shallow groundwater mainly between 0 and 5 years

Figure 4 | (a) Modelled vs. measured annual average discharge in a logarithmic scale, (b) Discharge absolute error vs. measured discharge (logarithmic scale) for each of the 15

stations.

Figure 3 | Location of monitoring stations used for water flow validation: 1-Soke,

2-Kocarlı, 3-Cine, 4-Aydın, 5-Yenipazar, 6-Kemer sonrası, 7-Nazilli, 8-Feslek,

9-Cubukdag, 10-Kızıldere, 11-Saraykoy, 12-Yenice, 13-Curuksu, 14-Adıguzel,

15-Bekilli.

535 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

indicate that the eastern part of the catchment has higher

values in comparison to the western part (Figure 6c).

The residence time in deep ground water, mainly situated

between 0 and 200 years, is also higher on the flat

landscapes in the eastern part of the basin (Figure 6d).

From above explanations, the modeled discharge values

appeared to be correct and the groundwater recharge

indexes were reasonable. Therefore, those results were

validated and used in the nutrients fluxes modeling.

Nutrients fluxes model

The nutrient flux model works on a five years time step.

Different parameters or factors are used to describe and

evaluate different paths of circulation for nutrients at the

surface of the soil and through the soil and groundwater.

Calibration for nutrient flux model was carried out to

estimate different parameters: the weighted effect of surface

runoff (sr), the maximum storage capacity of the soil (pms),

the weighted effect of groundwater recharge (gr), the loss

in the river network (rn1), the loss in the river network (rn2).

Calibration was made using measurements of total N and P

concentrations in the main rivers of the Buyuk Menderes

basin. For the sr parameter it was assumed that emissions to

the surface water and nutrient contents of the soils in the

Buyuk Menderes basin were close to those used in the

original model by De Wit (1999). Therefore the sr parameter

for N and P model was accepted as 0.00025. pms parameter

was not a local parameter and thus the value used by De

Wit (1999) could apply to the Buyuk Menderes basin since it

was pollutant specific. pms values for N and P model

chosen as 1.75 and 100, respectively. gr value, Qgw effect on

the leaching of nutrients, was increased slightly to calibrate

the model. The assigned gr value for N and P model was

230. rn1 and rn2 parameters, which govern loss, retention

and decay through the river network, appeared to be the

key for calibration. The Buyuk Menderes basin is largely

spread with alluvial farm lands, therefore, these parameters

were raised compared to values used by De Wit (1999).

The transport model for both N and P is the same except

that denitrification from soil and groundwater is taken into

consideration for N but not for P since this phenomenon

does not occur for this nutrient. As demonstrated in

Figure 7, measured N and P concentrations at 15 different

stations along the Buyuk Menderes basin are very variable

(Altınbas et al. 1999; Guven 2004; DSI data). The reason for

this is that water samples do not always stand for the

Figure 5 | (a) Discharge relative error vs. measured discharge (logarithmic scale) for each of the 15 stations, (b) Modelled groundwater recharge index vs. reference groundwater

recharge index.

Figure 6 | (a) Total groundwater recharge index, (b) Deep groundwater recharge index,

(c) Residence time in shallow ground water in years, (d) Residence time in

deep ground water in years.

536 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

average concentration of the entire river discharge at a

certain point in the river.

For each monitoring station on the 1999–2004 period,

an annual average nutrient concentration was calculated

using the monthly gross values of the measurement data.

Calculated concentration value was multiplied by the

modeled annual average water discharge at the location of

the corresponding station where measured data were not

always available. This process produces a measured or

observed average amount or load of nutrients that passes by

a station in one year. Figure 8 demonstrates total N and P

loads in the river network for the period of 1999–2004.

As shown in the figures, there is an overestimation of the

nutrient loads in the Upper Buyuk Menderes basin (Yenice

Reg., Adıguzel, Bekilli, Curuksu stations) which is especially

visible concerning phosphorus. This might be because of

the transmission factor tf from one cell to the downstream

one (during the transport process) is higher in the east part

of Buyuk Menderes basin than the west part. The reason

for that is that tf depends among other factor on the slope,

which is more important in the east region. In turn a

higher transmission factor leads to less retention and more

nutrients flowing into the streams. The results were

however considered as satisfactory and validated since

they reflected the measurements at most of the monitoring

stations.

In the Middle Buyuk Menderes basin (Nazilli, Feslek,

Kızıldere, Saraykoy, Cubukdag stations), there are signifi-

cant differences between modeled and observed values of

the nutrient loads, which is more prominent in nitrogen.

This might partly be due to fact that the city of Denizli and

surrounding cities have high population and great industrial

sectors and none of those cities have an active waste water

treatment plant. The estimation of point sources has been

Figure 7 | (a) Average measured N concentrations along the Buyuk Menderes river basin, (b) Average measured P concentrations along the Buyuk Menderes river basin.

Figure 8 | (a) Modelled vs measured N load (T. year21) in a logarithmic scale, (b) Modelled vs measured P load (T. year21) in a logarithmic scale.

537 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

based on average relations between emission and popu-

lation and waste water treatment connection and efficiency

rates, which gave reasonable results in the past for densely

populated areas (De Wit 1999; De Wit 2001; Mourad & van

der Perk 2004). Moreover, Kızıldere geothermal power

plant, which produces 57–80MW/year of electricity and

discharges 6million tones of wastewater annually to the

Greater Menderes river, is one of the main direct source of

nutrient pollution of the middle part of the basin. Discharge

of the wastewater into the river causes both ecological and

environmental problems because the discharge temperature

of the water is approximately 1408C (Durdu 2005). Due to

discharges from power plant and other thermal spring into

the river, there are indicative differences between measured

and modeled N and P loads. Figure 8a indicates that the

total N load in the Lower Buyuk Menderes Basin (Nazilli,

Aydın, Soke stations) is underestimated with PolFlow.

The reason is that there are extensive agricultural practices

in this region. Especially chemical pollution by pesticides

and fertilizers, conversion of native and forest lands to

agriculture and soil loss due to agricultural practices are

main diffuse sources of pollution. Figure 8b shows that the

total P loads at each sampling locations are underestimated.

Figure 9 is the map of nutrient loads and it appears that

nutrients accumulate along rivers resulting in an increasing

load moving to the downstream direction.

DISCUSSION

During the past decade, it has become more and more

obvious that environmental research has to consider how to

upscale results from the field scale to a larger scale in order

to enforce local environmental planning. Recent attention

has therefore focused on how to combine simple empirical

and more process-oriented models for hydrology, diffuse

nutrient losses and nutrient turnover in integrated models

(Kronvang et al. 1999). Integrating with a GIS environment,

such models can be significant tools for environmental

management, monitoring and research since they provide

insight into the coupling between the location and devel-

opment of deriving forces and pressures that determine

nutrient loss within a catchment and the resulting environ-

mental impact. The model demonstrated in this study linked

PolFlow model with PCRaster environment for diffuse

nutrient loss to surface water and nutrient turnover in

stream channels and riparian areas with a hydrodynamic

model and a GIS platform.

This study indicated that the distribution characteristics

of N in the Buyuk Menderes river were similar. The

concentration increased in the middle reaches and kept

relatively lower in the upper reaches (Figure 7a). This

distribution pattern of N in the Buyuk Menderes river was

mainly due to the differences in human activities. There is

limited anthropogenic disturbance in the upper reaches due

to the low population density (1–20 people per square

kilometer). In the middle reaches of Buyuk Menderes basin,

population density increases to 20–100 people per square

kilometer, which results in intensive anthropogenic disturb-

ance, through industrial and agricultural production, stock-

breeding, municipal swage, etc. Also a geothermal power

plant near Saraykoy (station 12) and geothermal touristy

hotels discharge their wastewater into the river. Therefore,

N concentrations increased dramatically in the middle

reaches. In the upper and lower reaches, N concentrations

are relatively low due to less differences in anthropogenic

disturbance. This results are also consistent with the Aydın

& Denizli Environmental Case Reports (2006). The study

also showed that the N concentrations in the tributaries

such as stations 3, 7 and 13, were generally clearly lower

than those in the mainstream water.

The distribution of P was different from N in the Buyuk

Menderes basin. In the middle and lower reaches, P

concentrations remained somewhat stable due to sus-

pended matter content (Figure 7b). Dissolved P is easily

adsorbed by suspended matter. The adsorption quantity is

depended to the physical and chemical properties of

suspended matter. In natural waters, P exists almost

exclusively in the form of phosphate ion. Therefore, P is

Figure 9 | (a) Annual modeled N load in the Buyuk Menderes basin in kilos for the

years 1999–2004 and river network, (b) Annual modeled P load in the

Buyuk Menderes basin in kilos for the years 1999–2004 and river network.

538 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

almost always in the particulate phase (Shen & Liu 2008).

The suspended matter in the middle and lower reaches of

the Buyuk Menderes river mainly comes from the upper

reaches and its tributaries. P concentrations in the Buyuk

Menderes river was related to the large amounts of soil

washed into the river by heavy rain, catching high

concentration of P in the form of particulate. With the

increasing runoff in the Buyuk Menderes river from the

upper reaches to lower reaches, the concentrations of N

and P also increased. This pattern in the lower reaches,

however is more obvious than in the middle and upper

reaches (Figure 10a, b). The continuous import of N and

P from the precipitation and surface water to the river

arises mainly from agricultural non-point sources. This

fact also supported by the Aydın & Denizli Environmental

Case Reports (2006).

In the Buyuk Menderes main stream, there are correla-

tions between N and runoff (Figure 10a). However, if we

consider the correlation in the lower reaches, we could find

that are clear positive linear correlation between the

concentrations of N and main stream runoff (r 2 ¼ 0.72,

p , 0.01). The difference in the correlations between N

concentration and runoff in the upper, middle and lower

reaches is related to the N sources of the Buyuk Menderes

river. N in the upper and lower reaches of the river water

mainly come from agricultural fertilizer, and soil erosion

etc. This fact also supported by Boyacioglu & Boyacioglu

(2007). Buyuk Menderes basin is a large agricultural region

where agricultural activities comprise the primary economic

activity. Increase of N concentration in the Buyuk

Menderes river has led to the increase N concentration in

the sea areas adjacent to the Buyuk Menderes estuarine.

This has resulted in abnormal nutrient ratios and important

changes in the ecological environment. There are significant

difference in scale and nature of human activities along

the middle reaches, which lead to the dramatic difference in

non-point sources N imported into the river. In the upper

reaches, however, the scale and nature of human activities

have little difference, the N concentration in the river

seems relatively low. There are no correlations between P

concentrations and runoff in the Buyuk Menderes main

stream, which shows that P concentration is mainly con-

trolled by suspended matter rather than runoff (Figure 10b).

Coastal eutrophication is becoming more and more

serious and tending to increase in the Aegean Sea adjacent

to the estuary of the Buyuk Menderes river, which has

become popular site for harmful algal bloom. Such trends

are presumably related to the nutrients transported by the

Buyuk Menderes river. It is important to construct a

possible dam to reduce the sediment load into the estuary

and improve the transparence of seawater, due to settle-

ment of suspended matter up to 60–70% in the reservoir.

However, it could be expected that, with a construction of a

possible dam, population increase and rapid economic

development in the Buyuk Menderes catchment, more

ecological and environmental issues might arise in the

estuary.

The model employed in this study aims to predict long-

term nutrient loads from a watershed. The mean absolute

(%) error of estimations of yearly nitrogen and phosphate

loads during the years 1999–2004 is 28% and 14% and

the corresponding r 2 values is 0.77 and 0.85, respectively.

Considering the simplicity of the model and the defective-

ness of input data this can be considered a quite good result.

Figure 10 | (a) The relationships between N concentrations and the river runoff r 2 ¼ 0.254, p , 0.01; (b) The relationships between P concentrations and the river runoff r 2 ¼ 0.146,

p , 0.01.

539 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

During the validation steps, the model underestimates

nitrogen loads by approximately 28% even though stream-

flow is quite well predicted during the study period

(Figure 8a). This is probable since the use of fertilizers

dropped dramatically in the beginning of the 200000s due

to the economical problems in Turkey. The model, as

implemented, produces good results for prediction of

phosphorus loads. The modeled phosphorus loads are

about 15% less than observed loads (Figure 8b). A possible

cause to this underestimation could be an underestimation

of phosphorus loads from erosion. The erosion factor,

the rainfall erosivity and the sediment delivery ratio are

dependent on parameters such as agricultural management

practices (Wallin 2005). Since no such data are available,

the chosen value is a very rough approximation. A more

probable explanation for the underestimation of phos-

phorus loads is that the observed values of point source

phosphorus loads in the basin are less than the actual values

and that phosphorus from anthropogenic sources. Water

from the Buyuk Menderes river basin is the main con-

tributor to the Aegean Sea region. The lack of sewage

treatment facilities in the basin contributes to very large

input of phosphorus to the sea.

The quality of data is of major importance for the result

of any modelling. In the Buyuk Menderes basin it is difficult

to find reliable data. In order for modelling results of

nutrient flows from the entire Aegean Sea basin to be

trustworthy for management and future scenario building,

more reliable data are therefore needed. It is important that

such data are easily accessed and homogenous in order to

support sustainable water management in the Aegean Sea

region. The current model helps to identify what data may

be most needed and at what locations.

CONCLUSIONS

A quantitative description of the water balance and nutrient

loads consistent with observations within the Buyuk

Menderes basin for the years 1999–2004 was illustrated

in this study. PolFlow model embedded in PCraster is

applied to the basin for quantifying water and substances

fluxes for the five year period. PolFlow has been developed

and calibrated for the Rhine basin, and it appeared also to

be applicable for the Buyuk Menderes basin. Results from

the study indicate that PolFlow model constitutes a

relatively simple and robust approach to quantifying water

flow and nutrient transport.

The integration of both surface and groundwater

transport within the same model is another asset that

gives also value to this approach. GIS based environment

makes the model outputs easier to analyze.

The modelled nutrient fluxes for the Buyuk Menderes

basin show a reasonable agreement with measured values.

Because of huge industrial sectors, thermal water tourism

and population, nutrient loads in the middle part of the

basin are increasing. In the lower part of the basin, there is

general perception that the emissions from point sources

have been remarkably stable, agriculture has become

the main source of nutrient pollution in the basin rivers.

Several aspects might however been highlighted as points

that could be improved and investigation that could be

carried out further. The quality, accuracy, and resolution of

the input data should be improved. In order to evaluate

better the consequences of climate change, population

increase, and agricultural practices on the nutrient load in

the Buyuk Menderes basin, several alternative scenarios

can now be developed using the presented modeling

methodology.

ACKNOWLEDGEMENTS

The research described in this paper was supported by the

Swedish-Turkish scholarships for European Studies, as part

of the Swedish Institute (SI) exchange program.

REFERENCES

Altınbas, U., Kurucu, Y., Bolca, M. & Turk, T. 1999 Orta ve Asagı

Buyuk Menderes Irmak Sularının Ayrımlı Sulama

Sureclerindeki Fiziksel ve Kimyasal Bilesiminin Sayısal Veriler

ile Bunların Cevre Kirliligine Olan Etkileri Uzerine

Arastırmalar, Report for the Ege University, Faculty of

Agriculture, Soil Science, Izmir, Turkey.

Aydın Environmental Case Report 2006 Ministry of Environment

and Forest report, http://www.cedgm.gov.tr/icd_raporlari/

aydinicd2006.pdf, (accessed 1 May 2008).

Boyacioglu, H., Boyacioglu, H. 2007 Surface water quality

assessment by environmetric methods Environmental

Monitoring and Assessment. 131, 371–376.

540 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009

Darracq, A., Greffe, F., Hannerz, F., Destouni, G. & Cvetkovic, V.

2005 Nutrient transport scenarios in a changing Stockholm

and Malaren valley region Sweden. Water Sci. Technol.

51 (3–4), 31–38.

Denizli Environmental Case Report 2006 Ministry of Environment

and Forest report, http://www.cedgm.gov.tr/icd_raporlari/

denizliicd2006.pdf, (accessed 1 May 2008).

De Wit, M. J. M. 1999 Nutrients fluxes in the Rhine and Elbe

basins. Utrecht: Koninklijk Nederlands Aardrijkskundig

Genootschap, Nederlandse Geografische Studies 259.

De Wit, M. J. M. 2001 Nutrient fluxes at the river basin scale. I: the

PolFlow model. Hydrological Processes 15, 743–759.

De Wit, M. J. M. & Bendoricchio, G. 2001 Nutrient fluxes in the Po

basin. The Science of the Total Environment 273, 147–161.

De Wit, M., Meinardi, C., Wendland, F. & Kunkel, R. 2000

Modelling water fluxes fort he analysis of diffuse pollution

at the river basin scale. Hydrological Processes 14,

1707–1723.

DSI, State Hydraulic Works web pages. http://www.dsi.gov.tr

(accessed 5 November 2005, 25 March 2006, 3 February

2007).

Durdu, O. F. 2005 Contamination of groundwater due to excessive

water abstraction from cold-water aquifers in the Greater

Menderes River basin, Turkey. Fresenius Environmental

Bulletin 14(2), 86–97.

Guven, G. 2004 Buyuk Menderes Nehri, Analysis of Some

Pollutants Parameters of Buyuk Menderes River in Aydın

Region, Master Thesis (in Turkish), Department of Chemical

Science, Adnan Menderes University, Aydın.

Kronvang, B., Svendsen, L. M., Jensen, J. P. & Dorge, J. 1999

Scenario analysis of nutrient management at the river basin

scale. Hydrobiologia 410, 207–212.

Mourad, D. & van der Perk, M. 2004 Modelling nutrient fluxes from

diffuse and point emissions to river loads: the Estonian part of

the transboundary Lake Peipsi/Chudskoe drainage basin

(Russia/Estonia/Latvia). Water Sci. Technol. 49(3), 21–28.

Meinardi, C., Beusen, A., Bollen, M. & Klepper, O. 1994

Vulnerability of diffuse pollution of European Soils and

Groundwater, Report 4615001002, National Institute of

Public Health and Environmental Protection (RIVM),

Bilthoven: RIVM.

Shen, Z. L. & Liu, Q. 2008 Nutrients in the Changjiang river.

Environ. Monit. Assess. Accepted online.

Wallin, A. 2005 Nutrient transport modeling in the Daugava river

basin. Master Thesis, Department of Earth Sciences, Uppsala

University, Uppsala, Sweden.

Wendland, F. 1992 Die Nitratbelastung in den

Grundwasserlandschaften der ’altenBundeslander (BRD).

Berichte aus der Okologischen Forschung, Band 8,

Forschungszentrum Julich (KFA), Julich.

541 O. F. Durdu and V. Cvetkovic | Modeling water and nutrients fluxes Water Science & Technology—WST | 59.3 | 2009