modelling through-soil transport of phosphorus to surface waters from livestock agriculture at the...

www.elsevier.com/locate/scitotenv

Science of the Total Environm

Modelling through-soil transport of phosphorus to surface waters

from livestock agriculture at the field and catchment scale

M.B. McGechana,T, D.R. Lewisa, P.S. Hoodab

aResearch Division, Scottish Agricultural College (SAC), Bush Estate, Penicuik, Edinburgh EH26 0PH, UKbSchool of Earth Sciences and Geography, Kingston University, Kingston upon Thames, Surrey KT1 2EE, UK

Available online 29 March 2005

Abstract

A model of phosphorus (P) losses in a small dairy farm catchment has been set up based on a linkage of weather-driven

field-scale simulations using an adaptation of the MACRO model. Phosphorus deposition, both in faeces from grazing livestock

in summer and in slurry spread in winter, has been represented. MACRO simulations with both forms of P deposition had been

calibrated and tested at the individual field scale in previous studies. The main contaminant transport mechanism considered at

both field and catchment scales is P sorbed onto mobile colloidal faeces particles, which move through the soil by macropore

flow. Phosphorus moves readily through soil to field drains under wet conditions when macropores are water-filled, but in dry

soil the P carrying colloids become trapped so losses remain at a low level. In the catchment study, a dairy farm is assumed to be

composed of fields linked by a linear system of ditches which discharge into a single river channel. Results from linked

simulations showed reasonable fits to values of catchment outflow P concentrations measured at infrequent intervals. High

simulated outflow P concentrations occurred at similar times of year to high measured values, with some high loss periods

during the summer grazing season and some during the winter when slurry would have been spread. However, there was a lack

of information about a number parameters that would be required to carry out a more exact calibration and provide a rigorous

test of the modelling procedure. It was nevertheless concluded that through soil flow of colloid sorbed P by macropore flow

represents a highly plausible mechanism by which P is transported to river systems in livestock farming catchments. This

represents an alternative to surface runoff transport, a mechanism to which high P losses from livestock farming areas have

often been attributed. The occurrence of high simulated levels of loss under wet conditions indicates environmental benefits

from avoiding slurry spreading on wet soil or during rain, and from some forms of grazing management.

D 2005 Elsevier B.V. All rights reserved.

Keywords: Water contamination; Phosphorus; Livestock; Grazing; Slurry; Catchments; Modelling

0048-9697/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.scitotenv.2005.02.015

T Corresponding author. Tel.: +44 131 535 3029; fax: +44 131

535 3031.

E-mail address: [email protected]

(M.B. McGechan).

1. Introduction

It is commonly thought that phosphorus (P)

pollution of surface waters arises mainly in arable

cropped fields, with much of the P transport asso-

ent 344 (2005) 185–199

M.B. McGechan et al. / Science of the Total Environment 344 (2005) 185–199186

ciated with soil-derived sediments detached and

mobilised as the result of tillage operations. However,

there is now a growing body of evidence that high

levels of P pollution can also arise from grassland

fields in pastoral farming areas. Hooda et al. (1997)

measured P losses to stream water in agricultural

catchments in Scotland, observing higher losses in

catchments with intensive dairy farming in the west

than in less intensively stocked/arable catchments in

the northeast. It was assumed at the time of this study

that the main transport route by which P moved to

streams was surface runoff (overland flow). However,

more recent field-scale modelling studies, supported

by measurements of P losses on a dairy farm (Hooda

et al., 1999) suggest that a form of through-soil

transport represents another route by which P can

pollute surface waters. This appears to arise both

during slurry spreading (McGechan et al., 2002;

McGechan, 2002), and during grazing (McGechan,

2003). In this paper the field-scale modelling

approach has been extended to the catchment scale,

in order to investigate whether this through-soil

transport mechanism can account for a proportion of

the high losses in dairy farming catchments measured

by Hooda et al. (1997).

2. Background to modelling soil phosphorus

transport

Whereas many models of field-scale soil nitrogen

(N) processes and dynamics have been developed,

with transformations between pools (representing

different forms of N) and transport of mobile (soluble)

forms in soil water to drainage systems, there are

comparatively few models of soil P processes. The

only known field-scale models which have their own

distinct representation of soil P dynamics (as reviewed

by Lewis and McGechan, 2002) are ANIMO (Groe-

nendijk and Kroes, 1997; Kroes and Rijtema, 1998),

EPIC (Jones et al., 1984; Sharpley and Williams,

1990) and CENTURY (Parton et al., 1987; Metherell

et al., 1993a,b). A further three models, CREAMS

(Knisel, 1988), GLEAMS (Leonard et al., 1987;

Knisel, 1993) and ICECREAM (Simes et al., 1998),

all have copies of the same set of basic equations for P

processes as EPIC. However, all these models were

initially developed as soil N dynamics models, but

additional routines have since been added to represent

P processes, generally by relatively simple adaptations

of the equations for soil N processes with representa-

tion of through-soil transport for soluble P only. Also

in EPIC and its associated models, there is an

emphasis on surface transport rather than through

the soil transport of P.

Catchment-scale modelling of P transport is

usually based on the use of Geographic Information

Systems, and has often been considered alongside N

transport. Models include AGNPS, which is an event-

based distributed parameter model that computes flow

and pollutant loadings for a single rainfall event.

AGNPS calculates runoff from agricultural watershed

and the transport processes of sediment, N and P

(Young et al., 1995). TOPCAT-P is a simplification of

TOPMODEL (Dayawansa, 2002) and uses soil

moisture stores and subsurface flow equations plus

representation of P processes as in EPIC. The

distributed catchment models SHETRAN (Ewen et

al., 2000) and SWAT (Arnold et al., 1994; Kirsh et al.,

2002) have also recently incorporated a P component

to predict event-based P loading to surface waters.

There are in fact significant differences between

the behaviour of N and P in the soil, particularly that

N compounds are very soluble to move with soil

water, whereas P has a lower solubility so is

transported extensively in particulate or colloidal

forms with various restrictions on their movements.

Also, P compounds are dreactive solutesT, so are

highly sorbed onto surfaces of soil components. In

early studies of soil P processes (e.g. Barrow, 1978,

1983, as reviewed by McGechan and Lewis, 2002a),

it was assumed its high sorbability would lead to P

being applied in mineral fertiliser or manure becoming

locked up in the soil, so it would not readily leach out

in water flows to drains. More recent studies have

drawn attention to the fact that a large proportion of

the area of sorbing surfaces in the soil is on the

smallest particles (such as clay), which readily

become detached from the soil matrix, and so become

highly mobile in soil water flows. It is now becoming

clear that sorption onto mobile suspended or colloidal

soil particles is an important mechanism leading to

leaching of P to drains. Slurry (liquid manure)

contains large quantities of finely divided insoluble

or colloidal particles of organic matter, onto which

inorganic compounds of P in the slurry tend to be

M.B. McGechan et al. / Science of the Total Environment 344 (2005) 185–199 187

attached by sorption. Also, faeces deposited on the

soil surface by grazing animals has similar character-

istics to slurry regarding the abundance of P-sorbing

colloidal particles. When slurry or dung is deposited

on the soil, some of these P carrying particles find

their way into water flows and to drains.

Particulate or colloidal material moves less freely

than solutes in moving soil water, since it can be

subjected to straining and filtration processes. How-

ever, restrictions on movements are greater for larger

particles and for flows through smaller pores (as

discussed in relation to literature on filtration theory

by McGechan and Lewis, 2002b). The largest soil

pores or dmacroporesT impose very little restriction on

movements of such particles, and also water flows in

macropores tend to be very fast leading to rapid

transport of pollutants to drains. High P loads have

been observed in field drains during the first week or

two following spreading slurry in winter, and also

during grazing by livestock in late autumn (Hooda et

al., 1999), which can only be accounted for by

macropore flow. There are a few other recent studies

from Europe reporting substantial through soil P

losses of a similar nature. Stamm et al. (1988)

observed raised concentrations of P during large

rainfall/drainage events in grassland soils. This

suggests mobilisation of colloid material, since if this

did not occur a lower concentration due to dilution of

the soil solution by incoming rain would be expected.

Similar raised P concentrations during rainfall events

have been observed in through soil leaching from

grassland by Hawkins and Scholefield (1996) and by

Haygarth et al. (1998), and in arable soils by Ulen

(1995) and Grant et al. (1996). These studies also

support the hypothesis that colloid facilitated transport

plays an important role in through soil leaching of P to

receiving waters, as represented by the modelling

exercise described in this paper.

The type of macropore flow considered here takes

place in very wet conditions in aggregated soils (as

opposed to that taking place in cracking clay under

dry conditions). Water-filled inter-aggregate spaces

provide an unrestricted route for rapid water move-

ment, and suspended colloidal particles also pass with

almost no restriction in such flows. Some simulation

models of soil water movement provide representation

of macropore flows, but with varying levels of

sophistication. Out of all known models, the MACRO

model (Jarvis, 1994) has the most comprehensive

treatment of macropore flow; it has representation of

distinct macropore and soil matrix pore regions (or

a domainsa ), with different equations governing water

movement and solute concentrations in each domain.

Recently developments of MACRO include a new

feature of colloid facilitated transport of reactive

(sorbed) contaminants (Jarvis et al., 1999). The main

application of MACRO up to now has been concerned

with movements of pesticides. However, some pesti-

cides exhibit behaviour similar to P, such as sorption

onto surfaces both on mobile colloidal particles and in

the static soil matrix. Parameterisation and testing of

the MACRO model to represent some of the

important soil P loss processes from grazed fields

and fields receiving slurry applications have been

described by McGechan et al. (2002) and McGechan

(2003). In this current paper the parameterised field-

scale model has been extended to carry out scenario

testing for a small catchment in a dairy farming area,

where cows are grazed during the summer months and

slurry is spread on various occasions throughout the

winter.

3. The MACRO model for representation of

phosphorus transport at the field scale

3.1. Basic features of model

The main features of the standard MACRO model,

including the basic equations for water and solute

transport in two ddomainsT, have been described in

detail by Jarvis (1994). A further evaluation of the

model and its equations has been presented by Larsson

and Jarvis (1999). All of these features are retained in

the special version of MACRO used in this study.

3.2. Hydrology

The MACRO model simulates water movements,

both vertically through the profile with a proportion

percolating to deep groundwater, and horizontally to

the back-filled trenches in which field (tile) drains are

located. The hydrological routines in MACRO are

similar to many other soil water models, with the

tension (water release) and hydraulic conductivity

relationships for specified layers in the soil profile


represented by mathematical functions, enabling

Richards’ (1931) equation to be solved at successive

timesteps. MACRO differs from other models in its

treatment of processes in the larger soil pores (macro-

pores) when capillary forces are very low, so water

movements can be assumed to be driven by gravita-

tional forces alone. The boundary between the macro-

pore and soil matrix (micropore) domains is

considered to occur at a soil water tension (ub) of

1.2 kPa, with corresponding water content (hb) and

hydraulic conductivity (Kb) values given by mathe-

matical functions (Fig. 1). In the soil matrix pore

region where both capillary and gravitational forces

must be considered, the hydraulic functions are as

described by Brooks and Corey (1964) and Mualem

(1976). In MACRO, contaminant transport is treated

separately with different concentrations in each of the

two domains. Calibration of the hydraulic processes,

including selection of parameters of the hydraulic

function equations for two soil types, has been

described by McGechan et al. (2002) and McGechan

(2002). Calibration also assumed that most of the

drainage water would pass to the field (tile) drains, but

a small proportion (estimated by a balance procedure

described by McGechan et al., 1997; McGechan,

2002) would pass to deep groundwater by percolation.

-1000Soil water tension, kPa

ϕ b

θs

θ b

Ks

Kb

Vol

umet

ric

soil

wat

er c

onte

nt

Log

hyd

raul

ic c

ondu

ctiv

ity

Soil matrix pores(micropores)

Macropores

Fig. 1. Two-domain representation of hydraulic properties assumed

in MACRO (Ks is the total saturated hydraulic conductivity, Kb is

the saturated conductivity of the soil matrix pores, hs is the total

saturated water content, hb is the saturated water content of the soil

matrix pores, and ub is the soil water tension at which the soil

matrix pores are saturated but the macropores contain no water);___, soil water tension; _ _ _ , hydraulic conductivity.

Upper boundary conditions with evapotranspiration

from a grass according to the Penman/Monteith

equation are represented using an option in the model

for a perennial crop.

3.3. Colloid facilitated contaminant transport

The special version of MACRO used in this study

has additional features to represent a contaminant

which is transported by attachment to mobile colloids

(Jarvis et al., 1999). Colloids can be soil-derived by

rainfall impact at the soil surface (not used here) or

added in an dirrigationT application of water to the soil.The so-called irrigation procedure is used here to

represent both winter slurry applications (McGechan

et al., 2002), and deposition of faeces by grazing

animals during summer months (McGechan, 2003).

Simulations are carried out describing transport both

of colloids and the contaminant (P in this case). The

concentrations of colloids and P in the applied

dirrigationT water must be specified.

3.4. Sorption

P is transported through the soil both in soluble

form and attached to mobile colloids by sorption. The

inorganic components of P in slurry (as in soil) are

particularly strongly sorbed onto mobile colloids as

discussed by McGechan and Lewis (2002a). In

MACRO, representation of sorption of the contami-

nant onto solid surfaces follows the Freundlich

equation:

Q ¼ KdCn ð1Þ

where Q is the quantity of P sorbed in g [P] kg�1

[soil], C is the concentration of P in solution in mg

l�1, Kd (or Kc for colloids) is the sorption coefficient

and n is an exponent indicating the non-linear form of

the equation.

In the special version of MACRO sorption can take

place onto both mobile colloids and static soil

components, with a separate set of coefficients in

Eq. (1) for each (and the exponent n constrained to a

value of unity for mobile colloids only). The same set

of coefficient values (specified according to layers for

the soil, Table 1) have been used here as selected in

previous studies for slurry (McGechan et al., 2002)

and faeces deposited by grazing animals (McGechan,

Table 2

Physical and hydraulic parameters of silty clay loam grassland soil

Layers Topsoil Subsoil

Surface Lower

Layer depth (m) 0–0.1 0.1–0.3 0.3–1.0

Porosity, hs (%) 52.6 49.9 46.2

Dry bulk density (kg m�3) 1.26 1.33 1.43

Pore size distribution index

(Brooks and Corey, 1964)

0.026 0.023 0.035

Residual water content

(% volumetric basis)

4.0 4.0 4.0

Macroporosity, hs–hb (%) 2.35 3.01 1.08

dBreak pointT tension, ub (kPa) 1.21 1.20 1.20

Saturated hydraulic conductivity,

Ks (mm h�1)

375 208 166.7

Hydraulic conductivity at break

point, Kb (mm h�1)

1.746 0.84 0.175

Drain spacing (m) 7.0

Drain depth (m) 0.65

Table 1

Colloid and phosphorus (P) transport parameters

Parameter Parameter

value

Reference filter coefficient for macropores, (m�1) 0.5

Reference flow velocity in macropores (m h�1) 50

Filtering exponent in macropores (dimensionless) 1.8

Filter coefficient in soil matrix pores (m�1) 40

Freundlich sorption coefficient (linear) for

P sorption on colloids, Kc (m3 g�1)

2.7

Initial soil sorbed P (Olsen test P) (mg P kg�1 soil) 38

Freundlich sorption exponent for P sorption onto

soil, n

1.65

Freundlich sorption coefficient for P sorption onto

soil, Kd (ln mg1�n kg�1)

soil layer depth 0–0.1 m 203

0.1–0.2 m 307

0.2–0.3 m 350

0.3–0.4 m 4500

0.4–0.5 m 5500

0.5–1.5 m 6500


2003). Initial values of soil inorganic P set at the start

of simulations, based on information from the ‘Olsen

P’ test (a commonly used agronomic test for soil P

fertility), as measured by Hooda et al. (1999, 2000),

were retained from the earlier slurry study.

3.5. Colloid trapping

Trapping and filtration equations, and their

parameters, in the special version of MACRO used

here are described in detail by Jarvis et al. (1999)

and McGechan et al. (2002). Filtration in soil matrix

pores is described by a simple filtration coefficient,

while the macropore filtration coefficient is specified

at a reference flow velocity, then adjusted to the

actual flow velocity according to a power relation-

ship. Coefficient values (Table 1), selected to

simulate trapping most of the colloids if they enter

soil matrix pores but allow almost completely free

movement in macropores, have been retained from

the previous studies (McGechan et al., 2002;

McGechan, 2003).

3.6. Field sites for field-scale model calibration

The MACRO model has previously been cali-

brated and tested at the field scale for two Scottish

soil types, including a silty clay loam (USDA

classification) under grass with hydraulic parameters

listed in Table 2. Simulated results were tested against

measurements of water flows and P concentrations in

water at fields with drainflow measurement equip-

ment, at Bush near Edinburgh, and at the Crichton

Royal Farm, Dumfries in the predominantly dairy

farming area of SW Scotland. Details and layouts of

hydrologically isolated plots with equipment for

measurement of drainflow quantities and contaminant

concentrations, are described by Vinten et al. (1991,

1992, 1994), McGechan et al. (1997) and Hooda et al.

(1998, 1999). Near saturation hydraulic conductivities

were measured using a tension infiltrometer method

(Jarvis and Messing, 1995), and water release

characteristics had also been measured (McGechan

et al., 1997).

3.7. Phosphorus and colloids in slurry

Slurry was assumed to be applied at a rate of 50

m3 ha�1, the maximum permissible under codes of

practice for environmental protection (SEERAD,

2002). This gave an application of 5 mm of water

in the ‘irrigation’ routine in MACRO (at a specific

spatial location), which was assumed to take place

over a 0.0003 h (1.0 s) period commencing at 10:00

h on the application day. Two alternative sets of

assumptions were made regarding slurry composi-

tion. In the first set, concentrations of 3700 g m�3


of faecal colloids and 178 g m�3 of inorganic

phosphate were assumed, corresponding to the

slurry application on 21 November 1994 at the

Dumfries site reported by Hooda et al. (1999) and

McGechan et al. (2002). This represents a fairly

dilute slurry (dry matter content around 3.5%)

arising where dirty water, dairy washings and other

farm wastes have been added. The second set

represented a more concentrated slurry (7% dry

matter content) with concentrations double those for

the dilute slurry. A concentration of inorganic P of

356 g m�3 corresponds to a typical value for the

davailableT P2O5 component of 7% dry matter

content dairy cow slurry reported by Dyson

(1992). The quantity of slurry of 7% dry matter

content produced by 190 dairy cows (see Section

4.2) over a 26-week housing period is given by Dyson

(1992) as 2610 m3, which would require 53 ha of land

area for spreading at 50 m3 ha�1. On a 50-ha farm, the

more concentrated slurry would thus require a second

application each year on a small part of the land area,

whereas the more dilute slurry would require exten-

sive repeated applications.

3.8. Phosphorus and colloids in faeces from grazing

livestock

The daily average quantity of inorganic P

excreted in faeces by a dairy cow was estimated as

26.2 g. This was based on figures presented for a

model dairy farm by Haygarth et al. (1998), which in

turn was based on a similar study for N cycling on a

dairy farm by Jarvis (1993). The model dairy farm

had an autumn calving herd so P excretion was

much higher in winter than summer, whereas for a

spring calving herd the reverse would be the case, so

for the current study the average daily figure was

assumed. A similar daily value of 23.6 g P was

estimated from data presented by Dyson (1992) for

slurry produced by winter housed dairy cows.

Assumptions about dung patch dimensions were

based on values presented by Lantinga et al.

(1987), although broadly similar figures (or ranges

of values in some cases) were also presented by

MacLusky (1960) and Richards and Wolton (1976).

Adult dairy cows were assumed to produce 12

defaecations of 1.0 l, each day in the field. A dung

patch covers 0.05 m2, giving an initial height of 20

mm for the patch. However, it was further assumed

that about 15 mm of liquid (containing much of the

colloidal material and colloidally attached P) would

rapidly infiltrate the soil surface, leaving a patch of

height 5 mm containing the larger particles of dry

matter. Assumed concentrations of colloids and P in

the water infiltrating the soil surface were adapted

from those previously estimated for slurry (McGe-

chan, 2002, 2003). This gave concentrations of 65.5 g

l�1 of colloids and 3.15 g l �1of inorganic P in the

infiltrating water.

Petersen et al. (1956), Richards and Wolton

(1976) and Hack-ten Broeke et al. (1996) have

discussed the probability of both dung or urine

patches being deposited over areas contaminated by

recent previous defaecations or urinations. Petersen

et al. (1956) fitted a binomial distribution to

experimental data describing spatial distribution of

excreta from grazing animals. For a stocking density

of 3.8 cows ha�1, 12 daily defaecations cover

0.024% of the land area, so double contamination

over a time interval of a few days is a relatively rare

occurrence and has been ignored in this study. In the

simulations, results estimated from the land area

under dung patches were combined with those from

areas receiving no dung on that day, to give a

weighted average daily P loss for a daily inorganic P

input of 314.4 g ha�1 (from 26.2 g cow�1 day�1

and a stocking density of 3.8 cows ha�1). In this

combination calculation, an adjustment was made to

base the P losses on the hydrology and drainflows

from areas receiving no dung (McGechan, 2003).

This procedure set losses to zero on days where

drainflow appeared to take place in simulations

representing the dung patch area (0.023% of the

field area) with a 15-mm water application, but no

drainflows would have taken place over the whole

field which is dominated by the hydrology with no

additional water applied.

3.9. Calibration of field-scale model

Calibration and testing of the field-scale model has

been described in detail by McGechan et al. (2002),

McGechan (2002, 2003). An example of simulated

against measured P losses is shown in Fig. 2, for a

grassland field which was grazed by dairy cows

(stocking rate 4 cows/ha) from 27 September to 12

Fig. 2. Comparison between simulated and measured cumulative phosphorus (P) flows to field drains at grassland site with autumn grazing

followed by a slurry application. Cumulation from the start of simulation period (1 October) and from slurry application date. ., measured total

inorganic P; _____ , simulated total inorganic P; _ . _ . _ . _ , simulated colloid-sorbed inorganic P; - - - - -, simulated dissolved inorganic P.


November 1994, and received a slurry application on

21 November 1994. This assumed parameter values as

listed in Tables 1 and 2, as well as P and colloid

quantities as described in Sections 3.7 and 3.8. Both

measured and simulated values illustrate the high

levels of P loss that can occur due to livestock faeces

deposition under wet soil conditions, and that the high

loss level lasts for only a short time following a slurry

application. Although measurements were only for

total inorganic P, simulated results indicate that losses

were predominantly of colloid sorbed P during

grazing and immediately after slurry spreading,

whereas the background loss occurring more than a

week after slurry spreading was almost entirely of

soluble P.

3.10. Field-scale predictive simulation procedure

Weather-driven simulation runs were carried out

using the special version of MACRO in predictive

mode to test the effect of spreading slurry each day

in October–March 1993–1994 and 1994–1995.

Similar simulations were carried out to test the

effect of a dung patch deposited each day in April–

September 1994 and 1995. Weather data included

daily rainfall recorded at the Caddell site where

catchment-scale P losses had been measured (Sec-

tion 4.2). However, in order to estimate evaporation

the MACRO model requires a number of other

weather parameters which had not been recorded at

the site, so use was made of values recorded at nearby

sites in Ayrshire (Prestwick Airport for of sunshine

hours and SAC Auchincruive for four other parame-

ters). Each simulation was run for a 3-month period

from the first of the month prior to the deposition date.

Simulated losses of soluble and colloid-sorbed inor-

ganic P were recorded for a period of 21 days

following deposition. A procedure was set up based

on DOS batch files to generate the MACRO

parameter files in the correct format and to run the

repeated paired simulations for all the deposition date

options. A separate single dhydrologicalT simulation

was carried out over the period October 1993–

September 1995 with water inputs as rainfall only,

to determine drainflows in the absence of additional

water in slurry or dung.


4. Catchment-scale modelling

4.1. General principles

A catchment-scale hydrological model operates in

a similar manner to a field-scale hydrological model

in many respects. However, most field-scale models

assume for simplicity that conditions are spatially

homogeneous over the area considered, but this is

not the case at the catchment scale. Water, which

originates as precipitation, moves along flowpaths

from high levels in the catchment to the receiving

water body. All along such a flowpath, additional

water containing soluble or colloidal contaminants is

added from branches in the system, so the concen-

tration of contaminant in the main flow can increase

or decrease. Water can be transported horizontally as

surface runoff, through field drains or ditches, and

through soil pores. One approach to catchment-scale

modelling is to set up a linkage between the

CADDELL

RESERVO

UPPE60% GG;

0 250 500 750m

NORTH

240+ +200

+120

+170

+240

106+

+120

Fig. 3. Cellular diagram o

drainage outputs of field-scale simulations of indi-

vidual fields or cells in the catchment. Here we

assume that the dominant mechanism for transport of

contaminants to a river system is through the field

drainage network. To be linked in this way, field-

scale models generally require some simplification,

particularly as it is not normally feasible to measure

detailed soil hydrological parameters at all positions

in the catchment. For the current study, a simple

linkage was made of results from field-scale

simulations with MACRO, and assuming that hydro-

logical parameters are the same for all soils in the

dairy farm area, as for one of the two soils (the silty

clay loam, Table 2) for which detailed parameters

were known.

4.2. Field site for model testing at the catchment scale

Out of six sites studied by Hooda et al. (1997),

the Caddell catchment (Fig. 3) was one of three

IR

R CATCHMENT 26% MOORLAND

+160

100+

+90

+75

95% GG

IN

80+

70+OUT

f Caddell catchment.


located in Ayrshire, one of the most intensive dairy

farming regions of the UK. Swards were rotatonally

grazed by dairy cows and cut for silage. Slurry was

spread periodically during winter when cattle are

housed. The total catchment area is 750 ha. Soils in

the catchment were described as non-calcareous,

calcareous or humic gleys. Some information was

recorded about nutrient status, but no detailed

hydrological soil parameters were measured. Stream

water flows and P concentrations were measured at

the upstream and downstream ends of a 1.5-km

length of stream, which received field drain outflows

from a 50-ha intensive dairy farming area (Fig. 3)

with a stocking density of 3.8 cows ha�1 (total 190

cows). Sampling took place at 3-week intervals

during the first year (October 1993–September

1994) and at weekly intervals during the second

year (October 1994–September 1995). Daily rainfall

was also recorded over the period 9 March 1993–30

September 1995.

S2

S5

S3 S3

S4

S5

S4

S5

S4

S5

S3

S4

S5

S1

S2

S1 F1

DF1

RF1

Fig. 4. Hydraulic routine of drainage between system of linear storage chan

unlabelled intermediate tiers follow in order.

4.3. Fluvial geomorphology and timing of water

movements through a catchment

Water leaving an individual field in a catchment

does not immediately enter the receiving stream or

river. The delay which occurs depends on a number of

factors, including the direct field to river distance, the

routing of the flow path, and the current flow rate

(which is related to soil wetness and hence rainfall

history). Artificial drainage networks can be consid-

ered using the classical concepts of fluvial geo-

morphology in natural river basins. The quantitative

analysis of channel networks and classification of

stream orders was began by Horton (1945), and

extended by Mock (1971).

In the current study, the farm drainage network is

assumed to be a maximum of five orders, with an

average network length equal to 200 m. This allows us

to classify fields into five tiers according to their distance

and hydrological routing from the outlet (Fig. 4).

DF1

RF1

DF2

RF2

DF3

RF3

DF4

RF4

DF5

RF5

RIVER

TIER 1

TIER 2

TIER 3

TIER 4

TIER 5

F1

F3

F2

F4

F5

nels from fields in five tier categories. Tiers 1 and 5 only are labelled,


Furthermore in order to simplify the network we assume

that the width function which defines the average

network link distance from the outlet is linear. Hydraulic

routing of the drainage simulation of MACRO, between

a linear system of storage channels, is taken according to

a modification of the Muskingum Method for reservoir

routing (Bras, 1990). Under this methodology drainflow

DFi and runoff RFi from fields Fi passes into storage

channels Si (of order i) and is accordingly routed

towards the river. A minimum storage Smin of 2 m3 per

channel is assumed below which there is zero output

flow. The use of a minimum storage capacity in channel

reservoirs particularly effects summer fluxes of con-

taminants to the river.

4.4. Calculating stream P concentrations in

catchment

A procedure was set up in an Excel spreadsheet to

add up from the results of the individual simulations

the contribution to P losses on each day October

1993a September 1995 from deposition of slurry or

dung. For the summer period, the area covered each

day by dung patches was calculated for 190 cows. For

the winter period, a daily area for slurry spreading was

allocated (which could be zero on some days) such

that the total area spread over each winter was 53 or

106 ha (depending on the slurry dry matter content).

Since losses associated with slurry or dung deposition

on a particular day were recorded for the contami-

Tier 5

Tier 10

0.5

1

1.5

2

2.5

3

3.5

4

0 2 4 6 8

Drain flow

Del

ay (

days

)

Fig. 5. Duration of transit time for contaminant passing through catchment

field drainage rate for five tiers of fields with different field to stream dis

nated area for 21 days following, the total loss on a

particular day was potentially for 21 such areas each

contaminated by deposition on one of the previous 21

days. For both summer and winter periods, the

concentration of P in contaminated water on a

particular day from the relevant total contaminated

area was combined with the quantity of uncontami-

nated water from the remaining field area (given by

drainflows from the hydrological simulation) to give

the overall P concentration in drainflows from the

whole farm area.

For both slurry and dung patch drainage, delays are

introduced between release from a field and P

reaching the river, based on the concepts described

in Section 4.3 (Fig. 5). As there was no a priori

information regarding the spatial input distribution of

P in the form area, a uniform overall distribution has

been assumed. Finally, a calculation was carried out

using the upstream river flows and P concentrations

combined with those from the farm area to give an

estimate of the downstream river P concentration.

5. Results and discussion

5.1. Losses from individual simulations

The pattern of loss occurrence following deposition

of P in slurry or dung has been discussed in papers on

previous studies (McGechan et al., 2002; McGechan,

10 12 14 16 18 20

(mm day-1)

drainage system from field to receiving stream. Shown in relation to

tances.


2002, 2003). In general, if deposition takes place when

soil is wet there is rapid transport with high P losses

over the first few days following deposition. After

about 3–4 days the rate of loss drops to a low level.

Losses are particularly high for the first few days if

rainfall occurs during that period. In contrast, if

deposition takes place when soil is relatively dry,

losses are generally at a low level. There is some rise in

loss levels if rain occurs soon after spreading, but not

to the same extent as if the soil had been wet at the time

of deposition. In general, high losses tend to be

predominantly of colloid sorbed P, whereas low losses

are predominantly of soluble P. This demonstrates the

importance of macropore flow of colloid-associated P

which occurs under wet soil conditions when macro-

pores are water-filled. In contrast, under dry conditions

only the smaller soil matrix pores (micropores) are

water-filled, so P carrying colloids become trapped

and P can only move in soluble form. Due to the high

sorbability and low solubility of P, soluble P losses

tend to be at a low level.

5.2. Stream outflow P concentration

Stream outflow P concentrations, calculated from

losses given by all the weather-driven simulations, are

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

93.8 94.0 94.2 94.4 94.6 9

Julia

Inor

gani

c P

conc

entr

atio

n (m

gP l -

1 )

Fig. 6. Simulated inorganic phosphorus (P) concentrations in Caddell str

during which slurry was spread. _____ , simulated concentration at outflow

inflow measurement point (sampled at weekly or three weekly intervals); xweekly or three weekly intervals).

shown alongside measured inflow and outflow P

concentrations in Fig. 6. Duplicate results were

determined based on simulations using soil hydraulic

parameters for each of the two soils for which they had

been measured in previous studies. In relation to slurry

spreading, results were also duplicated according to

whether the dilute or more concentrated slurry

composition had been assumed. Some manipulation

to improve fits (of predicted to measured outflow

concentrations) was carried out at this stage of areas

and days over and on which slurry was spread. This

was considered to be an acceptable procedure, since no

information had been recorded about when this slurry

spreading had taken place on the farm area, other than

the estimate of the total quantity which would have to

be disposed of.

Results based on simulations with hydraulic

parameter values for a clay loam soil show simulated

outflow P concentrations very much in excess of

measured values. In contrast, results based on the

hydraulic parameter values for the silty clay loam

soil with parameters listed in Table 2 show simulated

concentrations of a similar magnitude to measured

values (Fig. 6). Comparing results (shown in Fig. 6)

assuming spreading of the more concentrated (7%

dry matter) slurry with those for the more dilute

4.8 95.0 95.2 95.4 95.6 95.8

n date

eam outflow during two summer grazing seasons and two winters

measurement point (daily values); _____ , measured concentration at

, measured concentration at outflow measurement point (sampled at


(3.5% dry matter) slurry, losses were slightly higher

with the more dilute slurry. High simulated outflow

P concentrations occurred at similar times of year to

high measured values, with some high loss periods

during the summer grazing season and some during

the winter when slurry would have been spread.

However, there was a much larger day-to-day

fluctuation in concentrations than had been meas-

ured. Since spot measurements had only been made

on particular days at weekly at three weekly

intervals, such fluctuations would not be apparent

in the measured data. These simulated results were a

reasonable fit to the measured data with a chi-

squared per degree of freedom value of 0.11 (Lewis

et al., 1997), especially given the lack of hard

Fig. 7. Simulated inorganic phosphorus (P) concentrations in Caddell stream

(b) after introduction of network stores. __________ , simulated concentratio

concentration at inflow measurement point (sampled at weekly or three w

point (sampled at weekly or three weekly intervals).

information about a number of aspects which would

influence the results.

5.3. Effect of network storages

The major effect of storage channels with a

minimum storage below which there is zero channel

output is to delay the smaller field drainage flows and

to produce larger P fluxes in subsequent high flow

periods. This is exemplified in Fig. 7, which focuses

on the 1994 summer grazing season, and gives a

comparison of stream outlet fluxes for a system with

and without network stores. Several high P fluxes in

this period are due to accumulated P stored in the

channels, from small drainage discharges, prior to a

outflow during the low-flow grazing season in 1984. (a) before and

n at outflow measurement point (daily values); _____, measured

eekly intervals); x, measured concentration at outflow measurement


large hydrologic event, which then flushes P out of the

system.

6. Conclusions

In this study, a model of P losses in a small dairy

farm catchment has been set up based on a linkage of

weather-driven field-scale simulations using the

MACRO model. P deposition, both in faeces from

grazing livestock in summer and in slurry spread in

winter, has been represented. The main contaminant

transport mechanism considered has been P sorbed

onto mobile colloidal faeces particles, which move

through the soil by macropore flow. Phosphorus

moves readily through soil to field drains under wet

conditions when macropores are water-filled, but in

dry soil the P carrying colloids become trapped so

losses remain at a low level. The MACRO simulation

procedure had previously been calibrated and tested

for two soil types, and also both for P deposition in

slurry and in faeces from grazing animals.

Results based on linked simulations assuming

hydraulic parameter values for one particular soil

show reasonable fits to values of catchment outflow P

concentrations measured at infrequent intervals. The

catchment-scale measured data showed some periods

with high outflow P concentrations during the summer

grazing season and some during the winter when

slurry would have been spread. Results from the

linked simulations also showed some high outflow P

concentrations during both summer and winter, with

timing generally occurring at similar times of year to

high measured values. However, there was one period

during summer 1995 where high measured outflow P

concentrations were not reproduced in the simula-

tions. During this period the MACRO hydrological

simulations did not produce sufficient drainage flows.

It is of course possible to calibrate the field-scale

model to more accurately represent the catchment

hydrological outputs, but in this study we persisted

with the use of a single field calibration to characterise

the dairy farm area.

In general, the catchment-scale modelling work

was hampered by the infrequency of loss measure-

ments and a lack of information about a number of

parameters, which would be required to carry out a

more exact calibration and provide a rigorous test of

the procedure. In view of this limitation, the approach

taken could be regarded as scenario testing at the

catchment scale, based on a model which has been

rigourously calibrated and tested at the field scale. It

was nevertheless concluded that through soil flow of

colloid sorbed P by macropore flow represents a

highly plausible mechanism by which environmen-

tally harmful quantities of P are transported to river

systems in livestock farming catchments. This

represents an alternative explanation to supplement

or replace surface runoff transport, a mechanism to

which high P losses from livestock farming areas

have often been attributed. The occurrence of high

levels of loss under wet conditions indicates

environmental benefits from avoiding slurry spread-

ing on wet soil or during rain. There may also be

benefits during wet field condition in what would

normally be considered to be the grazing period,

from either housing the animals or moving them to

fields remote from the river or stream which drains

the catchment.

Acknowledgements

The author wishes to thank Professor Nick Jarvis

of the Swedish University of Agricultural Sciences for

assistance with adapting and using the MACRO

model. The Scottish Executive Environment and

Rural Affairs Department provided funds to carry

out the work.

References

Arnold JG, Williams JR, Srinivasan R, King KW, Griggs RH.

SWAT (Soil and Water Assessment Tool) User’s Manual.

Temple, USA7 USDA, Agricultural Research Service, Grass-

land, Soil and Water Research Laboratory; 1994.

Barrow NJ. The description of phosphate adsorption curves. J Soil

Sci 1978;29:447–62.

Barrow NJ. A mechanistic model for describing the sorption and

desorption of phosphate by soil. J Soil Sci 1983;34:733–50.

Bras RL. Hydrology: An Introduction to Hydrologic Science. USA7

Addison-Wesley Publishing; 1990.

Brooks RH, Corey AT. Hydraulic Properties of Porous Media.

Hydrology Papers, vol. 3. Fort Collins7 Colorado State

University; 1964. 27 pp.

Dayawansa NDK. The development of nutrient pollution models for

use at the catchment scale. PhD thesis, University of Newcastle;

2002.


Dyson PD. Fertiliser allowances for manures and slurries. SAC

Technical Note 309, Fertiliser Series No. 14, SAC Edinburgh;

1992.

Ewen J, Parkin G, O’Connell PE. SHETRAN: distributed river

basin flow and transport modeling system. J Hydrol Eng

2000;5:250–8.

Grant R, Laubel A, Kronvang B, Andersen HE, Svendsen LM,

Fugslang A. Loss of dissolved and particulate phosphorus from

arable catchments by subsurface drainage. Water Res 1996;30:

2633–42.

Groenendijk P, Kroes JG. Modelling the nitrogen and phosphorus

leaching to groundwater and surface water. ANIMO 3.5.

Report 144, DLO Winand Staring Centre, Wageningen; 1997.

143 pp.

Hack-ten Broeke MJD, De Groot WJM, Dijkstra JP. Impact of

excreted nitrogen by grazing cattle on nitrate leaching. Soil Use

Manage 1996;12:190–8.

Hawkins JMB, Scholefield D. Molybdate-reactive phosphorus

losses in surface and drainage waters from permanent grassland.

J Environ Qual 1996;25:727–32.

Haygarth PM, Hepworth L, Jarvis SC. Forms of phosphorus transfer

in hydrological pathways under grazed grassland. Eur J Soil Sci

1998;49:65–72.

Hooda PS, Moynagh M, Svoboda IF, Thurlow M, Stewart M,

Thomson M, et al. Soil and land use effects on phosphorus in six

streams draining small agricultural catchments in Scotland. Soil

Use Manage 1997;13:196–204.

Hooda PS, Moynagh M, Svoboda IF, Anderson HAA. Comparative

study of nitrate leaching from intensively-managed grass and

grass-clover pastures. J Agric Sci Camb 1998;131:267–75.

Hooda PS, Moynagh M, Svoboda IF, Edwards AC, Anderson HA,

Sym G. Phosphorus loss in drainflow from intensively-managed

grassland soils. J Environ Qual 1999;28:1235–42.

Hooda PS, Rendell AR, Edwards AC, Withers PJA, Aitken MN,

Truesdale VW. Importance of soil phosphorus sorption and

desorption indices in assessing phosphorus loss in agricultural

runoff. J Environ Qual 2000;29:1166–71.

Horton RE. Erosional development of streams and their drainage

basins: hydrophysical approach to quantitative morphology.

Bull Geol Soc Am 1945;56:275–370.

Jarvis SC. Nitrogen cycling and losses from dairy farms. Soil Use

Manage 1993;9:99–105.

Jarvis N. The MACRO model-Technical description and sample

simulations. Reports and Dissertations 19, Swedish University

of Agricultural Sciences, Department of Soil Sciences, Uppsala;

1994. 51 pp.

Jarvis NJ, Messing I. Near-saturated hydraulic conductivity in soils

of contrasting texture as measured by tension infiltrometers. Soil

Sci Soc Am J 1995;59:27–34.

Jarvis N, Villholth KG, Ulen B. Modelling particle mobilization and

leaching in macroporous soil. Eur J Soil Sci 1999;50:621–32.

Jones CA, Cole CV, Sharpley AN, Williams JR. A simplified soil

and plant phosphorus model, I. Documentation. Soil Sci Soc

Am J 1984;48:800–5.

Kirsh K, Kirsh A, Arnold JG. Predicting sediment and phosphorus

loads in the Rock River Basin using SWAT. Trans ASAE 2002;

45:1757–69.

Knisel W., editor. CREAMS: A Field-Scale Model for Chemicals,

Runoff and Erosion from Agricultural Management System-

sConservation Research Report, vol. 26. Dordrecht7 US Depart-

ment of Agriculture; 1988.

Knisel W. editor. GLEAMS: groundwater effects of agricultural

management systems, Version 2.1. UGA-CPES-BAED Pub.

No. 5. Tifton; University of Georgia, Tifton; 1993.

Kroes JG, Rijtema PE. ANIMO 3.5, Users Guide for the ANIMO

Version 3.5 Nutrient Leaching Model. DLO Technical Docu-

ment, vo. 46. Wageningen: Winand Staring Centre; 1998. p. 98.

Lantinga EA, Keuning JA, Groenwold J, Deenan PJAG.

Distribution of excreted nitrogen by grazing cattle and its

effects on sward quality, herbage production and utilization.

In: van der Meer HG, Unwin RJ, van Dijk TA, Ennik GC,

editors. Animal Manure on Grassland and Fodder Crops

Fertilizer or Waste? Dordrecht7 Martinus Nijhoff Publishers;

1987. p. 103–7.

Larsson MH, Jarvis N. Evaluation of dual porosity model to predict

field-scale solute transport in a macroporous soil. J Hydrol

1999;215:153–71.

Leonard RA, Knisel WG, Still DA. GLEAMS: Groundwater

Loading Effects of Agricultural Management Systems. Trans

ASAE 1987;30:1403–18.

Lewis DR, McGechan MB. Review of field scale phosphorus

dynamics models. Biosyst Eng 2002;82:359–80.

Lewis DR, Williams RJ, Whitehead PG. Quality simulation along

rivers (QUASAR): an application to the Yorkshire Ouse. Sci

Total Environ 1997;194/195:399–418.

MacLusky DS. Some estimates of the areas of pasture fouled by the

excreta of dairy cows. J Br Grassl Soc 1960;15:181–8.

McGechan MB. Effects of timing of slurry spreading on leaching of

soluble and particulate inorganic phosphorus explored using the

MACRO model. Biosyst Eng 2002;83:237–52.

McGechan MB. Phosphorus pollution implications of extending the

grazing season into late autumn. Grass Forage Sci 2003;58:

151–9.

McGechan MB, Lewis DR. Sorption of phosphorus by soil: Part 1

Principles, equations and models (review paper). Biosyst Eng

2002;82:1–24.

McGechan MB, Lewis DR. Transport of particulate and colloid-

sorbed contaminants through soil: Part 1. General principles

(review paper). Biosyst Eng 2002;83:255–73.

McGechan MB, Graham R, Vinten AJA, Douglas JT, Hooda PH.

Parameter selection and testing the soil water model SOIL. J

Hydrol 1997;195:312–34.

McGechan MB, Jarvis NJ, Hooda PS, Vinten AJA. Parameterisation

of the MACRO model to represent leaching of colloidally

attached inorganic phosphorus following slurry spreading. Soil

Use Manage 2002;18:61–7.

Metherell AK, Cole CV, Parton WJ. Dynamics and Interactions of

Carbon, Nitrogen and Sulphur in Grazed Pastures. Proceedings of

the XVII International Grassland Congress. 1993a. p. 1420–1.

Metherell AK, Harding LA, Cole CV, Parton WJ. CENTURY Soil

organic matter model environment. Technical Documentation.

Agroecosystem version 4.0. Great Plains System Research Unit

Technical Report No. 4. Colorado, USA: USDA-ARS, Fort

Collins; 1993b.


Mock SJ. A classification of channel links in stream networks.

Water Resour Res 1971;7:1558–66.

Mualem Y. A new model for predicting the hydraulic conductivity of

unsaturated porous media. Water Resour Res 1976;12:513–22.

Parton WJ, Schimel DS, Cole CV, Ojima DS. Analysis of factors

controlling soil organic matter in Great Plains grasslands. Soil

Sci Soc Am J 1987;51:1173–9.

Petersen RG, Lucas HL, Woodhouse WW. The distribution of

excreta by freely grazing cattle and its effect on pasture fertility:

I. Excretal distribution. Agron J 1956;48:440–4.

Richards LA. Capillary conduction of liquids in porous mediums.

Physics 1931;1:318–33.

Richards IR, Wolton KM. The spatial distribution of excreta under

intensive cattle grazing. J Br Grassl Soc 1976;31:89–92.

SEERAD. Prevention of environmental pollution from agricultural

activity. Code of good practice. Scottish Executive Environ-

ment, Agriculture and Rural Affairs Department; 2002.

Sharpley AN, Williams JR. editors. EPIC, Erosion/Productivity

Impact calculator: 1. Model Documentation. USDA Technical

Bulletin No. 1768; 1990. 235 pp.

Simes K, Yli-Halla M, Tuhkanen HR. Simulation of the

phosphorus cycle in soil by ICECREAM. Proc OECD Work-

shop dPractical and Innovative Measures for the Control of

AgriculturalPphosphorus Losses to WaterT. Northern Ireland7

Antrim; 1998. p. 38–9.

Stamm C, Fluhler H, Gachter R, Leuenberger J, Wunderli H.

Preferential transport of phosphorus in drained grassland soils. J

Environ Qual 1988;27:515–22.

Ulen B. Episodic precipitation and discharge events and their

influence on losses of phosphorus and nitrogen from tile drained

arable fields. Swed J Agric Res 1995;25:25–31.

Vinten AJA, Howard RS, Redman MH. Measurement of nitrate

leaching losses from arable plots under different nitrogen input

regimes. Soil Use Manage 1991;7:3–14.

Vinten AJA, Vivian BJ, Howard RS. The effect of nitrogen fertiliser

on the nitrogen cycle of two upland arable soils of contrasting

textures. Proc Int Conf., Cambridge. Peterborough7 The

Fertiliser Society; 1992.

Vinten AJA, Vivian BJ, Wright F, Howard RS. A comparative study

of nitrate leaching from soils of differing textures under similar

climatic and cropping conditions. J Hydrol 1994;159:197–213.

Young RA, Onstad CA, Bosch DD. AGNPS: an agricultural

non point source model. In: Singh VP, editor. Computer

Models of Watershed Hydrology. Water Resource Publica-

tions. 1995. p. 1001–20.

modelling through-soil transport of phosphorus to surface waters from livestock agriculture at the...

Documents