Estimating future scenarios for farm-watershed nutrient
fluxes using dynamic simulation modelling — Can on-farm
BMPs really do the job at the watershed scale?
MR. Rivers*,'”, D.M. Weaver'”, K.R.J. Smettem', P. M. Davies’.
* Corresponding author
1. Centre for Ecohydrology, School of Environmental Systems Engineering, The
University of Western Australia
2. Department of Agriculture and Food, Western Australia
3. Centre for Natural Resource Management, Albany Campus, The University of
Western Australia
The University of Western Australia, Nedlands, W.A. 6009. Australia
mrivers @agric.wa.gov.au
Abstract
A dynamic model of Phosphorus (P) movement through the Peel-Harvey Watershed in
South Western Australia was developed using STELLA® dynamic modelling software.
The model was developed to illustrate watershed P flux and to predict future P loss
rates under a range of management scenarios. Input parameters were sourced from
surveys of local agricultural practices and regional soil testing data. Model P-routing
routines were developed from the known interactions between the various watershed P
compartments and fluxes between various P stores. P-retention characteristics of a
variety of management practices were determined from field trials where available and
published values where not. The model simulated a 200 year time frame to reflect 100
years to the present day since initial land development, and forecast 100 years into the
future. Although the watershed has an annual P loss target of 70 tonnes per annum
(tpa), the measured present day loss is double this amount (140 tpa) and this is
projected to rise to 1300 tpa if current land management practices continue. Even if
broad-scale BMP implementation occurs, P losses are likely to increase to
approximately 200 tpa. This has significant implications for both future land use and
subsequent water quality in the watershed.
Introduction
The Peel-Harvey Watershed and agricultural nutrient management
Nutrient losses from land to water have accelerated globally over the last 50 years due
to landscape development for agricultural and urban pursuits (Reynolds and Davies
2001). These pursuits bring with them inputs of nutrients, predominantly phosphorus
(P) and nitrogen (N), which are used to boost plant production. The subsequent increase
in soil nutrient content beyond levels which can be utilized productively may lead to
increased losses of nutrients to local and regional waterways and subsequent
eutrophication (Nair et al. 2004, Behrendt and Boekhold 1994, Sharpley and Smith
1990). In Australia this problem is exacerbated by the fact that many waterways were
naturally oligotrophic meaning that excess nutrient inputs impact on the natural aquatic
ecosystems more readily and with greater consequence.
The study area for this research is the Peel-Harvey watershed in south Western
Australia which has a long history of nuisance and toxic algal blooms. The watershed
lies approximately 70km south of Perth, the State Capital, and covers an area of
approximately 3072km2. Approximately 190,000 ha of this is the coastal portion of the
watershed which has been cleared for land development and which contributes the
majority of nutrients into regional waterways. The region experiences a Mediterranean
climate, characterised by warm dry summers and cool wet winters (Seddon 1972).
About 90% of the region’s total annual rainfall occurs between May and September and
varies between 700 and 1100mm. Average daily temperatures range from 17°C to 30°C
in summer and from 6°C to 17°C in winter.
Land use in the 190,000ha of the coastal region of the watershed is dominated by
agriculture with grazing for beef production the most common agricultural activity
(Lavell et al. 2004).
Nutrient-transport modelling
Developing an understanding of the complexity of water or nutrient transport through
farms and watersheds can be a very difficult and resource-intensive task. To develop a
through understanding of nutrient transport on only one farm may involve the
establishment of multiple soil and water sampling points and their management over a
number of years. The fact that every farm and watershed is different then makes it
difficult to justify transfer of results from one monitored site to other sites — even within
the same locality. This makes the development of more generic principles which can
cross between locations, scales and time difficult but necessary because important land
management decisions need to be made at farm and watershed scales. One approach to
relate measured data from one point to others, or of applying well understood, if not
well measured, principles to a range of locations is through the use of models. These
allow us to apply measured relationships or systematic understanding developed in one
location or time-frame to others. It is extremely important to note, however, that “As
with any tool, the answers they give are dependent on how we apply them, and the
quality of these answers is no better than the quality of our understanding of the system”
(Butcher et al. 1998).
This study was designed to combine measured and surveyed data at the farm and
watershed scales with a watershed-scale dynamic model of nutrient fluxes to assess
future scenarios for the Peel-Harvey watershed.
Although many conceptual and process-based models describe dynamic natural
systems, they are often, in themselves, less dynamic in terms of describing the delays
and feedback loops common in the natural environment. System dynamics modelling
proposes a solution to this issue.
Figure | illustrates the relationship between the model methodology used in this paper
and the more conventional model types used more commonly. Although the model is
dynamic and utilises the specific benefits of this type of modelling approach, it
combines the benefits of both process-based and export-coefficient models as these are
the data-types generally encountered in this field and certainly the case in this study.
_————
Conceptual Model
LC
Numerical
mathematical Model
—————___
a + ™
‘a >)
[ Export coefficient ] Process-based [ Empirical ]
MM, a
System Dynamics model of
phosphorus movement through
watershed
Figure 1: Preferred model methodology
Previous work has been completed on the development of watershed and nutrient-
transport models as a means of both relating small-scale management changes to large-
scale impacts, and as a way of modelling water and nutrient policy scenarios (Cassell et
al. 2001, Young et al. 1989, Beasley and Huggins 1982, Heidtke and Auer 1993, Poiani
and Bedford 1995). However, many of these approaches are complicated, and their
widespread use may be restricted by computational complexities, particularly in
landscapes that are spatially and temporally heterogeneous.
This paper discusses a simulation model which estimates P transport through major
source, sink and flow sectors of the watershed. It attempts to track changes in stores and
flows of P over a 200 year time horizon to match watershed development and associated
nutrient inputs and outputs to the present day (100 years of development) and project a
further 100 years into the future. The model is a lumped, whole-of-watershed model
which allows for nutrient storage, assimilation and release from the major components
of the watershed environment. It is designed to explore long-term, large-scale changes
and is not designed to investigate detailed nutrient transport mechanisms within
individual watershed components.
Model Description
Overall model design
The overall, simplified conceptual design of the model is shown in Figure 2.
Essentially, the model has four major components: inputs, which direct the flow of P
from outside of the model boundary into the model as fertiliser, feed or precipitated P;
the soil store of P; the sediment store of P, and; subsequent P lost to the receiving water
body.
Phosphorus loss control through fertiliser, soil, plant and drainage
Best Management Practices
5
oa
oO
Runoff direct to
drains
(a : Loss to estuary
Phosphorus Soil P store Drain sediment P Estuary P store
inputs ino §=$£. > =—> ore —>
model
Leaching to
groundwater
Figure 2: Conceptual P transfer model
Annual P StoPd in Soil T
ert input T
S
Edge of paddfck loss KEHA
Pfloss rate
‘otal P into Gh
Area HA
Soil P los; PRI factor
/,
\L re
‘Stream|P loss
Stream stgre
Curiulatiye stream P store T
Estuary
Runoff direct tgrayams KGHA
Loss to’estuly
KGHA
Precipitation KGHA
Zaching to groundwater KGHA
Imporfing feed KGHA
GroyhdwatenP
n KGI
BO e
SoilP loss eaching ratio
‘Soil PRI factor
AreaHA Feed P input KGHA
Product P output T
Figure 3: Simplified STELLA® P transfer model
Annual edge of paddock losses T
7 'S
Annual/oss to estuary T
{) Graph 1
Groundwater P loss rate
ay
Annual stream storage
Between each of these P stocks are flows that govern the rate at which P is transferred
through the model. These are, in turn, controlled by a series of converters and
controllers which are variable through time and which can be highly complex. Finally,
overlain on the model are a series of P-mitigation strategies which have been employed
in the past in this region to manage P loss.
The structure of the actual model in detail as programmed into STELLA® but without
BMP detail is illustrated in Figure 3 (BMP details are discussed later and the more
complex form of the model including BMP details is attached as Appendix 1).
Detailed model description and initial parameterisation
Phosphorus inputs (flows)
Phosphorus input into the soil store occurs via three pathways: Precipitation, fertilising
and importation of feed.
Precipitated P inputs are set at 0.09 kg ha''yr! (adapted from Chen et al. 1985 and
McLaughlin er al. 1992) while P inputs from fertiliser and feed are varied through time
to reflect agricultural development in this region. Fertiliser P inputs are varied as shown
in Figure 4 and feed inputs as shown in Figure 5.
Graphical Function
15,00
Fertilsing
Years KCHA
0.000 4.500
20.00 6.450
Fertisn goo fits
g KCHA 80.00 13.73
100.00 14.40
120.00 14.70
140.00 14.63
160.00 14.63
180.00 14.70
000 200.00 14.75
_——
Ly) 2.000 200.00
Years Data Points: 11
dit Outp:
T i Del a Ca y
Figure 4: Model fertilser P inputs over time
Graphical Function
4.000 Importing
Yeats feed KGHA
0.000 0.000
20.00 0.440
Importin 40.00 1.500
g feed 69.00 2540
80.00 840
KCHA 100.00 3.050
120.00 3.050
140.00 3.050
160.00 3.050
180.00 3.050
2000 } : 200.00 3.050
[2.000 200.00
Years Data Points: 11
To Equat Delete Graph Cancel
Figure 5: Model feed P inputs over time
Soil P store (stock)
This stock describes the net accumulation of P in the soil component of the watershed. It
is the sum of the inputs as described above minus the sum of the losses from soil via
productive use of P and loss via overland runoff and vertical leaching.
P utilization (flow)
Graphical Function
3,000 Pi : : Utilisation
KGHA
0.140
Utilisatio
n KGHA
0,000
bD Years Data Points: 11
Figure 6: Model productive P utilisation over time
This parameter is described by a temporally-varying function in a similar manner to the
inputs of P from fertilser and feed (Figure 6). It is assumed that there was a very low
initial loss of P from the watershed which increased until the surveyed, present-day
levels of 2.00 kg ha"'yr’.
Runoff direct to streams (flow)
The algorithm used to calculate the loss of P via overland flow direct to the regional
drainage network (but not including BMP modification) is a function of a number of
inter-related parameters:
P 1OSS gunotr = P loss ¢,i; * soil PRI factor * (runoff ratio/100)
The net P lost as runoff is the “Soil P loss” modified by the function describing soil P-
retention capacity “Soil PRI factor” and the function describing the soil drainage
pathway (“Runoff ratio”: proportional runoff or leaching).
The “soil P loss” factor is itself a function of the total P in the soil store and the
consequent rate at which P is lost from the soil. The “Soil P loss rate” is a variable
function which is described in Figure 7.
As the net soil P store increases from 0 kg ha! to a maximum value of 1250 kg ha",
then the soil P loss rate increases from a very low value (high ability to retain P) to 1
(zero ability to retain P).
Graphical Function’
1.000
Soil'stores ol 8s
rate
0.000 0.050
125.00 0.060
250.00 0.070
mae 375.00 0.090
loss rate 500.00 0.115
625.00 0.160
750.00 0.215
875.00 0.305
1000.00 0.455
vow Ha) (GSE
| |
a 0.000 1250.00
Soil_store sui Polat | 11
To Equatio Delete Grap el ¢
Figure 7: Graphi to P in the soil
store
Expressed textually, the net result of the amalgamation of these functions is that the rate
at which P is lost from soil as runoff direct to the stream network is controlled by: the
amount of P in the soil at any given time; the consequent ability of that soil to hold
more P applied to it at that time; the inherent P-retention capacity of the soil, and; the
drainage characteristics of the soil.
Leaching to groundwater (flow)
The algorithm used to calculate the loss of P via vertical leaching (which still
subsequently reaches the regional drainage network but via sub-surface flow) is a
function of the same parameters described above for P lost as runoff direct to streams,
but modified by the leaching ratio which is 1-runoff ratio.
Although the algorithms for runoff P loss and leach P loss are identical, they are
calculated separately as different P-loss modifying BMPs act on the two different P-loss
pathways.
Stream sediment store (stock)
The streams represent the off-farm drainage system, and the destination of P loss from
edge-of-field at the farm-scale but prior to receipt by a receiving waterbody. This stock
describes the net accumulation of P in the stream sediment component of the watershed.
It is the sum of the inputs from runoff and leaching minus the losses from sediment to
the final receiving waterbody.
Loss to Estuary (flow)
This is calculated in a similar manner to that undertaken for “Runoff direct to streams”
and “Leaching to groundwater”. The net P lost to the estuary is the sum of “Stream P
loss” and “Groundwater P loss” modified by the functions describing the rate at which P
is lost from sediment: “Stream P loss rate” and; “Groundwater P loss rate”. Both of
these are variable functions described in Figure 8.
Graphical Function
1,000 Stream
sediment
store
0.000
37.50
Pp 75.00
isu 112.50
loss rate 150.00
187.50
225.00
262.50
300.00
337.50
375.00
Stream Ploss
rate
0,000
—
Ly 9.000 375.00 '—_____l
Stream_sediment_store D ql
Figure 8: Graphicarroncooroseu to uescrive Tstananreous scumentr_toss as related to P in the
sediment store
Estuary (stock)
Although defined as “Estuary” in this model, this stock represents the ultimate point of
the watershed to which all of the “streams” exit. Phosphorus in water flowing from the
streams to the receiving waterbody will undergo one or more of assimilation, sorption,
desorption or productive use whilst in the streams. This is identified as “Stream loss” in
the model and is regulated by “Stream loss rate” (a graphical function, itself varying as
in-stream P varies) and by the incoming P losses. This function then leaves a final
amount of P to flow into the final receiving waterbody.
Model initiation
Table 1 shows the major model components shown in Figure 3 and their initial values
unmodified by BMPs.
jor model component
Key Model component Model function Initial value Subsequent values
a Fertilising Flow 1.00 Kgha"yr' Graphical function
b Precipitation Flow 0.09 Kghayr' Graphical function
c Importing feed Flow 0.00 Kgha"yr' Graphical function
d Utilisation Flow 0.00 Kghayr' Graphical function
e Soil store Stock 300 Kg ha 3 ((atb+e)-(d+k+m)),
f Total P into soil store Converter NA atb+c
g Soil P loss rate Converter 150 Graphical function
h Soil P loss Converter NA fg
i Soil PRI factor Converter 05 Slider (input variable)
j Runoff ratio Converter 30 Slider (input variable)
k Runoff direct to streams Flow NA h¥i*(j/100)
1 Leaching ratio Converter 70 100-j
m Leaching to groundwater Flow NA h*i*(1/100)
Key Model component Model function Initial value Subsequent values
a Stream sediment store Stock 0 3 (k+m)-s),
° Stream P loss rate Converter 180 Graphical function
P Stream P loss Converter NA k*o
q Groundwater P loss rate Converter 0 Graphical function
t Groundwater P loss Converter NA m*q
s Loss to estuary Flow NA o+r
t Estuary Stock 0 xs,
Model Validation
As has been stated previously, the time-frame over which this model has been
developed to run is a 200 year period commencing 100 years ago at approximately the
time of European settlement of this region, and then 100 years into the future from the
present day. Such a lengthy time frame was selected because: watershed-scale response
to more localized management of soil and water resources is known to take long periods
to become apparent (Meals et al., 2010), and; the 100 year point in the model (present
day) provides an accurate model validation point as water quality monitoring data is
available to verify the load entering the “Estuary” component of the model. This is a
validation to observed data approach or historical behaviour test approach (Ford 2010).
Model calibration and validation was completed by initially loading the model P input
parameters and native soil and sediment P content and assimilation attributes with data
obtained from sampling of native (unfertilized) soils and with known natural watershed
nutrient input and output rates. The model was then run for an initial 100 year period
which aligns with the period of agricultural development for this watershed.
Results
Base run
In order to initially verify the overall efficacy of the model it was loaded with test data
of which the key data sources were the data obtained during the Peel-Harvey Coastal
Catchments Initiative (CCI) Projects (Lavell et al. 2004, Weaver et al. 2004, Neville et
al. 2004, Keipert et al. 2008) and regional soil P and P retention test results (Weaver
and Wong 2011).
The CCI survey data was used to determine the present-day inputs of P into, and outputs
from the watershed (Figure 9). This indicated that there is an annual “P surplus” of 2070
tonnes, or 80%, of the applied P, every year which is not being used productively. Soil
P and P-retention test results indicate that approximately 1200 tonnes is stored by the
watershed soils every year (but this capacity is declining) resulting in 870 tonnes a year
being lost to streams and groundwater. Stream P storage accounts for a large proportion
of these losses (again, declining) resulting in a net loss of 140 tonnes of P to the estuary
annually.
The consequent, present-day nutrient loss rate to the estuary was modelled to be 138
tPpa (Figure 10) which compares well with the value of 140 tPpa obtained from long-
term water quality monitoring (EPA 2008). This represents a validation point for
estuary P export as well as for watershed-compartment P content and loss rates.
catchmen
ww) oo
Cattle for Dairy
Cattle for Beef
Mixed Grazing streams
Horticulture:
Urban/Peri-urban
Horses
Other - feediots, piggery ete
inlet
Figure 9: Phosphorus input, transformation, storage and loss for the Peel-Harvey Watershed
(After Keipert et al. 2008)
me i
2000 + 10
———— Annual P loss to estuary (tonnes)
— — -— Annual stream P storage (tonnes)
——- Cumulative soil P store (kg per ha)
----- Annual edge of paddock P loss (kg per ha)
‘\
eer Vv =
Pr tee tT pg Fh Gd ik
eceoeeoeosoeoeoeeses9eeoe9a2
SSFanSHRSOESRSFARDTHRSRGDSS
AMAAMAAARMARSSeSesSsescoeeooceor
rrrrrrrr Tr TF NNNNNNN NNN
Year
Figure 10: P losses and storages for watershed compartments over a 200 year simulation
Best Management Practice implementation scenarios
BMP effectiveness and application
The figures in Appendix 1 illustrate the STELLA® model following division into
discrete sectors which are more manageable in programming terms. A summary of the
effects of the various BMP sectors in terms of mitigating the P loss through the model is
also shown below in Table 2.
through u: gement BMP:
Best Management Practice Proposed maximum Location within model and P-
P-loss reduction transport system
potential
Planting of perennial pasture species 10% P runoff from soil
Re-planting of riparian vegetation 10% P runoff from soil
Stock exclusion from waterways 30% P runoff from soil
Use of P-retentive soil amendments 40% P leaching through soil
Use of low-solubility P fertilizers 30% P leaching through soil
“Best practice” fertiliser management — | 10% P input through fertiliser
appropriate fertiliser only applied to
soils requiring additional P and at the
correct rates and times
The absolute effectiveness of these BMPs is not certain as reported values for their
effectiveness are variable, as is their effectiveness under different land use and
hydrological conditions. However, there is a significant amount of locally-derived
information for BMP effectiveness in the Peel-Harvey region (Regeneration
Technology Pty Ltd 2006, McKergow et al. 2002, Cronin 1998, Steele et al. 2009,
Department of Agriculture and Food WA 2008, Steele 2006, Summers 2004).
Individual BMPs are applied to those particular components of the model to which they
specifically apply (Table 2). For example, the proposed P-loss mitigation effect of the
use of best practice fertiliser management actually reduces the amount of P being
applied to farmland and, therefore in modelling terms, acts directly on P inputs.
Conversely, management of riparian zones acts predominantly on P runoff over the soil
surface. This is an important factor as multiple BMP actions on P as it moves through
the model / system act in series, with BMPs applied earlier in the system effectively
multiplying the effectiveness of those applied later.
Results of simulations
No-change scenario
When the initial model base run is allowed to run the course of the full 200 year
simulation (Figure 10 and Table 3) releases of P from the soil store reach their maxima
in approximately 70 years from now and will not reduce from this value. That is, the
soil P “storage” components of the watershed are already “leaking” P, and their ability
to buffer will be almost exhausted in 70 more years if current practices continue.
Concomitant with the reduction in soil P storage capacity, is a maximization of the
capacity of the watershed streams to store P in around 20 years time. From this point
onwards, an amount of P approaching the entire present day P farm budget surplus will
be released into the regional waterways.
BMP-implementation scenarios
The effectiveness of the implementation of BMPs generally increases the earlier in the
P-transport pathway that they are applied. The implementation of comprehensive
fertiliser-management practices which act at the very start of the P-transport pathway
and effectively reduces actual P imports into the watershed, has the potential to produce
a net P loss into the estuary of approximately 291 tonnes per annum (tpa) after 50 years
and 541 tpa after 100 years. Conversely, BMP actions implemented later in the P-
transport pathway, such as the use of perennial pastures and stock exclusion from
waterways, have a much lower effectiveness. Perennial pastures allow net P-losses of
1146 tpa and 1301 tpa at 50 years and 100 years respectively, and stock exclusion
allows net P losses of 1039 tpa and 1217 tpa at the 50 and 100 year points. Neither of
these results differ significantly from the expected P-losses under the “no change”
scenario.
Implementation of all BMPs, which effectively attempts to improve on all areas of
inefficiency in the farm to watershed P-transport system, can potentially lead to net P
losses to the estuary of 184 tpa and 345 tpa at the 50 and 100 year points respectively.
Even this scenario does not reduce P losses from current day levels and certainly does
not reach the target P-loss rates of 70 tpa.
Scenario P export to estuary (tonnes)
50 years from present 100 years from present
Current P export 140 tonnes per
annum
Target P export 70 tonnes per
annum
No change in management 1200 1342
Planting of perennial pasture 1173 1322
species to 50% of appropriate
land
Planting of perennial pasture 1146 1301
species to 100% of appropriate
land
Stock exclusion from all 1039 1217
watershed waterways
Improved management of 1039 1217
riparian vegetation to all
watershed waterways (complete
stock exclusion and vegetated
buffer).
Scenario P export to estuary (tonnes)
50 years from present 100 years from present
All “biological” BMPs — stock 1001 1187
exclusions, riparian management
and use of perennial pastures
Use of P-retentive soil 833 1051
amendments applied to all sandy
soils at 10 tonnes ha
Use of P-retentive soil 465 745
amendments applied to all sandy
soils at 20 tonnes ha"!
Low-solubility P fertiliser 833 1051
“Best Practice” fertiliser 707 958
management
All “chemical” BMPs — low- 291 S41
solubility fertilizers, “best”
fertiliser management and soil
amendmnent with P-retentive
materials.
All BMPs adopted 184 345
Discussion and conclusions
If nutrient input rates into the Peel-Harvey watershed continue at current levels then
this, combined with the expected reduction in buffering capacity of the soils and
streams, will have major environmental implications for a watershed and associated
waterways which are already under severe stress. Not only will the target P-loss rate of
70 tpa not be achieved, but over the course of the next 100 years, P losses will increase
by a factor of more than 9 times the current rate (from 140 tpa to approximately 1300
tpa). Broad-scale, comprehensive implementation of BMPs which address all
components of the P-transport pathway can, at best, produce annual P losses of 184
tonnes and 345 tonnes at the 50 and 100 year points respectively.
The importance of the location of P-management strategies along the P-transport
pathway has also been shown. Maximum BMP-effectiveness is achieved by the
application of BMPs both at the earliest possible point in the P-transport pathway and/or
at the most critical point in the watershed. Those BMPs which most reduce P imports
into the watershed (best-practice fertiliser management) and which attempt to address
the most critical issues in terms of P loss (the use of P-retentive soil amendments which
target extremely poorly P-retentive soils) are most effective at reducing P loss.
Implementation of a series of BMPs, each of which addresses the P-losses which were
not attenuated by the previous BMP in the pathway maximize potential P retention.
However, there are still a number of large uncertainties both in the model as used in this
study and in the watershed system itself. Phosphorus transport processes are almost
entirely hydrologically driven. Whilst the current model inherently contains some
hydrological information through the use of P-balance data as an input source, it does
not cater for spatial or temporal variations in hydrological regime. It is not known, for
example, what happens in terms of P loss in dry versus wet years, or what would
potentially happen to P stored in the soils and sediments of the watershed if a wet year
occurs after a long series of dry years (as is currently the case). The forms of P stores in
the terrestrial and hydrological components of the watershed (and subsequently the
model), and how these vary temporally are also poorly understood, as is the rate of P
“utilization” within soils and drainage systems by biota other than that used in
agricultural production.
Modelling indicates that nutrient levels in the estuary and waterways will increase
significantly over the next 50 to 70 years unless major efforts are made to reduce losses
at source. It is unlikely that symptomatic interventions at downstream points will be
able to successfully manage nutrient accumulation rates in the future without major re-
design of agricultural systems or re-engineering of soil and drainage systems if the
present agricultural paradigm remains.
References
Beasley DB and Huggins LF (1982) ANSWERS (Areal Nonpoint Source Watershed
Environmental Response Simulation) User's Manual. Chicago: U.S. Environmental
Protection Agency Report No. 905/9-82-001.
Behrendt H and Boekhold A (1994) Phosphorus saturation in soils and groundwaters.
Land Degradation and Rehabilitation. 4, 233-243.
Butcher J, Shoemaker L, Clements JT and Thirolle E (1998) Watershed modeling online
training module. U.S. Environmental Protection Agency, Watershed Academy Web.
Available from: http://www.epa.gov/watertrain/modeling/
Carpenter SR, Caraco NF, Correll DL, Howarth RW, Sharpley AN, Smith VH (1998)
Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecological
Applications. 8 (3), 559-568
Cassell EA, Kort RL, Meals DW, Aschmann SG, Dorioz JM and Anderson DP (2001)
Dynamic phosphorus mass balance modelling of large watersheds: long-term
implications of management strategies. Water Science & Technology. 43 (5), 153-162.
Chen L, Arimoto R and Duce RA (1985) The sources and forms of phosphorus in
marine aerosol particles and rain from northern New Zealand. Atmospheric
Environment. 19 (5), 779-787.
Cronin D (1998) The Effectiveness of Steamlining in Improving the Water Quality of
Agricultural Drains in the Peel-Harvey Catchement, Western Australia. Unpublished
Honours Thesis Murdoch University, Western Australia.
Department of Agriculture and Food Western Australia (2008) Bulleting 4689.
DairyCatch — Environmental best practice guidelines.
http://www.agric.wa.gov.au/PC_91777.html
Department of Natural Resources and Environment. (1996) Blue green algae and
nutrients in Victoria: A resource handbook. Department of Natural Resources and
Environment, Melbourne, Australia.
Djodjic F, Montas H, Shirmohammadi A, Bergstrom L and Ulen B. (2002) A decision
support system for phosphorus management at a watershed scale.. Journal of
Environmental Quality. 31, 937-945.
Environmental Protection Authority (2008) Water quality improvement plan for the
rivers and estuary of the Peel-Harvey system - phosphorus management. Government of
Western Australia.
Ford A (2010) ‘Modelling the Environment — 2™ Edition.’ Island Press. Washington
DC.
Heidtke TM and Auer MT (1993) Application of a GIS-Based Nonpoint Source
Nutrient Loading Model for Assessment of Land Development Scenarios and Water
Quality in Owasco Lake, New-York. Water Science and Technology. 28 (3-5), 595-604.
Keipert N, Weaver D, Summers R, Clarke M and Neville S (2008) Guiding BMP
adoption to improve water quality in various estuarine ecosystems in Western Australia.
Water Science and Technology, 57, 1749-1756
Lavell K, Summers R, Weaver D, Clarke M, Grant J. and Neville S (2004) ‘An audit of
the uptake of agricultural nutrient management practices in the Peel Harvey catchment.’
Proceedings of the 7" International River Symposium, Brisbane. Qld. Aug 31st — 3rd
Sept 2004.
McLaughlin MJ, Fillery IR and Till AR (1992) Operation of the Phosphorus, Sulphur
and Nitrogen Cycles. Bureau of Rural Resources Proceedings No. 14, 67-116.
McLaughlin A and Mineau P (1995) The impact of agricultural practices on
biodiversity. Agriculture, Ecosystems and the Environment. 55, 201-212.
Meals DW, Dressing SA and Davenport TE (2010) Lag time in water quality response
to best management practices: A review. Journal of Environmental Quality. 39, 85-96.
Nair VD, Portier KM, Graetz DA and Walker ML (2004) An environmental threshold
for degree of phosphorus saturation in sandy soils. Journal of Environmental Quality.
33, 107-113.
Neville SD, Weaver DM, Lavell K, Clarke MF, Summers RN, Ramsey H and Grant J
(2004) Nutrient balance case studies of agricultural activities in south west Western
Australia. Proceedings of the 7" International River Symposium, Brisbane. Qld. Aug
31st — 3rd Sept 2004.
Pike R (2003) ‘Approaching watershed modelling.’ Streamline: Watershed management
bulletin. 7(3). Fall 2003.
Poiani KA and Bedford BL (1995) GIS based nonpoint source pollution modelling:
Considerations for wetlands. Journal of Soil and Water Conservation. 50, 613-619.
Regeneration Technology Pty Ltd (2006) Draft best management practices for rural
drains in the South West and South Coast drainage districts of Western Australia.
Reynolds CS and Davies PS (2001) Sources and bioavailability of phosphorus fractions
in Freshwaters: A British perspective. Biological Reviews of the Cambridge
Philosophical Society. 76 (1), 27-64.
Seddon G (1972) ‘Sense of Place.’ University of Western Australia Press. Perth.
Sharpley AN and Smith SJ (1990) Phosphorus transport in agricultural runoff: the role
of soil erosion. Soil erosion on agricultural land. Proceedings of a workshop sponsored
by the British Geomorphological Research Group, Coventry, UK, January 1989 [Ed.
Boardman J, Foster IDL and Dearing JA]. 1990, 351-366 Chichester, UK. John Wiley
& Sons.
Steele J J (2006) Management of diffuse water quality issues in the Peel-Harvey coastal
drainage system — a literature review. Report to the Peel-Harvey Catchment Council.
Steele JJ, Keipert N and Summers RN (2009) Monitoring and Evaluation of the
Effectiveness of Fencing and Riparian Vegetation in the Peel Harvey Catchment.
Department of Agriculture and Food Western Australia.
Summers RN (2004) A review of Best Management Practices for achieving reductions
in nutrient and sediment loads to receiving waterways. Department of Agriculture and
Food Western Australia.
Weaver D M, Neville SN, Summers RN, Deeley DM and Clarke MF (2004) Reducing
nutrient discharge from agriculture through the implementation of BMPs — how far can
we go? Proceedings of the 7" International River Symposium, Brisbane. Qld. Aug 31st
— 3rd Sept 2004.
Weaver DM, Neville SN, Ovens R, Keipert N, Summers RN and Clarke MF (2008)
Farm gate nutrient balances in south west Western Australia — understanding nutrient
loss risk within agricultural landuses. 12th International Conference on Integrated
Diffuse Pollution Management (IWA DIPCON 2008). Research Center for
Environmental and Hazardous Substance Management (EHSM), Khon Kaen University,
Thailand ; 25-29 August 2008.
Weaver DM and Wong MTF (2011) Stuck in topgear: overdue downshift from
phosphorus build-up to maintenance in areas of Australian crop, dairy and meat
industries. In Prep.
Young RA, Onstad CA, Bosch DD and Anderson WP (1989) AGNPS, a non-point
source pollution model for evaluating agricultural watersheds. Journal of Soil and
Water Conservation. 44, 168-173.
