A catchment study of sources and sinks of nutrients and sediments in south-east Australia

Journal of Hydrology 515 (2014) 166–179

Contents lists available at ScienceDirect

Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol

A catchment study of sources and sinks of nutrients and sediments in south-east Australia R. Adams a, Y. Arafat b, V. Eate b, M.R. Grace b, Sh. Saffarpour a, A.J. Weatherley c, A.W. Western a,⇑ a

The University of Melbourne, Department of Infrastructure Engineering, Parkville, VIC 3010, Australia Monash University, School of Chemistry, Clayton, VIC, Australia c The University of Melbourne, School of Land and Environment, Parkville, VIC 3010, Australia b

a r t i c l e

i n f o

Article history: Received 6 December 2012 Received in revised form 3 February 2014 Accepted 14 April 2014 Available online 30 April 2014 This manuscript was handled by Laurent Charlet, Editor-in-Chief, with the assistance of Ewen James Silvester, Associate Editor Keywords: Land use Water quality Sediments Nitrogen Phosphorus Diffuse source pollution

s u m m a r y A study of nutrient and sediment generation was carried out on a dairy farm in south eastern Australia over a period of two years. The study collected nitrogen (N), phosphorus (P) and total suspended sediment (TSS) samples from three headwater subcatchments totalling 4.4 km2 in area, which had streamflow measurement sites located at their outlets. In addition to these three sites, samples were collected from farm tracks, the discharge from a farm dam (water storage pond) and a rainfall sampler. Over 20 runoff events were monitored in addition to background flows in the catchment. The data enabled nutrient exports to be estimated from the total catchment and two of the three subcatchments. The atmospheric nutrient loads were also estimated and found to be similar to other studies in the region, and were low compared with the nutrient export. The nutrient load estimates indicated that a large proportion of the catchment Total Nitrogen (TN) and Total Phosphorus (TP) loads originated from critical source areas on the farm associated with dairy operations, and that the catchment TN and TP export rates were high compared to other studies. The sediment data also indicated that TP concentrations were closely related to TSS in samples collected from surface runoff. Catchment TN loads in baseflow were high relative to the total load, probably due to indirect effluent and direct groundwater discharges into watercourses, with high concentrations of soluble N (nitrate plus nitrite and ammonium). The farm dam was found to trap a high percentage (>60%) of the influent TP load; however, the retention of TN by the dam was seasonally varying and close to net neutral. Ó 2014 Elsevier B.V. All rights reserved.

1. Introduction An understanding of sediment and nutrient dynamics from headwater catchments to receiving water bodies such as lakes, estuaries and coastal waters is important as high export rates can have detrimental effects on water quality and the aquatic environment (Young et al., 1996; Harris, 2001; Sutton et al., 2011). Pastoral agriculture, particularly intensive dairy farming, has been identified as a high export source in many places (Drewry et al., 2006; Haygarth et al., 2004; Holz, 2010; Sharpley et al., 2008; Wilcock et al., 2007). Smith et al., (2013) showed that intensification of the dairy industry in south-western Victoria was associated with increasing nutrient (N and P) concentrations in water bodies at the catchment scale. Dairy farms are an inter-related, complex mix of potential sources and sinks of N and P. Studies typically assess the role of agricultural fields (paddocks) as the source of these nutrients to water bodies; however, farm buildings, ⇑ Corresponding author. Tel.: +61 3 8344 7305. E-mail address: [email protected] (A.W. Western). http://dx.doi.org/10.1016/j.jhydrol.2014.04.034 0022-1694/Ó 2014 Elsevier B.V. All rights reserved.

farmyards, barns and dairy sheds have the potential to export nutrients, particularly organic forms of N and P, from animal wastes (Dougherty et al., 2004; Edwards et al., 2008). Export from farmyards remains a largely overlooked research topic (Edwards et al., 2008). Pionke et al., (2000) introduced the concept of Critical Source Areas (CSA) as high runoff generating areas of the landscape that also have the potential to generate high nutrient loads. The farm features described above fit into this definition. The CSA concept has been used more extensively for P generation (Heathwaite et al., 2005a; Müller et al., 2010) than for N generation (Pionke et al., 2000). Farm tracks are one important potential CSA, however their contribution to nutrient and sediment export from a catchment is often overlooked (Monaghan and Smith, 2012). A study in the Bog Burn catchment (NZ) by Monaghan and Smith (2012) found that overland flow from tracks and laneways was highly enriched in both soluble forms of N and P, and particulate forms of P. Using estimates of urine and dung loadings per animal from the literature, Eckard et al. (2007) visually observed N deposition on laneways and in sheds due to dairy cows moving around a farm

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

and estimated that deposition accounted for 33–52% of the overall N losses from the farm (using published data on loading rates). The role of impervious surfaces such as farm tracks and unsealed roads has been investigated in European catchments for sediment transport. Gruszowski et al. (2003), found that 30% of the suspended sediment transported to the River Leadon (UK) was contributed from roads. However, P export from the Sem (Southern England) catchment studied in Deasy et al. (2008) was dominated by drainflow, from one field (tile) drain constructed by the farmers to relieve waterlogging, and P fluxes were two orders of magnitude higher in the drain than in the overland flow from the tracks. It is clear that farm tracks and other farm infrastructure containing CSAs can be important pollutant sources. Management of these sources requires the identification of the CSAs (Heathwaite et al., 2005a) and selecting appropriate mitigation measures, for example fencing off waterways to prevent stock access and constructing riparian buffer strips to trap particulate nutrients (as used in the Toenepi catchment in New Zealand (Wilcock et al., 2007)). In the late 1990s, best management practices (BMPs), including the diversion of dairy shed effluent from the stream to land discharge were also introduced in the Toenepi and other New Zealand dairy farming catchments. These changes led to a considerable improvement in stream water quality over 5– 10 years of nutrient monitoring (Wilcock et al., 2007). Monaghan and Smith (2012) recommended sediment traps and bunding track drainage to ensure that this runoff does not enter watercourses. Fields are a potentially significant nutrient source and intensively grazed dairy fields have been extensively studied in Australia and elsewhere for P (Nash and Murdoch, 1997; Nash and Halliwell, 1999; Fleming and Cox, 2001; Barlow et al., 2005; Weatherley et al., 2011) and for N (Ridley et al., 2001; Eckard et al., 2004, 2007; Barlow et al., 2007). For an intensively-grazed runoff plot in Darnum, Victoria, Haygarth et al. (2004) found that overland flow and interflow generated concentrations of filterable reactive P (FRP) were relatively constant during events compared with the variable runoff from the plot (although at peak runoff some dilution was observed). They suggested that the Darnum results were typical of small-scale runoff plots, which were much smaller than the catchments in other studies. In the Jubilee catchment (UK), Deasy et al. (2009) found that surface runoff sources (from a field) contributed 0.1% and 1.0% of the catchment runoff and sediment load respectively. Surface runoff was limited and restricted by low connectivity between the hillslope and stream. In regions of the world with highly-variable climates (prone to drought or extended dry seasons), including south-eastern Australia, farmers tend to store water on-farm using farm dams (small storage reservoirs). Farm dams have also been largely overlooked in the literature as a source or sink of sediments and nutrients, and their effect on water quality is not well known in Australia (Brainwood et al., 2004), or internationally (Maxted et al., 2005; Fairchild and Velinsky, 2006). In the USA, Lessard and Hayes (2003) studied small ponds in a humid temperate climate zone for SS and TP removal, but did not find any evidence of TP removal at downstream monitoring sites. In New South Wales, Australia, (Brainwood et al., 2004) found that water quality in three farm dam monitoring sites was related to the land use of the impounded catchment but did not evaluate nutrient removal. Additionally, on dairies farmers often use ponds or tanks to store effluent to recycle nutrients on farm, and these may overflow during wet spells and pollute watercourses (Müller et al., 2010), in which case they become a source rather than a sink of nutrients. Instream concentrations of dissolved and particulate nutrients typically respond to flow variations due to the changing importance of different water sources, the biogeochemical dynamics of those sources, and differences in steam processes such as erosion. The role of storm events in generating high sediment and nutrient

167

loads has been studied in some detail, particularly as a source of particulate P (PP) (Pionke et al., 1999, 2000; Correll et al., 1999; Haygarth et al., 2004, 2005b; Deasy et al., 2008; Sharpley et al., 2008), and less frequently as a source of N, especially nitrate (Correll et al., 1999; Pionke et al., 1999; Poor and McDonnell, 2007), and particulate N (PN), which is often included in TN instead of being separately analysed. Often the great majority of annual export is in the largest couple of runoff events due both to increased concentrations and the volume of runoff involved (e.g. Pionke et al., 1999; Nash and Murdoch (1997). Studies involving intensive (temporal and spatial) monitoring of storm events can quantify runoff and sediment exports at different spatial scales (e.g. Deasy et al., 2011). Johnes (2007) analysed a set of catchments in the UK with daily P time series data. She found that the catchments having a low baseflow index (BFI) tended to export mostly PP during events, and that a large percentage of the annual TP export was from events rather than low-flow (baseflow) periods. A low BFI indicates that event runoff tended to dominate their hydrographs. Haygarth et al. (2004) reported that 69% of the annual TP load was exported from a 1.8 ha paddock in a single runoff event. Typically the concentrations of sediment associated materials such as PP and PN increase during runoff events (a ‘‘concentration’’ effect) and there is some evidence that common hydrological pathways are involved (Kato et al., 2009; Correll et al., 1999). Kato et al. (2009) observed that a ‘‘first flush’’ mechanism was apparent in the export of TP, PP, PN and ammonium during events. They stated that particulates were more likely to be exported in the first flush than dissolved nutrients since a readily mobilised source was available on the farmland soil surface, and in the substrate of dry drainage canals. Concentrations of dissolved material such as nitrate often decrease (a ‘‘dilution’’ effect) as flow increases during an event (Holz, 2010; Poor and McDonnell, 2007). Sometimes other factors can control temporal variation. For example, Holz (2010) found that ammonium varied as a function of grazing events rather than flow. We have identified several clear knowledge gaps based on a review of the research in worldwide regions that may have similar conditions (climate and agricultural activities) to the south east region of Australia. The first is the importance of tracks in conveying sediments and nutrients to watercourses during storm events. The second is the role of farm dams in potentially mitigating nutrient export. Thirdly, many previous studies have concentrated on carefully designed field experiments at small spatial scales. This study investigates nutrient (N and P) export from headwater subcatchments containing a working dairy farm in south-eastern Australia. It also examines the effect of farms tracks and a farm dam on those loads. This paper aims to quantify the role of various catchment elements as sources and/or sinks of nutrients in a dairy farming catchment, to examine the differences in sources between storm event periods and baseflow periods and to quantify atmospheric deposition of nutrients. 2. Methods 2.1. Description of study area The monitoring sites were situated on a 188 ha dairy farm located in the 4.4 km2 Poowong East headwater catchment (PEC) of the Lang Lang River located at (38.33°S, 145.83°E) in Victoria Australia, approximately 130 km south east of Melbourne (Fig. 1). The land use in the PEC is almost entirely intensively-grazed pasture, mostly associated with dairy cattle and some beef cattle, particularly in the upper SW part of the catchment (Table 1). Monitoring was carried out at the farm between August 2009 and November 2011.

168

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

Fig. 1. Map of catchment and location of study area in south eastern Australia. Terrain data is from DSE (2008).

2.1.1. Description of farm operations The farm has a dairy (milking shed and hardstanding washdown area) located on the west bank of the main stream (Fig. 2). To access the dairy for milking, on most days between 150 and 200 dairy cows cross the stream from the eastern paddocks via a bridge up to four times a day. The number of lactating cattle varies by season. The mean stocking rate (total head of cattle divided by farm area)  1 cow/ha, the stocking density (typical head of cattle in each paddock divided by paddock area)  100 cows/ha. Some paddocks are used seasonally for hay and silage production and there are small areas of forage or fodder crops at certain periods (for simplicity the land use ‘‘pasture’’ encompasses all of these activities). Animal wastes (faeces and urine) are deposited on this bridge and associated connecting tracks. During high rainfall Table 1 PEC subcatchment characteristics. Land use

UPH area (km2)

UDP area (km2)

DU area (km2)

Pasture Tracks Woodland Farm buildings Water Farmyard Intensive use areas

0.495 0.009 0.064 0.004 0.008 0.008 0.037

2.33 0.036 0.14 0.004 0.017 0.013 0.008

1.17 0.008 0.031 0.001 0.01 0.002 0.014

Total area

0.60

2.58

% Impervious

3.6%

2.1%

1.24 (note DL = 1.29) 0.9%

events, surface runoff washes these into the stream flowing under the bridge via a series of drainage culverts and gutters (Adams et al., 2010). During wet periods, dairy effluent overflows from the farm’s effluent recycling system (next to the farm buildings indicated in Fig. 2) into the stream. Details on fertiliser applications (typically N in the form of ammonium-based fertilisers, chicken manure and urea; P in the form of superphosphate and DAP) were provided by the farmers on a weekly basis. Additional distributed nutrient loads originated from both the grazing livestock and effluent (manure) spreading activities, in addition to the point sources due to dairy and track overflows. Between May and December 2010 the applications to the paddocks totalled 98 kg N/ha and 33 kg P/ha. The information on nutrient sources from the neighbouring properties was not available so that it was not possible to determine a complete nutrient budget for the PEC subcatchments. Typical fertiliser application rates from studies on nearby farms in the Gippsland region are 200 kg N/ha/year in the form of urea or ammonium nitrate (Eckard et al., 2007) and 44 kg P/ha/year in the form of inorganic phosphate (Nash and Halliwell, 1999).

2.1.2. PEC Subcatchments Fig. 2 shows the landscape features within the farm boundary (the black outline in Fig. 1), and the streamflow water monitoring sites that define the outlets of the three PEC subcatchments (Dam Upper (DU) 1.2 km2, Upper Dam Paddock (UDP) 2.6 km2 and Upper Possum Hollow (UPH) 0.6 km2). DU was named such as it was upstream of the farm dam; the other two were named after the paddocks containing the sites. Streams in the PEC are first to fourth

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

169

Fig. 2. Detailed map of farm area showing monitoring sites and landscape features. The red crosses denote monitoring sites on Track 1 referred to in the text. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

order and according to the farmer flow intermittently. Areas of riparian woodland (dark green) exist where the farmer has implemented Best Management Practices (BMPs) by fencing off the streams to prevent livestock access and planting native trees and other vegetation. Table 1 shows the areas of each land use. The total percentage of ‘‘impervious’’ area in each subcatchment is the sum of tracks, farm buildings and farmyards (including hardstanding areas). The DU subcatchment captures runoff from a branch of the main stream (Fig. 1). Downstream of the DU monitoring site is a large farm dam constructed in the early 1970s, which supplies water to the main dairy for ‘‘stock and domestic’’ uses, typically milking shed wash-down (Figs. 1 and 2). Over time, an overflow channel formed on a bedrock exposure on the west side of the dam wall, which is the location of the Dam Lower (DL) site. The subcatchment area between DU and DL is only 5 ha. The (unnamed) stream from the farm dam subcatchment containing DU and DL joins the main stream just downstream of UDP. The UPH subcatchment flanks the main (unnamed) stream between the lower gauging station (UPH) and the UDP gauging station. The UPH subcatchment contains the farm buildings including the dairy operations, described above. The bridge enabling the stock to cross to the dairy is shown in Fig. 2 along with the lower Track 1 monitoring site (TL) described below. The hydrological setting of the study area is the humid, cool temperate zone of south-eastern Australia. Mean annual rainfall (1961–1990) was approximately 1100 mm (BOM, 2009), with a similar annual areal potential evapotranspiration of 1040 mm (BOM, 2005). Monthly rainfall totals are distributed fairly evenly across the year with the wetter months in late winter and early spring according to historical data. The PEC is situated between

133 and 321 m elevation in fairly steep hill country that was cleared of native forest in the early 1900s. The average slope across the PEC is 14°, ranging from 0° to 47.5° (derived from a 10 m DEM described in the supplementary material). The local soils are derived from the underlying Cretaceous mudstones and are classified as Brown Dermosols in the Australian Soil Classification (Isbell, 2002). The profiles are relatively well-drained grading from a fine sandy clay loam in the surface to medium clay over bedrock at approximately 1 m depth. Surface soils on the hillslope site (UPH subcatchment) were found to be highly acidic (pHw, 5.6). 2.2. Terrain and feature analysis A map of the land use in the PEC was obtained by digitizing aerial photographs from the internet www.bing.com/maps (at a resolution of approximately 2 m) in ARC-GIS and identifying significant features on the ground from these (Fig. 2). A major track connecting the eastern side of the property to the main farm buildings (Track 1, shown in Fig. 2) was surveyed using a handheld GPS unit. This track is important hydrologically since it is steeply sloping and has drainage gutters on both edges that discharge runoff from the lower 10 m of the track into the main stream near the dairy effluent overflow. Several fields also drain into Track 1 due to the steep terrain (Fig. 1). Additional information on the terrain analysis can be found in the Supplementary Material (S1.1). 2.3. Hydrological and meteorological monitoring data 2.3.1. Meteorological data From April 2010, a Campbell Scientific automatic weather station (AWS) was installed between UDP and DU. The station

170

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

recorded precipitation using a Hydrological Services Australia tipping-bucket raingauge (0.2 mm tips). Rainfall from a tipping bucket raingauge (site ULL) maintained by the local catchment management authority, located at Ranceby 2 km SW of the PEC boundary, was also used for some of the analysis. Hourly data were obtained from this raingauge (source Melbourne Water) for the period 1/5/2010 to 30/11/2011. 2.3.2. Runoff The three PEC monitoring sites all had flow measurement gauges installed in April 2010. Prior to this, water level was recorded using a logged capacitance probe at the UPH site installed in July 2009. The UDP and UPH gauges were operational until the end of 2011. The three gauging stations all recorded flow at ten minute intervals and had different equipment installed based on site characteristics (see Supplementary Material (S1.2) for more information on runoff measurement). 2.4. Water quality sampling Water quality samples were collected both manually and with an autosampler. During 2009–2010, ISCO 6712 (Teledyne-ISCO, Nebraska USA) autosamplers were installed at the three gauging stations (DU, UDP and UPH) and programmed to collect (up to 24) samples during events, triggered by rising stage in the streams. Sites were visited as soon as possible (<48 h) after events so that filtered subsamples for dissolved nutrient analysis could be obtained in the field, with the assumption made that concentrations of FRP in particular would not change measurably between sampling and collection times. This assumption was based partly on a study by Kotlash and Chessman (1998) who determined that FRP (and dissolved forms of N) concentrations would not decrease significantly providing samples were collected within 48 h of pumping. More information on the sampling methods can be found in the Supplementary Material (Section S1.3). Samples were field filtered using 0.2 lm Advantec™ filters and syringes to obtain a 20 mL subsample for the analysis of dissolved nutrients NOx (nitrate + nitrite), NH4 (ammonium) and FRP. A pre-filter (0.45 lm) was used to help to prevent clogging during the filtration process. In some cases it was not possible to obtain dissolved samples, particularly from UPH, as suspended sediment clogged the filters immediately. Samples were transported back to the laboratory on ice and refrigerated prior to analysis, which took place within two weeks, except for the filtered subsamples which were frozen for up to one month. Filtering using 0.2 lm instead of 0.45 lm samples has been recommended to remove: (i) any bacteria that may subsequently have an effect on the dissolved P speciation; and (ii) 0.2–0.45 lm colloidal particles that have attached P (Heathwaite et al., 2005b), since these forms of P are not strictly ‘‘dissolved’’ P. It appears that the use of 0.45 lm filters is largely a consequence of these being readily available and adopted by most monitoring agencies as the norm (Haygarth et al., 2005a). Manual sampling occurred from mid-2009 at the gauging stations described above using grab sampling with a sampling arm to access the middle of the channel, and at DL from July 2009 to September 2011 either when samples were collected from the autosamplers or weekly if no events were sampled (during 2011 sampling switched to bi-weekly). Table S1 (in the Supplementary Material) provides a list of what measurements and sampling took place at each monitoring site. Grab samples were also collected from Track 1’s gutter (at sites TL and AR – Fig. 2) during selected runoff events in 2009 and 2011 to enable track loads to be estimated. A batch of samples collected in 2009 and 2011 were analysed for TSS, for which the largest possible sample volume (up to 1 L) was collected.

In 2010, an additional ISCO 3700 autosampler was installed at the DL site, and programmed to sample every 12 h (giving 12 days of samples per run). A second ISCO 3700 was installed at an additional site DS (‘‘Dam South’’, 30 m upstream of the dam perimeter) in March 2011, enabling collection of a continuous higher-frequency dataset measuring the water quality (Total P and N only) both entering and leaving the dam. 2.5. Laboratory analysis Nutrients were measured using the methods indicated below in the NATA (National Association of Testing Authorities)-accredited Water Studies Centre Laboratory, at Monash University using Flow Injection Analysis (FIA) (QuickChem 8000, Lachat, Loveland, CO, USA). (i) Dissolved nutrients were measured in the filtered samples. Ammonium was determined using the phenate method (APHA, 2005). Nitrate was converted to nitrite in a cadmium reduction column, and NOx measured colourimetrically following the procedure in APHA (2005). FRP was determined using the phosphomolybdenum blue method of Murphy and Riley (1962). (ii) Total Nutrients were measured after alkaline persulphate digestion and analysed using the colorimetric FIA methods for FRP and NOx as described above. Measurement errors from the FIA analysis methods (i) and (ii) were reported as being the same as the lower limits of detection for each nutrient. Where nutrient concentrations exceeded the upper measurement limit, dilutions were made to lower the concentration within the measurement range. (iii) TSS. (Total Suspended Sediment) concentrations were measured in the laboratories at the University of Melbourne using the standard gravimetric method (APHA, 2005). A known volume of sample was filtered under a vacuum with the residue collected on a Whatman™ #5 filter paper. The filter paper was weighed prior to this step. The filter paper plus residue was then oven dried at 105 °C. After drying the paper was weighed again and the TSS concentration calculated by the difference in mass before and after the filtration and drying steps divided by the sample volume. 2.6. Analysis of events and loads In this paper a number of different aspects of nutrient loads are addressed. Loads directly analysed were: the event loads and interevent loads at each of UPH, DU and UDP; loads derived from track away from the vicinity of the dairy shed; and changes in loads through the dam. Annual loads were calculated at DU and UPH, but not at UDP due to insufficient event sampling. In addition, annual loads are differenced between the downstream and two upstream sites, that is UPH less the combination of DU and UDP, where the yield from UDP was assumed to be the same as the yield from DU. This difference is attributed to the UPH subcatchment and consists of: (a) diffuse load derived from fields; (b) loads derived from tracks; (c) loads derived from intensively trafficked and hard standing areas around the dairy shed itself (including any effluent overflows) and (d) the influence of the dam on loads. Of these (a) can be estimated from assumptions about similarity with the upstream catchments (which are dominated by fields), (b) is estimated based on sampling of track runoff concentrations and estimation of track runoff coefficients, and (d) is determined from measurements as described below. It was not possible for practical reasons to sample the discharge from (c) so it forms an

171

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

unknown term in the loads estimated at UPH. Prior to the analysis, the bulk of the loads from tracks (b) were thought to enter the stream during events, since during field visits in dry periods flow in track gutters was not observed. Therefore these will assumed to be incorporated into the UPH event loads. The load estimation method at each of the three flow monitoring stations involved first separating the hydrograph into events and baseflow, estimating the event loads, estimating the interevent baseflow loads and then combining all these. Events were defined as follows (with baseflow periods between each one): The start of an event was defined as a clear rise in stage (s) following the onset of rainfall. The end of the event was defined as either (in order of priority): (a) the time when s decreased back to its initial value at the start of the event; (b) the time when s started to rise again following at least 0.2 mm rainfall, thus defining the start of the next event (in this case the previous time step would be taken as the end time); (c) The time when there had been (i) a dry period of at least 24 h and (ii) s was observed to be decreasing at less than 5 mm/h. Event runoff volume was then calculated by integrating the instantaneous (10 min) flows over the event and converted to runoff depth by dividing by catchment area (Table 1). Event loads. Depending on data availability (see Section 2.4 for a description of the event monitoring), event loads were calculated by either multiplying the observed event flow weighted mean concentration (CFWM) and event runoff or by assuming a CFWM and multiplying by event runoff. Where three or more samples existed for an event, CFWM was calculated using:

C FWMi ¼

X

CiQ i

.X

Qi

ð1Þ

where Ci and Qi are the sampled concentration and the flow rate at the time of the sample, respectively. Where two samples were taken, CFWM was assumed to be the arithmetic average of event sample concentrations. For events with only one sample (e.g. Event 13 at DU), the single sample was assumed to represent the entire event. For events with no samples, CFWM was taken as the mean of CFWM for all sampled events. Grab samples taken within 72 h of the event start were also used in the analysis. Due to limitations of the sampling methods (both limited autosampler capacity and occasional equipment failures) event samples sometimes did not encompass the entire event (from onset of event runoff until cessation of the falling limb). The total event load was calculated as the event discharge volume multiplied by CFWM, which scales up the ‘‘observed’’ load and assumes the samples are representative of the whole event. The standard errors (SE) of the event load estimates Let were calculated using the method outlined by Fox (2005), based on the assumption that the Q and C data were correlated in natural log space. This method was applied to events with at least 3 samples. Uncertainty bounds of ±20% were applied to those events with less than 3 samples, based on the estimated accuracy of Q, due to the

flow gauging methods used (most of the errors in the loads would come from errors in the flow measurements). Baseflow loads. The baseflow load Lb (i.e. inter-event load) can be estimated by (2) which states that on a general basis:

Lb ¼ C b V b

ð2Þ

where Cb is an estimate of the baseflow concentration of the particular nutrient species (values in Table 2 below) and Vb the total runoff volume during the baseflow periods. At UPH the baseflow periods were further subdivided into ‘‘low’’ and ‘‘high’’ baseflow periods as discussed above, with each of these assigned an appropriate mean Cb. Annual event load. Using a time series of loads calculated for each of the 2010 discrete events it was possible to estimate the annual event load at UPH and DU. It was decided not to estimate the load at UDP as only 4 events were sampled there. This method was used to estimate P loads partitioned between storms and baseflow periods (Sharpley et al., 2008). There were 20 monitored events between May and December 2010, with two or more samples collected at DU, UPH or UDP. There were 4 events where no samples were taken. The nutrient loads entering and leaving the dam were also calculated. Annual total load. The total loads over May–December 2010, at UPH (LUPH) and DU (LDU), were estimated as the sum of loads for all events and baseflow periods. The methods described in Fox (2005) could not be used to estimate the load variance associated with the baseflow samples due to a lack of correlation between Cb and Q even after log transformation of the variables. Therefore, the Lb term had an uncertainty associated with: (a) errors in the Q data (discussed above) and (b) the values of Cb used in Eq (2). Therefore, uncertainty bounds were estimated by adding the flow uncertainty (±10% Q) and the Cb uncertainty (±2 the SE of Cb). Track loads (LTR) in UPH only were estimated using a runoff coefficient model for each event combined with an estimate of the nutrient concentration in track runoff. The track load calculations make the assumptions that: (i) the 2.1 ha of tracks, farmyards and farm buildings in the UPH subcatchment are runoff generation features directly connected to the stream and (ii) the concentrations measured on the eastern side of the stream at TL monitoring site (Fig. 2) are representative of all of these features (the Track 1 (TL) subcatchment area was 0.36 ha). Since the streamflow sampled at UPH included the contribution of the upstream tracks, the estimated total UPH event loads included the track loads (LTR).

LTR ¼ C tm At Rc ðPe  Pi Þ for Pe > P i

ð3Þ 2

where At is the total impervious area (2.1 km ; from Table 1), Ctm is the median concentration from track samples of the nutrient being calculated, Rc is the track runoff coefficient (0.66 was used based on a study of tracks in forested catchments in Victoria by Sheridan and Noske (2007)), Pe is the total event rainfall and Pi the constant initial loss (0.1 mm was used based on field observations during events). In general Pi is the initial infiltration plus the depth of rainfall

Table 2 Summary of event and baseline nutrient data. Dataset

Mean (SD)

DU baseflow DU events UDP baseflow UDP events UPH baseflow UPH events Rainfall Tracks (TL &AR)

No. of samples

67 162 30 39 47 116 63 5

Concentration (mg/L) TN

TP

1.76(0.74) 4.38(3.41) 2.23(2.06) 4.76(2.78) 6.10(2.90) 7.92(5.94) 0.67(0.56) 25.0(19.1)

0.11(0.15) 0.87(1.21) 0.12(0.92) 1.00(1.26) 0.50(0.25) 1.70(2.46) 0.04(0.06) 6.33(6.24)

No. of samples (dissolved)

42 50 34 18 43 37 0 3

Ratio (–) TN:TP

NOx:TN

NH4:TN

FRP:TP

48.1 10.1 23.6 10.7 14.1 9.9

0.45 0.27 0.57 0.47 0.38 0.41

0.04 0.03 0.04 0.02 0.17 0.07

0.28 0.07 0.43 0.15 0.33 0.16

Insufficient

Data

172

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

required to fill any depression storage on the tracks prior to runoff commencing. The load uncertainty was estimated from the concentration (Ct) uncertainty, since this was probably the principal source. This was calculated by evaluating the uncertainty in Ct, using the full range of track concentration data, due to a limited number of samples for a full statistical analysis. Atmospheric sources of N and P were also considered in this study. A sequential rainfall sampler located in the riparian zone near DU was used to collect up to 10,250 mL bulked samples for rainfall chemistry analysis (each bottle representing 6.6 mm of rainfall). Eq. (4) can be used to estimate the atmospheric deposition (La) of either TN or TP, where the mean concentration of each nutrient in the rainfall samples is given by CP (Values are shown in Table 2 below). The rainfall over the study period was P and Ac is the total catchment area (440 ha).

below in Section 3.8. If the 7 day period contained an event at the start or end of the week then the period was shortened or lengthened so that the event was contained in one period. Some weeks were missing due to insufficient sampling. In 2010, weekly DLd2010 for TP and TN can also be estimated using Eq. (6). The weekly DU load estimates (LDUi, where i denotes the week number) contain additional uncertainty due to measurement errors described above, and these bounds can be used to estimate the corresponding uncertainty bounds for DLdi. For weeks with missing data, mean concentrations from the baseflow samples at DU (there were no events in these weeks) and all the samples at DL were used in Eq. (5) to estimate DD. Over the 2010 monitoring period DLd2010 can be expressed as a percentage P removal, R, of the incoming weekly DU loads LDUi,, representing the net removal of nutrients by the dam in Eq. (7)

La ¼ C P PAc

R ¼ 100  DLd2010

ð4Þ

2.6.1. Limitations of analysis Sufficient dissolved nutrient data for analysis using the methods above were only available from the three monitoring sites from some of the 2010 events (summarised in Section 3.2 below), so it was not possible to estimate annual dissolved nutrient loads. However, the ratios of dissolved to total nutrient concentrations can be used to estimate these loads over the entire period at each site for each species. In 2011 only four events at DU were sampled for TP, TN and TSS and it was not possible to estimate the total load over a longer period of time during 2011 (several large events were not sampled due to equipment problems). 2.7. Dam water and nutrient balance To enable an estimate of the nutrient loads entering and leaving the farm dam, and the residence time in the impoundment, a water balance was first performed. The physical characteristics of the dam are shown in Table 5 in Section 3.8. A weekly water balance incorporating stream inflows, precipitation, evaporation, pumping, overflows and storage changes was conducted. This showed that the precipitation, evaporation, pumping and storage changes were all small (<1% of the water balance) and thus the dam was assumed to be at steady state with stream inflows equalling overflows. This simplified the dam nutrient balance calculations that follow and accounts for periods of missing data. The weekly time step also allows for the different sampling intervals during 2010 and 2011 at two different inflow and one outflow monitoring sites (see Section 3.8) and accounts for the lagging effect of the dam on the streamflow. For a constant flow rate, the proportional change in nutrient load through the dam (DD) equates to the proportional change in nutrient concentration between the outflow (CDL) and inflow (CIN) and is given on a weekly basis by Eq. (5)

DD ¼ ðC DL  C IN Þ=C IN

ð5Þ

Also on a weekly basis the change in load across the dam DLd is given by

DLd ¼ LDU  DD

ð6Þ

It should be noted that DD has a lower bound of 1. The two sampling sites immediately upstream of the dam DU and DS were used to provide CIN values. There were insufficient DU samples collected in 2011 to allow weekly average concentrations to be estimated (some weeks had no sampling), and samples from the DS site were used instead to estimate DD using Eq. (5), by assembling a time series of CIN and CDL values. A weekly arithmetic average concentration was calculated from all water quality samples collected in each 7 day period, and the results are discussed

.X

LDUi

ð7Þ

3. Results and discussion 3.1. Hydroclimatic conditions The daily rainfall (recorded at the catchment raingauge) and runoff for the UPH site between May 2010 and November 2011 are shown in Fig. 3 with runoff events indicated by . Over this period, it was found that the AWS gauge recorded 1908 mm versus 2301 mm at the ULL raingauge. The AWS rain gauge was calibrated, and also compared against a co-located collecting gauge, and the Melbourne Water (ULL) gauge was operated according to their hydrometric measurement protocols. It is clear from this that there is considerable spatial variability in rainfall in the study area. The ULL gauge is also around 30 m higher in elevation than the AWS and situated on a more exposed ridgeline, which may partly explain this difference. It should be noted that the runoff coefficient for this period was 0.53 (based on AWS rainfall). The daily mean UPH flow was 7.7 ML/day, equivalent to 1013 mm over the period 1/5/2010 to 30/11/2011. Between May and December 2010, 49% of the UPH runoff of 476 mm was in events. During the same period the DU runoff was 427 mm and the UDP runoff only 320 mm. 3.2. 2010 Individual events at all sites In hydrological terms, runoff at all 3 sites was extremely variable between events. Based on the AWS rainfall (which as mentioned above is probably an underestimate of the true catchment rainfall particularly for the DU subcatchment), runoff coefficients varied over the course of 2010 from 0 to >1 and tended to increase during wet periods between late autumn and early spring as the catchment wetted up. Runoff coefficients greater than 1 probably represent events where higher rainfall occurred over the catchment than measured at AWS and/or where there was significant antecedent flow. If data from the nearby ULL raingauge is used (not shown here for brevity), the runoff coefficients are less than 1. Runoff from the monitored events at UDP was lower than the other two catchments for all except one (Event 3). The runoff from individual events at the three subcatchments is summarised in the Supplementary Material (Table S2). The mean (standard deviation in parentheses) event runoff in 2010 was 10.4 (9.80) mm, 12.4 (8.89) mm and 13.2 (8.86) mm, respectively at DU, UDP and UPH. It should be noted that in Table 2, the mean concentration of TP and TN from DU events appears to be similar to UDP events, however, these statistics include some DU 2011 events where much higher concentrations of TP and TN relative to TN and TP in the 2010 events were measured at DU (Section 3.3). Values of both CFWM (TN) and CFWM (TP) measured at UDP were nearly twice as high as

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

173

Fig. 3. Daily rainfall and runoff at UPH between May 2010 and November 2011(events denoted by black crosses).

those measured at DU in some events (3 and 14). These observations are supported by Fig. 4 (left pane), which shows a summary of the measured concentrations of TP and TN classified into baseflow and event samples using box-whisker plots. There were insufficient track samples to create these plots, so concentrations are summarised in Table 2 instead. The right pane of Fig. 4 depicts the yields calculated from up to 20 individual events at each site (including track loads – UPH (Tracks)) using box-whisker plots, these will be discussed in more detail below in Section 3.4. Uncertainty bounds were calculated for individual events at the four sites. The uncertainty bounds associated with the UPH track yield were quite large compared to yields estimated from events at the three subcatchment monitoring sites, reflecting the large range of concentration values in the track samples.

some of the highest event concentrations observed at DU for both nutrients. A CFWM (TN) of 14.9 mg/L and CFWM (TP) of 4 mg/L was calculated from medium-sized Event 26 in July 2011. Later in the year when Event 27 was observed, concentrations of both TN and TP decreased, probably due to the wet spring experienced by the catchment in 2011, and associated dilution effects caused by increased saturated areas close to the stream. Antecedent runoff over the previous 14 days was high, up to 52 mm (Event 27b). The higher nutrient concentrations at DU in 2011 compared with 2010 were probably due to the samples being collected at higher flows only (due to the repositioning of the autosampler trigger switch).

3.3. 2011 Events monitored at DU

The monitored runoff between 1st January and 30th April 2010 was only 17 mm (out of 493 mm i.e. 3.4% of total flow, indicating ephemerality). Therefore, nearly all the annual TN and TP loads will be included in these estimates, so subsequently these will be

In 2011, a subset of events was monitored at DU only. These events occurred during the winter and spring periods and recorded

3.4. 2010 Catchment loads

Fig. 4. Left plot panes show a box-whisker plots of TP and TN concentrations measured at the three monitoring sites during events and baseflow periods. Right plot panes show a box whisker plots of the mean 2010 event TP and TN yields estimated from the same sites and UPH (Tracks).

174

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

referred to as ‘‘2010 catchment loads’’. The results of the catchment loads for 2010 are shown in the right pane of Fig. 4 above, in terms of the TN and TP yields (Load/Subcatchment area) from DU, UDP, UPH and UPH(Tracks). Loads from UPH (Tracks) LTL were calculated using the methodology described above (Eq. (3)) and converted to yields by dividing by the area of tracks in UPH. Subcatchment yields (kg/ha) from UPH in Fig. 4 were clearly much higher (2–3) than both UDP and DU for both TN and TP based on the above event analysis. The TN and TP yields from the UPH (Tracks) were larger than the UPH subcatchment yields reflecting the high concentration of TN and TP observed in track runoff samples (Table 3). Observations during storm events in 2009 showed that some of the runoff conveyed by Track 1 originated from the fields which became hydraulically connected to the track gutters, however (visually) the runoff from the lower part of the track (near TL in Fig. 2) was conveying a greater sediment load than the field runoff. Calculating the loads at UPH allows an estimate of the total yield of the catchment to be made for comparison with other studies. The results from the 2010 load analysis (total and event) with associated uncertainty bounds are presented in Table 3 (as yields). The subcatchment contribution from DU can also be calculated, and based on the data in Fig. 4 was, in terms of yield, much lower than the entire catchment. At UPH and DU in 2010 sufficient data were collected to enable a breakdown of loads into baseflow and event loads (using Eq. (4)). The layout of the catchment (Fig. 2) ensures that the dam (DL) outflow combines with the main streamflow just downstream of UDP, where some event data were available to estimate the CFWM (Table 2) and yields from four 2010 events (the range is shown in Fig. 4). The main points are as follows. At UPH the ratio of event yield: baseflow yield for TN was only 1.1 (uncertainty range 0.75–1.7), indicating that a significant quantity of TN was exported from the catchment during baseflow periods (Table 3) largely due to dissolved N (ammonium plus NOx) loadings. The ratio of event runoff to inter-event runoff was 0.95 at UPH, which is close to the TN load ratio, indicating similar concentrations in baseflow and during events, on average. At DU the ratio for TN was also low (1.7; range 1.2–2.2) indicating a significant export of TN during baseflow periods from the headwater subcatchment also. The ratio of event runoff to inter-event runoff was 0.74 at DU, which was surprisingly low, that is less than half the TN load ratio, reflecting a significant increase in TN concentration during events at DU. At both DU and UPH the ratio of event yield : baseflow yield for TP was >3, much higher than for TN, indicating that events are much more important for exporting TP than TN,

which corresponds to the findings of similar studies where both TP and TN have been evaluated (e.g. Holz, 2010). These results indicate significant variation in the relative importance of different flow pathways in transporting components of TP and TN. The proportion of the TP load transported by events at both sites was similar. Total track loads comprised only 2.0% and 2.1% respectively of the monitored event loads of TN and TP at UPH, based on concentrations in samples taken at TL and AR. This may be due to: (i) the fairly small area of tracks in the UPH subcatchment; (ii) the fact that some of the badly contaminated areas of the farm on the western side of the stream and near the stream on the east side were not sampled due to logistical constraints; and (iii) the high runoff coefficient of the study area. There appears to be a major nutrient source in the UPH subcatchment that has not been captured by our track load estimates. DU subcatchment had yields of N and P of 12 and 1.5 kg/ha respectively (Table 3). Given the farm layout this yield is mainly from paddocks. Fig 5 shows the flow-concentration relationships at DU and UDP are very similar, allowing an assumption of similar yields between these two subcatchments. The ‘‘missing’’ loads can then be estimated by scaling up DU and subtracting the combined loads from UDP, DU and the tracks from the UPH load in Table 3, giving 8000 kg TN and 1300 kg TP. That is the estimated TN and TP load increased by 190 and 240% in the UPH subcatchment, which is only 15% of the area. Similarly, measured and estimated (DL outflow and UDP flow) inflows to the UPH subcatchment account for 1350 ML of the 2090 ML of runoff, a 55% increase. Assuming that the expected runoff from UPH subcatchment would total 270 ML (from scaling the runoff measured at DU by the ratio of the subcatchment areas), there is an unaccounted for inflow of 470 ML, which must be a sum of groundwater inflows (from upper to lower subcatchments) and errors (either under-estimated inflows and/or over-estimated outflows) in the runoff data. It is likely there is some bias in the flow measurements as the 55% gain in flow seems implausibly large from 15% of the catchment area. Further, in 2010, the measured runoff from the whole catchment was 470 mm, from DU was 420 mm (with leakage problems around this weir – see Supplementary material), but from UDP it was only 320 mm. The UDP flow seems quite low and the weir was located in an area with alluvial fill and there could have been some flow bypassing the weir through the subsurface. The flow weighted mean concentration is insensitive to consistent flow measurement bias and these increase by 120% and 158% for TN and TP, respectively between the DU and UPH stations, providing strong evidence for a major source between these stations. Field observations would suggest

Table 3 Summary of nutrient loads and fluxes 1/5/2010 to 31/12/2010 (range in parentheses). Load (kg) TN

UPH (all) UPH (Tracks)a DU ‘‘Missing’’ at UPH Atmospheric (catchment) Farm dam (net removal) Farm dam R (%) a

Yield (kg/ha) TN

Yield (kg/ha) TP

Total

Events

Total

Events

Total

Events

Total

Events

12,700 (10,300–15700) 127 (64.1–264)

6740 (5770–8150)

1920 (1390–2640)

1430 (1040–1970)

15.3 (13.1–18.5)

911 (747–1110) 1300 141 118 (81.7–189.6) 70.4 (67.1–72.5)

4.37 (3.15–6.00) 14.2 (8.07–27.6) 1.55 (1.04 –2.55)

3.2 (2.4–4.5)

1440 (1170–1750) 8000 2200 148 (112–181)

28.9 (23.5–35.7) 63.3 (21.3–87.9) 12.0 (9.73 –14.6)

11.2 (10.1–11.8)

Load (kg) TP

All loads and yields from UPH (Tracks) occurred during Events.

28.5 (12.3–83.6) 186 (126–305)

163 (108–277)

5.0 1.15 (0.87–1.41)

0.32 0.91 (0.63–1.47)

7.60 (6.23–9.22)

1.36 (0.90–2.31)

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

175

Fig. 5. Concentration (C) vs. Runoff (Q) plots at monitoring sites (log scales).

that these loads come from a combination of tracks connecting farm buildings, areas where cows stand for a significant period and the dairy shed complex. Further analysis showed that loads estimated from events 14 and 15, where monitoring data were available at all three sites, indicated that the UPH subcatchment generated between 41% and 43% of the TP load and 43% to 45% of the TN load at the catchment outlet (Table 3), from only 14% of the total catchment area (Table 1). Runoff from the UPH subcatchment comprised 15% of the total runoff from these events. The track loads (from Track 1 data; about 2 kg of P and 9 kg of N per event) are insignificant compared to the total UPH loads; therefore it is assumed that the dairy operations (including very intensively used track areas close to the dairy shed) are contributing a major point source of nutrients to the PEC catchment.

3.5. C–Q relationships at each site Fig. 5 shows the C–Q relationships for TN and TP at all three sites. Additional dissolved nutrient C-Q plots are shown for UPH, for NOx, FRP and NH4. These plots can be used to identify possible flow pathways and support the discussion of relative importance of event and baseflow loads at both DU and UDP, described above. The C–Q plots in Fig. 5 for UDP and DU are similar for both TN and TP. In contrast, the relationships at UPH are quite different from the upstream sites and more complex. A more detailed analysis of the UPH dataset is therefore required in order to provide a satisfactory explanation (Section 3.9.2). The weak relationship of both TN and TP with Q during baseflow periods allowed baseflow loads to be calculated using a constant value of Cb (using Eq. (3) with values from Table 3) at DU and UDP (constant high and low baseflow values were used at UPH). These are shown as yields (loads per unit area) in Table 3 and are indicated by the (E) after the site name in the box-whisker plots (Fig.4 left hand pane). A lack of correlation between C and Q may indicate a constant source of TN and TP, e.g. a shallow groundwater discharge into the stream and/or a point source. The increase in C at low baseflow conditions between the upstream and

downstream sites indicates that there must also be a constant source in the reach, which could be either in-stream, such as the mobilisation of nutrients from the streambed, or an inflow. Moreover, in terms of similarity, DU and UDP subcatchments appear to be functioning similarly in terms of their behaviour as typical headwater sources of N and P, where events export the vast majority of the annual nutrient loads (especially P), in contrast to UPH where nutrient export is much higher and events are less strongly dominant in the overall export particularly of N. Table 3 shows the mean event TP and TN C statistics from events at both subcatchments, that support this finding.

3.6. Dissolved nutrients The relationships between dissolved nutrient C and Q at both UDP and DU were weak and are not shown graphically. However, at UPH interesting relationships were identified, with both ammonium and FRP showing a dilution effect at high flows (Fig. 5 lower panes). However, the relationship between C(NOx) and Q (Fig. 5 lower panes) was quite complex, with the highest concentrations observed during high baseflow and low event (around 0.1 mm/h) conditions, which could be a result of more active leaching from a high concentration source. At DU and UDP, NOx concentrations were highest during baseflow and diluted at higher flows (Table 3). Shallow groundwater may be an important source of NOx for these catchments (one sample collected from a bedrock seep at DU measured a concentration of NOx of 17.9 mg/L in September 2011). The behaviour of NOx at DU and UDP matches other studies, e.g. Poor and McDonnell (2007) that found nitrate (from their agricultural subcatchment) tends to decrease during runoff events due to a dilution effect from surface runoff containing lower concentration water than the baseflow (during certain periods of the year). This did not appear to be the case at UPH except for the largest events. Ammonium concentrations at UPH decreased significantly during events, while increasing slightly at DU and UDP (Table 2). In UDP and DU, increased surface runoff during events could be transporting ammonium originating from animal urine from the

176

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

catchments into the watercourses. Holz (2010) reported elevated ammonium concentrations in a 12 ha grazed subcatchment in NW Tasmania, following grazing events. FRP concentrations did not differ significantly between event and baseflow conditions at DU and UDP, and were relatively low at both sites (0.02 mg/L and 0.04 mg/L respectively). FRP was much more variable between events than in baseflow. At UPH, FRP was highest and most variable at high baseflow and for small events and it diluted significantly at higher flows. Dissolved nutrient measurements from track samples (Table 2) were quite sparse, partly because the grab samples were extremely difficult to filter arising from the high particulate concentration. Due to runoff entraining both solid and liquid animal excreta (Edwards et al., 2008), concentrations were elevated compared to the stream samples for both NOx (maximum 21 mg/L) and FRP (maximum 1.6 mg/L), while ammonium concentrations (1.3– 1.7 mg/L) were comparable to UPH concentrations. 3.7. Sediments In total there were insufficient TSS samples (36) collected to enable event or annual estimates of the TSS export load from the PECs to be computed. However, an analysis of the variability of TSS concentrations from the different sites at different times (i.e. comparing event vs. baseflow samples), shown in Table 4 helps explain the dynamics of particulate forms of N and P. TSS samples were collected from both dam inflow and outflow monitoring sites (DU and DL) so that the sediment attenuation capacity of the farm dam could be (qualitatively) estimated. Sufficient TSS data was collected at DU to enable a sediment rating curve (C–Q relationship), where Q is in L/s and CTSS is the TSS concentration (Eq. (8)) and a strong relationship between TP (CTP) and TSS concentration to be identified (Eq. (9)) with a correlation between CTSS and CTP of 0.99.

C TSS ¼ 39:41Q 0:8645 C TSS ¼ 1321:6C TP

R2 ¼ 0:66

ð8Þ

R2 ¼ 0:99

ð9Þ

The role of sediments in exporting particulate forms of P has been discussed above (Section 3.4), and it is apparent that the DU subcatchment is exporting both TSS and particulate P. The

Table 4 Summary of TSS Data (all Concentrations in mg/L) (N/A indicates that no samples were available for calculating statistic).

Median (Grab) Median (Event) Max (Grab) Max (Event) n

UPH

UDP

DU

DL

Tracks

136.7 224 746.7 504 11

121.3 N/A 702.9 N/A 10

68.9 1455 593.4 8164 15

11.5 N/A 92.5 N/A 4

N/A 1835 N/A 6990 4

Table 5 Farm dam properties. Property

Units

Value

Mean catchment inflow (DU) Net precipitation Pumping rate Dam surface area Residence time 2010 Removal of TN (DU data) 2010 Removal of TP (DU data) 2011 Removal of TN (DS data) 2011 Removal of TP (DS data)

(ML/day) (ML/day) (ML/day) m2 Days (min, median) % Weekly range % Weekly range % Weekly range % Weekly range

1.9 0.01 0.025 5870 0.52, 8.6 65 to +33 124 to +87 13 to +44 83 to +56

correlations between TSS and other nutrients were variable, in decreasing order: TN (0.99); Ammonium (0.94); FRP (0.58); NOx (0.19), indicating that the flow pathways transporting the latter were unlikely to be similar to those transporting sediments. The high correlation between TSS and TN may be due to generally high concentrations of organic and particulate N during events (results indicated that ammonium plus NOx comprised between 10% and 50% of TN during events based on the FWM concentrations at DU and UPH). 3.8. Influence of farm dam Fig. 6 (upper) compares the concentrations of TP and TN collected at all three sites close to the dam (DU, DS and DL) in 2010–2011. In 2011 sampling at DS took place at regular 12 h intervals, whereas at DU event and routine grab sampling (usually bi-weekly in 2011) were carried out. Unfortunately, equipment problems meant that most of the four events monitored at DU in 2011 were not monitored at DS, so it was not possible to compare event samples for these two sites upstream of the dam. However, it is possible to compare baseflow samples and they were similar for both TP and TN, where the samples were taken within a few hours of each other. Fig. 6 (upper) indicates that weekly TN concentrations at DS and DL in 2011 were very similar, except for some high flow periods where the DS concentrations were higher than at DL (e.g. February and late-April). Therefore, the dam acted as a sink of TN during high flow weeks but during lower flow periods it was a sink at some times and a source at others. During events the reduction in TP through the dam was similar to TN, with lower concentrations measured at DL than at DS (in February and April 2011). Periods when the dam was a source of P tended to coincide with low TP concentrations in baseflow entering the dam at DS, e.g. March, early May and late June 2011. Measurements of TSS in the dam outfall (DL) indicated that concentrations of TSS were probably reduced due to deposition in the dam itself (DL range 2–95 mg/L, n = 4; Table 4, with too few measurements to assess the statistical validity of this finding). A discussion of the rate of sedimentation in the dam is outside the scope of this study. A weekly time series of DD values (Fig. 6 (lower)) was computed using Eq. (5) and summary statistics are given in Table 5. From Eq. (5), a positive value of DD indicates the dam was a source, while a negative value of DD indicates that the dam was a sink, during that particular week. The residence time (Tr) was estimated from the total storage and inflow (QDU) data and was highly variable, with a median of 8.6 days, decreasing to less than 1 day during runoff events (Table 5). Fig. 6 (lower) shows the relationship between dam inflow/volume (i.e. 1/Tr) vs. DD values for TN and TP in both years. At low inflows, values of DD for TN were less variable in magnitude and closer to zero on average than for TP implying that there was less net gain or removal of N than of P. Nevertheless input and output loads of TN often differed by more than 20%. For TP there were 14 out of 48 weeks where more than 50% on the influent TP was removed. Both positive and negative values of DD exist for both TN and TP, indicating the dam can act as both a source and sink in the short term, particularly at low inflows. Fig. 6 (lower) shows that during weeks containing events (i.e. high inflow/volume), the dam was often removing a large percentage of the TP inflow load (DD approaching the lower bound of 1). Removal of TN was less pronounced during events than TP (DD was between 0.5 and 0). In terms of loads, values of DLd2010 were calculated using Eq. (6) above in 2010 only, using weekly values of LDU as the influent loads to the dam (Table 3). The overall impact on the catchment nutrient budget was to remove 70.4% (67.1–72.5%) of the influent TP from

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

177

Fig. 6. Upper plot shows a weekly time series of TN and TP concentrations at all three monitoring sites around the farm dam (log y scale). Lower plot shows the relationship between proportional change in nutrient concentration through the dam (DD) and ratio of Inflow:Volume (the inverse of residence time) for TN (left) and TP (right). Negative DD values indicate removal by the dam.

the DU branch of the stream network in 2010. The impact on N was less variable, however the dam was sometimes a source of TN, as an increase in TN downstream of the dam was observed during some weeks. The net removal of TN loads by the dam was 11.2% (10.1–11.8%). In Table 5 using the DU nutrient load data calculated above, a comparison can be made between the removal by the dam and the DU subcatchment loads over the May–December 2010 period. Lastly, these results support the hypothesis that the dam mitigated nutrients (especially P) in the stream below DU. 3.9. Summary and comparison with other studies The results imply that transport of particulates during events is important for transporting P, and is a factor in the dominance of events in the TP loads. For TN it may be that N loads are a combination primarily of organic forms during events and mineral (nitrate leaching) forms of N from soils during high baseflow which may explain the more muted role of events in N transport. The following discussion places the subcatchment yields in the context of other similar studies worldwide. The catchment TP and TN yields (for UPH and DU, in 2010) are shown in Table 3 along with an estimate of the missing portion (i.e. not captured by the monitoring – Section 3.4) of the TP and TN loads. Clearly, yields from DU were lower than from UPH for the reasons discussed above. Drewry et al. (2006) reported a range of 7–9 kg N/ha/year (TN) exported from catchment scale Australian studies comprising a mix of dryland and irrigated dairying (exports from plot to field scale experiments were higher than these figures). They reported that annual TP export varied between 0.8 and 1.04 kg P/ha/year (from the same catchments). The exports of TP (1.6 kg/ha) and TN (12 kg/ha) (Table 3) from DU (and, by inference from similar average concentrations, UDP; Fig. 4 left panes) are higher than these catchment scale studies. However, Wilcock et al. (2007) reported that the 15 km2 Toenepi catchment in New Zealand exported 35.0 kg N/ha/year and 1.16 kg P/ha/year, in 1995–1997, indicating that high nitrogen export rates from dairy catchments are not uncommon. Müller et al. (2010) reported that over the most recent monitoring period at Toenepi, TP exports were

0.6 kg/ha/year, indicating a decline since the mid-1990s due to a combination of a drier climate and the successful introduction of BMPs (Wilcock et al., 2007). Sharpley et al. (2008) reported that the small FD36 research catchment in Pennsylvania (USA) exported 0.64 kg P/ha/year over a decade of monitoring and 80% of TP was exported during storm events. Dissolved P (in this study FRP measured by 0.2 lm filtration) exports were not as high as might be expected from other studies in the region, e.g. Nash and Murdoch (1997), although their study was carried out at the paddock not the catchment scale where attenuation of dissolved P could occur. The dominant form of P exported from the catchment was particulate, which is similar to the findings of Sharpley et al. (2008) and others. The exports from the PECs (at UPH) of both TP and TN (Table 3) are much higher than other catchment scale studies. At all three sites the ratio between TN and TP (Table 3) decreased during runoff events compared with baseflow, by as much as a factor of 5 (DU). This was also observed in the Rhode River Watershed (USA) by Correll et al. (1999), in monitoring data from their mixed use and to a lesser extent, cropland watersheds. This presumably relates to the much stronger dominance of surface flow pathways for TP export than for TN export. Deasy et al. (2008) reported TP concentrations from sampling tracks (no loads were given) in the Sem catchment ranging from 0.79 to 3.7 mg/L P (mean 1.57 mg/L P), however other (UK) studies (reviewed by Deasy et al. (2008)) had in fact recorded higher concentrations of P in track runoff. Monaghan and Smith (2012) measured TP concentrations from the tracks in Bog Burn catchment varying from 7.6 to 16.3 mg/L (seasonal means), and TSN (total soluble N) concentrations varying from 6.5 to 22.9 mg/L. In comparison, the median TP and TN concentrations in this study from Track 1 were respectively 5.55 mg/L P and 24.7 mg/L N. Edwards et al. (2008) measured concentrations of 100 and 5 mg/L of N and P respectively from areas of farms such as tracks and hard standings with substantial manure loadings from stock traffic, which indicated similar P concentrations but higher N concentrations compared to this study. Monaghan and Smith (2012) estimated track yields (from 2770 m2 of tracks), expressed as the percentage of

178

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

total catchment yields of 12% and 1% of TP and TN respectively, compared to estimates of 2% for both TP and TN from this study (which assumed a higher runoff coefficient from the tracks). Farm tracks in the PECs exported higher yields (based on elevated concentrations) of other dissolved species of N and P in 2009–2011 than the subcatchments themselves. 3.9.1. Sediments and their relationship with nutrients Results from this study indicated that TP is transported, particularly from the DU subcatchment, in a particulate form attached to sediments mostly during runoff events (Section 3.4), a finding similar to many other studies (e.g. Heathwaite et al., 2005a; Sharpley et al., 2008). FRP concentrations from this study were obtained from 0.2 lm filtration which compared to other studies using 0.45 lm filtration may measure lower concentrations due to the removal of P attached to colloids and bacteria (Heathwaite et al., 2005b). No other studies in Australia have used this filter size so comparisons with other studies are difficult. Sediment loads could not be estimated at the three monitoring sites due to insufficient TSS data; however, a sediment rating curve established using data from DU (Eq. (7)) indicated high sediment concentrations were associated with high flows during events (TSS event C > 1000 mg/L). High concentrations of TP were also associated with high concentrations of TSS (Eq (8)) leading to a high export of TP (DU TP subcatchment yield – Table 3=1.6 (range 1.0–2.5) kg P/ha/year). Monaghan and Smith (2012) also measured TSS concentrations in excess of 1000 mg/L from tracks in the Bog Burn catchment in summer and autumn. Furthermore, there was a strong correlation between TP and TSS concentrations at DU (Eq. (8)). FRP and TSS were not correlated. This suggests that particulate forms of P are attached to soil, and washed into the streams by high-intensity rainfall episodes, as found in other studies (Haygarth et al., 2004, 2005b; Heathwaite et al., 2005a; Sharpley et al., 2008). These results indicate that at DU particulate P is more important than FRP, in contrast to the findings of Nash and Murdoch (1997). 3.9.2. Relative sources and sinks Few studies provide an estimate of the impact of small impoundments on nutrient fluxes (Brainwood et al., 2004), so this study helps to fill this knowledge gap. In the PEC, scaling up the percentage removal figures above over the entire monitored period in 2010, the farm dam on the DU branch removed 118 (82–190) kg of P in 2010 and 148 (112–181) kg of N in 2010, the uncertainty originating from that in the DU load values. The impact of the dam was much larger on P than on N in terms of yield. The removal rates were of a similar order of magnitude from the load estimated to be exported from the monitored tracks in the UPH subcatchment during events when they become hydraulically connected to the streams. The total UPH (Tracks) event loads were estimated as 29 kg (12–84 kg) P and 126 kg (64–264 kg N)); however, comparison of the total UPH loads with loads estimated by scaling up the yields observed at DU cast some doubt on our track load estimates, due to a ‘‘missing’’ load component discussed above in Section 3.4. The concentration data (discussed above in Section 3.5) from samples at UDP, DU and UPH support the findings that there is a major source of pollution probably originating from at or near the farm’s dairy shed operations. In terms of potential mitigation measures, attenuating and/or storing runoff from the CSAs, identified as tracks and farmyard (dairy) areas, and using BMPs such as (riparian) buffer zones (e.g. Wilcock et al., 2007; Monaghan and Smith, 2012) to attenuate particulate forms of N and P would be desirable for the overall catchment health of the Lang Lang River and the receiving water body, Western Port Bay. Atmospheric loads were insignificant (Table 3) compared to other sources in the catchment. The atmospheric N load of

5 kg N/ha/year was at the lower range of the values in Eckard et al. (2007) of 4.2-7.9 kg N/ha/year. This range was similar to 5.3 kg N/ha/year figure for the entire Richmond River catchment estimated by McKee and Eyre (2000). Our estimate of the atmospheric P load of 0.32 kg P/ha/year at Poowong was similar to the Richmond River catchment average (0.36 kg P/ha/year) in that same study. These are very small compared with typical fertilizer application rates in the Gippsland region: 200 kg N/ha/year (Eckard et al., 2007) and 44 kg P/ha/year (Nash and Halliwell, 1999). 4. Conclusions In summary, the sediment and nutrient monitoring undertaken on a dairy farm in Poowong East, Victoria has indicated that the PEC catchment containing the farm exported high loads of N and P. Events were important in exporting loads as concentrations increased (especially for TP) and they accounted for over half the runoff, probably due to both markedly increased hydrologic connectivity in general during the runoff season, particularly during events, and to critical source areas such as tracks being connected to watercourses during events. Exports of both dissolved and particulate nutrients from the UPH subcatchment that contains the farm buildings, with associated intensively used milking facilities, stock access tracks, effluent treatment and overflow systems, were much higher than other areas of the catchment for reasons that were not entirely quantifiable but appeared to be related to effluent overflow and track runoff. The track monitoring did not capture the entire load associated with the dairy operations. That additional load was estimated to be 1200 kg P and 8100 kg N in 2010, which was over 50% of the total N and P load exported from the entire catchment. Dissolved sources of N, comprising both nitrate and ammonium as well as organic N were also exported in large amounts, while dissolved P exports were minor compared with TP exports. The farm dam did not remove a significant amount of N from the stream draining the DU subcatchment. However, the removal of P mostly in the form of particulate attached P during events was very significant, accounting for around 2/3 of the estimated influent load to the dam. The layout of the dairy farm in this study and its proximity to watercourses would suggest that implementing BMPs to reduce N and P loadings to the catchment would be difficult given that over 50% of the load originated from this locality. In general, siting dairy operations away from watercourses would be preferable. Given the nature of the catchment, it is likely that different flow pathways from the source to the waterways probably exist both for (a) baseflow loads compared to event loads (b) dissolved forms of N and P compared to sediment-attached particulate forms. Acknowledgements The project was funded by the Australian Research Council (grant DP987738). The farmers, Peter and Wilma Mackay and their colleagues, are thanked for providing access to their farm for the study. Many people helped in the field, especially Rodger Young, Lucas Dowell, Olaf Klimczak and Jacqui Lloyd. Dr. Jim Freer (University of Bristol), assisted with the preparation of the data used in Figs. 1 and 2. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jhydrol. 2014.04.034.

R. Adams et al. / Journal of Hydrology 515 (2014) 166–179

References Adams, R., Dowell, L., Lloyd, J.E., Weatherley, A.J., Grace, M.R., Western, A.W., 2010. The effect of engineered features on sources and sinks of nutrients on a dairy farm in south-eastern Australia. In: Kirkby, C. (Ed.), BHS 3rd International Symposoium, Managing Consequences of a Changing Global Environment, BHS, Newcastle Upon Tyne, UK, pp. 1–6. APHA, 2005. Standard Methods for the Examination of Water and Wastewater, 21st ed. American Public Health Association, New York. USA. Barlow, K., Nash, D.M., Grayson, R.B., 2005. Phosphorus export at the paddock, farmsection and whole farm scale on an irrigated dairy farm in south-eastern Australia. Aust. J. Agric. Res. 56, 1–9. http://dx.doi.org/10.1071/AR04166. Barlow, K., Nash, D.M., Hannah, M., Robertson, F., 2007. The effect of fertiliser and grazing on nitrogen export in surface runoff from rain-fed and irrigated pastures in south-eastern Australia. Nutr. Cycl. Agroecosyst. 77, 69–82. http:// dx.doi.org/10.1007/s10705-006-9046-6. BOM, 2005. Australian Climate Averages – Evapotranspiration (Climatology 1961– 1990). (Last Accessed 02.04.12). BOM, 2009. Australian Climate Averages – Rainfall (Climatology 1961–1990). (Last Accessed 02.04.12). Brainwood, M.A., Burgin, S., Maheshwari, B., 2004. Temporal variations in water quality of farm dams: impacts of land use and water sources. Agric. Water Manage. 70, 151–175. http://dx.doi.org/10.1016/j.agwat.2004.03.006. Correll, D.L., Jordan, T.E., Weller, D.E., 1999. Transport of nitrogen and phosphorus from Rhode River watersheds during storm events. Water Resour. Res. 35 (8), 2513–2521. Deasy, C., Heathwaite, A.L., Brazier, R.E., 2008. A field methodology for quantifying phosphorus transfer and delivery to streams in first order agricultural catchments. J. Hydrol. 350, 329–338. http://dx.doi.org/10.1016/ j.jhydrol.2007.10.037. Deasy, C., Brazier, R.E., Heathwaite, A.L., Hodgkinson, R., 2009. Pathways of runoff and sediment transfer in small agricultural catchments. Hydrol. Process. 23, 1349–1358. http://dx.doi.org/10.1002/hyp.7257. Deasy, C., Baxendale, S.A., Heathwaite, A.L., Ridall, G., Hodgkinson, R., Brazier, R.E., 2011. Advancing understanding of runoff and sediment transfers in agricultural catchments through simultaneous observations across scales. Earth Surf. Proc. Land. 36, 1749–1760. http://dx.doi.org/10.1002/esp.2197. Dougherty, W.J., Fleming, N.K., Cox, J.W., Chittleborough, D.J., 2004. Phosphorus transfer in surface runoff from intensive pasture systems at various scales: a review. J. Environ. Qual. 33 (6), 1973–1988. http://dx.doi.org/10.2134/ jeq2004.1973. Drewry, J.J., Newham, L.T.H., Greene, R.S.B., Jakeman, A.J., Croke, B.F.W., 2006. A review of nitrogen and phosphorus export to waterways: context for catchment modelling. Mar. Freshw. Res. 57, 757–774. http://dx.doi.org/10.1071/MF05166. DSE, 2008. Vicmap Elevation DTM20m and DTM10m Product Description V1.24pp. State Government of Victoria, Melbourne, Australia. Eckard, R.J., White, R.E., Edis, R.S., Smith, A., Chapman, D.F., 2004. Nitrate leaching from temperate perennial pastures grazed by dairy cows in south-eastern Australia. Aust. J. Agric. Res. 55, 911–920. http://dx.doi.org/10.1071/SR96039. Eckard, R.E., Chapman, D.F., White, R.E., 2007. Nitrogen balances in temperate perennial grass and clover dairy pastures in south-eastern Australia. Aust. J. Agric. Res. 58, 1167–1173. http://dx.doi.org/10.1071/AR07022. Edwards, A.C., Kay, D., McDonald, A.T., Francis, C., Watkins, J., Wilkinson, J.R., Wyer, M.D., 2008. Farmyards, an overlooked source for highly contaminated runoff. J. Environ. Manage. 87, 551–559. http://dx.doi.org/10.1016/ j.jenvman,2006.06.027. Fairchild, G.W., Velinsky, D.J., 2006. Effects of small ponds on stream water chemistry. Lake Reserv. Manage. 22 (4), 321–330. http://dx.doi.org/10.1080/ 07438140609354366. Fleming, N.K., Cox, J.W., 2001. Carbon and phosphorus losses from dairy pasture in South Australia. Aust. J. Soil Res. 39 (5), 969–978. http://dx.doi.org/10.1071/ SR00042. Fox, D.R., 2005. Protocols for the Optimal Measurement and Estimation of Nutrient Loads: Error Approximations, Australian Centre for Environmetrics, Melbourne, Victoria, Australia. Gruszowski, K.E., Foster, I.D.L., Lees, J.A., Charlesworth, S.M., 2003. Sediment sources and transport pathways in a rural catchment, Herefordshire, UK. Hydrol. Process. 17, 2665–2681. http://dx.doi.org/10.1002/hyp.1296. Harris, G.P., 2001. Biogeochemistry of nitrogen and phosphorus in Australian catchments, rivers and estuaries: effects of land use and flow regulation and comparisons with global patterns. Mar. Freshw. Res. 52, 139–149. http:// dx.doi.org/10.1071/MF00031. Haygarth, P., Turner, B.L., Fraser, A., Jarvis, S., Harrold, T., Nash, D., Halliwell, D., Page, T., Beven, K., 2004. Temporal variability in phosphorus transfers: classifying concentration-discharge event dynamics. Hydrol. Earth Syst. Sci. 8 (1), 88–97. Haygarth, P.M., Condron, L.M., Heathwaite, A.L., Turner, B.L., Harris, G.P., 2005a. The phosphorus transfer continuum: linking source to impact with an interdisciplinary and multi-scaled approach. Sci. Total Environ. 344, 5–14. http://dx.doi.org/10.1016/j.scitotenv.2005.02.001. Haygarth, P.M., Wood, F.L., Heathwaite, A.L., Butler, P.J., 2005b. Phosphorus dynamics observed through increasing scales in a nested headwater-to-river

179

channel study. Sci. Total Environ. 344, 83–106. http://dx.doi.org/10.1016/ j.scitotenv.2005.02.007. Heathwaite, A.L., Quinn, P.F., Hewett, C.J.M., 2005a. Modelling and managing critical source areas of diffuse pollution from agricultural land using flow connectivity simulation. J. Hydrol. 304, 446–461. http://dx.doi.org/10.1016/ j.jhydrol.2004.07.043. Heathwaite, A.L., Haygarth, P.M., Matthews, R., Preedy, N., Butler, P., 2005b. Evaluating colloidal phosphorus delivery to surface waters from diffuse agricultural sources. J. Environ. Qual. 34 (1), 287–298. http://dx.doi.org/ 10.2134/jeq2005.0287. Holz, G.K., 2010. Sources and processes of contaminant loss from an intensively grazed catchment inferred from patterns in discharge and concentration of thirteen analytes using high intensity sampling. J. Hydrol. 383, 194–208. http:// dx.doi.org/10.1016/j.jhydrol.2009.12.036. Isbell, R., 2002. The Australian Soil Classification. CSIRO Publishing, Collingwood, Victoria, Australia, 152pp. Johnes, P.J., 2007. Uncertainties in annual riverine phosphorus load estimation: Impact of load estimation methodology, sampling frequency, baseflow index and catchment population density. J. Hydrol. 332, 241–258. http://dx.doi.org/ 10.1016/j.jhydrol.2006.07.006. Kato, T., Kuroda, H., Nakasone, H., 2009. Runoff characteristics of nutrients from an agricultural watershed with intensive livestock production. J. Hydrol. 368, 79– 87. http://dx.doi.org/10.1016/j.hydrol.2009.01.028. Kotlash, A.R., Chessman, B.C., 1998. Effects of water sample preservation and storage on nitrogen and phosphorus determinations: implications for the use of automated sampling equipment. Water Res. 32 (12), 3731–3737. Lessard, J.L., Hayes, D.B., 2003. Effects of elevated water temperature on fish and macroinvertebrate communities below small dams. River Res. Appl. 19, 721– 732. http://dx.doi.org/10.1002/rra.713. Maxted, J.R., McCready, C.H., Scarsbrook, M.R., 2005. Effects of small ponds on stream water quality and macroinvertebrate communities. NZ J. Mar. Freshw. Res. 39 (5), 1069–1084. http://dx.doi.org/10.1080/00288330.2005.9517376. McKee, L.J., Eyre, B.D., 2000. Nitrogen and phosphorus budgets for the sub-tropical Richmond River catchment, Australia. Biogeochemistry 50, 207–239. http:// dx.doi.org/10.1023/A:1006391927371. Monaghan, R.M., Smith, L.C., 2012. Contaminant losses in overland flow from laneways in southern New Zealand. Agric. Ecosyst. Environ. 159, 170–175. http://dx.doi.org/10.1016/j.agee.2012.07.022. Müller, K., Srinivasan, M.S., Trolove, M., McDowell, R.W., 2010. Identifying and linking source areas of flow and P transport in dairy-grazed headwater catchments, North Island, New Zealand. Hydrol. Process. 24, 3689–3705. http://dx.doi.org/10.1002/hyp.7809. Murphy, J., Riley, J.P., 1962. A modified single solution method for the determination of phosphorus in natural waters. Analyt. Chem. Acta 12, 31–36. Nash, D.M., Halliwell, D.J., 1999. Fertilisers and phosphorus loss from productive grazing systems. Aust. J. Soil Res. 37, 403–429. http://dx.doi.org/10.1071/ SR98087. Nash, D.M., Murdoch, C., 1997. Phosphorus in runoff from a fertile dairy pasture. Aust. J. Soil Res. 35, 419–429. http://dx.doi.org/10.1071/SR96039. Pionke, H.B., Gburek, W.J., Schnabel, R.R., Sharpley, A.N., Elwinger, G.F., 1999. Seasonal flow, nutrient concentrations and loading patterns in stream flow draining an agricultural hill-land watershed. J. Hydrol. 220, 62–73. http:// dx.doi.org/10.1016/S0022-1694(99)00064-5. Pionke, H.B., Gburek, W.J., Sharpley, A.N., 2000. Critical source area controls on water quality in an agricultural watershed located in the Chesapeake Basin. Ecol. Eng. 14, 325–335. http://dx.doi.org/10.116/S0925-8574(99)00059-2. Poor, C.J., McDonnell, J.J., 2007. The effects of land use on stream nitrate dynamics. J. Hydrol. 332, 54–68. http://dx.doi.org/10.1016/j.jhydrol.2006.06.022. Ridley, A.M., White, R.E., Helyar, K.R., Morrison, G.R., Heng, L.K., Fisher, R., 2001. Nitrate leaching loss under annual and perennial pastures with and without lime on a duplex (texture contrast) soil in humid southeastern Australia. Eur. J. Soil Sci., 237–252. http://dx.doi.org/10.1046/j.1365-2389.2001.00387. Sharpley, A.N., Kleinman, P.J.A., Heathwaite, A.L., Gburek, W.J., Folmar, G.J., Schmidt, J.P., 2008. Phosphorus loss from an agricultural watershed as a function of storm size. J. Environ. Qual. 37, 362–368. http://dx.doi.org/10.2134/ jeq2007.0366. Sheridan, G.J., Noske, P.J., 2007. A quantitative study of sediment delivery and stream pollution from different forest road types. Hydrol. Process. 21, 387–398. http://dx.doi.org/10.1002/hyp.6244. Smith, A.P., Western, A.W., Hannah, M.C., 2013. Linking water trends with land use intensification within dairy farming catchments. J. Hydrol. 476, 1–12. http:// dx.doi.org/10.1016/j.jhydrol.2012.08.057. Sutton, M.A., Oenema, O., Erisman, W., Leip, A., van Grinsven, W., Winiwarter, A., 2011. Too much of a good thing. Nature 472, 159–161. Weatherley, A.J., Quin, B.F., Dassanayake, K.B., Rowarth, J.S., 2011. Runoff losses from irrigated dairy pastures treated with phosphorus fertilisers of differing solubility in south-eastern Australia. Aust. J. Soil Res. 49, 633–641. http:// dx.doi.org/10.1071/SR11156. Wilcock, R.J., Monaghan, R.M., Thorrold, B.S., Meredith, A.S., Betteridge, K., Duncan, M.J., 2007. Land–water interactions in five contrasting dairying catchments: issues and solutions. Land Use Water Resour. Res. 7, 21–210. Young, W.J., Marston, F.M., Davis, J.R., 1996. Nutrient exports and land use in australian catchments. J. Environ. Manage. 47, 165–183. http://dx.doi.org/ 10.1006/jema.1996.0043.