Sources of phosphorus in two subcatchments of the Manawatu River, and discussion of mitigation measures to reduce the phosphorus load

Abstract

Dissolved phosphorus (P) from various sources contributes to periphyton growth in rivers. The Manawatu catchment is in the southern North Island, and since it flows through pastoral farmland with soft rock geology, it carries nitrogen, P and sediment that lead to growth of periphyton. We estimated long-term average sediment P in rivers in the catchment above Palmerston North (PNth) (3910 km²) with the NZeem® model, and also measured the total P on the bed of the river. The annual loss of P in eroded sediment made up 95% of the P flux, (1500 t P yr⁻¹ at PNth); 96% of the erosion occurred under pastures and 4% under forest. Dissolved P was measured at Hopelands and at PNth, and its origins—point sources, forests, the bed of the river and farmland—were assessed. The annual average dissolved P was estimated to be 75 t P at PNth, with 34 t P coming from sheep and beef farms, and this could be reduced with targeted planting of trees; 19 t P came from dairy farms, and could be reduced with changes to management of effluent, limiting soil Olsen P to the optimum agronomic range, and excluding cows from streams, banks and channels. Dissolved P from point sources (15 t P) could be decreased with changes to management of effluent. During low flows, sediment on the bed of the river released approximately 1.3 t dissolved P. These particulate-bound P losses could be reduced by targeted planting of trees on highly erodible land. The Upper Manawatu Water Management Zones (a priority catchment for mitigation) above Hopelands Bridge were also modelled. Council programmes to reduce soil erosion and total P, and to reduce dissolved P during low flow have the potential to reduce P loads in the river. Monitoring of P in the river should be carried out to define a more precise base line, and to monitor improvements to water quality as the programmes progress.

Keywords:

• dissolved organic-P
• DRP
• phosphate
• suspended particles
• fertilizer

Introduction

Phosphorus (P) is lost from land to waters in three forms—P in suspended sediment particles (PP), dissolved reactive P (DRP) and dissolved organic-P (DOP) (McDowell & Wilcock 2008). PP is generated during major storms when soil is eroded on steep land. Some sediment is retained in storage in the channel bed and may subsequently move downstream in future storms (Ballantine et al. 2009). Sediment (and the associated P) may also be deposited on floodplains where it may be stored for periods of 10–1000 years or longer (e.g. Ballantine et al. 2006) and transferred to the channel through stream bank erosion (McDowell & Wilcock 2007). DRP is generated by dissolution of P in rocks, soils and fertilizer, and mineralization of P in soil organic matter; it is readily available to periphyton in rivers. DOP is produced by dissolution of soil organic matter, is less available than DRP and requires biological processing to access this P, and then mineralize the DOP into DRP (McDowell et al. 2009). DRP and DOP are produced from both diffuse agricultural sources and point sources (Ledein et al. 2007; Clothier et al. 2007; Parfitt et al. 2009; Roygard et al. 2012). DOP is also generated in waters when P is taken up by biological activity. Point sources mainly include industries and waste water treatment plants (WWTP) at townships.

Nutrient mass balances for nitrogen (N) and P have been obtained for large rivers flowing into the Atlantic, with the largest flux of P coming from the Amazon Basin (236 kg P km⁻² yr⁻¹), and that mainly arises from sediment eroded from the Andes; for many other rivers a large part of the nutrient load comes from towns and cities (Howarth et al. 1996). In contrast, in the Manawatu catchment above Palmerston North (Teacher's College), there are only five small townships, but there is significant erosion in the soft rock geology of the hill country (Dymond & Shepherd 2006).

The Horizons Region comprises 22,200 km² in the southern North Island (Parfitt et al. 2008a) (Fig. 1). There is widespread public concern about the health of rivers in the region, and this is being addressed in the One Plan, in the National Land and Water Forum, and in the Sustainable Dairying Water Accord. Horizons Regional Council's sustainable land use initiative (SLUI) addresses erosion on farms and associated sediment in rivers including the Manawatu. A more recent Horizons initiative—nutrient management plans (NMPs)—targets reductions in nutrient loss (mainly N and P) from intensive land uses in priority catchments. Both of these initiatives, which include recommended best management practices (discussed later), form part of the Horizons One Plan. The One Plan, a combined regional policy statement, regional plan and coastal plan, emphasizes integrated catchment management. The One Plan seeks to provide a mechanism to improve water quality, and to reduce N and P concentrations that contribute to the growth of periphyton in the river.

Figure 1  Manawatu Catchment above Teachers' College and the Upper Manawatu Catchment above Hopelands. Scale is 0–24 km. Major sampling sites are Palmerston North Teachers' College; Hopelands; Ta, Tamaki; T, Tiraumea; K, Mangatainoka; H, Mangahao; P, Pohangina; ○, waste water treatment plants.

Little is known about either the amounts or the sources of total-P and dissolved P in the Manawatu River, or about the quantitative effectiveness of mitigation practices on total-P and dissolved P in the river (Parfitt et al. 2004; Parfitt 2006). In New Zealand catchments, grazing stock, soil-P, fertilizer and forests all contribute to dissolved P in rivers (Hart et al. 2004: Elliott et al. 2005; McDowell 2006; McDowell & Wilcock 2007, 2008; McGroddy et al. 2008; McDowell et al. 2009) and sediment is generated from hill country during major storms (Dymond et al. 2010). In a preliminary study (Parfitt et al. 2008b) with Horizons, the NZeem® and Overseer® models were used to generate data on sediment P and soluble P in the upper catchment in a first attempt to determine the amounts of P from diffuse sources, and the likely merits of proposed mitigation strategies.

Overseer® is an agricultural management tool that assists farmers and their advisers to examine nutrient use and movements within a farm to optimize production and environmental outcomes (AgResearch 2009). The computer model calculates and estimates the nutrient flows in a productive farming system and identifies risk for environmental impacts through nutrient loss, including runoff and leaching, and greenhouse gas emissions. It has a P component that predicts the annual loss of P from farmed land (McDowell et al. 2005). NZeem® is an empirical model that predicts mean annual sediment yield from a given catchment, based on annual rainfall, type of terrain and percentage of woody vegetation cover; it estimates the average sediment generated in a catchment over an extended period (about 50 years) (Dymond et al. 2010) including those major storms that cause erosion to occur on a decadal timescale. If the P concentration in the sediment is known, then the P generated by erosion can be estimated on top of those annual losses estimated at the farm scale by Overseer®. The use of a model such as NZeem® may be required in highly erodible catchments such as the Manawatu River where the P generated by erosion from major storms on a catchment scale (e.g. via slips, etc.) could be missed at the farm scale (Parfitt et al. 2008b).

Here we seek to estimate the sources and inputs of P to the Manawatu river, upstream from Teachers' College and Hopelands Bridge, and the losses of P at these major gauging stations. The measured loads from point sources, mainly WWTP, are also reported here. The sediment and PP deposited on the bed of the river after each storm, and the amounts of DRP released to the water were estimated as part of this study. Here we further develop methods to measure the amount of PP on the bed of the river for 2010 and estimate the amount of DRP released. We also measure the load of DRP at the gauging stations for 1989–2009 and assess likely sources of P using Overseer® in each stretch of the river. The quantitative effects of SLUI and NMPs were estimated for the Upper Manawatu Water Management Zones (UMWMZ) (a priority catchment for mitigation) above Hopelands.

Materials and methods

Study site and sampling

Two subcatchments were selected for study: the Manawatu catchment (391,000 ha) upstream from Teachers' College (this is upstream from the Palmerston North WWTP) and the Upper Manawatu catchment (127,000 ha) upstream from Hopelands Bridge, where long-term monitoring data were available. The study area is shown in Fig. 1; the land use areas were estimated from databases including Agribase, Land Cover Data Base2 and Horizons Dairy/Land-Parcels (Clark & Roygard 2008), and are reported in Table 1.

Table 1  Land use (ha) from Horizons databases.

Land use	Manawatu catchment above Teachers' College (ha)	Upper Manawatu catchment above Hopelands (ha)
Sheep and/or beef	241,108	87,400
Dairy	51,891	20,300
Native forest	79,749	12,700
Exotic forest	13,044
Built up/parks	1909
Horticulture – vegetables	169
Horticulture – other	198
Cropping	727
Water body	761
Other	1841	6600
Total	391,397	127,000

Area of river

The wetted area of the rivers in the catchments was measured using Google Earth. Most of the satellite images were taken during low flow, but a small part was taken during a flood, which would slightly overestimate the wetted area at low flow. The ruler provided by Google Earth was used to measure the entire length of the main channel and width selected at regular intervals along it. The wetted area of each river was estimated by the product of length and mean width (estimated from c. 50 cross sections for each tributary). Summation gave the total area of wetted channel in the Manawatu catchment.

Flow data

Flow data for the Manawatu River have been recorded by Horizons for 1989–2009. Hamill (2008) suggests that three times the median flow is approximately the flow needed to turn over river gravels, effectively clearing the stream bed of periphyton, therefore we counted the average number of these events during November–April (three events in summer) and May–October (four events in winter). We assumed that the river bed needs to be stable for at least 20 days to release available P at a maximum rate (Biggs 2000).

Sediment and water samples

Depth integrated triplicate samples of sediment in the river were taken at Teachers' College during flood events on 21 and 23 January 2004, 16 February 2004, 21 June 2010, 11 October 2010 and 6 September 2010. Duplicate samples from the bed of the river were taken on 31 March 2010 and 22 April 2010, both during low flow at the end of the summer. No flood occurred between the two samplings. The autumn was dry, and permitted two samplings under similar conditions. Generally, sampling was carried out twice at a site. The Manawatu River was sampled at five different sites and the Tamaki stream near the Ruahine Ranges at two sites. The Tamaki stream was sampled once on 30 March 2010. In total, more than 50 samples of sediment were obtained. Sites were chosen where there were no rapids, no stagnant water and no boulders. GPS positions were recorded.

Bed sampling followed the method of Lambert & Walling (1988). A metal cylinder (height 60 cm; diameter 39 cm) was pushed into the river bed in order to create a seal. The top 10 cm of the river bed was then stirred with a trowel with the water inside the cylinder to remobilize the fine sediments. A representative sample of the turbid water (about 2 L) was collected in a large-neck white plastic bottle. The volumes of water in the plastic bottle and the cylinder were both measured.

Fine sediment samples were flocculated by freezing the suspensions, and then thawing in the dark, and removing water with a siphon. The suspension was dried at 35 °C overnight, and then sieved through a 1 mm sieve and weighed before analysis.

The amount of fine sediment stored per unit surface of river bed (Br, g m⁻²) was calculated from:

where Cs is the sediment concentration (g L⁻¹), V is the volume of water enclosed in the sampling cylinder, calculated as the product of the depth of water in the cylinder and the cylinder surface area, and A (m²) is the cross-sectional area of the cylinder.

Samples of water from the river were taken at Teachers' College on 75 days in 2004, filtered through a 0.2 µm filter, and DRP was measured, together with DOP on 21 days. Both Teachers' College and Hopelands are long-term monitoring sites for the National River Water Quality Network (NRWQN) and there is a monthly record of DRP in the river since 1989.

Sediment and water analyses

The sediment samples were analysed for inorganic P and total P by the methods of Landcare Research (2010). Organic P was obtained from the difference between the total P and inorganic P. Bicarbonate extractable P used 0.5 M sodium bicarbonate, at pH 8.5, with a 1:20 soil:extractant ratio of 1:20, and with 30 minutes shaking).The 0.1 M NaOH-extractable P fraction can be used to estimate both short- and long-term algal available P in sediments according to Dorich et al. (1985). This extraction method was carried out with an extractant:solids ratio of 500:1 (v/w) for 17 h as described Dorich et al. (1985). P was analysed by the ammonium molybdate colorimetric method on a flow injection analyser. Total P in waters was measured after oxidation with persulphate, and DOP obtained from the difference between the total P and DRP.

Calculation of loads in river bed and in waters

NZeem® (Dymond et al. 2010) was used to estimate the suspended sediment load in the Manawatu River from various land uses. These loads were multiplied by the average total P concentration to produce the loss of total P in eroded sediments. The P concentration of the sediment on the bed of the Manawatu River was in the relatively narrow range of 520 to 550 mg kg⁻¹. The average P concentration of sediment samples taken in major flood events was 545 mg kg⁻¹, and we have used this value, assuming that this is representative of the sediment.

The measured data for DRP and total P in the Manawatu River under various flow conditions at Hopelands were supplied from long-term records held by Horizons. For Teachers' College, data for total P and DRP were obtained from the NRWQN data set together with daily mean flows, to determine the average P export between 1989 and 2009. Rating curves were generated by using trend analysis. The function

was used to estimate the daily concentration of DRP for <95 percentile flow (<320 m³ s⁻¹). For the upper 5 percentile flow, the mean concentration was 18.8 mg m⁻³ (SEM = 1.5). For the major flood events of 23 July 1992 and 16 February 2004, the concentration was 12 mg m⁻³. For total P, there were two linear relationships between flow and total P, for flows below and above 250 m³ s⁻¹: P = 0.5*flow and P = 1.4*flow, respectively. Data inputs for the Overseer® nutrient budget model (12/2011 version) for the catchment above Hopelands were obtained from Clothier et al. (2007) and Mackay et al. (2008).

Overseer® was used to estimate the ‘background’ P losses (kg P ha⁻¹ yr⁻¹) from soil profiles for average dairy farms and sheep and beef farms. These data were based on interviews with a range of representative farmers and their advisers. Overseer® was used for each farm, and these data were then amalgamated, and the means taken, to obtain data for an ‘average’ dairy farm and an ‘average’ sheep and beef farm in the catchment. The data on animal stock units ha⁻¹ and milk solid production for dairy farms in the catchments above Teachers' College were obtained through the Agribase® (2007 data) and Dairy NZ databases (Dymond et al. 2013; Ausseil et al. 2013).

Forests with greywacke/sandstone/mudstone geology have spring waters that contain concentrations of close to 5 mg DRP m⁻³ and 5 mg DOP m⁻³, and assuming annual leaching of 600 mm, were assumed to lose 0.06 kg P ha⁻¹ yr⁻¹ to headwater streams (McGroddy et al. 2008). Since the area of forest has been measured from satellite data, the loss of soluble P from forests (in tonnes) can be estimated. Losses from point sources (tonnes DRP) were obtained from river water samples taken above and below water treatment plants and industries. The losses from forests and point sources were subtracted from the annual flux of dissolved total-P in the rivers, and the remaining tonnes of P in the river (Pf) was apportioned to dairy and sheep/beef farms. The Overseer® model gave a P loss in kg ha⁻¹ for ‘average’ dairy and ‘average’ sheep/beef farms. The ratio (dairy to sheep/beef) of P loss was maintained in the calculations for all the dairy and sheep/beef land (kg ha⁻¹), and this ratio together with the tonnes of P from farm land (Pf) was used to solve the equations to apportion P loss (in the river) from dairy and sheep/beef land.

The quantitative effects of changes to management of farms, that lead to reduced P loads in the rivers, are based on established methods reported in the literature. These are reported in the discussion section.

Results

Dissolved P in the river at Hopelands and Teachers' College

The long-term flow data and DRP concentrations in the river gave an average flux of 23 t DRP per year moving past Hopelands (Roygard et al. 2012) and 49.5 t DRP at Teachers' College. About half the DRP moved during the 10 percentile high flow. DOP concentrations in the Manawatu River were generally about 50% of DRP concentrations (Table 2). When DOP was included, the total dissolved P flux was estimated to be approximately 34.6 t yr⁻¹ moving past Hopelands, and 75 t yr⁻¹ at Teachers' College. The standard error of the mean for these data were 5 and 12 t, respectively.

Table 2  DRP and DOP concentrations of 21 samples collected at various flows at Teachers' College between 2004 and 2005.

	DRP	DOP
	(mg m^−3)	(mg m^−3)
Mean	17	10
Median	16	8
SEM	2	2
SEM, standard error of mean.

P in suspended sediment particles at Hopelands and Teachers' College

The long-term mean for suspended sediment arising from erosion in the catchments, calculated from NZeem®, was 981,000 t yr⁻¹ for Hopelands and 2,750,000 t yr⁻¹ for Teachers' College (Table 3). Most of the erosion occurred on land under sheep and beef farming. The mean total P content of sediment in the Manawatu River was 545 mg kg⁻¹ (Table 4). This led to an estimated flux of P in sediment of 535 t P yr⁻¹ at Hopelands and 1500 t P yr⁻¹ at Teachers' College (Table 3). The average loss of PP per unit area was 4.4 kg P ha⁻¹ yr⁻¹ from all land above Hopelands and 3.9 kg P ha⁻¹ yr⁻¹ from all land above Teachers' College. Losses of PP per unit area from different land uses were 5.5 kg P ha⁻¹ yr⁻¹ for sheep/beef, 2.2 kg P ha⁻¹ yr⁻¹ for dairy, and 0.6 kg P ha⁻¹ yr⁻¹ for land under forest.

Table 3  Long-term average annual mass of sediment and PP load at Hopelands and Teachers' College modelled by NZeem®.

	Hopelands			Teachers' College
	Sediment (t)	PP (t)	P load (kg ha^−1)	Sediment (t)	PP (t)	P load (kg ha^−1)
Sheep/Beef	886,000	483	5.5	2,445,000	1332	5.5
Dairy	81,000	44	2.2	207,000	113	2.2
Forest	13,000	7	0.6	99,000	54	0.6
Total	981,000	535	4.4	2,750,000	1500	3.9
PP, phosphorus in suspended sediment particles.

Table 4  P contents of samples of sediment collected during six flood events at Teachers' College.

	Total P	Inorganic P	Organic P
	(mg kg^−1)	(mg kg^−1)	(mg kg^−1)
Mean	545	403	142
Median	535	395	143
SEM	21	18	7
SEM, standard error of mean.

P in sediment particles on the river bed

The area of the river bed containing fine sediment was 1,530,000 m² above Hopelands, and 9,880,000 m² above Teachers' College (Table 5). The store of the sediment to 10 cm depth ranged from 0.27 kg m⁻² in the Tamaki Stream to 3.1 kg m⁻² near Hopelands. This resulted in an estimate of 3662 t fine sediment on the bed of the river above Hopelands and 14,774 t above Teachers' College.

Table 5  Length, width, area of wetted area of the river, and mean concentration of fine sediment on the river bed, and total amount of fine sediment on the river bed to a depth of 10 cm. Mean samples taken on 30 March and 22 April 2010.

Location		Length (km)	Width (m)	Area *10^3 (m^2)	Conc FS (kg m^−2)	Total FS (t)
Above Hopelands	Tamaki	26	9	235	0.27	67
	Above Tamaki Confluence	21	20	564	2.35	1336
	Tamaki Confluence—Hopelands	28	35	732	3.10	2268
	Total			1531		3662
Below Hopelands	Hopelands—Mangahao	18	45	806	3.10	2498
	Mangahao—Teachers' Coll	35	68	2350	1.73	4065
	Pohangina	46	24	1100	1.72	1897
	Mangahao	52	28	1470	0.50	739
	Mangatainoka	67	25	1645	0.84	1379
	Tiraumea	69	10	693	0.46	320
	Makuri	21	13	283	0.76	214
	Total			8351		1112
Total above Teachers' College		381		9882		14,774
FS, fine sediment.

The total P in the fine sediment above Hopelands was 1850 kg P and above Teachers' College was 7702 kg P (Table 6). This was mainly inorganic P, except in the Mangatainoka and Makuri rivers where the organic P made up about half of the total P, possibly because topsoil erosion made a larger contribution to this fine sediment.

Table 6  Mean total P concentration, kg P,% inorganic P and organic P in fine sediment in the river bed to a depth of 10 cm. Samples taken on 30 March and 22 April 2010.

		Total P		Inorganic P	Organic P
Location		(mg kg^−1)	(kg)	(%)	(%)
Above Hopelands	Tamaki	527	35	86	14
	Above Tamaki Confluence	470	623	77	23
	Tamaki Confluence—Hopelands	525	1191	70	30
	Total		1850
Below Hopelands	Hopelands—Mangahao	525	1311	70	30
	Mangahao—Teachers' College	543	2207	71	29
	Pohangina	400	759	84	16
	Mangahao	474	350	75	25
	Mangatainoka	675	930	53	47
	Tiraumea	660	211	76	24
	Makuri	389	83	48	52
	Total		5852
Total above Teachers' College			7702

The bicarbonate-P concentrations for the fine sediment (FS) ranged from 15 to 47 mg kg⁻¹ (Table 7). One sample from the Tamaki East stream had an anomalously high value of 106 mg kg⁻¹. The NaOH extractable P concentrations ranged from 35 to 171 mg kg⁻¹. The Tamaki East stream had a value of 304 mg kg⁻¹ and a total P value of 3220 mg kg⁻¹. This could have arisen from recent fertilizer additions to the adjacent paddock together with cows that have access to the stream. These anomalous data have been excluded from the totals.

Table 7  Bicarbonate-P and P extracted by 0.1M NaOH from the fine sediments in the river bed.

		BiCarbonate P		0.1 M NaOH P
Location		(mg kg^−1)	(kg)	(mg kg^−1)	( kg)
Above Hopelands	Tamaki	15	1	37	3
	Above Tamaki Confluence	47	62	94	125
	Tamaki Confluence—Hopelands	42	96	80	181
	Total		159		309
Below Hopelands	Hopelands—Mangahao	42	106	80	199
	Mangahao—Teachers' College	23	91	82	333
	Pohangina	22	42	59	111
	Mangahao	18	13	53	39
	Mangatainoka	41	56	171	236
	Tiraumea	42	14	107	34
	Makuri	18	4	35	7
	Total		326		960
Total above Teachers' College			485		1269

Sources of dissolved P in the river

The sediment on the river bed above Hopelands contained 0.16 t bicarbonate-P and 0.31 t of NaOH extractable P (Table 7), and at Teachers' College, the sediment on the river bed contained 0.48 t bicarbonate-P, and 1.3 t NaOH extractable P.

The other sources of dissolved P in the river were point sources and diffuse sources (farmland and forest). The annual load of DRP from point sources above Hopelands was measured as 5.8 t DRP per year, and assuming DOP was half of DRP, there were 2.9 t DOP from point sources giving a total estimate of 8.7 t of dissolved P (Table 8). Therefore, the dissolved P from diffuse sources was 25.6 t P yr⁻¹. This has to be apportioned to land under dairy, sheep and beef, and forest. For Teachers' College, 15 t P yr⁻¹ arises from point sources (Table 8), therefore the dissolved P from diffuse sources was 58.7 t P yr⁻¹.

Table 8  Mean dissolved P, measured and estimated, from township waste water treatment plants (WWTP) and industries measured in 2010–12.

Location	DRP (t y^−1)	DOP^1 (t y^−1)	Total P (t y^−1)
Dannevirke	5.8	2.9	8.7
Eketahuna	0.2	0.1	0.3
Pahiatua	1.0	0.5	1.5
Woodville	0.2	0.1	0.3
Ashhurst^2	2.6	1.3	3.9
Industries	0.2	0.1	0.3
Total	10.0	5.0	15.0
DOP, dissolved organic P; DRP, dissolved reactive P.
^1 Estimated.
^2 Estimated from the census population.

The dissolved P from forest was estimated at 0.8 t P yr⁻¹ at Hopelands and 5.6 t P yr⁻¹ at Teachers' College. The dissolved P from diffuse sources on agricultural land was therefore 24.8 t P yr⁻¹ at Hopelands and 53.1 t P yr⁻¹ for Teachers' College. There were negligible losses from cropping farms and horticulture in the catchments.

Overseer® calculations gave ‘background’ losses of P from pastures to headwater streams in years when there was no major erosion event. On dairy farms in UMWMZ, the predicted losses were 1.8 kg P ha⁻¹ yr⁻¹. On sheep and beef farms, the estimated losses were 0.6 kg P ha⁻¹ yr⁻¹ in gently sloping hill country and 0.8 kg P ha⁻¹ yr⁻¹ in steep hill country. We therefore assume that the average losses from dairy farms were 1.8 kg P ha⁻¹ yr⁻¹ and for sheep and beef farms were 0.7 kg P ha⁻¹ yr⁻¹. Maintaining this proportion of 1.8:0.7, an attenuation value of about 80% was required, between land and Hopelands and Teachers' College to give estimated totals of 24.8 and 53.1 t P yr⁻¹, respectively. For Hopelands, estimated yields to the river are 0.18 kg P ha⁻¹ yr⁻¹ from sheep and beef farms, and 0.46 kg P ha⁻¹ yr⁻¹ from dairy farms (Table 9). Multiplying by the land areas gives estimated P losses at Hopelands of 15.6 t P for sheep and beef, 9.3 t P for dairy, 0.8 t P for forest, and a total from diffuse sources of 25.6 t P yr⁻¹ (Fig. 2). For Teachers' College, the yields to the river were 0.14 kg P ha⁻¹ yr⁻¹ from sheep and beef farms, and 0.36 kg P ha⁻¹ yr⁻¹ from dairy farms. The losses were 34.2 t P for sheep and beef, 18.9 t P for dairy, 5.6 t P for forest, and a total from diffuse sources of 58.7 t P yr⁻¹ (Table 10; Fig. 2).

Figure 2  Sources of dissolved P in Manawatu River at Hopelands (black) and Teachers' College (grey).

Table 9  Estimate sources of various forms of P at Hopelands per year from 1989–2009.

	P fraction	Sheep/Beef	Dairy	Forest	Point sources	From SS in bed of river	Total	Reference
Erosion storms (t) (long-term average)	SSP	483	44	7			535	NZeem®
Annual average (t)	DRP	10.4	6.2	0.4	5.8	0.3	23.0	Horizons data
Annual average (t)	DOP	5.2	3.1	0.4	2.9		11.6	50% of DRP
Dissolved P (t)	DRP + DOP	15.6	9.3	0.8	8.7	0.3	34.6	Our estimate
Dissolved P river (kg ha^−1)	DRP + DOP	0.18	0.46	0.06^1				Our estimate
Hectares		87,400	20,300	12,700
DOP, dissolved organic P; DRP, dissolved reactive P; SSP, suspended sediment P.
^1 McGroddy et al. (2008).

Table 10  Sources of various forms of P at Teachers' College per year from 1989 to 2009.

	P fraction	Sheep/Beef	Dairy	Forest	Point sources	From SS in bed of river	Total	Reference
Erosion storms (t)
(long-term average)	SSP	1332	113	54			1500	NZeem®
Annual average (t)	DRP	22.8	12.6	2.8	10	1.3	49.5	NIWA & Horizons data
Annual average (t)	DOP	11.4	6.3	2.8	5		25.5	50% of DRP
Dissolved P (t)	DRP + DOP	34.2	18.9	5.6	15	1.3	75.0	Our estimate
Dissolved P in river (kg ha^−1)	DRP + DOP	0.14	0.36	0.06^1				Our estimate
Hectares		241,000	51,900	92,800
DOP, dissolved organic P; DRP, dissolved reactive P; SSP, suspended sediment P.
^1 McGroddy et al. (2008)

Discussion

For the first time in New Zealand we have estimated both the total and dissolved P losses and the likely sources of P for a large catchment, by using the Overseer® nutrient budget and NZeem® models in tandem. The losses as PP were estimated to be 20 times greater than from dissolved P (1500 vs 75 t P yr⁻¹ at Teachers' College). The largest source of soluble P in the Manawatu River is pasture land, and the smallest sources were forests and sediment on the bed of the river. The P concentration of spring waters in forests is about 10 mg P m⁻³, which equates to soluble P losses of about 0.06 kg ha⁻¹ yr⁻¹ (McGroddy et al. 2008). Forest land in the Manawatu also loses 0.6 kg P ha⁻¹ yr⁻¹ by erosion (Dymond et al. 2010); this is within the range of 0.2–3 kg P ha⁻¹ yr⁻¹ for P lost by weathering and erosion of soils and rocks in New Zealand (Parfitt et al. 2008a). Erosion losses equate to losses of 5.5 kg P ha⁻¹ yr⁻¹ from sheep and beef pasture land, and 2.2 kg P ha⁻¹ yr⁻¹ from dairy pasture land. These losses are generally larger than for overseas catchments (Withers & Jarvie 2008) and arise from the soft rock geology, steep slopes, pasture cover and intense storms. The total flux, including sediment, is 493 kg P km⁻² yr⁻¹, which is greater than that from the Amazon river (236 kg P km⁻² yr⁻¹) and for all Atlantic rivers (Howarth et al. 1996), and larger than for the whole Horizons region of 22,200 km (158 kg P km⁻² yr⁻¹) estimated by Parfitt et al. (2008a).

Total annual P losses from sheep farms in Manawatu hill country were 0.7–1.5 kg P ha⁻¹ yr⁻¹, and just 15% of this was DRP (Lambert et al. 1985). However, later data from Gillingham & Gray (2006) showed that larger losses occurred during some storm events. They found that during 2001 two large storms caused 1.2 kg P ha⁻¹ to move as sediment from land to streams. There were also losses during 15 other smaller events that year, with 0.7 kg P ha⁻¹ yr⁻¹ moving as PP from dung, sheep tracks and other areas of exposed soil. They reported total P losses of 1.8 kg P ha⁻¹ yr⁻¹ where the Olsen-P value was 6, and 4.1 kg P ha⁻¹ yr⁻¹ where the Olsen-P was 26. The DRP concentrations were 0.019 and 0.037 g m⁻³, DOP values were 0.018 and 0.019 g m⁻³, and PP were 0.024 and 0.053 g m⁻³, respectively (Gillingham & Gray 2006).

The 0.3 and 1.3 t P released from sediment on the bed of the river is derived from the P extracted with NaOH (Dorich et al. 1985). There is considerable uncertainty in these numbers because Dorich et al. (1985) report that about half of the P extracted by NaOH was taken up by algae in 2–14 days. On the other hand, there are about seven flood events each year that expose fresh sediment surfaces in the Manawatu river, so that a single extraction with NaOH may underestimate the P release (Ekholm & Krogerus 2003; Jarvie et al. 2005). Our estimation does not allow for sorption by sediment, and is only a snapshot for two samplings in 2010. Others have suggested using a bioassay or extraction with dilute CaCl₂ to estimate P dissolution from sediment; these methods give a smaller dissolution (Hedley 1978; Lucci et al. 2010). Therefore, our data probably only give an order of magnitude for the release of P, and this gives a relatively small source of DRP compared with the annual losses from pasture (after attenuation).

Potential mitigation gains in Upper Manawatu Water Management Zones (above Hopelands)

Point sources contribute about 20%–25% to the load of soluble P in the river. A large part of this comes from detergents in effluent and sewage (Stow et al. 2001). In recent years, effluent has been treated (with alum) by some townships, effluent ponds are being lined to prevent leakage, and effluent is being applied to land, and this has generally reduced the input to the river, so the P load from these sources is decreasing (Anon. 2012). If P-free detergents were used this would considerably reduce the P load in the river. At Hopelands, we estimate point sources will contribute about 6.5 t dissolved P by 2020 compared with 8.7 t P in 2011, a reduction of 25%, as a result of planned mitigation practices (Fig. 3).

Figure 3  Modelled estimates of sources of dissolved P (DRP + DOP) in the Manawatu River at Hopelands in 2010 (black), and loads achievable by 2020 (grey) if recommendations were implemented (t P yr⁻¹). Note: Some of the 511 t of PP remains on the bed of the river and generates a maximum of 0.3 t of dissolved P yr⁻¹.

For dairy farming, gains could be made from changes to management, such as removing cows from streams, banks and channels (McDowell & Wilcock 2007), and retaining all milking platform P losses in effluent ponds and improved effluent management. Losses due to cows in streams were about 0.5 kg P ha⁻¹ yr⁻¹ (McDowell & Wilcock 2007) and there were 20,300 ha of dairy land at Hopelands, so fencing streams on 10% of the farms gives gains totalling 1 t P. Gains can be made from diverting drainage from tracks that link to streams, and bridging stream crossings. For dairy farms this could reduce the losses by 0.1 kg P ha⁻¹ yr⁻¹. If this applies to half the dairy farms, this could reduce the load in the Manawatu River at Hopelands by 1 t P yr⁻¹.

Shifting from pond discharge, for farm dairy effluent management, to an effective land-based effluent treatment system (that includes ensuring the effluent block is of sufficient size, and sufficient pond storage capacity to hold effluent when soils are wet) offers enormous scope for reducing farm P losses (Houlbrooke et al. 2004; Monaghan & Smith 2004; Monaghan et al. 2008). Gains can be made, therefore, from irrigating farm dairy effluent according to deferred irrigation criteria where applications are made only to soil that has a sufficient soil water deficit to store applied volumes. Furthermore, when soil infiltration limitations exist or preferential flow of applied effluent is likely, further gains can be made using low-rate irrigation technology (Houlbrooke et al. 2006). For dairy farms this could reduce the losses by 1 kg P ha⁻¹ yr⁻¹ on at least 10% of the milking platforms. This would reduce the load in the Manawatu River at Hopelands by 2 t P yr⁻¹. The consented effluent loads at Hopelands have already been reduced from a peak of about 3 t DRP yr⁻¹ in 1998 to 0.5 t DRP yr⁻¹ in 2006 (Roygard & McArthur 2008).

Gains can be made from using reactive-phosphate-rock (RPR) on pastures with acid soils rather than more soluble P fertilizers (Hart et al. 2004), since soluble fertilizers can move short distances to streams (McDowell et al. 2010). The gains depend on weather conditions, such as storms occurring shortly after applying fertilizer. Assuming that the loss from soluble P fertilizers during a large storm is an average of 1 kg P ha⁻¹ yr⁻¹ (Gillingham & Gray 2006) and from RPR is 0.5 kg P ha⁻¹ yr⁻¹, then the loss over 1000 ha would be 1 and 0.5 t P, respectively. Most fertilizer P, however, is retained in soils by incorporation in organic matter and by reaction with iron and aluminium (Parfitt et al. 2008a).

Limiting P fertilizer inputs to maintenance amounts, and keeping levels within FertResearch recommended guidelines, reduces farm P losses (McDowell & Wilcock 2007). Monaghan et al. (2007) found targeting fertilizer inputs to maintain Olsen values in the optimum agronomic range offered the greatest saving and a predicted reduction in P losses by 30%–37% in two dairy catchments. The agronomic optimum can be attained through use of Overseer® together with soil test information and production levels for each major land unit on the farm. When too much P fertilizer is used on low anion storage capacity soils, P losses increase exponentially; therefore P fertilizer use here should be reduced. On the other hand, P fertilizer use could increase if farm management intensifies in UMWMZ, leading to increased P losses. We assume these increases and reductions in P loss will cancel each other in the catchment budget, but there is uncertainty in this assumption.

The gains overall from dairy farms could possibly be 4 t dissolved P, and 3 t for sheep and beef farms (Fig. 3).

The greatest gains in reducing total export of P from land to rivers at high flows can be made by applying best management practices on erodible land under pasture. Loss of PP at Hopelands is 535 t P yr⁻¹, with 527 t P yr⁻¹ from steeper land on farms. The SLUI whole farm plan could reduce the PP losses from pasture land from 527 to 263 t P yr⁻¹ (Fig. 4) with targeted planting of highly erodible land, and by implementing whole farm plans on approximately 10% of farms with the highest areas of highly erodible land (Dymond & Shepherd 2006; Schierlitz et al. 2006; Ausseil et al. 2013). It takes several years after planting for tree roots to bind the soil, so the achievement of these gains will lag behind the implementation of whole farm plans, especially if they are staggered over a number of years (Douglas et al. 2008). Targeting the SLUI whole farm plans at the highest priority farms is the best way to achieve gains in the shortest term. By reducing the introduction of fresh sediment into the bed of the Manawatu River there would also be reductions in P released from the river bed.

Figure 4  Modelled estimates of sources of particulate P in Manawatu River at Hopelands in 2010 (black), and loads achievable by 2020 (grey) if recommendations were implemented on all affected land on day 1 (t P yr⁻¹).

The sum of all these gains could reduce the average DRP + DOP (dissolved P) load at Hopelands from 34.6 to 25.4 t P yr⁻¹; similar proportional reductions could be made for the Manawatu catchment above Teachers' College. The ANZECC (2000) recommended guideline for slightly disturbed lowland ecosystems for DRP has been set at 10 mg P m⁻³. Current concentrations at Hopelands are usually >10 mg P m⁻³, except when periphyton actively strip DRP in summer. With these mitigation measures it should be possible to reduce current DRP concentrations towards the guideline concentration.

There is uncertainty in our estimates, arising from uncertainties in the Overseer® nutrient budget (Ledgard 2009) and NZeem® models (Dymond et al. 2010), and there are statistical errors associated with taking the average annual data. The Overseer® model, however, can be used with more certainty to assess changes to management of land (Ledgard 2009). With these changes to management of land and WWTP, improvements in water quality are expected to occur, and monitoring of P in the Manawatu River should be carried out on a regular basis to both measure these changes and give more precise numbers.

Conclusions

The total and dissolved P losses and the likely sources of P have been estimated for the Manawatu catchment using a modelling approach. Losses of soluble P come from both point sources and diffuse sources; the order being: sheep and beef farms > dairy farms > point sources >forest > river sediment. Mitigation measures that would improve water quality during times of low flow include reducing losses from point sources, removing animals from stream banks and water ways, and better management of dairy effluent and of P fertilizer inputs. The long-term losses of PP occur mainly during major rainstorms, and give large total P losses from land to rivers. Most of these losses come from sheep and beef pastures on hill country. Decreasing the sediment deposition on the bed of the river will also improve water quality during times of low flow, since less P will be released from the sediment in the river bed, and the habitat for aquatic organisms will also improve.

Acknowledgements

We were grateful to Anne-Gaelle Ausseil (Landcare Research), to the Landcare Research Environmental Chemistry Laboratory and to NIWA for data, and to colleagues at AgResearch, Brent Clothier (Plant and Food), Troy Baisden (GNS) and Deborah Ballantine (NIWA) for guidance. The funding for this project was from the Foundation for Research, Science and Technology through the technology transfer Envirolink programme and from Horizons Regional Council.