Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils

STOTEN-20445; No of Pages 11 Science of the Total Environment xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils Rachel Cassidy ⁎, Donnacha G. Doody, Catherine J. Watson Agri-Environment Branch, Agri-Food and Biosciences Institute (AFBI), Newforge Lane, Belfast, BT9 5PX, Northern Ireland

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Overland flow and drainage were monitored in plots across a soil Olsen P gradient. • Soil Olsen P was unrelated to measured concentrations in drainage or overland flow. • Observed concentrations exceeded environmental quality standards for all plots. • Higher antecedent soil moisture deficits result in greater P losses. • Alternative management strategies may be needed to reduce P risk to freshwater.

a r t i c l e

i n f o

Article history: Received 6 May 2016 Received in revised form 7 July 2016 Accepted 9 July 2016 Available online xxxx Editor: Jay Gan Keywords: Legacy phosphorus Water quality, overland flow Drainage Grassland Soil moisture

a b s t r a c t Rates and quantities of legacy soil phosphorus (P) lost from agricultural soils, and the timescales for positive change to water quality, remain unclear. From 2000 to 2004 five 0.2 ha grazed grassland plots located on a drumlin hillslope in Northern Ireland, received chemical fertiliser applications of 0, 10, 20, 40, 80 kg P ha−1 yr−1 resulting in soil Olsen P concentrations of 19, 24, 28, 38 and 67 mg P L−1, respectively, after which applications ceased. Soil Olsen P and losses to overland flow and drainage were monitored from 2005 to 2011 on an event and weekly flow proportional basis, respectively. Soluble reactive P and total P time series were synchronised with daily rainfall and modelled soil moisture deficits. From 2005 to 2011 soil Olsen P decline was proportional to soil P status with a 43% reduction in the plot at 67 mg P L−1 in 2004 and a corresponding 12% reduction in the plot with lowest soil P. However, there was no significant difference in the flow-weighted mean concentration for overland flow among plots, all of which exceeded 0.035 mg L−1 in N98% of events. Strong interannual and event variations in losses were observed with up to 65% of P being lost during a single rainfall event. P concentrations in drainage flow were independent of Olsen P and drain efficiency was potentially the primary control on concentrations, with the highest concentrations recorded in the plot at 38 mg L−1 Olsen P in 2004 (up to 2.72 mg L−1). Hydrological drivers, particularly antecedent soil moisture, had a strong influence on P loss in both overland and drainage flow, with higher concentrations recorded above a soil moisture deficit threshold of 7 mm. This study demonstrates that on some soil types, legacy P poses a significant long term threat to water quality, even at agronomically optimum soil P levels. © 2016 Elsevier B.V. All rights reserved.

⁎ Corresponding author. E-mail address: [email protected] (R. Cassidy).

http://dx.doi.org/10.1016/j.scitotenv.2016.07.063 0048-9697/© 2016 Elsevier B.V. All rights reserved.

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

2

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

1. Introduction Phosphorus (P) is a limiting nutrient for plant growth and has long been applied to agricultural soils to raise grass and crop productivity. In freshwater aquatic ecosystems P availability constrains primary production, with excessive levels contributing to eutrophication and the continued degradation of many waterbodies (Tunney et al., 1998; Uhlmann and Lothar, 1994; Smith and Schindler, 2009). The detrimental impact of P on water quality has been the subject of extensive study and review (e.g. Hart et al., 2004; Ulén et al., 2007; Withers et al., 2014; Jarvie et al., 2013b) and, cognisant of this, strategies to prevent both diffuse and point source losses to water have been implemented through voluntary and regulatory measures in many countries (Kleinman et al., 2015; McDowell et al., 2015). Diffuse nutrient loss from the agricultural landscape poses a particular challenge as a multiplicity of factors including soil type, climate, connectivity to water courses and land management practices strongly influence the potential for loss and lead to significant temporal and spatial variability (Withers et al., 2014). While advances in the resolution and accuracy of monitoring and research in agricultural catchments (e.g. Jordan et al., 2012; Mellander et al., 2012; Wade et al., 2012; Owen et al., 2012; Dupas et al., 2015) are providing insights into the complexity of the processes involved, many aspects of the “how much”, “where” and “when” of nutrient delivery to surface waters remain ambiguous (Jarvie et al., 2013a). Failure to improve water quality is a major concern for regulatory authorities where success is linked to achieving a target within a set timeframe. While residence times of nitrogen (N) in soil are short, P has considerable potential for storage and retention, with soil sorption capacities many times that of the corresponding P concentration in solution (Frossard et al., 2000; Sharpley et al., 2013). While controls on nutrient applications and implementation of mitigation measures (Amery and Schoumans, 2014) might serve to reduce inputs and P loss from the system, legacy soil P (Kleinman et al., 2011) is likely to be contributing to the failure to detect significant changes in water quality across Europe and elsewhere (e.g. Bradley et al., 2015; Sharpley et al., 2013). The time required to ‘draw down’ P in agricultural soils is difficult to ascertain, given the complex pathways, long residence times and physico-chemical processes involved in mobilisation and transportion (e.g. Sharpley et al., 2013; Haygarth et al., 2014; Kleinman et al., 2011; Jarvie et al., 2013a). The relationship between soil P status, usually defined in terms of plant availability, and P concentrations in both drainage and overland flow has been investigated in a number of studies, with many reporting a correlation between soil test P and P levels in overland flow (e.g. Tunney et al., 1998, 2002; Pote et al., 1996; Smith et al., 2003; Daly and Casey, 2005). Combining data for a study on 4 grassland fields in Ireland with data from studies by Brookes et al. (1997); Pote et al. (1996); Sibbesen and Sharpley (1997) and Torpey and Morgan (1999); Tunney et al. (2002) found a strong (r2 = 0.92) correlation between soil test P and soluble P in overland flow. Under more controlled conditions in a long-term grassland plot experiment Watson et al. (2007), assessed the relationship between soil P status and P losses in drainage and overland flow and demonstrated that rate of P application impacted on the soluble reactive P (SRP) and total P (TP) concentrations in overland flow and drainage. However, losses were highly variable and confounded by hydrological variability. Others, however, found that larger rates of mineral P application did not necessarily result in greater losses in leachate, with soil P sorption capacity (PSC) and degree of P saturation shown to be of greater influence than soil P as estimated using agronomic tests such as Olsen P (Leinweber et al., 1999). Common to many of these studies, is that nutrient applications, either as chemical fertiliser or animal manures, were ongoing throughout the period of observation. As such, relationships between soil P and overland flow or drainage concentrations may be obscured by

incidental losses of fertiliser or slurry where applications were followed by rainfall (e.g. Tunney et al., 2002; Preedy et al., 2001; Haygarth and Jarvis, 1997; Hahn et al., 2012). Other drivers, such as soil moisture and rainfall intensity, have also been shown to add to the risk of manures and fertilisers being lost to water following application (e.g. Vadas et al., 2011). Disentangling these diffuse sources and isolating losses relating to soil status alone is challenging and few studies have examined P loss under conditions of declining soil P. An exception is Schärer et al. (2007) who examined P loss from irrigated plots over a 2 year period following cessation of P applications and found no significant change in available P in the soil or in runoff concentrations. In an attempt to provide some further insights into water quality responses following cessation of P fertiliser application, we investigate soil P decline and corresponding trends in P loss in overland flow and drainage from 5 field-scale grassland plots with different soil Olsen P concentrations over a 7 year period. The study investigated: (1) The extent to which P fertilisation history and Olsen P concentrations affect P loss to surface waters following cessation of P fertiliser application. (2) The temporal variation in P losses in overland flow and drain flow following the cessation of fertiliser applications. (3) The soil and hydrological drivers of P loss in overland flow and drain flow following the cessation of fertiliser application. (4) The implications for agricultural mitigation strategies aimed at reducing eutrophication risks in landscapes with similar soil types.

2. Methodology 2.1. Site description Five grassland experimental plots were established in 1987, on a hill slope site on the Agri-Food and Bioscience Institute (AFBI) Research Farm near Hillsborough, Co. Down (54° 27.212′ N; 6° 5.010′ W). The site is underlain by Silurian greywacke, which is weakly metamorphosed with low matrix porosity. The overburden is glacial till; a gleyed sandy clay-loam soil composed of 48% sand, 31% silt, 21% clay and 12% organic matter (Watson et al., 2007) classed as a Surface Water Gley (or Dystric Gleysol according to the FAO Classification) (Cruickshank, 1997; Doody et al., 2010). Soil hydraulic conductivity was measured at 0.2 m d−1 (Watson et al., 2000), which corresponds to a Hydrology of Soil Type (HOST) Class 24, a slowly permeable mineral soil, gleyed within 40 cm, over impermeable bedrock (Lilly, 2010) and accounting for approximately 54% of the land cover of Northern Ireland (Jordan and Rawlins, 2007). Each 0.2 ha slope parallel plot (14 × 143 m) on the side of the drumlin was hydrologically partitioned, using a PVC membrane to separate the plots at depth (N 1 m) and raised earth along the plot edges to direct any surface water flow. Sub-surface drainage was installed consisting of perforated lateral drains of PVC pipe laid at 10 m intervals across each plot at a depth of 0.8 m dipping to 1.0 m at the western edge of each plot where they connect to a main collector pipe draining to the base of the plot (further detail is provided in Watson et al. (2000)). Overland flow was collected in a surface drain at the base of each plot. The drainage flow and overland flow from each plot were directed to separate v-notch weirs and flows monitored continuously over a range of 0.01 to 7.0 L s− 1 to an accuracy of ± 4% (Watson et al., 2007). Following partitioning and installation of field drains the site was ploughed and re-seeded with perennial ryegrass, so that the hillslope plot and land cover are representative of typical farming conditions in the region.

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

2.2. Plot treatment histories From 1987 onward a number of experiments ran consecutively on the site and the associated P and N fertiliser treatments (summarised in Table 1). Maintenance dressings of potassium (K) and sulphur (S) were applied to the plots as a single annual application over all years. Plots were grazed between April and October in line with normal practices in this region where cattle are housed through the winter period and fed silage. Grazing periods and stocking rates (2 beef steers per plot) were identical across the plots. The dates of stock release and housing varied slightly year on year, however, due to weather conditions and differences in grass growth and trafficability, and stock were removed once grass length fell below 7 cm. Chemical P amendments to the plots ceased in September 2004 and subsequently only N, K and S were amended annually while grazing and management practices continued as previously. The period 2005–2011 is the focus of this study. The plots which until October 2004 received P fertiliser applications of 0, 10, 20, 40, and 80 kg P ha− 1 yr− 1 are referred to as P0, P10, P20, P40 and P80, respectively, in the rest of the paper. 2.2.1. Monitoring Full details of the monitoring set-up and instrumentation are given in Watson et al. (2000). In brief, drainage flow from each plot was recorded at 15-min intervals and a daily composite sample of four 6 hourly samples were collected in a refrigerated autosampler. The daily drainage flow samples were then combined to a weekly composite for analysis. Overland flow was collected from a shallow trench running across the width of each plot at the lowest end and ran under gravity to a mini v-notch weir with an autosampler which triggered when flow was detected. Once triggered the samplers were programmed to collect 24 × 200 mL water samples at 20 min intervals into a composite collection bottle which was collected daily and analysed within 24 h in the laboratory. SRP was determined on filtered samples (through 0.45 μm cellulose Millipore membrane) using the acidic molybdate method of Murphy and Riley (1962). TP was determined on unfiltered samples through digestion with potassium persulphate and sulphuric acid and analysed as for SRP (Eisenreich et al., 1975). Daily overland flow and drainage volumes were paired with the daily (overland flow) and weekly (drainage) composite sample concentrations for all P fractions for each plot. 2.2.2. Meteorological data Daily meteorological data were obtained from the British Atmospheric Data Centre for Hillsborough Weather Station (Station ID 1489), which is located within 500 m of the field site. Average annual rainfall at the site over the period 2000–2012 was 923 mm (range: 717 mm–1230 mm). Estimated potential evapotranspiration over the same period was 647 mm. For 45 instances (33 days in 2006 and

3

12 days in 2010) where daily records were missing for the Hillsborough site (Station ID 1489), values were replaced with those from the nearest synoptic station in Belfast, Newforge Lane (ID 1475) 13 km NNE of the site. Rainfall varies seasonally with highest rainfall and greatest interannual variation recorded during July to October. Over the period 2005– 2011 the wettest year was 2008 with 1067 mm rain and the driest was 2005 with 807 mm. The soil moisture deficit (SMD) for the plots over the duration of the study was calculated using the model of Schulte et al. (2005). This has been shown in other studies (e.g. Kerebel et al., 2013) to accurately represent soil moisture over a range of well to poorly drained Irish soils. Previously at this site Doody et al. (2010) demonstrated a good correlation (r2 = 0.74) between volumetric soil moisture (VSM) and the modelled SMD, though with significant scatter at low SMD values making it less accurate in determining the threshold when the soil is at field capacity (56% VSM in this site) and at increased potential for overland flow. In the SMD simulations undertaken here a moderately-drained, rather than poorly-drained, classification was used as with sub-surface drainage the plots generally returned to field capacity soon after rainfall events. Inputs to the model include the maximum and minimum daily temperatures (°C), rainfall (mm), wind speed (m s−1) and solar radiation (J m−2 day−1). Outputs include evapotranspiration (mm), drainage (mm) and SMD (mm).

2.2.3. Soil P data Soils (0–7.5 cm depth) in the top, middle and bottom sections of each 143 m long plot were sampled (repeated samples in a Wconfiguration) and tested for plant available Olsen P (Olsen et al., 1954) on an approximately weekly basis from 1987 to 2009 and with reduced frequency thereafter. Olsen P is the standard metric in soil tests for agronomic purposes in Northern Ireland, England and Wales (DEFRA, 2010) and the basis by which farmers determine nutrient need and target fertiliser applications. The change in soil Olsen P relative to the defined agronomic P indices was examined, where Index 0 is the range 0–9 mg Olsen P L−1, Index 1 is the range 10–15 mg Olsen P L−1, Index 2 is the range 16–25 mg Olsen P L−1, Index 3 the range 26–45 mg Olsen P L− 1 and Index 4 the range 46–70 mg P L−1. Soil total P was analysed annually in February.

2.2.4. Physical characteristics - topography A high resolution LiDAR data acquisition at a scan density of 40 points per metre (ppm) was completed for the Hillsborough Research Farm in 2015. This was resolved to a 0.25 m grid digital terrain model (DTM), representing the ground surface with vegetation removed. The DTM was used to examine the topographic controls on overland flow across the plots, applying hydrological connectivity modelling (using the TauDEM Toolbox in Arc GIS, (Tarboton, 2015)) to derive

Table 1 Fertiliser and grazing treatment history of the experimental plot sites to December 2011. Date

N regime kg ha−1 yr−1 (CAN)

P regime (superphosphate) kg ha−1 yr−1

K and S regime kg ha−1 yr−1

03/1989 to 02/1999 03/1999 to 02/2000

0–500 different rates/plot 250 6 applications Mar–Aug 250

10

Single annual K (as muriate of potash) application 19 kg K 18 kg S

03/2000 to 02/2005 03/2005 to 01/2012

250

0

0, 10, 20, 40, 80 (6 applications Mar–Sep) 0

Lime t ha−1

Grazing regime

Grazed April–October by 2 beef steers (7–12 months) to maintain a 7 cm sward 4.9 in April 1999 (to raise pH from 5.8 to 6.2)

20 kg K 18 kg S 18 kg K 34 kg S

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

4

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

the topographic wetness index (TWI),  TWI ¼ ln

α tanβ

 ð1Þ

where α is the upslope contributing area for surface water flow to each point and the gradient of the surface, tan(β), controlling how quickly it drains (Beven and Kirkby, 1979). High values of the topographic wetness index indicate accumulation of flow at a grid cell due either to large upslope contributing areas flowing into that cell or a low gradient, which slows runoff. The network index models of Lane et al. (2004) and the SCIMAP modelling framework of Reaney et al. (2011) and Lane et al. (2006) build on this base, conceptualising river catchments as a network of flow paths that accumulate various diffuse contaminant species (such as sediment or P) that would pose a threat to water quality if hydrologically connected to receiving surface waters, an approach that has been applied in many recent studies (see Lane, 2008; Milledge et al., 2012; Reaney et al., 2011; Reid et al., 2007; Thomas et al., 2016; Thompson et al., 2013). 2.2.5. Phosphorus concentration and load estimates Time series data for SRP and TP concentrations in overland flow and drainage water were analysed separately. Seasonal trends were examined monthly using the flow-weighted mean concentrations (FWMC) for overland and drainage flow. Further, higher resolution, analyses used event and weekly data which were estimated as indicated below; (1) Overland flow episodes were partitioned into events separated by a minimum of 24 h quiescence without flow. For each event the FWMC was calculated. (2) Drainage flow and concentrations were analysed as a weekly FWMC. Drainage flow was relatively continuous during much of the year, so identifying discrete events was not feasible. To examine the hydrological drivers of P loss from the plots the synchronised meteorological data was analysed to determine the antecedent SMD and rainfall in the days prior to and during each overland flow event and week of drain flow. Annual loads were calculated as the total flux in overland flow for each plot. 2.2.5.1. Statistical analysis. A linear and mixed modelling approach examined the explanatory variables and differences among the five plots in the study. The starting point was to assess the need for a multilevel model to describe the interactions between the response variable (the FWMC of TP or SRP) and the explanatory variables. Such models are applicable where the influence of a treatment may be discerned in repeated observations; in this case the potential influence of different soil Olsen P concentrations (with plot as the variable) on losses. The modelling protocol followed that of Zuur et al. (2009). Multiple linear regression was applied to evaluate the relationships between individual variables (using the lm function in the nlme package in R (Pinheiro et al., 2015, R Development Core Team, 2005)) and identify necessary transformations. Residuals versus fitted values for the models were examined graphically for heterogeneity, to ensure the spread of residuals was constant. Collinearity was assessed using the variance inflation factor (VIF) statistics, to indicate any predictors having a strong linear relationship with the other predictors (VIF N 1) (Field et al., 2012). Following these steps logarithmic transformations were applied to the flow-weighted P concentration time series. A baseline linear regression model was fitted to the data (using the gls function in R library nlme) considering the flow-weighted mean concentration as a function of all identified explanatory variables; (1) time, (2) the soil moisture deficit in the preceding 7 days, (3) calendar month and (4) total event rainfall. The next stage in analysis applied linear mixed effects modelling (using the lme function in R library nlme) to include both fixed effects

(e.g. time, antecedent SMD, event rainfall) and random effects that account for (1) variations in the intercept among plots and (2) variations in the intercept and slope among plots. A maximum likelihood (ML) function was used as a measure of fit to allow comparison of models with changing fixed effects using a chi-square likelihood ratio test and Akaike's Information Criterion (AIC). The AIC measures the relative quality of the model as an estimate of the goodness of fit against the model complexity, with a lower value indicating a better model relative to another. A first order autoregressive covariance structure was used to allow for greater correlation between adjacent points in the time series. 3. Results and discussion 3.1. Soil Olsen P status Following the cessation of P applications in September 2004, the soil Olsen P concentrations across the plots reflected the gradient in applied fertiliser loads. In 2005 P80 had a median range for the whole plot of 53.4–67.7 mg L−1, which corresponds to high Index 4 in terms of soil fertility. By comparison the control plot, P0, which received no P fertiliser additions from 2000, had a median range of 13.9–20.0 mg L−1 (Fig. 1) and was between Index 1 and 2 in terms of fertility. By the latter half of 2009 (1st July 2009) and up to 2011 (a longer time period was used due to a lack of samples in 2010–11) there was a clear decline in Olsen P across the plots, most notably in P80, which declined to 30.1–45.6 mg L−1 for all samples taken between July 2009 and the end of 2011. P0 also declined slightly to a median range between 11.0 and 16.7 mg L−1, with concentrations for the middle and bottom of the plot in Index 1. Greatest Olsen P reductions occurred in P80, and the fall in Olsen P concentrations reflected the magnitude of the initial P loading of the plots, in agreement with Schulte et al. (2010). The time series for the decline of Olsen P from 2004 to 2011 (Fig. 1, inset) were examined, with each observation composed of the mean of records from the top, middle and bottom of each plot. Rapid increases in concentrations across the fertilised plots during 2000–2004 (not shown, but discussed in Watson et al. (2007)) were followed by a slower decline once fertiliser applications ceased. Using the same approach as in Schulte et al. (2010) the fit to P80 was used to predict the time at which the starting concentrations (taken as the mean concentration in September–December 2004) in plots P40, P20, P10 and P0 occur and the respective time series offset accordingly (Fig. 1, inset). For example, the mean starting concentration of 15.1 mg L−1 in P0 was predicted to occur on day 6570 (27/08/2022) using the fit to P80 and thus the P0 time series was offset accordingly. The resulting decay time series was best fit (r2 = 0.89) by a negative natural logarithmic decay curve of the form polsen ¼ a þ b: ln ðT Þ

ð2Þ

where Polsen is the soil Olsen P concentration at time, T, and the parameter b describes the rate of decline. Based on the decay function the projected time for a soil at a concentration of ~ 70 mg L− 1 (high Index 4) to reach 16 mg L−1 (threshold Index 1–2) is ~21 years and to reach 10 mg L−1 (threshold Index 0–1) is ~33 years. Other studies have observed an exponential decay in Olsen P concentrations with time (Dodd et al., 2012; Johnston et al., 2016; Schulte et al., 2010). In the current study, an exponential decay curve also produced a reasonable fit to the offset data (r2 = 0.86) but failed to capture the initial sharp decline in Olsen P from ~80 mg Olsen P L−1 in the first months following cessation of applications to 40 mg L−1 after only 3.3 years. In apportioning the changes in available soil P observed during this study there are a number of potential mechanisms to consider; (1) losses in overland flow and drainage (both soluble P and particulate), (2) transformations with the soil profile to less available forms, (3) uptake by grass (where subsequently removed as cattle body

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

5

Fig. 1. Soil Olsen P across the top (suffix T), middle (suffix M) and bottom (suffix B) of each plot comparing 2005 concentrations with those of July 2009–2011 (paler colours). Inset: transformed Olsen P time series, offsetting initial P concentrations for each plot to predicted occurrence time based on logarithmic fit to P80 time series (equation: polsen = −12.79 ln(T) + 130.21, r2 = 0.89). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

mass) and (4) loss to groundwater bypassing the sub-surface drainage system. While uptake by grass and off-take by cattle were similar across the plots, losses to groundwater were not measured during this study but were assumed to be low due to the impeded nature of the soil. However, transformation within the soil profile is likely to have played a significant role in the changes in soil P availability observed during this study. In general, the proportion of soluble and plant available P relative to the total soil P store in agricultural soils is relatively small (Brady and Weil, 2009). A temporal decline in available soil P occurs as P complexes to less available forms (with Fe, Al and Mn) and becomes occluded within oxide minerals and immobilised into organic matter (OM) (Brady and Weil, 1996; Sinclair et al., 2012). In the current study the proportions of Olsen P relative to total soil P reflect this with total soil P concentrations for the plots (sampled annually in February) ranging from 1042 mg P kg−1 in P0 to 1573 mg P kg−1 in P80. This compares to the plant available Olsen P fraction which ranged from 16.2–24.5 mg kg−1 in P0 to 44.2–67.0 mg kg−1 in P80 (samples from July 2009– 2011). Soils vary in their ability to buffer changes in available soil P (Herlihy et al., 2004) with soil chemical and physical properties controlling binding and sorption characteristics of the soil and the potential for mineralisation and dissolution/desorption. Soils with high clay content have the capacity to accumulate P due to their potential to retain P bound to oxides of iron (Fe) and aluminium (Al), with the potential for desorption increasing as the PSC is approached (Sibbesen and Sharpley, 1997). Daly et al. (2015) show a strong correlation between binding energies and sorption buffer capacities and soil pH and the Al content in Irish soils. The differential binding energy of Al compared to Fe for P retention in soils has been shown by Jordan et al. (2005) in a comparison of P loss from soils in 2 contrasting Irish catchments, with Fe bound P more easily remobilised. This is supported by recent work by Mellander et al. (2016) who showed increased potential for P loss to groundwater in a grassland soil where soil Fe content, and soil OM, were high. In the current study, soils have, on average, 3 times higher concentrations of oxalate extractable Fe (4356–6344 mg kg−1) than oxalate extractable Al (1332–2318 mg kg−1). With a greater proportion of the soil P likely to be bound to Fe oxides at this site, this potentially facilitates P being brought back into soluble forms and can sustain a long term flux to surface waters from the plots, and particularly under reduced conditions following prolonged saturation (e.g. Ann et al., 1999; Sallade and Sims, 1997; King et al., 2015). The rate of mineralisation of the soil organic P pool will also impact on P availability

to plants and losses to water. However, further work is needed to investigate the contribution of the organic P pool in these plots. 3.2. Overland flow losses 3.2.1. Magnitude and source areas for overland flow Soil Olsen P concentrations at the top of each plot were consistently higher than in the middle and bottom of the plots (Fig. 1) with differences between the middle and bottom of the plots less pronounced. Given this variation any comparison of P losses in overland flow should establish that hydrologically the plots are similar in terms of propensity for saturation and generation of overland flow. If the source areas for mobilisation of P during rainfall events vary then, given the differences in Olsen P concentrations, it would be difficult to justify comparisons of concentrations in overland flow. In previous work by Thompson et al. (2012) overland flow was measured along the length of a single plot (P80) and compared with modelled hydrological connectivity based on a dGPS survey of elevations. Modelled areas of high saturation potential were largely coincident with areas where overland flow occurred and were concentrated in the lower part of each plot. To refine this and extend to all plots a 0.25 m resolution LiDAR DTM of the entire site was used to model hydrological connectivity. This process, described in Appendix 1 (Supplementary material), demonstrated that the plots were almost identical (Appendix 1, Fig. A1) with greatest connectivity, and therefore susceptibility to erosion and P loss, at the middle and bottom of the plots. In contrast, the top of the plots have low TWI, a reduced propensity for saturation and thus the frequency and magnitude of overland flow events are expected to be less. These differences may have contributed to the more rapid decline in Olsen P in the lower sections of each plot. Other factors that may have influenced the soil P distribution within the plots include the unpredictability of animal grazing and rest patterns, congregation around drinking troughs which were located at the top of the plots and “socialising” across the fences between plots which may have led to uneven dung deposition. These effects were not quantified, although anecdotal evidence regarding dung deposition does not support the hypothesis that this would account for the consistent variation in Olsen P from the top to bottom of the plots. Soil sampling avoided any areas with visible evidence of manure deposition and the plots were carefully managed to avoid overgrazing or trampling. The similarity among plots is supported by comparisons of daily flow volumes from each plot over the duration of the study, which shows no

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

6

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

differences in the median flows (median flow ± confidence intervals (CI) = 1256 ± 738 L, P0; 2033 ± 845 L, P10; 1364 ± 910 L, P20; 1108 ± 654 L, P40; 1570 ± 759 L, P80). 3.2.2. Event based P loss in overland flow Overland flow was examined on an event basis for both TP and SRP time series, following the same methodology for each. The time series for event FWMC SRP concentrations for P0, P40 and P80 over the period 2005–2011 are shown in Fig. 2. Modelling was initiated with a linear regression model (function gls in R) with all potential explanatory variables, including time (days), calendar month, event rainfall (mm), and average SMD (mm) over 7 days prior to the event (Model 1, Table 2(a) for SRP; Model 1, Table 2(b) for TP). logFWMC¼TimeþSMDþMonthþRain

ð3Þ

Random structure for within plot variation was then introduced using mixed effects models with random intercept (Model 2, Table 2) and random intercept and slope (Model 3, Table 2). For SRP, there was no improvement in the model with intercept as a random effect (Model 1 compared to Model 2: L-ratio = 2.01, p = 0.1566) or including both intercept and slope as a random effect (Model 2 compared to Model 3: L-ratio = 0.00, p = 1). Fixed effects were evaluated in turn using a backwards stepwise elimination approach (all models shown in Table A1, Appendix 2, Supplementary information). The exclusion of calendar month as an explanatory variable significantly worsened the fit (L-ratio = 124.7, p b 0.0001) with the AIC increasing from 201.01 to 315.71. The exclusion of event rainfall also raised the AIC (poorer fit), but was only weakly significant at the 95% confidence limit (L-ratio = 4.05, p = 0.0443). Overall the optimum model for SRP was a linear model including all four explanatory variables (Model 1), although event rainfall only weakly, though significantly at the 95% confidence level, improved the model. Similarly, for TP (Table 2(b)) comparison of the models using a likelihood ratio test showed no improvement in the model with intercept as a random effect (Model 1 compared to Model 2: L-ratio 0.85, p = 0.3573) or with intercept and slope as random effects (Model 2 compared to Model 3: L-ratio 0.05, p = 0.9775). Fixed effects were evaluated in turn. The exclusion of calendar month as an explanatory variable significantly worsened the fit (L-ratio = 153.30, p b 0.0001) with the AIC increasing from −345.81 to −214.51. For TP, in contrast to SRP, the exclusion of event rainfall improved the model fit AIC (from −214.51 to − 216.03) but the improvement was not significant (L-ratio = 0.48, p = 0.4887). Model coefficients for the optimum model for TP and SRP are given in the Supplementary material, Appendix 2, Table A2. A key observation was that there was no significant difference in the concentration time series among plots, indicating that the Olsen P

content of the plots has no bearing on P mobilisation in overland flow, and was unexpected given the differences in Olsen P across the plots. Among the fixed variables, antecedent soil moisture deficit and calendar month have a marked influence on overland flow concentrations, while event rainfall is only weakly significant. There is a clear interannual pattern in SRP concentrations in the time series (Fig. 2), with peak concentrations tending to coincide with the summer months. In the model the calendar month variable can represent a range of influences which vary annually, such as seasonal variations in growth, the impact of animal grazing (and excrement) and rainfall intensities. Variations in grass growth from season to season can, particularly at this scale, affect the bare soil surface area of the plots and have an impact on P mobilisation in events (Bilotta et al., 2007; Butler et al., 2008). Cattle trampling, grazing patterns and preferred resting locations, in addition to manure deposits, are also likely to have an impact (Kurz et al., 2005; Pietola et al., 2005; Doody et al., 2014). Examination of P fractions shows the proportion of TP as SRP in overland flow varied seasonally. While, on average, SRP concentrations were 48% those of TP, SRP fractions were lowest between March and June (on average 25–40%), consistently highest in July across plots (70–80% across P0, P40, P80) and between 40 and 60% in other months (Fig. A2, Appendix 4, Supplementary material). The variation in the proportions of SRP in TP may be linked to uptake of SRP by plants in the growing season and to the role of various soil processes in the release and transformation of P from the soil matrix into the available P pool. These include redox effects releasing SRP during periods of soil saturation (in wet autumn-winter periods) and the release of P associated with rewetting of dry soils (June–July rains), discussed further in Section 3.5. For all plots, SRP concentrations in overland flow events were well in excess of the Environmental Quality Standard (EQS) for freshwaters (Fig. 2) used by the Irish EPA. A level of 0.025 mg SRP L−1 for high status in waterbodies was exceeded by between 98.7 and 99.4% of events across plots and a good to moderate status indicator of 0.035 mg SRP L−1 was exceeded by 98.1–99.3% of events. Although the EQS is a metric for regulatory monitoring of river sub-catchments and so may not be suitable for a direct comparison with field-scale plots, it nonetheless provides an indication of the potential for overland flow to reduce water quality at larger scales. Under a scenario where all fields in a catchment are under grassland management, have soil P indices of 2 and above, and lack inputs of low nutrient ground water for dilution, the long-term prognosis for water quality in the receiving streams and lakes would be poor based on the observations in this study. This aligns with findings from other studies on the impact of withdrawing P fertiliser, such as that of Dodd et al. (2013) and Dodd et al. (2012) who estimated a 23–44 year period, following the cessation of P applications to grassland soils in New Zealand, before water extractable P concentrations declined to a limit of 0.02 mg SRP L−1 in overland flow and required curtailed farm productivity to do so. From the Olsen P

Fig. 2. Time series of SRP concentrations in overland flow showing, for clarity, only P0, P40 and P80. The environmental quality standard for freshwaters in Ireland is shown for reference on the SRP time series (good status should not exceed 0.035 mg L−1; High status should not exceed 0.025 mg L−1).

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

7

Table 2 Models for (a) SRP and (b) TP FWMC time series analysis. Fit was estimated using a maximum likelihood method and including a regressive autocorrelation structure (CorCAR1 in R). Lratio is the log-likelihood ratio statistic used to calculate the p-value and the significance of difference between compared models. Optimum model in bold. (a)

Model SRP overland flow P0, P10, P20, P40, P80

AIC

L-ratio

Model comparison

p-Value

df

Model 1 Model 2 Model 3

Log FWMC ~ Time + EventRain + PreSMD + Month Model 1 + random effects (intercept) Model 1 + random effects (intercept & slope)

212.01 216.02 220.02

2.01 0.00

1 vs. 2 2 vs. 3

p = 0.1566 p = 1.0000

17 18 20

(b)

Model TP overland flow P0, P10, P20, P40, P80

AIC

L-ratio

Model comparison

p-Value

df

Model 1 Model 2 Model 3

Log FWMC ~ Time + EventRain + PreSMD + Month Model 1 + random effects (intercept) Model 1 + random effects (intercept & slope)

−345.81 −342.97 −339.01

0.85 0.05

1 vs. 2 2 vs. 3

p = 0.3573 p = 0.9775

17 18 20

decay function (Fig. 1, inset) the estimated time for a decline from high soil Index 4 (~ 70 mg L− 1) to 16 mg L− 1 (threshold Index 1–2) is ~ 21 years and even with this, based on the findings here P losses in overland flow and drainage may not show any significant change. This effect may be a factor in the lack of improvements in water quality noted by Campbell et al. (2015) in a similar landscape and soil type in Co Monaghan who found no clear link between decreased high soil test P and improved water quality. 3.3. Annual loads in overland flow As for P concentrations, the variability in annual loads from the plots also reflects the hydrological controls on P transfers (Table 3 and Fig. A4, Appendix 5, Supplementary material) and the significance of differences in plot responses to single rainfall events during the year. Though all plots respond to rainfall occurrence, the magnitude of P transfer varies considerably and this is likely to reflect the dynamic nature of the expansion and contraction of the area generating overland flow during an event and small scale variations in the susceptibility to mobilisation linked to grass cover or other factors within each 0.2 ha plot (Thompson et al., 2012). An analysis limited to the 3 years following the cessation of fertiliser P applications would suggest a decline; however in subsequent years loads increased, confounding this interpretation. In 2011 the load from plot P0, receiving no P amendments since 2000, exceeded the load from P80. The P load and annual FWMC (total flux/total flow) varies markedly (Table 3), both in the periods before and following the cessation of P fertiliser amendments. For 2003 and 2004 there is a link between P loading and losses, with highest FWMC and loads in P40 and P80. Higher FWMC in 2005 may represent a residual pool of available P from fertiliser applications in September, but given the high concentrations in P0, it may also relate to prolonged dry periods succeeded by rain releasing P into solution (discussed in Section 3.5), as 2005 was the driest year in the study. From 2006, there are clear indications that the cessation of P fertiliser applications effected a reduction in the annual FWMC (max. FWMC TP 2003–05 = 2.77 mg L−1; max. FWMC TP 2006–11 = 0.7 mg L−1). Comparison of the timing of large events prior to 2005 with dates of P application Table 3 Total Annual Loads and flow-weighted mean concentrations (FWMC) of TP in overland flow, 2005–2011, all plots. Horizontal line indicates the cessation of P fertiliser applications from September 2004. Loads TP kg ha−1 yr−1

Annual FWMC TP mg L−1

P0

P10

P20

P40

P80

P0

P10

P20

P40

P80

2003 2004

0.05 0.23

0.07 0.07

0.10 0.26

0.07 0.27

0.40 0.24

0.36 0.81

0.38 0.70

0.67 0.80

0.72 1.33

2.77 0.96

2005 2006 2007 2008 2009 2010 2011

0.33 0.30 0.11 0.23 0.27 0.23 0.37

0.47 0.18 0.26 0.57 1.13 0.17 0.25

0.62 0.24 0.25 0.64 0.38 0.29 0.24

0.25 0.31 0.19 0.22 0.35 0.24 0.27

0.34 0.24 0.18 0.39 0.47 0.26 0.14

1.05 0.66 0.73 0.29 0.36 0.34 0.63

1.31 0.49 0.53 0.51 0.44 0.31 0.39

1.32 0.49 0.68 0.47 0.37 0.38 0.39

0.99 0.70 0.64 0.33 0.40 0.47 0.58

1.04 0.62 0.54 0.36 0.43 0.43 0.44

indicate that much of the load prior to 2005 may be due to incidental loss of dissolved fertiliser. 3.4. P losses in drain flow In contrast to overland flow where volumes of water from the plots were similar over the period of observation, and magnitudes of flows from the plots in any event were generally comparable, drainage discharge was highly variable. Median weekly flows from P20 and P40 (300 ± 78 L and 134 ± 27 L, respectively) were significantly lower, when compared using the CI around the median, than for the other plots (794 ± 177 L in P0, 1155 ± 211 L in P20, 956 ± 184 L in P80) and an order of magnitude different between P40 and P10. In addition, weekly concentrations are inversely related to flow rates, with higher concentrations coincident with lower flow (median flow ± CI = 134 ± 27 L; median FWMC TP = 0.118 ± 0.018 mg L−1, P40; median flow = 1155 ± 211 L; median FWMC TP = 0.044 ± 0.005 mg L−1, P10). This pattern is apparent in the concentration time series (Fig. 3) with P40 having elevated concentrations compared to all other plots. Although the drainage system in each plot was installed to the same specification there is uncertainty regarding drain efficiency, particularly the extent to which the PVC pipes may have collapsed, blocked or deteriorated over the period since installation (1989). As drainage flow was persistent over long periods the time series were examined on a weekly interval rather than as events. As for overland flow the selected variables included the antecedent SMD in the 7 days prior to the week of observation, which provides an indication of soil conditions prior to drainage. In addition, the total rainfall during the week of observations was determined. Modelling was initiated with a linear regression model (function gls in R) with all potential explanatory variables, which included time (days), calendar month, weekly rainfall (mm), and average SMD (mm) over 7 days prior to the week of observation (Model 1, Table 4a for SRP; Model 1, Table 4b for TP). logFWMC ¼ Time þ SMD þ Month þ Rain

ð4Þ

For SRP, comparison of the models showed a significant improvement in the model fit with intercept as a random effect (Model 1 compared to Model 2: L-ratio = 120.3, p b 0.0001) but no improvement with random intercept and slope (Model 2 compared to Model 3: Lratio = 0.0000, p = 1). This indicates that although the concentration levels differ by plot the change in concentrations over time are no different. Fixed effects were evaluated in turn (Supplementary material, Appendix 3, Table A3 (a)). The exclusion of calendar month significantly worsened the fit (L-ratio = 51.82, p b 0.0001). The exclusion of total weekly rainfall (Model 11) reduced the AIC to 403.09 but the improvement was not significant. Overall the optimum model for SRP was the mixed effects model (Model 2, Table 4a) with random intercept. The trend for drainage concentrations was close to zero (Supplementary material, Appendix 3, Table A4). For TP, comparison of the models (Table 4b) showed a significant improvement with the intercept as a random effect (Model 1 compared

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

8

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

Fig. 3. Weekly drainage concentration time series for SRP showing, for clarity, only P0, P40 and P80. The environmental quality standard for freshwaters in Ireland is shown for reference (good 0.035 mg L−1; high 0.025 mg L−1).

to Model 2: L-ratio 93.71, p b 0.0001) but no improvement with random intercept and slope (Model 2 compared to Model 3: L-ratio 1.165, p = 0.559). Fixed effects were evaluated in turn. The exclusion of month as an explanatory variable (Supplementary material, Appendix 3, Table A3b) significantly worsened the fit (L-ratio = 111.71, p b 0.0001). The exclusion of total weekly rainfall (Model 11) increased the AIC to 125.24, significantly worse at the 95% confidence limit (Model 11 compared to Model 2: L-ratio = 34.681, p b 0.0001). As for SRP, the overall optimum model for TP was the mixed effects model (Model 2, Table 4 b) with random intercept and including all four explanatory variables. The drainage concentration time series showed marked differences to those of overland flow. Weekly comparisons of TP and SRP time series using a linear and mixed effects modelling approach indicated significant differences in intercepts among plots, no difference in slopes and no apparent change in concentrations with time. However, although the concentration time series are significantly different across plots, the differences are unrelated to soil test P. Highest concentrations occurred in Plot P40 (intercept = −0.913, Table A4). The concentrations in P20 and P10 were lower (intercept = −1.123 and − 1.232 respectively), while the time series for P80 and P0 had the lowest concentrations and were almost indistinguishable (intercept = − 1.275 and − 1.279 respectively). As for overland flow, antecedent SMD in the 7 days prior to the observations was a key explanatory variable for concentrations, along with time and calendar month. Elevated concentrations in the plots with lowest flows/impeded drainage systems may be linked to the retention of water within those plots, raising the water table and encouraging reduced conditions under which P is released. A number of laboratory studies have shown the release of P due to iron reduction under anaerobic conditions following prolonged saturation (e.g. Scalenghe et al., 2014; Ann et al., 1999; Sallade and Sims, 1997 and references in the review of King et al., 2015). In field studies Loeb et al. (2008) found P concentrations in the pore water of riverine floodplains to increase following inundation and linked this to iron-bound P release. Comparison of the fraction of TP as SRP across the plots (Supplementary material, Appendix 4, Fig.

A3) shows a marked difference in P40 compared to the other plots, with SRP fractions more than twice those in P0 and P80 during June to September and between 10 and 20% higher in the other months. This seems consistent with higher SRP release under anaerobic conditions resulting from prolonged saturation, compared to that in plots where the drain system is functioning better. In terms of water quality, drainage concentration exceedances of the EQS for good, and moderately to good status were lower than for overland flow. A notable exception again was plot P40 which exceeded 0.025 mg L− 1 90.2% of the time and 0.035 mg L−1 80.1% of the time. Lowest exceedances occurred in P80 (N0.035 mg L−1, 11.5%; 0.025 mg L−1, 18.6%) and in P0 (N 0.035 mg L−1, 13.3%; 0.025 mg L−1, 19.9%). P 10 and P20 exceeded good status on 34.2 and 42.5% of weeks and good to moderate status on 21.8 and 35.4% of weeks, respectively. The difference in drainage efficiency and flows from each plot (Fig. 3) adds considerable uncertainty to the calculation of annual loads and comparisons among plots. Calculated loads only reflect the water captured by the drainage system and not total infiltration. The probable deterioration of drains in P40 over time, and associated redox effects from prolonged saturation, make load estimation using either effective rainfall or drainage flow unreliable and as such were not undertaken. 3.5. Antecedent soil moisture For both overland flow and drainage flow antecedent soil moisture conditions were significant explanatory variables in the prediction of concentrations for both TP and SRP concentrations (Fig. 4) with a strong positive correlation between soil moisture in the period prior to rainfall and concentrations in subsequent overland flow or drainage events, particularly above a SMD of 6–10 mm. The release of P during rewetting of soils has been observed in a number of studies, at both laboratory and field scales. In soil cores Turner and Haygarth (2001); Venterink et al. (2002) and Daly and Casey (2005) showed solubilisation of P to be greater following a sequence of drying and rewetting soils. McDowell and Sharpley (2002) showed antecedent soil moisture conditions to affect P concentrations in overland flow experiments at both laboratory

Table 4 Models for (a) SRP and (b) TP in drainage (full model lists in Appendix 3, Supplementary material). Significantly different model comparisons are highlighted in bold. Optimum model in bold. (a) Model 1 Model 2 Model 3

Model SRP drainage P0, P10, P20, P40, P80 Log FWMC ~ Time + WeeklyRain + PreSMD + Month Model 1 + random effects (intercept) Model 1 + random effects (intercept & slope)

AIC 523.34 405.04 409.04

L-ratio

Model comparison

p-Value

120.30 1e−7

1 vs. 2 2 vs. 3

p b 0.0001 p =1

(b) Model 1 Model 2 Model 3

Model TP drainage P0, P10, P20, P40, P80 Log FWMC ~ Time + WeeklyRain + PreSMD + Month Model 1 + random effects (intercept) Model 1 + random effects (intercept & slope)

AIC 184.27 92.56 95.40

L-ratio

Model comparison

p-Value

93.71 1.17

1 vs. 2 2 vs. 3

p b 0.0001 p = 0.559

df 17 18 20 df 17 18 20

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

9

Fig. 4. Relationship between antecedent soil moisture deficit (SMD) and TP concentrations in overland and drainage flow.

and field scales, with lower suspended sediment mobilised from boxes and plots with a higher soil moisture content than in those with a deficit. Differences were attributed to soil aggregate breakdown, slaking and dispersion as dry soils are subjected to rainfall. The duration of inter-event dry periods was shown by Majdalani et al. (2008) to have a major impact on particulate mobilisation in soils, where they observed a 15-fold increase in the mass of leached particles when inter-event durations were increased from 1 h to 4 days. Changes were correlated with soil moisture and the effects linked to breakdown of the soil matrix as a result of differential capillary stresses in the drying process. The stronger relationship between antecedent SMD and drainage concentrations compared to overland flow concentrations (Fig. 4) may reflect the difference in processes controlling P mobilisation in each. In drainage, the measured concentrations reflect the infiltration and movement of water through the soil horizon, in conditions where mobilisation of P from soil is more influenced by soil biogeochemical processes (potentially influenced by SMD) than by other physical drivers. In overland flow, physical drivers such as rainfall intensity, exposed soil surface area and surface hydrological connectivity also impact on mobilisation, so the signal associated with SMD alone may be more difficult to discriminate. In the most recent Irish Meteorological Office report on future climate in Ireland (Dunne et al., 2008) simulations indicate that winters are becoming wetter with a 5–10% increase in rainfall expected by 2050. In contrast, summers are expected to become drier (5–10% decrease in rainfall over 2012–2060). Longer antecedent periods with higher SMD will, potentially, lead to increased P losses from land in overland flow and drainage when rain comes. Conversely, wetter winters, with prolonged soil saturation, will potentially allow for increased release of P into solution due to redox effects. 4. Conclusions This study aimed to assess the extent to which soil legacy P influences P losses to surface waters during a period following cessation of P fertiliser applications. For observed patterns in P loss to overland flow and drainage we examined the drivers involved, the significance of hydrology and the implications for agricultural mitigation strategies to reduce eutrophication risks in landscapes with similar soil types. Key findings are that: 1. Following the cessation of P fertiliser applications soil Olsen P declined across all plots. Soils with median Olsen P concentrations of N60 mg L−1 (P80) declined to 38 mg L−1 over the monitoring period and soils at 18 mg L− 1 (no P amendments from 2000) declined slightly to 13 mg L−1.

2. Cessation of P fertiliser applications impacted P loss from the plots with the positive correlation between P loading and loss to water breaking down and FWMC decreasing in all fertilised plots. 3. From 2005 to 2011 no link was found between soil Olsen P concentrations and water quality in plots ranging from Index 1 to high Index 4, with FWMC time series for overland flow almost identical. 4. Total soil P levels in the plots are orders of magnitude larger than plant available forms and in this soil type the rate of dissolution, desorption and mineralisation linked to both redox effects and drying/ wetting cycles appear to be high enough to maintain the water soluble P pool at levels that pose a long-term threat to water quality. Similar hydrological conditions in the plots mean that the available P pool, and therefore mobilisation, is likely to be similar during events. 5. SRP concentrations in overland flow were in excess of the EQS for good water quality on almost all occasions, highlighting the continued risk to water quality even 11 years (for plot P0) following the cessation of P amendments, under grazing management. 6. Despite differences in concentrations among plots for drainage no link was found to earlier P loading. It is hypothesised that differences instead relate to deterioration of the drain system (P40) and increased P release through redox effects linked to saturation. 7. Antecedent soil moisture has a strong influence on FWMC of TP and SRP with drier antecedent conditions linked to greater losses once the soil is rewetted. The implications of these findings in light of future climate change need to be considered further. 8. Soil buffering capacity requires greater attention if more accurate estimates of potential risks of P mobilisation to surface waters in Irish catchment are to be made. As such, a measure of the buffering capacity of agricultural soils, potentially in combination with soil test P, may provide a better estimate of risk to water quality than soil test P alone. 9. To effect reductions in P loss from land under similar management in the HOST 24 soil class, alternative management interventions may be necessary. Soil Olsen P status alone does not indicate risk to water quality. Given that this study suggests that soils at the recommended Olsen P status for grass production are capable of losing significant loads and concentrations of P to water, the implications for achieving WFD requirements within required timescales may be considerable. Acknowledgements We thank the staff in the Agri-Environment Branch in AFBI for their excellent technical support in site maintenance, sample collection and analysis over the duration of this study. We also thank the farm staff at Hillsborough for management of the CENIT site. We appreciate the

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

10

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx

time taken by the anonymous reviewers to review this manuscript and for their valuable comments and suggestions. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2016.07.063. References Amery, F., Schoumans, O.F., 2014. Agricultural Phosphorus Legislation in Europe. Institute for Agricultural and Fisheries Research (ILVO), Merelbeke. Ann, Y., Reddy, K.R., Delfino, J.J., 1999. Influence of redox potential on phosphorus solubility in chemically amended wetland organic soils. Ecol. Eng. 14, 169–180. Beven, K.J., Kirkby, M.J., 1979. A physically based, variable contributing area model of basin hydrology/Un modèle à base physique de zone d'appel variable de l'hydrologie du bassin versant. Hydrol. Sci. Bull. 24, 43–69. Bilotta, G.S., Brazier, R.E., Haygarth, P.M., 2007. The impacts of grazing animals on the quality of soils, vegetation, and surface waters in intensively managed grasslands. In: Donald, L.S. (Ed.), Advances in Agronomy vol. 94. Academic Press, pp. 237–280. Bradley, C., Byrne, C., Craig, M., Free, G., Gallagher, T., Kennedy, B., et al., 2015. In: Byrne, C., Fanning, A. (Eds.), Water Quality in Ireland 2010–2012. Environmental Protection Agency, Aquatic Environment Office of Environmental Assessment. Brady, N.C., Weil, R.R., 1996. The Nature and Properties of Soils. Prentice-Hall Inc. Brady, N.C., Weil, R.R., 2009. Elements of the Nature and Properties of Soils. Prentice Hall. Brookes, P.C., Heckrath, G., Smet, J., Hofman, G., Vanderdeelen, J., 1997. Losses of phosphorus in drainage water. In: Tunney, H., Carton, O.T., Brookes, P.C., Johnston, A.E. (Eds.), Phosphorus Loss from Soil to Water. Proceedings of a Workshop, Wexford, Irish Republic, 29–31 September 1995, pp. 253–271. Butler, D.M., Ranells, N.N., Franklin, D.H., Poore, M.H., Green Jr., J.T., 2008. Runoff water quality from manured riparian grasslands with contrasting drainage and simulated grazing pressure. Agric. Ecosyst. Environ. 126, 250–260. Campbell, J.M., Jordan, P., Arnscheidt, J., 2015. Using high-resolution phosphorus data to investigate mitigation measures in headwater river catchments. Hydrol. Earth Syst. Sci. 19, 453–464. Cruickshank, J.G., 1997. Soil and Environment: Northern Ireland. Agricultural and Environmental Science Division, DANI and The Agricultural and Environmental Science Dep., The Queen's University Belfast, Newforge Lane, Belfast, UK. Daly, K., Casey, A., 2005. Environmental aspects of soil phosphorus testing. Ir. J. Agric. Food Res. 44, 261–279. Daly, K., Styles, D., Lalor, S., Wall, D.P., 2015. Phosphorus sorption, supply potential and availability in soils with contrasting parent material and soil chemical properties. European Journal of Soil Science]–>Eur. J. Soil Sci. 66, 792–801. DEFRA, 2010. In: Department of Environment FaRAD (Ed.), Fertiliser Manual (RB209). The Stationary Office, UK. Dodd, R.J., McDowell, R.W., Condron, L.M., 2012. Predicting the changes in environmentally and agronomically significant phosphorus forms following the cessation of phosphorus fertilizer applications to grassland. Soil Use Manag. 28, 135–147. Dodd, R.J., McDowell, R.W., Condron, L.M., 2013. Changes in soil phosphorus availability and potential phosphorus loss following cessation of phosphorus fertiliser inputs. Soil Res. 51, 427–436. Doody, D.G., Ferris, C.P., Watson, C.J., 2014. Nutrient losses from two contrasting dairy cow grazing systems. Biol. Environ. Proc. R. Ir. Acad. 114B, 61–70. Doody, D.G., Higgins, A., Matthews, D., Foy, R.H., Pilatova, K., Duffy, O., et al., 2010. Overland flow initiation from a drumlin grassland hillslope. Soil Use Manag. 26, 286–298. Dunne, S., Hanafin, J., Lynch, P., McGrath, R., Nishimura, E., Nolan, P., et al., 2008. In: McGrath, R., Lynch, P. (Eds.), Ireland in a Warmer World Scientific Predictions of the Irish Climate in the Twenty-First Century. Community Climate Change Consortium for Ireland (C4I), Dublin. Dupas, R., Gruau, G., Gu, S., Humbert, G., Jaffrézic, A., Gascuel-Odoux, C., 2015. Groundwater control of biogeochemical processes causing phosphorus release from riparian wetlands. Water Res. 84, 307–314. Eisenreich, S.J., Bannerman, R.T., Armstrong, D.E., 1975. A simplified phosphorus analysis technique. Environ. Lett. 9, 43–53. Field, A., Miles, J., Z.F., 2012. Discovering Statistics Using R. SAGE Publications, London. Frossard, E., Condron, L.M., Oberson, A., Sinaj, S., Fardeau, J.C., 2000. Processes governing phosphorus availability in temperate soils. J. Environ. Qual. 29, 15–23. Hahn, C., Prasuhn, V., Stamm, C., Schulin, R., 2012. Phosphorus losses in runoff from manured grassland of different soil P status at two rainfall intensities. Agric. Ecosyst. Environ. 153, 65–74. Hart, M.R., Quin, B.F., Nguyen, M.L., 2004. Phosphorus runoff from agricultural land and direct fertilizer effects. J. Environ. Qual. 33, 1954–1972. Haygarth, P.M., Jarvis, S.C., 1997. Soil derived phosphorus in surface runoff from grazed grassland lysimeters. Water Res. 31, 140–148. Haygarth, P.M., Jarvie, H.P., Powers, S.M., Sharpley, A.N., Elser, J.J., Shen, J., et al., 2014. Sustainable phosphorus management and the need for a long-term perspective: the legacy hypothesis. Environ. Sci. Technol. 48, 8417–8419. Herlihy, M., McCarthy, J., Breen, J., Moles, R., 2004. Effects over time of fertiliser P and soil series on P balance, soil-test P and herbage production. Ir. J. Agric. Food Res. 43, 147–160. Jarvie, H.P., Sharpley, A.N., Spears, B., Buda, A.R., May, L., Kleinman, P.J.A., 2013a. Water quality remediation faces unprecedented challenges from “legacy phosphorus”. Environ. Sci. Technol. 47, 8997.

Jarvie, H.P., Sharpley, A.N., Withers, P.J., Scott, J.T., Haggard, B.E., Neal, C., 2013b. Phosphorus mitigation to control river eutrophication: murky waters, inconvenient truths, and “postnormal” science. J. Environ. Qual. 42, 295–304. Johnston, A.E., Poulton, P.R., White, R.P., Macdonald, A.J., 2016. Determining the longer term decline in plant-available soil phosphorus from short-term measured values. Soil Use Manag. (n/a-n/a). Jordan, C., Rawlins, B., 2007. The soils of Northern Ireland and their environmental significance. TELLUS Conference 17-18 October 2007. Jordan, P., Melland, A.R., Mellander, P.E., Shortle, G., Wall, D., 2012. The seasonality of phosphorus transfers from land to water: implications for trophic impacts and policy evaluation. Sci. Total Environ. 434, 101–109. Jordan, P., Menary, W., Daly, K., Kiely, G., Morgan, G., Byrne, P., et al., 2005. Patterns and processes of phosphorus transfer from Irish grassland soils to rivers—integration of laboratory and catchment studies. Nutirent Mobility within River Basins: A European Perspective. 304, pp. 20–34. Kerebel, A., Cassidy, R., Jordan, P., Holden, N.M., 2013. Soil moisture deficit as a predictor of the trend in soil water status of grass fields. Soil Use Manag. 29, 419–431. King, K.W., Williams, M.R., Macrae, M.L., Fausey, N.R., Frankenberger, J., Smith, D.R., et al., 2015. Phosphorus transport in agricultural subsurface drainage: a review. J. Environ. Qual. 44. Kleinman, P., Sharpley, A., Buda, A., McDowell, R., Allen, A., 2011. Soil controls of phosphorus in runoff: management barriers and opportunities. Can. J. Soil Sci. 91, 329–338. Kleinman, P.J., Sharpley, A.N., Withers, P.J., Bergström, L., Johnson, L.T., Doody, D.G., 2015. Implementing agricultural phosphorus science and management to combat eutrophication. Ambio 44 (2), 297–310. Kurz, I., Coxon, C., Tunney, H., Ryan, D., 2005. Effects of grassland management practices and environmental conditions on nutrient concentrations in overland flow. J. Hydrol. 304, 35–50. Lane, S.N., 2008. What makes a fish (hydrologically) happy? A case for inverse modelling. Hydrol. Process. 22, 4493–4495. Lane, S.N., Brookes, C.J., Kirkby, M.J., Holden, J., 2004. A network-index-based version of TOPMODEL for use with high-resolution digital topographic data. Hydrol. Process. 18, 191–201. Lane, S.N., Brookes, C.J., Louise Heathwaite, A., Reaney, S., 2006. Surveillant science: challenges for the management of rural environments emerging from the new generation diffuse pollution models. J. Agric. Econ. 57, 239–257. Leinweber, P., Meissner, R., Eckhardt, K.U., Seeger, J., 1999. Management effects on forms of phosphorus in soil and leaching losses. European Journal of Soil Science]–>Eur. J. Soil Sci. 50, 413–424. Lilly, A., 2010. A hydrological classification of UK soils based on soil morphological data. 19th World Congress of Soil Science, Soil Solutions for a Changing World. IUSS, Brisbane, Australia, pp. 1–4. Loeb, R., Lamers, L.P.M., Roelofs, J.G.M., 2008. Prediction of phosphorus mobilisation in inundated floodplain soils. Environ. Pollut. 156, 325–331. Majdalani, S., Michel, E., Di-Pietro, L., Angulo-Jaramillo, R., 2008. Effects of wetting and drying cycles on in situ soil particle mobilization. European Journal of Soil Science]– >Eur. J. Soil Sci. 59, 147–155. McDowell, R.W., Sharpley, A.N., 2002. The effect of antecedent moisture conditions on sediment and phosphorus loss during overland flow: Mahantango Creek catchment, Pennsylvania, USA. Hydrol. Process. 16, 3037–3050. McDowell, R.W., Dils, R.M., Collins, A.L., Flahive, K.A., Sharpley, A.N., Quinn, J., 2015. A review of the policies and implementation of practices to decrease water quality impairment by phosphorus in New Zealand, the UK, and the US. Nutr. Cycl. Agroecosyst. 1-17. Mellander, P.E., Jordan, P., Shore, M., McDonald, N.T., Wall, D.P., Shortle, G., et al., 2016. Identifying contrasting influences and surface water signals for specific groundwater phosphorus vulnerability. Sci. Total Environ. 541, 292–302. Mellander, P.-E., Melland, A.R., Jordan, P., Wall, D.P., Murphy, P.N.C., Shortle, G., 2012. Quantifying nutrient transfer pathways in agricultural catchments using high temporal resolution data. Environ. Sci. Pol. 24, 44–57. Milledge, D.G., Lane, S.N., Heathwaite, A.L., Reaney, S.M., 2012. A Monte Carlo approach to the inverse problem of diffuse pollution risk in agricultural catchments. Sci. Total Environ. 433, 434–449. Murphy, J., Riley, J.P., 1962. A modified single solution method for the determination of phosphate in natural waters. Anal. Chim. Acta 27, 31–36. Olsen, S.R., Cole, C.V., Watanabe, F.S., 1954. Estimation of Available Phosphorus in Soils by Extraction with Sodium Bicarbonate. Washington, USDA. Owen, G.J., Perks, M.T., Benskin, C.M.H., Wilkinson, M.E., Jonczyk, J., Quinn, P.F., 2012. Monitoring agricultural diffuse pollution through a dense monitoring network in the River Eden Demonstration Test Catchment, Cumbria, UK. Area 44, 443–453. Pietola, L., Horn, R., Yli-Halla, M., 2005. Effects of trampling by cattle on the hydraulic and mechanical properties of soil. Soil Tillage Res. 82, 99–108. Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., R Core Team, 2015. NLME: Linear and Nonlinear Mixed Effects Models. R Package Version 3.1–121. Pote, D.H., Daniel, T.C., Moore, P.A., Nichols, D.J., Sharpley, A.N., Edwards, D.R., 1996. Relating extractable soil phosphorus to phosphorus losses in runoff. Soil Sci. Soc. Am. J. 60. Preedy, N., McTiernan, K., Matthews, R., Heathwaite, L., Haygarth, P., 2001. Rapid incidental phosphorus transfers from grassland. J. Environ. Qual. 30. R Development Core Team, 2005. R: A Language and Environment for Statistical Computing, Reference Index Version 2.15.0. R Foundation for Statistical Computing, Vienna, Austria. Reaney, S.M., Lane, S.N., Heathwaite, A.L., Dugdale, L.J., 2011. Risk-based modelling of diffuse land use impacts from rural landscapes upon salmonid fry abundance. Ecol. Model. 222, 1016–1029.

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063

R. Cassidy et al. / Science of the Total Environment xxx (2016) xxx–xxx Reid, S.C., Lane, S.N., Montgomery, D.R., Brookes, C.J., 2007. Does hydrological connectivity improve modelling of coarse sediment delivery in upland environments? Geomorphology 90, 263–282. Sallade, Y.E., Sims, J.T., 1997. Phosphorus transformations in the sediments of Delaware's agricultural drainageways: II. Effect of reducing conditions on phosphorus release. J. Environ. Qual. 26. Scalenghe, R., Edwards, A.C., Barberis, E., Ajmone-Marsan, F., 2014. Release of phosphorus under reducing and simulated open drainage conditions from overfertilised soils. Chemosphere 95, 289–294. Schärer, M., Stamm, C., Vollmer, T., Frossard, E., Oberson, A., Flühler, H., et al., 2007. Reducing phosphorus losses from over-fertilized grassland soils proves difficult in the short term. Soil Use Manag. 23, 154–164. Schulte, R.P.O., Diamond, J., Finkele, K., Holden, N.M., Brereton, A.J., 2005. Predicting the soil moisture conditions of Irish grasslands. Ir. J. Agric. Food Res. 44, 95–110. Schulte, R.P.O., Melland, A.R., Fenton, O., Herlihy, M., Richards, K., Jordan, P., 2010. Modelling soil phosphorus decline: expectations of water framework directive policies. Environ. Sci. Pol. 13, 472–484. Sharpley, A., Jarvie, H.P., Buda, A., May, L., Spears, B., Kleinman, P., 2013. Phosphorus legacy: overcoming the effects of past management practices to mitigate future water quality impairment. J. Environ. Qual. 42, 1308–1326. Sibbesen, E., Sharpley, A.N., 1997. Setting and justifying upper critical limits for phosphorus in soils. In: Tunney, H., Carton, O.T., Brookes, P.C., Johnston, A.E. (Eds.), Phosphorus Loss from Soil to Water. Proceedings of a Workshop, Wexford, Irish Republic, 29– 31 September 1995, pp. 151–176. Sinclair, A.H., Crooks, B., Edwards, T.C., Lumsdon, D., Shand, C., 2012. Assessing the potential risks to water quality from phosphate leaching. In: Consulting, S. (Ed.), SEPA Contract Number R11123STA. SEPA, Aberdeen, p. 82. Smith, V.H., Schindler, D.W., 2009. Eutrophication science: where do we go from here? Trends Ecol. Evol. 24, 201–207. Smith, R.V., Lennox, S.D., Bailey, J.S., 2003. Halting the upward trend in soluble phosphorus transported from a grassland catchment. J. Environ. Qual. 32, 2334–2340. Tarboton, D.G., 2015. TauDEM Toolbox in ArcGIS (TauDEM 5.3). Utah State University. Thomas, I.A., Jordan, P., Mellander, P.E., Fenton, O., Shine, O., Ó hUallacháin, D., et al., 2016. Improving the identification of hydrologically sensitive areas using LiDAR DEMs for the delineation and mitigation of critical source areas of diffuse pollution. Sci. Total Environ. 556, 276–290.

11

Thompson, J., Cassidy, R., Doody, D.G., Flynn, R., 2013. Predicting critical source areas of sediment in headwater catchments. Agric. Ecosyst. Environ. 179, 41–52. Thompson, J.J.D., Doody, D.G., Flynn, R., Watson, C.J., 2012. Dynamics of critical source areas: does connectivity explain chemistry? Sci. Total Environ. 435, 499–508. Torpey, P., Morgan, M.A., 1999. Phosphorus profile of water, soil and sediment in a farm drainage system. In: O′Kiely, P. (Ed.), Agricultural Research Forum. University College Dublin, pp. 25–26. Tunney, H., Carton, O., O'Donnell, T., Fanning, A., 1998. Phosphorus Loss from Soil to Water. End of Project Report. Teagasc, 19 Sandymount Avenue, Ballsbridge, Dublin 4. Tunney, H., Daly, K., Kurz, I., 2002. The Role of Agriculture in Phosphorus Eutrophication of Surface Water 17th WCSS 14-21 August 2002, Thailand. pp. 1–10. Turner, B.L., Haygarth, P.M., 2001. Biogeochemistry: phosphorus solubilization in rewetted soils. Nature 411, 258. Uhlmann, D., Lothar, P., 1994. Causes and effects of “nitrate saturation” in phosphatedeficient water bodies. Water Sci. Technol. 30, 281–288. Ulén, B., Bechmann, M., Fölster, J., Jarvie, H.P., Tunney, H., 2007. Agriculture as a phosphorus source for eutrophication in the north-west European countries, Norway, Sweden, United Kingdom and Ireland: a review. Soil Use Manag. 23, 5–15. Vadas, P.A., Jokela, W.E., Franklin, D.H., Endale, D.M., 2011. The effect of rain and runoff when assessing timing of manure application and dissolved phosphorus loss in runoff1. JAWRA J. Am. Water Res. Assoc. 47, 877–886. Venterink, H.O., Davidsson, T.E., Kiehl, K., Leonardson, L., 2002. Impact of drying and rewetting on N, P and K dynamics in a wetland soil. Plant Soil 243, 119–130. Wade, A.J., Palmer-Felgate, E.J., Halliday, S.J., Skeffington, R.A., Loewenthal, M., Jarvie, H.P., et al., 2012. Hydrochemical processes in lowland rivers: insights from in situ, highresolution monitoring. Hydrol. Earth Syst. Sci. 16, 4323–4342. Watson, C.J., Jordan, C., Lennox, S.D., Smith, R.V., Steen, R.W.J., 2000. Inorganic nitrogen in drainage water from grazed grassland in Northern Ireland. J. Environ. Qual. 29, 225–232. Watson, C.J., Smith, R.V., Matthews, D.I., 2007. Increase in phosphorus losses from grassland in response to Olsen-P accumulation. J. Environ. Qual. 36, 1452–1460. Withers, P.J.A., Neal, C., Jarvie, H.P., Doody, D.G., 2014. Agriculture and eutrophication: where do we go from here? Sustainability 6, 5853–5875. Zuur, A.F., Ieno, E.N., Walker, N.J., Saveliev, A.A., Smith, G.M., 2009. Mixed Effects Models and Extensions in Ecology with R. Springer, New York.

Please cite this article as: Cassidy, R., et al., Impact of legacy soil phosphorus on losses in drainage and overland flow from grazed grassland soils, Sci Total Environ (2016), http://dx.doi.org/10.1016/j.scitotenv.2016.07.063