JEQ Grow Your Career With ASA
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 1 March 2008
Published in J Environ Qual 37:417-428 (2008)
DOI: 10.2134/jeq2007.0049
© 2008 American Society of Agronomy, Crop Science Society of America, and Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Dougherty, W. J.
Right arrow Articles by Lawrie, R. A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Dougherty, W. J.
Right arrow Articles by Lawrie, R. A.
Agricola
Right arrow Articles by Dougherty, W. J.
Right arrow Articles by Lawrie, R. A.
Related Collections
Right arrow Water Quality
Right arrow Phosphorus
Right arrow Nutrient Management

TECHNICAL REPORTS

Surface Water Quality

Phosphorus Fertilizer and Grazing Management Effects on Phosphorus in Runoff from Dairy Pastures

Warwick J. Doughertya,*, Paul J. Nichollsc, Paul J. Milhama,b, Euie J. Havilahd and Roy A. Lawriea

a New South Wales Dep. of Primary Industries, Locked Bag 4, Richmond, NSW, Australia 2753
b Centre for Plant and Food Science, Univ. of Western Sydney, LB 1797, Penrith South DC, NSW, Australia 1797
c New South Wales Dep. of Primary Industries, Elizabeth Macarthur Agricultural Inst., PMB 8, Camden, NSW, Australia 2570
d New South Wales Dep. of Primary Industries, Pasture Research Unit, PO Box 63, Berry, NSW, Australia 2535

* Corresponding author (warwick.dougherty{at}dpi.nsw.gov.au).

Received for publication January 29, 2007.

   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 Materials and Methods
 Results and Discussion
 Discussion
 Conclusions
 REFERENCES
 
Fertilizer phosphorus (P) and grazing-related factors can influence runoff P concentrations from grazed pastures. To investigate these effects, we monitored the concentrations of P in surface runoff from grazed dairy pasture plots (50 x 25 m) treated with four fertilizer P rates (0, 20, 40, and 80 kg ha–1 yr–1) for 3.5 yr at Camden, New South Wales. Total P concentrations in runoff were high (0.86–11.13 mg L–1) even from the control plot (average 1.94 mg L–1). Phosphorus fertilizer significantly (P < 0.001) increased runoff P concentrations (average runoff P concentrations from the P20, P40, and P80 treatments were 2.78, 3.32, and 5.57 mg L–1, respectively). However, the magnitude of the effect of P fertilizer varied between runoff events (P < 0.01). Further analysis revealed the combined effects on runoff P concentration of P rate, P rate x number of applications (P < 0.001), P rate x time since fertilizer (P < 0.001), dung P (P < 0.001), time since grazing (P < 0.05), and pasture biomass (P < 0.001). A conceptual model of the sources of P in runoff comprising three components is proposed to explain the mobilization of P in runoff and to identify strategies to reduce runoff P concentrations. Our data suggest that the principal strategy for minimizing runoff P concentrations from grazed dairy pastures should be the maintenance of soil P at or near the agronomic optimum by the use of appropriate rates of P fertilizer.

Abbreviations: AN, application number • BM, pasture biomass • DgP, dung P since fertilizer • DRP, dissolved reactive P • FP, fertilizer P rate • NSW, New South Wales • PP, particulate P • TDP, total dissolved P • TP, total P • TSF, time since fertilizer application • TSG, time since grazing


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
Materials and Methods
 Results and Discussion
 Discussion
 Conclusions
 REFERENCES
 
ELEVATED concentrations of phosphorus (P) in surface waters can contribute to increases in algal growth and consequently a decline in the environmental health and amenity of waterways. Nutrient concentrations in surface waters of coastal catchments in New South Wales (NSW), Australia are often above desirable levels (Anonymous, 1998). In response to the declining surface water quality in several catchments, the Healthy Rivers Commission set a broad objective to decrease inputs of P into waterways from the agricultural areas of these catchments (Anonymous, 1998).

The NSW dairy industry is an intensive, pasture-based enterprise located predominantly in coastal catchments in the southeast of Australia (Fig. 1 ). The intensity of dairy production is being increased partly by greater use of P fertilizer to increase pasture production. As a consequence of increasing use of P fertilizers, dairy farms are often net accumulators of P (Nash and Halliwell, 1999; White and Gourley, 2001; Lawrie et al., 2004). The majority of this excess P accumulates in the topsoil of grazing paddocks (Sharpley, 2003; Lawrie et al., 2004; Dougherty et al., 2006). Because surface runoff is the dominant pathway for P export from soils used for dairying in southeastern Australia (Fleming and Cox, 1998; Stevens et al., 1999), an accumulation of P in the topsoil is likely to increase the pool of mobilizable P and hence increase runoff P concentrations and consequently increase P exports (Sharpley, 1995; Pote et al., 1999). For example, increases in soil P resulting from fertilizer and manure inputs under dairy pastures in southeast Australia resulted in severalfold increases in P concentrations in runoff from simulated rainfall (Dougherty et al., 2006).


Figure 1
View larger version (27K):
[in this window]
[in a new window]

  		Fig. 1. Location of the research site in relation to the main dairying regions (shaded) in southeast Australia.

 
The concentration of P in runoff from dairy pastures may be affected by factors other than the quantity of mobilizable P in the topsoil, including P contained in dung and plants (Sharpley, 1981; Mundy et al., 2003; McDowell et al., 2007), the time since fertilizer application (TSF) and time since grazing (TSG) (Nash et al., 2005; McDowell et al., 2007), and runoff hydrological characteristics (Bush and Austin, 2001; Fleming et al., 2001). Studies in southeast Australia have examined a number of these factors in grazing situations (Nash and Halliwell, 1999; Fleming and Cox, 2001; Fleming et al., 2001; Mundy et al., 2003; Nash et al., 2005). Given the potential effects of these various factors, it is notable that a number of these studies have considered only limited combinations of these factors. Exceptions are the recent studies of Nash et al. (2005) and McDowell et al. (2007) that illustrate the combined effects of a few of these factors and their relative importance. However, the importance of changes in soil P status or fertilizer P rates were not explicitly considered in these studies.

Given that there is generally little opportunity to influence hydrology and runoff volumes, and consequently P exports from these production systems (Nash et al., 2005), reducing or limiting increases in the concentrations of P in runoff is the main management strategy to minimize the impact of runoff on waterways. To most efficiently reduce runoff P concentrations, an understanding of the range of factors influencing runoff P concentration and their relative importance is required (McDowell et al., 2007).

The objectives of this research were to (i) assess the effect of increasing rates of fertilizer P on runoff P when the fertilizer P is applied at regular intervals and (ii) with fertilizer P accumulating in the soil at various rates, to assess the effects on runoff P of several factors associated with management, namely the number of previous fertilizer applications, TSF, TSG, dung P deposited since fertilizer application (DgP), and pasture biomass (BM). The interactions of these management-associated factors with fertilizer P rate are also examined.


   Materials and Methods
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and Methods
Results and Discussion
 Discussion
 Conclusions
 REFERENCES
 
Location and Soil
Six adjacent runoff plots (50 x 25 m) were located within a grazing-systems experiment at Camden, NSW (Fig. 1). The objective of the grazing-systems experiment was to evaluate the effect of P fertilizer application at various rates on pasture production while a target milk production was maintained. The runoff plots were grazed simultaneously as part of the rotational grazing in the broader study.

Average annual rainfall at the site is 793 mm; rainfall is slightly summer dominant and is highly erratic. The soils are Brown Chromosols (Isbell, 1997) or Haploxeralfs (Soil Survey Staff, 1999). The A1 horizon is 12 to 20 cm deep and is a moderately aggregated, gray-brown, clay-loam that sets hard when dry. Selected chemical properties of the surface soil (0–10 cm) at the start of the experiment are listed in Table 1 . The mean initial Olsen P for all six plots was 17 mg kg–1, which is within the suggested agronomic optimum of 14 to 17 mg kg–1 (Gourley et al., 2007).


View this table:
[in this window]
[in a new window]

  		Table 1. Summary of soil chemical properties (0–2 and 0–10 cm) for the study site at the start of the experiment.

 
Fertilizer and Grazing Management
Phosphorus fertilizer was broadcast in two equal applications in autumn and spring of each year to achieve annual application rates of 0, 20, 40, or 80 kg ha–1 P (P0, P20, P40, and P80, respectively). The fertilizer was a mixture of single and triple superphosphate—both types being highly water soluble (Nash and Halliwell, 1999)—applied in ratios to balance calcium inputs to the plots. Plots were stocked at a rate of 2.2 cows ha–1. Cows were fed concentrates while being milked and fed silage on a feed-pad (as needed to maintain a target annual milk production of 13.2 kL ha–1) before returning to pasture. Pasture biomass was measured before and after each grazing and after runoff events using a calibrated rising plate meter (Fulkerson and Slack, 1993). The pastures were mixtures of perennial ryegrass (Lolium perenne L) and white clover (Trifolium repens L). Irrigation was used opportunistically to support pasture growth during dry periods.

Runoff Plots
Of the six runoff plots, there was one each of the P0 and P80 treatments and two each of the P20 and P40 treatments, with treatments allocated to the plots at random. The runoff plots were set up early in January 1999, and any disturbed areas were stabilized by re-sowing pasture. Monitoring of P in runoff commenced in June 1999 and continued until the end of December 2002. The plots were 50 m long x 25 m wide, with an average slope of 5% (±0.5%), and were located in a mid-slope position. They were isolated from surface run-on by plastic strips (0.5 cm thick and 40 cm wide) buried vertically ~30 cm into the soil. An earthen drainage bank ~25 m above the plots intersected the B-horizon to ~30 cm and minimized throughflow inputs to the plots. Surface runoff from each plot was intercepted using a semi-circular PVC drain (30 cm diam.) that ran the full width of the plot. Water from the drains then flowed through 15-cm RBC flumes (Clemmens et al., 1984). Depth in the flumes was measured and logged at 5-min intervals. Water samples were collected from the flume outlets at evenly spaced set volumes using autosamplers.

Runoff Events
Runoff events occurred as a result of rainfall (n = 12) or irrigation (n = 7). The artificial (irrigation) runoff events were generated to provide supplementary information on the effects of factors influencing runoff P concentrations. They were generated by applying water at ~8 mm h–1 for ~10 h using overhead sprinklers and represented rainfall events with an average recurrence interval of 5 yr (Pilgrim, 1987). The coefficient of uniformity of application of water by this method was >60%. A runoff event was defined as an event where there was a time period greater than 12 h between cessation of runoff from a prior runoff event or recommencement of runoff in a subsequent event.

Analysis of Runoff Samples
Runoff samples were subsampled, filtered (<0.45 µm), and analyzed for dissolved molybdate-reactive P (DRP) within 12 h of collection. Total P (TP) and total dissolved P (TDP) were determined on unfiltered and filtered samples, respectively, that were stored frozen (–20°C, maximum 1 wk) until they were thawed at room temperature and were then digested using acidic persulfate (APHA, 1995). Phosphorus was determined colorimetrically (Murphy and Riley, 1962). Particulate P (PP) was calculated as TP – TDP.

Concentration and Load Calculation
For each runoff event and plot, the flow-weighted concentration of each of the P forms was calculated. Total P concentration in artificial runoff was at least eight times greater than that in the irrigation water (<0.1 mg TP L–1); consequently, no adjustment was made to the runoff P concentrations. Average annualized loads (kg ha–1 yr–1) of P in runoff were calculated for each plot by summing the load for all years and dividing by 3.5 (the number of years of the experiment). Only the loads from the natural runoff events were used in these calculations.

Soil Sampling and Analysis
Soils were sampled (0–10 cm and 0–2 cm) before the initial application of fertilizer at the start of the experiment and again at the end of the experiment (6 mo after the final application of fertilizer). For each plot, 30 cores (2.5 cm diam.) were collected along a permanent sampling transect within the plot and composited. Undecomposed plant debris was removed from the surface before sampling, and dung and urine patches were avoided. Soil samples were dried at 40°C, then ground and passed through a 2-mm sieve to remove stones and plant debris. The samples were stored at 4°C before analysis.

Soil pH in deionized water and in 0.01 mol L–1 CaCl2 and electrical conductivity were determined in 1:5 extracts. Organic carbon was estimated using the method of Walkley and Black (1934). Plant-available P was determined using the method of Olsen et al. (1954). A measure of readily desorbable P (CaCl2–P) was made by determining molybdate reactive P (Murphy and Riley, 1962) in a soil extract that was prepared by a 30-min extraction at a 5:1 solution:soil ratio in 10 mmol L–1 CaCl2 followed by centrifugation and filtration (<0.45 µm).

Dung Phosphorus Budgets
Dung P budgets were calculated because differences in pasture production between P treatments meant that cows grazing the lower P plots were fed more supplementary feed, which had a higher concentration of P than the pastures they grazed. As a result, the P content of dung on the lower P plots was higher than the P content of the pastures alone would suggest (there was a significant positive relation between fertilizer rate and biomass P concentration; data not shown). The dung P budgets were calculated for each plot using pasture and feed consumption data. The total P consumed was partitioned between animal maintenance, milk production, and excretion as dung. Dung P was apportioned to the plots in proportion to the time spent on the plots. These calculations were made for all six runoff plots for each grazing event. For each runoff event, dung P on each plot was accumulated over all the grazing events since the most recent fertilizer application.

Statistical Methods
General
There are two features of our concentration data that are potential sources of dependence: Among the runoff events, observations for each P rate and replicate were made on the same plot, and among the plots, observations were made within the same event. The statistical methods that permit data subject to such dependencies to be analyzed correctly are termed linear mixed models (Verbyla et al., 1999). With plot effects included in the linear mixed model as random effects, the occurrence of a significant plot variance component would indicate the presence of correlation among the repeated observations within plots. Similarly, with effects of the events included as random effects, a significant event variance component would indicate the presence of correlation among the simultaneous plot observations within events. In linear mixed models for data like ours, the effects of interest are generally the fixed effects (e.g., an effect of fertilizer), and inclusion of appropriate random effects ensures that the fixed effects are assessed appropriately in the presence of correlations such as those described previously.

Effect of Fertilizer P Rate on Runoff P Concentrations and Forms within Runoff Events
A single linear mixed model analysis was made of the 17 postfertilizer runoff events. In this model, with events included as fixed effects, log10 TP was regressed on fertilizer P rate (FP) for each of the individual events, random effects for each of the six plots were included, and deviations from the regressions were pooled into a single residual error term. In an alternative mixed model for the same data, again with events included as fixed effects, the individual regressions were replaced with an overall linear regression of log10 TP on FP and interactions between the events and the intercept and slope of the regression. In addition, a random cubic spline term for assessing curvature of the overall regression, a fixed effect of runoff volume, and random effects of the plots were included. The mixed modeling approach, including the use of a cubic smoothing spline to assess curvature, was based on the methods of Verbyla et al. (1999), and all analyses were performed using ASReml (Gilmour et al., 2006). The components DRP and PP were expressed as proportions of TP and, after a logit transformation (log of [%PP/(1 – %PP)]), were analyzed in the same way as log10 TP.

In a comparison of the linear relations between log10 TP and FP for two pairs of artificial and natural runoff (events 9 and 10 and events 18 and 19), tests of parallelism and coincidence of the regressions of log10 TP on FP were made between the events within each pair using a mixed model analysis of the data from the four events that included random plot effects.

Effects of Management-Associated Factors on Runoff P Concentrations
A linear mixed model analysis was applied to the TP data combined from the 17 postfertilizer events. The model consisted of the fixed linear effects of FP (kg ha–1 yr–1), number of previous fertilizer applications (AN), TSF (days), TSG (days), BM (kg ha–1), and DgP (kg ha–1), together with the interactions of FP with each of the other five effects and an effect of event type (natural or artificial). Random effects included cubic splines for each fixed effect except FP, interactions of the type FP x spline (AN) corresponding to each fixed interaction, a lack of fit term for each fixed effect, and plot effects. For the purpose of investigating the presence of dependency within events in this analysis, events on consecutive days (21 and 22 Mar. 2000 and 5–7 Feb. 2002) were treated as subsamples of the one event, so there were two events. Random effects of the resulting 14 events were added to the model. The possibility of trends between days within the multiday events was also considered, so fixed effects for the days within the two relevant events were also added to the model. The initial model was reduced to a final model by deleting nonsignificant fixed effects, spline terms, and interactions (P > 0.05), but all nonzero lack of fit terms were retained irrespective of their level of significance. The modeling method and the software used were as cited for the earlier analyses. Dissolved reactive P and PP were expressed as proportions of TP and, using a logit transformation, were modeled in the same way as log10 TP.


   Results and Discussion
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
Discussion
 Conclusions
 REFERENCES
 
Soil
Phosphorus fertilizer application increased Olsen P and CaCl2–P for both sampling depths (0–2 and 0–10 cm), as illustrated by the data at the end of the experiment (Fig. 2 ). Relatively large increases in soil test P were observed in the 0- to 2-cm layer, which is consistent with the surface application of P, the strong affinity of soil for P, and the absence of cultivation in these pasture systems (Sharpley, 2003; Dougherty et al., 2006). Because the concentration of P in runoff is controlled by soil P concentration at the immediate soil surface (i.e., typically a depth <2 cm) (Sharpley, 1985; Dougherty, 2006), the observed preferential accumulation of P at the soil surface would exacerbate the effects of P application on runoff P (compared with the added P being distributed evenly throughout the top 10 cm).


Figure 2
View larger version (13K):
[in this window]
[in a new window]

  		Fig. 2. Soil P concentrations at start of the experiment (T0 = average for all plots) and after 3 yr of P application for the 0- to 10-cm and 0- to 2-cm soil depths for each of the treatments (bars represent SE). The effect of P rate on CaCl2–P and Olsen P after 3 years was significant (P < 0.01).

 
Runoff Events
General Characteristics
Annual rainfall totals during the experiment were 629, 517, 466, and 480 mm for 1999, 2000, 2001, and 2002, respectively, and were well below the 50-yr average of 793 mm. There were 12 naturally occurring runoff events and seven artificial runoff events during the monitoring period. After the start of fertilizer application, there were 17 runoff events (Table 2 ; Fig. 3 ). Total P in runoff (flow-weighted concentrations over the postfertilizer application monitoring period) from the P0 plot was 1.94 mg L–1 (ranging from 0.86 to 3.01 mg L–1), whereas total P in runoff from the P80 plot was 5.57 mg L–1 (ranging from 1.78 to 11.13 mg L–1).


View this table:
[in this window]
[in a new window]

  		Table 2. Timing and main features of the runoff events, both natural and artificial (characteristics are averages for all six plots).

 

Figure 3
View larger version (22K):
[in this window]
[in a new window]

  		Fig. 3. Mean runoff total phosphorus (TP) concentration for each of the treatments for all runoff events. SE bars are presented. Artificial runoff events are shown in light gray.

 
Runoff events occurred over a wide range of cumulative P fertilizer levels and a large range of management conditions (Table 2). Among the runoff events on individual plots, TSF ranged from 3 to 209 d, and TSG ranged from 0 to 95 d. Estimated BM ranged from 543 to 2731 kg ha–1. Dung P accumulated on the plots since the last fertilizer application ranged from 0.0 to 12.4 kg P ha–1.

Prefertilizer Runoff Events
Two runoff events occurred (1 and 14 July 1999) before fertilizer treatments were applied. The average TP concentrations in these events were 3.16 ± 0.16 and 3.05 ± 0.18 mg L–1, respectively. For these events, there was no significant association between runoff TP concentrations of the plots and their subsequent fertilizer P rate. Similarly, there was no significant association between initial soil P concentration and subsequent fertilizer P rate. These observations provide evidence that there was negligible bias in the postfertilizer relations between runoff TP and applied P rate due to the allocation of treatments to plots.

Effect of Event Type and Event Characteristics
On two occasions, a natural runoff event occurred within a short time of an artificial event (on 11 and 27 July 2001 and on 5 and 10 Dec. 2002; Fig. 3). On each occasion, the P concentrations were very similar for the artificial and natural events. There were no significant differences (P > 0.30) in the slopes and the intercepts of the relation between log10 TP and FP for these events. Furthermore, inclusion of an "event type" term (artificial or natural) in the linear mixed model for all events that investigated the management effects found this term to be not significant (P > 0.5). We have interpreted these results as indicating that, for each of the seven artificial runoff events, runoff TP values were comparable with those had a natural event occurred, which supports our use of the combined data from natural and artificial events for examining effects of the management factors.

The effect of runoff event volume on runoff log10 TP concentration was tested in the alternative linear mixed model and was found to be nonsignificant (P > 0.50). Furthermore, inspection of plots of runoff P concentration within an event against time, runoff volume, and runoff rate revealed no consistent trends in P concentration for any of the events or any relationship between event runoff volume and fertilizer P rate. Consequently, it was assumed that P concentration was independent of the hydrological characteristics of events, and the influence of these factors was not considered further. The finding that P concentration is independent of the hydrological characteristics supports the use of runoff P concentration as the independent variable in our analysis and is consistent with the findings of Nash et al. (2005).

Effect of Fertilizer P Rate on Runoff P Concentration and Form within Runoff Events
Runoff P Concentration
Additions of fertilizer P increased the concentration of P in runoff. However, there were highly significant (P < 0.001) differences among events in the slope and intercept of the relation with FP. Among the 17 postfertilizer events, the slopes of the relations between log10 TP and FP ranged from 0.0018 ± 0.0011 (P < 0.20, event 5) to 0.0108 ± 0.0011 (P < 0.001, event 17), with an average of 0.0057 ± 0.0006 (P < 0.001). There was no curvature (P > 0.5) about the overall linear effect of fertilizer P rate on runoff log10 TP.

The mean TP values of each FP rate for each of the runoff events shown in Fig. 3 illustrate that variation in the relation with FP among events is not only due to the influence of increasing cumulative fertilizer P applications but also results from other conditions at the time the events occurred. For example, the relatively high slope of the relation between log10 TP and FP for events 7 and 17 (16 Nov. 2000 and 21 Nov. 2002, respectively) is likely to be at least partly associated with the short time since fertilizer application (5 and 15 d, respectively) as reported in the literature ( Nash et al. [2005] and Preedy et al. [2001]). However, further inspection of the data suggested that other factors influenced runoff P concentrations. Consequently, the effects on log10 TP of five factors associated with cumulative P fertilizer and grazing management were examined as outlined in the Statistical Methods, and the results are presented and discussed in the following sections.

P Forms
Dissolved reactive P accounted on average for 86% of the runoff P (53–100%). Eighty-five percent of runoff samples had more than 75% of P present as DRP. In similar pasture systems, Nash and Murdoch (1997) reported that DRP accounted for >90% of TP, whereas Fleming and Cox (1998) reported DRP percentages as low as 50%. The lower proportions of DRP in this latter study were due to greater proportions of PP caused most likely by relatively poor ground cover.

Within four of the events there was a significant relation (P < 0.05) between logit DRP and P rate (i.e., 0.0234 ± 0.0095 for event 3, 0.0216 ± 0.0095 for event 4, 0.0191 ± 0.0095 for event 11, and –0.0249 ± 0.0095 for event 6). When tested across all events, differences among the slopes and intercepts of the relations were significant (P < 0.05 and P < 0.001, respectively). However, the slope of the overall relation between logit DRP and P rate (0.0056 ± 0.0041) was nonsignificant (P > 0.20). Dougherty et al. (2006) reported an increasing proportion of DRP with increasing soil P status from intensively managed pastures. Across all runoff events in our study, PP accounted for only 8% of TP on average. The average slope of the relation between logit PP and P rate was –0.011 ± 0.005 (P < 0.05).

In our study, the occurrences of a relatively low percentage of P as DRP on several occasions were generally not associated with particular events or plots. On those occasions when the DRP percentage was low, it was not clear why this was so because the plots exhibited no obvious features that would predispose them to particulate P movement, such as a short TSG and/or low BM (McDowell et al., 2003; Kleinman et al., 2005; McDowell et al., 2007). The dominance of DRP over other forms of P in the runoff highlights the potential difficulty in decreasing P concentrations in runoff from dairy pastures using methods such as buffer strips that are designed to primarily reduce P transport by trapping particulate P.

Modeled Combined Fertilizer P and Management Effects on Runoff TP
The effects of fertilizer P and the management factors on runoff log10 TP were jointly estimated using a linear mixed model as described in the Statistics section of Materials and Methods. In the final model, all the management-associated factors had a significant linear effect or a significant interaction with FP (P < 0.001), and TSF also had significant curvature about the linear effect (P < 0.05). Estimated coefficients of the terms in the final model are shown in Table 3 (except for the TSF spline, which is nonparametric). These coefficients estimate the partial effect of each factor or interaction after adjusting for any correlated effects of the other fixed terms in the model. The sizes of the lack-of-fit variance components for each factor are also shown in Table 3, relative to the residual error variance. All but two lack-of-fit variances were non-zero, and in those cases the error variance used for testing the effect of the factors consisted of the residual error plus lack-of-fit. The limited number of runoff events available for estimating the effects of the factors in our analysis necessitates that care be exercised in extending our coefficient estimates to other sites. Nevertheless, the effects our analyses are generally similar to those identified by others (Austin et al., 1996; Mundy et al., 2003; Nash et al., 2005; McDowell et al., 2007), lending weight to our findings. The estimate of residual error variance in the model was 0.00063, and the plot variance component was 1.74 times that of the residual error, which indicates that there was a degree of correlation among observations within plots. The variance component for the random effects of events that was added to the model was zero, indicating there was no dependence among plot observations within runoff events. The effects of days within the multi-day events were nonsignificant (P > 0.20).


View this table:
[in this window]
[in a new window]

  		Table 3. Estimates of the linear effects and interactions of factors influencing runoff log10 total phosphorus concentration and sizes of the lack-of-fit variance components relative to the residual error variance.

 
The estimated trends in runoff TP for each of the factors in the model, together with their 95% confidence limits, were back-transformed from the logarithmic scale and are shown in Fig. 4 –6. The partial residuals (i.e., the data with the effects of other model terms subtracted) are included in these figures. For any particular factor, the estimates were evaluated at the mean values of the other factors, namely: FP, 33.3 kg P ha–1 yr–1; AN 4.6; TSF, 81 d; BM, 1463 kg DM ha–1; DgP, 7.7 kg ha–1; and TSG, 51 d.


Figure 4
View larger version (14K):
[in this window]
[in a new window]

  		Fig. 4. Modeled effects of application number on runoff total phosphorus (TP) concentration for each rate of P (thin lines are 95% confidence intervals). Effects were estimated at the mean values of the other factors included in the model and back-transformed from the logarithmic scale. Application number refers to an application of half the rates of P fertilizer because fertilizer was applied as two split applications per year. Partial residuals, which subtract the effects of the other model terms from the TP data, are plotted for each rate of P.

 
Fertilizer P Rate and Application Number
The estimate of the overall effect of FP on runoff log10 TP in the model with the management effects included was positive (P < 0.001) (Table 3), which is consistent with the effect of FP found in our earlier analysis. The overall effect of AN was also positive, but there was a positive and highly significant FP/AN interaction (P < 0.001) (Table 3 and Fig. 4), which indicates that not only did runoff TP tend to increase with the application of fertilizer P, but also that the increase was greater the larger the P rate. This is consistent with a decrease in soil P buffering capacity with increased cumulative application of P (Barrow et al., 1998).

Time since Fertilizer
The linear component of the effect of TSF on log10 TP was negative (Table 3), and there was curvature about the linear trend as indicated by the significant TSF spline term (P < 0.05) (Table 3). The rate of decrease in runoff TP with increasing TSF was more pronounced sooner than later after fertilizer application (Fig. 5 ). Furthermore, there was a negative and highly significant interaction between the linear effects of FP and TSF (P < 0.001) (Table 3), indicating that the magnitude of the linear component of the decrease in runoff log10 TP with increasing TSF was greater the higher the P fertilizer rate.


Figure 5
View larger version (17K):
[in this window]
[in a new window]

  		Fig. 5. Modeled effects of time since fertilizer on runoff total phosphorus (TP) concentration for each rate of P (thin lines are 95% confidence intervals). Effects were estimated at the mean values of the other factors included in the model and back-transformed from the logarithmic scale. The TP scale varies among the five plots. Partial residuals, which subtract the effects of the other model terms from the TP data, are plotted for each rate of P.

 
Although the large effect that P fertilizer can have on runoff P concentrations when runoff occurs soon after application has been widely observed (Preedy et al., 2001; Nash et al., 2005), there has been little description of an interaction between fertilizer P rate and TSF. The data of Austin et al. (1996) also support the existence of such an interaction. They reported that between 1 and 8 d after fertilizer application, the TP concentration decreased from ~60 to ~7 mg P L–1 for a 50 kg P ha–1 treatment, whereas TP decreased from ~100 to ~13 mg P L–1 for a 100 kg P ha–1 treatment.

The high concentration of P observed in runoff occurring soon after fertilizer application is the result of mobilization of P that is poorly equilibrated with the soil (Haygarth and Sharpley, 2000). Within the first 24 h after the application of P fertilizer, most of the P has moved from the fertilizer granules into the soil (Lawton and Vomocil, 1954). Thereafter, the decline in the availability of freshly added fertilizer P to be mobilized in runoff is related to the rate at which it is rendered unavailable by reaction with the soil matrix. Like other chemical reactions, the rate of reaction of P with soil is concentration dependent (Bramley et al., 1992), and on this basis the FP/TSF interaction we report is to be expected.

A continuing decline in runoff TP beyond ~20 d after fertilizer application has generally not been detected in other research. For example, analyses of the effect of increasing TSF on P concentrations in runoff from grazed sites in southeast Australia by Nash et al. (2000; 2005) revealed little apparent decline beyond 20 d after fertilizer application. The continuing decline in runoff TP concentrations beyond ~20 d revealed by the analysis of our data is likely to be the result of the continuing relatively slow reaction of fertilizer P with the soil (Bramley et al., 1992; Burkitt et al., 2002). Furthermore, our analysis suggests that by 200 d after fertilizer application, there was relatively little difference in runoff TP concentrations among the fertilizer P rates (Fig. 5).

Dung P
There was a positive effect of DgP (P < 0.001) on log10 TP (Table 3 and Fig. 6a ). Increasing the load of dung P has previously been shown to increase the concentration of P in runoff at short times since dung application (Ebeling et al., 2002; Mundy et al., 2003). These authors also found that the effects of dung declined with time. However, the limited data in our study precluded an examination of this aspect of the effect of DgP. It is possible that dung P content and hence cumulative dung P loadings could be reduced in dairy production systems by manipulation of livestock dietary rations (Ebeling et al., 2002), which may reduce the concentrations of P in runoff. Conversely, the trend toward increased stocking rates in grazing systems in the Australian dairy industry (ABARE, 2006) may increase the cumulative dung P loadings and consequently increase runoff P concentrations.


Figure 6
View larger version (12K):
[in this window]
[in a new window]

  		Fig. 6. Modeled effects of dung P, pasture biomass, and time since grazing on runoff total phosphorus (TP) concentration (thin lines are 95% confidence intervals). Effects were estimated at the mean values of the other factors included in the model and back-transformed from the logarithmic scale. Partial residuals, which subtract the effects of the other model terms from the TP data, are plotted in each figure.

 
Effect of Pasture Biomass
There was a significant negative effect (P < 0.001) of BM on log10 TP (Table 3 and Fig. 6b). The P contained in biomass has been variously proposed to affect runoff P concentrations (Bromfield and Jones, 1972; Nash and Halliwell, 1999; McDowell et al., 2007). Although there is good evidence that rain leaches P from plant biomass, Sharpley (1981) showed that subsequent interaction of rainfall/runoff that contains the biomass P with the soil surface means that it is ultimately the soil that controls runoff P concentration. Therefore, we infer that P leaching from the biomass in our pastures was not directly influencing runoff P concentrations.

There was no significant effect (P > 0.05) of BM on the proportion of PP, indicating that the negative effect of BM on TP was not the result of a decreasing mobilization of PP. One possible explanation for thenegative relationship found between BM and TP is the influence of microbial activity on soluble P in soil. Perrott et al. (1990; 1992) found that soil microbial P could vary by up to 40 mg kg–1 between seasons. Our measures of CaCl2–P, which are an estimate of the potentially mobile pool of P, were ≤27 mg kg–1; therefore, it is conceivable that microbial activity could substantially affect the size of this potentially mobile pool of P and hence runoff P concentrations. We did not measure soil microbial P in our study, but it is likely that greater pasture biomass was associated with periods when soil moisture and temperature favored soil microbial activity, which may have decreased the amount of labile P in the surface soil resulting in lower runoff P concentrations.

Time since Grazing
The effect of TSG on log10 TP was significant (P < 0.05) (Table 3 and Fig. 6c). A declining effect of TSG on runoff P concentrations has been reported by other authors. A weak negative effect of TSG was reported by Nash et al. (2000), who observed that increasing TSG from 0 to 20 d decreased runoff TP by <2 mg L–1. Similarly,McDowell et al. (2007) reported decreasing concentration of P in runoff as TSG increased. A negative effect of TSG may derive from a diminishing effect of P leakage from tissues damaged during grazing (Mundy et al., 2003; McDowell et al., 2007), physical disruption to the soil surface by stock during grazing (Sharpley and Syers, 1976; McDowell et al., 2007), and/or a declining availability of P in dung with time (Ebeling et al., 2002; Mundy et al., 2003). We had insufficient data to adequately investigate this last possible cause, which would be expected to be expressed as an interaction between DgP and TSG. There was a tendency for the proportion of TP present as PP to decline with increasing TSG, indicating that at least part of the TSG effect we observed may be associated with physical disturbance of the soil surface by cattle treading.

Effect of P Rate on Loads of P in Runoff
The average annualized loads of TP in runoff from each of the plots are presented in Table 4 . In a linear regression analysis involving all six plots, the relation between the average annualized load of TP (kg P ha–1 yr–1) and fertilizer P rate was not significant (P > 0.10). However, the volumes of runoff from one of the P40 treatments were often atypically low (<50% of the other plots), and when the data from this plot were excluded, the relation between fertilizer P rate and annualized load of TP was significant (P = 0.012). From the P0 treatment the average TP loss in runoff was 0.399 kg of P ha–1 yr–1, and from the P80 treatment the average TP loss in runoff was 1.112 kg of P ha–1 yr–1. Similar loads of P in runoff from intensively managed pastures were reported by Fleming and Cox (1998). In contrast, an average annual loss of 5.8 kg P ha–1 was reported by Nash et al. (2000), although their average TP concentration (10.5 mg L–1) was much higher than that reported here and elsewhere. The difference between the load lost from the P0 and P80 treatments provides an estimate of the effect of P fertilizer application. This difference (0.713 kg of P ha–1 yr–1) represents <1% of the P applied on the P80 plot, which is agronomically insignificant, thus limiting the economic imperative to more efficiently manage fertilizer P.


View this table:
[in this window]
[in a new window]

  		Table 4. Average annualized loads{dagger} (kg TP ha–1 yr–1) of P in runoff for each of the runoff plots.

 

   Discussion
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
 Discussion
Conclusions
 REFERENCES
 
The concentration of P in runoff from our intensively managed pastures was high relative to catchment water quality standards even when no fertilizer was applied and soil P status was at or below the agronomic optimum. The addition of fertilizer P further increased the concentration of P in runoff. However, the concentration of P in runoff and the effect of fertilizer P application varied as the result of variations in a number of management-related factors. A number of authors (e.g., Haygarth and Jarvis [1999] and Nash et al. [2005]) have defined a two-component conceptual model comprising systematic and incidental components of P in runoff to explain variations in runoff P concentrations. The incidental component of runoff P arises typically when discrete events such as fertilizer or manure application occur shortly before runoff and results in much higher concentrations than are normally observed. The systematic component arises as the result of contributions from all other sources of runoff P. This model presumes that there is relatively little opportunity to modify the systematic component of P loss and that it is the incidental component that is most easily manipulated or prevented (Preedy et al., 2001; Nash et al., 2005). We suggest a refinement to that conceptual model by proposing three components that can be used to explain variations in runoff P concentrations similar to those we observed. Each of these components presents different opportunities for manipulation. First, there is a "baseline" component, which arises solely from the activity of growing and grazing a pasture on a soil with a basal soil P status (represented in our study by the P0 plot, excluding incidental grazing effects). Second, the "fertility" component comprises the effect of fertilizer P in raising soil P status and consequently runoff P above the baseline (in this study the P20, P40, and P80 plots) excluding incidental fertilizer effects. Third, the "incidental" component comprises P that is poorly equilibrated with the soil when runoff occurs soon after fertilizer application or dung application during grazing. The first two components of our conceptual model are a refinement of the "systematic" component identified by other authors, whereas the "incidental" component is the same as that identified by other authors (e.g., Nash et al. [2005], Preedy et al. [2001], Hart et al. [2004]).

The baseline component (component 1 of our conceptual model) is important because it defines the lower limit of runoff P concentrations that can be expected from grazing of intensively managed pastures and is that component over which the land manager has little direct control. The P in runoff from our P0 plot approximates this component and arises from the labile P pools in the grazing system, including soil, plant biomass, and dung. The runoff TP concentration approximating this baseline component ranged from 0.86 to 3.01 mg L–1 (flow-weighted average of 1.94 mg L–1). Our P0 plot concentrations will have included part of component 3 of our model associated with grazing (see discussion of component 3 below). When those events in which grazing had occurred in the 20 d preceding runoff events are not included in the calculation of a flow-weighted average for the P0 plot, the flow-weighted runoff P concentration representing purely the baseline component of our model is 1.84 mg P L–1. This runoff P concentration resulted from grazing at a stocking rate of 2.2 cows ha–1, which is slightly below the Australian average of 2.5 cows ha–1 (ABARE, 2006), on soil with Olsen P (0–10 cm) that decreased from 17 to 10 mg kg–1 during the monitoring period, which is at or below the suggested agronomic optimum of 14 to 17 mg kg–1 (Gourley et al., 2007). Similar runoff TP concentrations (~1–6 mg L–1) have been reported from other Australian sites that had similarly moderate Olsen P concentrations (Greenhill et al., 1983; Austin et al., 1996; Nash et al., 2005). Although the Olsen P (0–10 cm) concentration in the soil on our P0 plot was at or below the agronomic optimum throughout the life of the experiment, suggesting that further addition of P is warranted for optimal pasture production, the concentration of P in runoff is well above typical water quality criteria of <0.05 mg P L–1 (Anonymous, 2004). However, we acknowledge that edge-of-field concentrations such as those we measured are moderated at the catchment scale by other landscape and in-stream processes.

We hypothesize that the substantial baseline concentration of P in runoff from the plots we examined is partly the result of P inputs to the soil surface from sources other than fertilizer (i.e., dung and senescing pasture). Despite no fertilizer P being applied to the P0 plot, dung P budgets indicated that approximately 12 kg P ha–1 yr–1 was being deposited on the soil surface in dung and a further 21 kg P ha–1 yr–1 in senescing pasture. These inputs of P are the minimal returns that occur in such intensive grazing systems and highlight the challenges involved in attempting to reduce the baseline runoff P concentration.

The addition of P fertilizer increased runoff P concentration above the baseline by increasing the pool of mobilizable P in the soil (these increases in runoff P concentrations arise from the fertility component of our conceptual model). Within nearly all individual runoff events (and overall), there was a significant effect on runoff P concentration of fertilizer P application. We propose that this component of runoff P is one over which land managers do have substantial control. Minimizing the inputs of P fertilizer to those required to maintain optimal production would limit the concentrations of P in runoff. For instance, at our site, 20 kg P ha–1 yr–1 was approximately the amount of fertilizer P required to maintain optimal production. The flow-weighted runoff P concentrations for the P20 and P80 plots over the monitoring period of this trial were 2.77 and 5.57 mg L–1, respectively. Approximately one third of the difference in average runoff P concentrations between these treatments was attributable to the third component of our model (i.e., incidental losses from fertilizer and grazing, estimated by not including those events with TSF and TSG <20 d in the calculation of flow-weighted averages). Hence, the fertility component of our conceptual model associated with the application of fertilizer at 80 kg P ha–1 yr–1 (about 4 times the agronomic requirement) resulted in approximately a 70% increase in runoff P concentration over that of the agronomic optimum. This illustrates the relatively substantial opportunity there was in our experiment to restrict runoff P concentrations if P fertilizer rates had been accurately matched to rates necessary for optimal production compared with the effect of eliminating incidental losses.

The high TP concentrations (up to ~11 mg P L–1) at short TSF and TSG (component 3 of our conceptual model) are the "incidental" components of P loss and have been identified by other authors as one of the key opportunities to reduce P losses (i.e., by avoiding broadcast fertilizer applications and grazing during or immediately preceding runoff events) (Nash et al., 2000; Withers et al., 2003; Hart et al., 2004). The ability to maximize TSF and TSG to decrease the contribution of the component to P exports depends on the reliability of forecasting rain and runoff events, which is most likely to be successful in environments with relatively predictable and distinctly seasonal rainfall that results in periodic soil saturation. These characteristics are not typical of the environment in which the NSW coastal dairy industry operates (Linacre and Hobbs, 1982). This unpredictability, coupled with the relatively small contribution of this component to runoff P concentration (i.e., approximately half that of component 2), suggests that there is limited opportunity to reduce runoff P concentrations in grazed dairy pasture systems in coastal NSW by maximizing TSF and TSG.

The significance of the negative TSF/FP interaction term in our model suggests that the soil/pasture system is better able to buffer the effects of small additions of P than large ones. In locations where incidental losses may comprise a substantial component of P export and are difficult to reduce by maximizing TSF, reductions in P export via this component may be achieved by applying more frequent, smaller applications of P—as is the case with nitrogen— rather than one large annual application. There is also anecdotal evidence to suggest that this strategy may improve production responses to fertilizer P application. However, more frequent applications of P fertilizer increase the risk of a coincidence of fertilizer application and runoff. The impact of these two potentially conflicting factors needs to be carefully considered.

Our data suggest that strategies to minimize or reduce P accumulation in the zone of soil–runoff interaction are the key to reducing runoff P concentrations and export. Such strategies may include more careful matching of P applications with production requirements, nutrient budgeting (with a view to reducing P excess), subsurface placement of fertilizer, de-stratification (i.e., mixing of the P-rich topsoil with lower P subsoil by cultivation), application of P fixing ameliorants, harvesting of P-rich biomass (to decrease the labile soil P pool), and dietary P manipulation (to reduce dung P inputs). Strategies such as using grassed buffer strips to reduce runoff P concentrations are likely to be of limited effectiveness because they rely primarily on physical trapping of sediment-associated P, which our and other studies have revealed is only a small proportion of P in runoff from dairy pastures.

The economics, logistics, and long-term effectiveness of the various strategies to reduce P export in runoff require further examination. Because the loss of P in runoff from dairy pastures is <1% of applied P, which limits the direct economic imperative to reduce P losses, the environmental, social, and political benefits of reduced P losses are the key drivers of any improvements in P management relating to runoff.

Finally, our observation of multifactorial influences on runoff P concentrations was possible because of the experimental design, the measurement of a wide range of parameters, and the use of advanced statistical techniques. These multifactorial effects should be considered in the planning of future experiments on runoff P and the choice of experimental designs.


   Conclusions
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
 Discussion
 Conclusions
REFERENCES
 
The concentration of P in runoff from our intensively managed pastures was high relative to water quality standards, even when no fertilizer was applied and soil P status was below the agronomic optimum. The addition of P fertilizer significantly increased runoff P concentration. However, the effect of fertilizer P addition on runoff P concentration was highly variable due to the effects of previous fertilizer applications, TSF, DgP, BM, and TSG (i.e., the influences on runoff P are multifactorial). Our results suggest that the greatest opportunity to reduce P in runoff lies in restricting fertilizer P inputs to the minimum necessary for pasture production. However, even when best practice is implemented, the minimum P concentrations in runoff from intensive pastures systems in southeast Australia are likely to be high when compared with water quality targets.


   ACKNOWLEDGMENTS
 
This research was funded by Dairy Australia (formerly Dairy Research and Development Corporation), Land and Water Australia, Incitec Pivot Limited (formerly Incitec), and the NSW Department of Primary Industries (formerly NSW Agriculture). We thank Helena Warren, Mark Holdsworth, Karen Christie, Rose Sinclair, Kris Riley and John Hamilton for their assistance in the conduct of the experiment. We also thank three anonymous reviewers for a number of helpful suggestions.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
 Discussion
 Conclusions
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 Materials and Methods
 Results and Discussion
 Discussion
 Conclusions
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Dougherty, W. J.
Right arrow Articles by Lawrie, R. A.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Dougherty, W. J.
Right arrow Articles by Lawrie, R. A.
Agricola
Right arrow Articles by Dougherty, W. J.
Right arrow Articles by Lawrie, R. A.
Related Collections
Right arrow Water Quality
Right arrow Phosphorus
Right arrow Nutrient Management

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Soil Science Society of America Journal Journal of Plant Registrations The Plant Genome