Direct losses of P from fertilizer have been classified under the term "incidental losses" (Haygarth and Jarvis, 1999), meaning derived from a discrete event, where application of fertilizer is coincident with a hydrological factor such as heavy rainfall events. However, the authors of this review believe that the term "incidental" can be misleading due to its common-English alternative meaning of "inconsequential." We would suggest the term "event-specific losses" as a better alternative. Several studies have indicated that DRP and TP concentrations in runoff may be correlated with soil test P values, generally increasing linearly as soil P fertility levels increase (Sharpley et al., 1993; Pote et al., 1999; Cox and Hendricks, 2000). Thus, the long-term overfertilization of soils is recognized as a potential contributor to eutrophication of surface waters (Sims, 1993; Frossard et al., 2000). However, the potential influence of event-specific losses directly from recently applied fertilizer has received relatively little study (Preedy et al., 2001; Macleod and Haygarth, 2003). In this section we review the effects of this P source, focusing mainly on pastoral land, and attempt to put them into perspective with overall losses of P in surface runoff.
Much of the work done in this area has been conducted in New Zealand and Australia, probably reflecting the relatively high importance of agriculture to both the economy and water quality in these two countries. For the same reason, most published work deals with runoff from grazed pastures, rather than from cultivated land. Sharpley and Syers (1976) compared P losses in surface runoff from grazed and ungrazed plots on permanent pasture at Massey University, New Zealand, on Tokomaru silt loam soil, a Typic Fragiaqualf developed in siliceous loess, with or without the application of single superphosphate (SSP) at the rate of 50 kg P ha–1, over a 4-mo period. Overall losses of DIP (molybdate-reactive P, the fertilized plots were approximately four to five times greater than those from the corresponding unfertilized plots. The proportion of added fertilizer lost in surface runoff from plots that were undrained ranged from 5.6 to 6.7%, of which between 48 and 52% was as TDP. For two plots that were mole-drained, 1.0% of fertilizer P was lost, of which 70% was as TDP. Fertilizer addition led to a significant increase in DIP and PP concentration in surface runoff, in the first week following fertilizer application. Mean DIP concentration in surface runoff increased to a peak of about 3.3 mg L–1 after 3 d, declining over the following 6 wk, though it remained higher than the background DIP concentration throughout the monitoring period. Mean PP concentration increased to a maximum value of 2.8 mg L–1 in the first week after fertilizer application, then decreased much more gradually than that of DIP from the same plot, and was still significantly higher than the background PP concentration almost 12 wk after the fertilizer was applied. This pattern of a sharp increase in P losses following soluble fertilizer applications, tailing off over a period usually of a few months, has been observed in every subsequent study.
Even greater increases in P concentration in runoff were measured at Massey University by Sharpley and Syers (1979a) following aerial application of SSP at 30 kg P ha–1 to a pasture catchment, and in further studies on the same runoff plots in a comparison of P runoff losses from solid and dissolved SSP at 50 kg P ha–1 (Sharpley and Syers, 1983). Amounts of total fertilizer P lost from all these studies ranged from up to about 3.8 to 11.5% of the amounts of the P applied, equivalent to exports of up to 3.25 to 7.1 kg P ha–1 yr–1. In a similar study, Sharpley and Syers (1979b) estimated that up to 4.8 and 8.8% of the P applied as SSP at 50 kg P ha–1 was transported in surface runoff as TDP and TP, respectively, or 2.4 and 4.4 kg P ha–1, over the course of a year. Some authors suggest caution should be exercised in extrapolating results from runoff plots to whole watersheds (Sharpley et al., 1978; Sharpley and Syers, 1979b), while others have found no significant difference when comparing P losses in simulated rainfall runoff from microplots compared with natural runoff from much larger paddock-scale plots (Cornish et al., 2002). Nevertheless, these data indicate that a substantial proportion of the P lost from pasture in surface runoff can be derived from recently applied fertilizer. In the context of typical background levels of P loss from pastoral land (e.g., around 1.3 kg P ha–1 yr–1 in New Zealand; Gillingham and Thorrold, 2000), these amounts are very large indeed.
McColl and Gibson (1979a)(1979b) measured P concentrations in runoff from unconfined collectors in two small (approximately 400 m2) hill-pasture enclosures grazed by sheep. Single superphosphate was applied in spring at about 51 kg P ha–1 to the top part of the enclosures. Fertilizer application caused a large increase in P concentrations in runoff from the enclosure that had recently been grazed (mean grass length 2.6 cm), but not from the ungrazed enclosure (mean grass length 10 cm). In the grazed enclosure, TP concentrations in runoff samples from the first post-fertilizer rain event were about 22 times higher than those in the runoff from the pre-fertilizer event, about 26 mg P L–1 compared with about 1.2 mg P L–1. The P concentrations decreased rapidly after the first post-fertilizer runoff event, 4 d after SSP application, but the fertilizer effects persisted until the third post-fertilizer event, which occurred 20 d after SSP application. This decline is somewhat faster than those typically measured in the studies of Sharpley and coworkers described above. However, those experiments were measuring P losses from plots to which fertilizer had been directly applied, whereas in the McColl and Gibson (1979b) experiment, fertilizer was deliberately not spread in the immediate vicinity of the collectors. The term "immediate vicinity" was not defined by these authors, but is assumed to mean within a meter or so. Nevertheless, the influence of fertilizer application was still very significant. The post-fertilizer runoff events accounted for less than 1% of the total annual downslope movement of water, but about 35% of the annual P movement, in the recently grazed pasture (McColl and Gibson, 1979c).
Gillingham et al. (1997) conducted an experiment comparing P runoff from microplots using simulated rainfall in the hill-country catchments at Whatawhata and at Waipawa in the Hawkes Bay region of New Zealand. At both sites, DRP in runoff 5 to 6 d following P fertilizer application (as SSP or monocalcium phosphate, MCP) was very high compared with that from the unfertilized plots. Mean DRP concentrations at Whatawhata increased from 0.06 to 6.4 mg L–1 while at Waipawa the mean concentrations increased from 0.46 to 31.5 mg L–1. The mean concentrations declined from these peaks quite rapidly, but the effects of the fertilizer applications on DRP in the runoff water were still apparent 90 to 110 d after the fertilizer was spread. This is much longer than that found by McColl and Gibson (1979a)(1979b). Obviously, local soil and other conditions will affect the length of time that recently applied fertilizer effects last; but from this review, a period of between around 30 to 50 d appears to be most typical for the most significant proportion of P losses. A notable point that arises from this observation is that there are relatively few areas in most temperate zones where at least one or two significant rainfall events would not occur during such a period, which would typically be in spring and/or autumn.
Other studies conducted outside of Australasia have found similar influences from recently applied fertilizer. Olness et al. (1980) compared nutrient losses from paired, gently sloping (3°) native grassland watersheds with Udertic Paleustoll and Udic Argiustoll silt loams soils, in Oklahoma, USA, under rotational or continuous grazing. Three days after ammonium phosphate fertilizer was broadcast at a rate of 75 kg P ha–1 on one of each paired watershed, an intense thunderstorm delivered about 94 mm of rain, generating between 3.6 to 5.4 cm of runoff from the watersheds. This led to an increase in TP in surface runoff of about 10- to 25-fold and in TDP of about 200- to 600-fold compared with losses from the unfertilized watersheds. Soluble P accounted for 79 and 65% of the TP exported in the first runoff event from the fertilized rotationally and continuously grazed watersheds, respectively. Runoff P concentrations decreased sharply with successive events, but after 2 mo and several runoff events the surface runoff P concentrations were still quite high, particularly the TDP fraction, being about 10 to 20 times greater than those from the unfertilized watersheds. Even 12 mo after fertilizer application, the runoff TDP concentrations were still 5 to 15 times greater than the long-term mean background concentrations. The authors estimated that over the course of a year, about 3 kg ha–1 of TP (79% TDP) and 5 kg ha–1 of TP (67% TDP) was lost in runoff from the fertilized rotationally and continuously grazed watersheds, respectively. The differences in P losses between the two grazing regimes over the course of the experiment were attributed to increased losses of PP associated with greater soil erosion under continuous grazing. As part of the experimental design, these watersheds were routinely overgrazed and generated on average about three times as much runoff as the rotationally grazed watersheds, and were severely eroded with active, steep-walled gullies.
A similar effect due to an intense rainstorm following shortly after fertilizer application was measured by Haygarth and Jarvis (1997), in an experiment conducted on 1-ha-sized grazed lysimeter plots on a clayey, noncalcareous pelostagnogley soil (Typic Haplaquept) in Devon, UK. Fertilizer (triple superphosphate, TSP) at the rate of 16 kg P ha–1 was added to some of the plots in spring. Over a period of 4 to 8 d following this application, more than 50 mm rainfall was recorded, which resulted in the highest P losses in runoff recorded during the experiment. Runoff in this experiment consisted of overland flow and subsurface flow to a depth of 30 cm. Mean DRP concentrations increased from 41 µg L–1 in a runoff event before fertilizer application to 252 µg L–1. A comparison of maximum values showed a 10-fold increase in DRP concentration from 131 to 1296 µg L–1. Comparisons of TP loading losses indicated even greater losses, with mean values of 92 compared with 4641 mg ha–1 h–1, with maximum values of 160 compared with 18510 mg ha–1 h–1. It was estimated, therefore, that for a storm that lasted about 30 h, losses may have been in excess of 0.5 kg P ha–1, 3% of the total P applied. The authors speculate that the high intensity rainfall may have caused erosion of PP, and thus contributed to the high losses of TP in relation to DRP. After excluding data from the runoff event immediately following the fertilizer application, TDP was found to contribute about 69% of TP in surface runoff and interflow to a 30-cm depth (Haygarth et al., 1998b).
Shuman (2002) measured DRP losses in simulated rainfall-runoff from 25-m2 turfgrass plots, on a Cecil sandy loam soil (Typic Kanhapludult) in Georgia, USA, with a 5° slope. Rainfall was applied and runoff collected 4, 24, 72, and 168 h after the application of monoammonium phosphate (MAP) fertilizer at 0, 5, and 11 kg P ha–1, in two successive years. Dissolved reactive P concentrations in the first runoff event were high, with an average over the two years of 0.75, 3.8, and 7 mg L–1 for the 0, 5, and 11 kg P ha–1 treatments, respectively. However, in this experiment the concentrations tailed off rapidly over the ensuing runoff events, and there was no statistical difference between treatments after 168 h, with average DRP concentrations being about 0.8 mg L–1. The large differences in length of time in the measurable effects of fertilizer applications on P loss in runoff in the experiments discussed above is another indication of the site-specific nature of P exports from agricultural land.
As noted previously, most studies on runoff from cultivated land indicate that most P losses in this situation are associated with the PP fraction. However, a few field studies have been conducted where fertilizer effects were measured against a control treatment, where this was not the case. Withers et al. (2001) compared P losses in surface runoff from 15 adjacent 32-m2 experimental plots with a 5° slope, on a silty clay loam soil (Argillic Albaqualf) developed over Old Red Sandstone in Herefordshire, UK, over a 2-yr period, with all plots being cropped to cereals each year. At the beginning of the experiment, TSP was surface-applied at a rate of 60 kg P ha–1. Three weeks following this, approximately 25 mm of rainfall fell on the site, producing approximately 16 L of surface runoff with an average DRP concentration of about 6.5 mg L–1, which constituted about 84% of the TP concentration. Subsequent (lower volume) runoff events in the month following fertilizer application produced much lower concentrations in runoff water. A cumulative load of about 82 mg P plot–1 was lost in surface runoff in the first three events, of which about 77% was TDP, compared with a total of 24 mg P plot–1 from the control treatments, of which only about 21% was TDP.
For the fourth monitoring period of the experiment the following year, treatments were applied to the seedbed surface, the preparation of which included rolling after being sown with winter barley (Hordeum vulgare L.). This had the effect of consolidating the soil surface, and meant that lower rainfall intensities were needed to generate surface runoff. In the first runoff event, which occurred within a few days of fertilizer application (TSP at 90 kg P ha–1), extremely high DRP concentrations were measured in runoff from the fertilizer-treated plots: 74 compared with 0.7 mg L–1 from the control plots. Treatment differences in P losses lasted for approximately 1 mo after application. Thus, even on conventionally cultivated soil, the presence of soluble P fertilizer on the soil surface can be the most influential determinant of the forms of P lost in surface runoff.
Zhang et al. (2003) compared DRP and TP losses from 30-m2 plots on two contrasting paddy soils under wheat (Triticum aestivum L.) in southeastern China, following application of inorganic P fertilizer (an N–P–K and SSP) at 0, 20, 80, and 160 kg P ha–1. One site (Anzhen) was on a loam clay soil with an impermeable clay layer down the soil profile, while the other site (Changshu) was on a silt loam soil with a relatively uniform and permeable profile. The time period between fertilizer application and the first runoff event was not recorded in the paper. However, significant differences were found both between sites and between rates of P application. Although the volume of runoff from the Anzhen site was approximately 25% greater than that from the Changshu site, P concentrations and loads were lower. In the control treatment, the majority of P exported was particulate (about 70–95% of TP). However, the proportion of P in runoff in the form of DRP increased with increasing rate of P application. At Anzhen, TP concentration in the first runoff event ranged from 1.13 mg L–1 in the control treatment to 4.13 mg L–1 in the 160 kg P ha–1 treatment, while DRP concentrations for these treatments were 0.26 and 2.76 mg L–1, respectively. The concentrations of TP gradually declined over the following runoff events, as did the proportion of TP in the form of DRP, in all treatments, which ranged from about 6 to 12% in the final, fifth runoff event, about 9 wk after the initial runoff event.
In contrast, concentrations and loads of runoff P from the fertilized plots were much greater at the Changshu site. While control plots values were similar to those at Anzhen, concentrations of TP and DRP in the first runoff event from the 160 kg P ha–1 treatment were 17.4 and 13.34 mg L–1, respectively. The concentrations remained higher than at the Anzhen site, even after six runoff events, and the proportion of DRP to TP remained relatively high throughout the experiment, in contrast to results from the other site. Overall, the proportion of TP exported from the fertilized plots at Anzhen in the form of DRP ranged from 9 to 34%, while that from the Changshu site ranged from 35 to 50%. These differences between the two sites were attributed to the difference in clay content and hence P adsorption between the two soils.
These results highlight the overriding influence that the application of soluble fertilizer can have on the characteristics of P runoff in the short term, and support the suggestion of Kleinman et al. (2002) that soil P contributes little to runoff P losses following recent surface application of labile P sources. They also support the suggestion that there is reduced contact between soil particles and P fertilizer granules under high rates of SSP application, allowing the fertilizer to remain a P source for a considerable time (Withers et al., 2003), although as demonstrated by Zhang et al. (2003), the extent of this effect will vary from site to site, depending in part at least on the soil characteristics, particularly clay content and degree of saturation of P sorption capacity.
Comparisons of Runoff from Inorganic Fertilizer and Manure Application
In many parts of the world, housing of stock over winter means that disposal of accumulated manures is a major problem, and as applications are usually made on the basis of N content, the amount of P that may be applied can be much greater than would be the case with an inorganic fertilizer application (Chadwick and Chen, 2002). Many studies have compared losses of P and other nutrients from inorganic fertilizers with those from organic manures and slurries. Nichols et al. (1994) measured losses of TDP and TP in simulated rainfall runoff from 9-m2 runoff plots under established tall fescue (Festuca arundinacea Schreb.), on a Captina silt loam (Typic Fragiudult) soil in Arkansas, USA, with a uniform 5° slope. Treatments were amendment with poultry litter or an N–P–K fertilizer at an equivalent rate giving 87 kg P ha–1, surface-applied, or with shallow incorporation to a depth of 2 to 3 cm by rotary tillage. Simulated rainfall was applied at 50 mm h–1, sufficient to produce 30 min of continuous runoff, 7 d after treatment application, and no significant difference was found in the volume of runoff coming from the various treatments. Total P concentrations in runoff were significantly lower from plots treated with poultry litter compared with inorganic fertilizer, with an average of 15.4 and 26.2 mg L–1, respectively. Approximately 67% of TP in the runoff from the poultry litter plots was in the form of TDP, whereas in runoff from the fertilizer treatment plots, >95% of TP was as TDP. The shallow incorporation method had no significant effect on the amounts of P lost from the fescue plots, due to inadequate turn-under of the surface-applied litter and fertilizer. As the depth of incorporation was very shallow, in an attempt to minimize damage to the fescue roots, it is probable that a significant part of the applied P that was incorporated into the soil was still at the effective depth of interaction for the overland flow (Sharpley, 1985), as well.
Heathwaite et al. (1998) compared N and P losses in runoff from sloping (15–20°), ungrazed grassland plots on weakly structured brown Cambisol soil of the Denbigh Association, receiving inorganic fertilizer, farmyard manure (FYM), and liquid cattle slurry. A compound N–P–K fertilizer supplying P as DAP was applied to some of the plots at the rate of 100 kg P ha–1. Simulated rainfall was applied to the plots at a rate of 22 mm h–1 for 35 min on four consecutive days, starting 3 d after the day of fertilizer application. Much greater TP concentrations were measured from the fertilizer plots (mean 15.3 mg P L–1) compared with those from the FYM and slurry treatments (1.76 and 0.82 mg P L–1, respectively). Total P export in surface runoff from the base of the 20-m hillslope plots receiving fertilizer averaged about 3.8 kg P ha–1, of which about 67% was in the form of DRP, compared with about 0.1 kg P ha–1 from the untreated control plots.
It could be said that applying artificial rainfall so shortly after fertilizer application is not a very realistic scenario, but in many if not most temperate parts of the world, rain events are unpredictable over any longer than a few days ahead, and so the coincidence of rainfall and fertilizer application is bound to happen from time to time. Preedy et al. (2001) conducted an experiment specifically designed to measure event-specific transfer of P from 30-m2 plots treated with inorganic and organic amendments, on the same site used by Haygarth and Jarvis (1997), over a 7-d period. Treatments were TSP, dairy slurry, and TSP + slurry, applied at the rate of 29 kg P ha–1, to grassland that was already wet, with further rainfall forecast. Water discharge collected from the plots came from overland flow and interflow to a depth of 27 cm, and accounted for 98% of the rainfall that fell on the plots over the 7 d. Concentration in runoff from the TSP and TSP + slurry treatments peaked at around 11 mg TP L–1 (67–68% as DRP), while those from the slurry treatment peaked at about 7 mg TP L–1 (20% as DRP), when rainfall was at its most intense (approximately 3 mm h–1) about 28 to 32 h into the experiment. This 4-h period accounted for 18% of rainfall and discharge, but produced 46% of the TP load from the TSP treatment, compared with 38% in the TSP + slurry, 33% in the slurry, and 27% in the control, respectively. It was estimated that over the course of the experiment, P loads from the TSP and slurry-treated plots were equivalent to losses of about 1.8 kg TP ha–1, while those from the TSP + slurry treatment were equivalent to about 2.3 kg TP ha–1 (approximately 6.2 and 7.9% of applied P, respectively), compared with only about 0.06 kg TP ha–1 from the control plots.
Gaudreau et al. (2002) measured nutrient concentrations in runoff from 6-m2 turfgrass plots, on a Boonville fine sandy loam (Chromic Vertic Albaqualf) soil in Texas, USA, treated with composted dairy manure applied at 50 and 100 kg P ha–1, and an unidentified inorganic P fertilizer applied at 25 and 50 kg P ha–1. Generally, differences in the TDP concentration between the control and the other treatments were relatively small, the latter being roughly one to three times greater. However, one runoff event happened to occur 3 d after treatment application, compared with a lag of 27 to 87 d for the other events. The control TDP concentration was 1.7 mg L–1, while those from the 50 and 100 kg P ha–1 manure treatments were 5.5 and 9.8 mg L–1, respectively, and those from the 25 and 50 kg P ha–1 fertilizer treatments were much greater at 16.6 and 30 mg L–1, respectively. Other field experiments reviewed in this paper have also demonstrated the coincidence of natural rainfall with fertilizer application (e.g., Olness et al., 1980; Nash et al., 2000b).
Given the expensive nature of field experiments and the unpredictability of natural rainfall, the use of rainfall simulators and laboratory microcosms is becoming more common in P transport research. Kleinman et al. (2002) compared losses of P in simulated rainfall-runoff from three different manures and DAP fertilizer applied at 100 kg P ha–1 to three soils packed into 0.2-m2 runoff boxes. The soils were a Buchanan (Aquic Fragiudult)–Hartledon (Typic Hapludult) association, Hagerstown (Typic Hapludalf), and Lewbeach (Typic Fragiudept), in two sets, one with low (12–26 mg kg–1 Mehlich-3 P) and the other with high background P levels (396–415 mg kg–1 Mehlich-3 P). All P amendments, with the exception of dairy manure on the Buchanan–Hartledon and Hagerstown soils, led to very significantly increased losses of P in runoff generated 72 h after treatment application. Dissolved reactive P in runoff water accounted for 64% of TP overall from the amended soils, compared with only 9% from the control soils. Losses of P were greatest from the DAP treatments, although they were generally not statistically different from those of the other amendments (p = 0.05). Losses of P were mostly, though in the case of dairy manure amendment, not always, greater from the high P soils compared with the low P soils. The differences between these two sets was most pronounced in the DAP treatments. If the average amounts of P lost from the DAP-treated soil boxes are scaled up directly, then the equivalent of 1.1 and 1.4 kg P ha–1 was lost as DRP and TP from the low P soils, and the equivalent of 4.1 and 6 kg P ha–1 as DRP and TP from the high P soils, respectively. Differences between the amounts of P lost from plots treated with inorganic P fertilizers compared with organic manures and slurries appear to be correlated with the relative amount of soluble P present in these P sources.