A rainfall simulation study on the relationships between soil test P versus dissolved and potentially bioavailable particulate phosphorus forms in runoff

Valumavesiin liuenneen fosforin lisaksi osa maaaineksenmukana pelloilta kulkeutuvasta fosforista kiihdyttaa vesistojen rehevoitymista. Koska maaainesfosforin rehevoittavien muotojen maarittaminen valumavesista on kallista, naiden rehevoittavien fosforijakeiden suuruus tunnetaan huonosti. Taman tyon tarkoituksena oli arvioida mahdollisuuksia kayttaa viljavuustutkimuksen fosforilukua rehevoittavien fosforijakeiden pitoisuuden arviointiin siten, etta tulokseksi saadaan rehevoittavan fosforin maara erodoitunutta maa-aineskiloa kohden. Yhdistamalla tama tieto mitattuihin eroosiomaariin, voitaisiin kenttakoetuloksista laskea arvioita rehevoittavista fosforikuormista nykyista luotettavammin. Tutkimuksen aineisto koostui 15 maanaytteesta, jotka oli haettu savimailta eri puolilta Etela- Suomea. Maiden savespitoisuudet vaihtelivat valilla 3163 %, pH-luvut 5,07,4, hiilipitoisuudet 1,54,2 % ja viljavuusuuton fosforiluvut 346 mg P/l. Maanaytteet pakattiin laboratoriossa vanerilaatikoihin ja kostutettiin vedella kyllastyneeseen tilaan, jollaisessa ne ovat usein talven ja kevaan valuntahuippujen aikana. Laatikot asetettiin 2,5 2,8 asteen kallistukseen ja niita sadetettiin sadesimulaattorissa 2 tuntia intensiteetilla 5 mm/h. Sadetuksen aikana laatikoista poistuneesta valumavedesta maaritettiin veteen liuenneen fosforin pitoisuus, seka veden mukana kulkeutuneen maa-aineksen fosforijakeista leville helposti kayttokelpoinen jae ja pelkistyneissa oloissa veteen liukeneva jae. Valumaveden liuenneen fosforin pitoisuus suureni selkeasti maan fosforiluvun kasvaessa (r2 0,84), kun taas maan fosforiluvun ja erodoituneen maa-aineksen sisaltaman rehevoittavan fosforin valinen yhteys oli heikompi (r2 0,560,58 rehevoittava maa-ainesfosfori suhteutettuna maa-aineskiloa kohden). Eroosioaineksen sisaltaman rehevoittavan fosforin pitoisuuden ennustaminen maan fosforiluvun perusteella nayttaisi olevan mahdollista. Ennusteiden luotettavuuden kannalta maan fosforiluvun ja eroosioaineksen sisaltamien fosforijakeiden valiset yhteydet olisi kuitenkin maaritettava erikseen useista, ominaisuuksiltaan melko yhtenaisista maa-aineistoista, koska muutkin maan ominaisuudet kuin pelkka fosforiluku vaikuttavat ennusteyhtaloihin. Esimerkiksi taman tyon tuloksiin pohjautuvaa ennusteyhtaloa kayttamalla valuman mukana poistuvan eroosioaineksen rehevoittavan fosforin pitoisuuksista olisi saatu liian pienia arvioita neljalta koekentalta.


Introduction
In areas where clayey soils are abundant, particulate phosphorus (PP) is the major form of P in field runoff (see Pietiläinen and Rekolainen 1991, Turtola and Paajanen 1995, Puustinen et al. 2005).Depending on soil characteristics and environmental conditions, a variable part of the runoff PP may become algal-available over time (Logan et al.Electronic version updated March 2006 Uusitalo, R. & Aura, E. Relationships between P pools in runoff versus soil test P 1979, Ekholm 1994) and accelerate eutrophication of surface waters in the same manner as dissolved molybdate-reactive phosphorus (DRP) does.
Data abound on the effects of different land management options on concentrations and losses of DRP and PP (the Finnish studies include Puustinen 1994, Turtola 1999, Uusi-Kämppä et al. 2000).From these work we learn that the total offsite P losses are often reduced by erosion control measures, such as vegetated buffers, but that this decrease may be counterbalanced by increased losses of DRP (see also Culley et al. 1983, Bundy et al. 2001).Because of the nonequivalent effects of DRP and PP on eutrophication -DRP is fully and immediately algal-available, whereas a variable part of the PP may solubilize -, the findings of studies on land management practices cannot be fully utilized in efficient eutrophication control.To be able to predict the net effects of field management on eutrophying P losses, we should be able to assess the pollution potential of PP in different situations.
Quantitative estimates of the losses of potentially bioavailable P forms have been made at four field sites in southern Finland (Uusitalo et al. 2003, Uusitalo 2004).At these sites desorbable PP (anion exchange resin-extractable PP, AER-PP) was found to make up as great a part of the P losses as did DRP; on average 10-20% of runoff PP was estimated to become algal-available at low ambient P concentration in an oxic water column.In case the eroded soil would end up in an anoxic sediment, solubilization of redox-labile PP (assessed by bicarbonate-dithionite extraction; BD-PP) could increase the bioavailability to 35-60% of runoff PP.At these four sites, potential bioavailability of runoff PP was the higher, the higher was the agronomic P status of the source soil.
The use of agronomic soil P tests in assessing the potential for phosphorus release from soil to runoff have been evaluated in a number of research papers (e.g.Sharpley et al. 1978, Heckrath et al. 1995, Pote et al. 1996).Soil test P has usually been found to correlate well with the concentration of DRP in runoff, whereas PP -in many cases the major form of runoff P -has received little attention (Jokela et al. 1998).Calculating direct corre-lations between PP concentration in runoff and soil test P is also meaningless because the concentrations of PP are mainly governed by the erosion rates.However, when normalized to soil loss, P contents and P solubility indices of the source soil do affect losses and concentration of bioavailable PP.This is because PP/TSS ratio increases with soil P content, and the erosion rate-normalized losses of algal-available PP increases with soil P phytoavailability (agronomic P status).Combined with soil loss data, soil P tests could, as example, help in assessing whether a management practice implemented at a particular site is to reduce the losses of those PP pools that are relevant for eutrophication more than they are known to increse the losses of DRP.
In this paper, we report results of a rainfall simulation study in which we tested the potential of acid ammonium acetate buffer, the agronomic soil test of Finland, and an estimate of soil P saturation, based on extraction using Mehlich-3 solution, for their abilities to predict the concentrations of (i) DRP in runoff and (ii) the content of potentially bioavailable P forms (i.e., desorbable and redox-sensitive PP/TSS) in sediment.The reason for selecting P Ac as a soil P-test is obvious: the large database of P Ac concentrations of Finnish agricultural soils.As a soil P saturation index we used a procedure based on Mehlich 3 extractant, a rapid method that is used as an agri-environmental P saturation index in Delaware, USA (see Sims et al. 2002).According to our unpublished data, DPS M3 also correlates well with the DPS index used in recent Finnish studies (Turtola andYli-Halla 1999, Peltovuori et al. 2002) where non-apatite soil P was related to oxalate-extractable Al and Fe.

Soils and soil analyses
During the summer of 2002, about 50-l soil samples were retrieved from the Ap horizons (0-20 cm Vol. 14 (2005): 335-345.depth) of 15 fields located in southern Finland (Fig. 1).Soil samples were taken with a spade, thoroughly mixed, and a 1-l subsample of each soil was taken for laboratory analyses.This subsample was air-dried at +35°C and ground to pass through a 2-mm sieve, whereas the rest of the sample was stored field-moist in a plastic bag in a storage (at outdoor temperatures) for one to four months prior to rainfall simulations.
An overview of the physical and chemical properties of the soils is given in Table 1.Particlesize distribution was determined by the pipette method of Elonen (1971), soil pH was measured in 1:2.5 (v/v) water suspension with a glass electrode, and total C was analyzed with a LECO (St.Joseph, MI, USA) CN-2000 analyzer.For oxalate-extractable Al and Fe, the method of Schwertmann (1964) was used; the concentrations of Al and Fe were determined with an inductively coupled plasmaatomic emission spectrometer (ICP-AES; Thermo Jarrell Ash, Franklin, MA, USA).The agronomic P status was assessed using an extraction with acid (pH 4.65) ammonium acetate buffer (P Ac ;Vuorinen and Mäkitie 1955), whereas extraction with Meh-   -----------------% -----------------% mg l -1 -mmol kg  (Mehlich 1984) was conducted for an assessment of soil P saturation (DPS M3 ; see, e.g.Sims et al. 2002).In calculating DPS M3 , the molar concentration of M3-extractable P was related to the molar concentrations of M3-Al and M3-Fe: Phosphorus, for P Ac and M3-P, was determined by molybdate colorimetry using a Bran + Luebbe (Norderstedt, Germany) autoanalyzer with 880 nm filter, and Al and Fe in the M3-extracts were determined with the ICP-AES.

Rainfall simulation
The rainfall simulator used was a drop former-type stationary laboratory simulator, a modification of the drip-type simulator described by Bowyer-Bower and Burt (1989).In the used equipment, deionized water is pumped to a column where water level above the drop former is kept at a constant level by regulating bypass flow; in this work, the water head height was maintained at 40 mm.From the column, water is conducted to 0.020-mm (inner diameter) capillary tubes which form the drops, and with the above noted water head setting, a single drop has an average weight of 37.5 mg.There are 96 drop-forming capillars attached to a onesquare meter steel frame.The drop fall height is adjustable, in this work set to 2.3 m.Kinetic energy of the simulated rain at impact on the sample surface is undefined and variable, as a screen, of 0.5-mm stainless steel wire and with 3-mm openings, breaks the drops 0.53 m below the drop former, or 1.8 m above the sample surface.
For the rainfall simulation experiments, conducted unreplicated, the soils were brought in a saturated state to simulate the conditions during the main runoff periods in winter and spring when most of the soils of south Finland are plowed, bare and wet.Field-moist soil was spread into a 40 × 60 cm plywood box in several about 1-cm layers.The addition of a layer of soil was followed by dripping it with deionized water until moist.
After a 5-7-cm soil layer was reached, the box was set on a 2.5-2.8°angle under the rainfall simulator, and rainfall with 5 mm h -1 intensity was applied until runoff was just about to start.The soil was then covered by a plastic sheet and left to stand for two or three days at +4-5°C in order to obtain uniform moisture.After this, the box with soil was again placed under the rainfall simulator, set on a 2.5-2.8°angle, and rainfall with an intended intensity of 5 mm h -1 (actual intensity varying from 4.4 to 6.0 mm h -1 ) was applied at the room temperature (about +20°C).Runoff was collected in 2-l polyethene containers which were after a 2-h runoff period thoroughly shaken, and two subsamples were taken in 0.5-l plastic bottles.These were stored at +4-5°C in the darkness until analyzed.

Runoff analyses
The P chemistry of the water samples obtained from runoff simulations is summarized in Table 2. Concentration of DRP in runoff was determined after filtration of a subsample through a 0.2 µm Nuclepore filter (Whatman, Maidstone, UK) and that of TP after an autoclave-mediated digestion (added K 2 S 2 O 8 and H 2 SO 4 , 120°C, 100 kPa, 30 min) of an unfiltered subsample.The concentration of PP was taken as the difference between TP and DRP.Modification of the molybdenum blue method (Murphy and Riley 1962) was employed in P analyses with a LaChat (Milwaukee, WI, USA) QC analyzer equipped with 880 nm wavelength filter.The concentration of total suspended solids (TSS) was estimated by weighing the evaporation residue of 50 ml of runoff.
To estimate the amount of readily algal-available PP in runoff (see Uusitalo and Ekholm 2003), 40-ml portions of each runoff sample were measured into three 50-ml capacity centrifuge tubes.Into each of the tubes, a nylon netting bag containing 1 g of Dowex (Fluka, Neu-Ulm, Germany) 1 × 8 strongly basic AER, saturated with HCO 3 -ions (see Sibbesen 1977Sibbesen , 1978)), was added.The tubes were capped and placed on an end-over-end shaker and shaken at 37 rpm.After 20 h of shaking, the Vol.14 (2005): 335-345.AER bag was removed from the sample and washed with deionized water.The AER with sorbed P was thereafter shaken for 4 h in 40 ml of 0.5 M NaCl to displace P from the AER into the solution.The NaCl solution with P was acidified with 1 ml of 6 M HCl and allowed to stand overnight before P determination; acidification was done to reduce CO 2 evolution during the P determination.To obtain an estimate for desorbable PP (AER-PP), DRP was subtracted from AER-extractable P.
To approximate redox-labile PP, runoff was amended with bicarbonate and dithionite solutions (detailed description of the method is given in Uusitalo and Turtola 2003).In short: triplicate 40ml subsamples of runoff were measured into 50-ml capacity centrifuge tubes, and 1 ml of 0.298 M NaHCO 3 and 1 ml of 0.574 M Na 2 S 2 O 4 solutions (dithionite prepared just before the extraction) were added in each tube.The tubes were capped and shaken for 15 min on an orbital shaker at 120 rpm.After shaking, the sample was immediately decanted into a suction filter device equipped with a 0.2 µm Nuclepore membrane.For colorimetric P determination, 10 ml of the filtrate was digested in an autoclave with peroxodisulfate and sulfuric acid, as in TP analysis.The digestion was conducted to eliminate disturbances (by dithionite and soluble Fe) during the P determination.The amount of BD-PP was calculated as BD-extractable P (BD-P t ) less DRP.Calculated this way, BD-PP also includes some unreactive dissolved P [strictly, total dissolved P (i.e., TP <0.2 µm ) should be used instead of DRP], but earlier findings suggest that in turbid runoff the error due to this incoherence is small [in a material of 49 field runoff samples, 6% overestimation in BD-PP was recorded by Uusitalo et al. (2003)].

Results and discussion
Agronomic P status and P saturation of the soils Ammonium acetate is a weak extract for soil P, the amounts of soil P extractable in the acetate buffer Uusitalo, R. & Aura, E. Relationships between P pools in runoff versus soil test P (in 1:10 soil-to-solution ratio) being approximately the same as those solubilized in 1:60 soil-to-water suspension (Yli-Halla 1989, Saarela et al. 2003).
The relationship between DPS M3 and P Ac (Fig. 2) curves at the higher DPS levels in a manner typical for quantity-intensity relationships.With increasing P saturation, P sorption occurs with decreasing bonding energy, and the concentration of easily soluble P (here, P Ac ) starts to increase more rapidly for each one percent increase in P saturation (see Olsen andWatanabe 1957, Ballaux andPeaslee 1975).
When the agronomic P status for clayey soils with 3-6% organic matter (1.8-3.5% C) is inferred from the amount of soil P Ac , the limits between "satisfactory" and "good" P status is at 14 mg P Ac l -1 , that between "good" and "high" at 23 mg P Ac l -1 , whereas the limit between "high" and "excessive" P status is at 40 mg P Ac l -1 (Viljavuuspalvelu Oy 2000).According to the equation in Fig. 2, these limits would correspond to DPS M3 values of about 5, 8, and 11%, respectively.For comparison, Sims et al. (2002) described the agri-environmental interpretation of DPS M3 used in Delaware, USA, in which soils having DPS M3 between 6 and 11% are classified as having "optimum P level", with "economic response to P unlikely".The soils having higher DPS M3 than 11% are "above optimum, no P recommended".Further, for soils having DPS M3 higher than 15% "improved P management to reduce potential for nonpoint P pollution" should be implemented (Sims et al. 2002).

Soil P saturation and runoff P forms
In our material, there were only two soils with DPS M3 about 11%, and these produced runoff with DRP concentrations above 0.5 mg l -1 (Fig. 3 and Table 2).According to our data, a DPS M3 value as low as 5% would be close to a critical point above which runoff DRP starts to rapidly increase with increasing DPS M3 .If economic response to P would be unlikely at DPS M3 above 6%, as is the case in Delaware soils (Sims et al. 2002), Fig. 3 further suggests that the environmental risks of surplus P additions start to accumulate after this DPS M3 level is exceeded.Assuming that the relationship between DPS M3 and P Ac would be applicable for several types of Finnish soils, the 5% DPS M3 would correspond to a typical P Ac concentration of the Finnish agricultural soils, 12-13 mg P Ac l -1 (Saarela et al. 2003).
Whilst the relationship between soil P saturation and the DRP concentration in runoff (Fig. 3) has a shape similar to many of the published curves between soil P and runoff DRP concentration (e.g.Heckrath et al. 1995, McDowell andSharpley 2001), the contents of AER-PP and BD-PP in eroded sediment were in our material linearly related with DPS M3 (Fig. 3).However, there are a limited number of sorption sites in a kilogram of sediment matter, and the concentrations of these P forms must level out at some higher P saturation level (that was not covered by our material; i.e., the relationship actually has a shape of a Langmuirtype curve).By comparing the paper of Raven and Hossner (1993) to that of Barrow and Shaw (1977), it seems that at least some soils require excessive P additions to show a saturation of surface sorption sites.The former authors reported linear increase in AER-extractable P upon relatively modest P additions of up to 100 mg P kg -1 .In turn, Barrow and Shaw (1977) could show saturation phenomenon by AER extractions, but the amounts of P they added were up to 1500 mg kg -1 soil.Vol. 14 ( 2005): 335-345.eroded matter.Therefore, we acknowledge that the theoretical basis for predicting runoff P forms from soil P Ac is somewhat slender (see Fried and Shapiro 1956).The equations for P Ac and runoff P forms presented in Table 3 are merely descriptions of the correlation between P Ac and the P forms in runoff, not causal relationships.However, the use of P Ac to assess the TSS-normalized content of potentially bioavailable runoff P is emphasized for practical reasons; the data of soil P Ac cover well the area of agricultural soils of Finland, there are long timeseries of soil P Ac concentrations, and these data are readily available.
The different P forms in runoff increased with the concentration of P Ac of the source soils (Fig. 4), and the linear equations for runoff P forms and soil P Ac are given in Table 3.Our material was not evenly distributed along the P Ac axis -two soils had higher P Ac content than the other soils, and applied leverage to the fitted lines.We therefore ran the fitting procedure twice, with and without these two soils, and there are two estimates for the slope and intercept values listed (in Table 3 indicated with suffixes 50_ and 25_).
Runoff DRP concentration correlated better with soil P Ac than did the two forms of runoff sediment-normalized PP.The slope value estimate for the P Ac versus 50_DRP relationship was essentially equal to that earlier reported by Uusitalo and Jansson (2002) for P Ac versus edge-of-field runoff DRP: 0.015-0.018mg l -1 increase in DRP concentration for each unit increase in soil P Ac .For the narrower range of P Ac (25_DRP), increase in runoff DRP concentration was more modest, 0.009 mg l -1 for a unit increase in P Ac ; also this slope was significantly different from zero, and the correlation coefficient had an equally high value as the one for the whole material (Table 3).Uusitalo and Jansson (2002) made an assessment that DRP concentration of edge-of-field runoff sampled from a small, about 200-ha, catchment could in 95% of the cases be inferred from soil P Ac data with an accuracy of about 0.1 mg l -1 .The wide deviation in slope values for the 50_DRP and 25_DRP data sets (and the C.I.s of the slope estimates) suggest that the accuracy in predicted DRP concentrations for the material of the present study is not better than that.2.

Agrononic P Ac test versus runoff P forms
Runoff P forms -DRP in runoff and AER-or BDextractable sediment P -are not outcomes of soil P Ac , but all of these variables reflect the P saturation density of P sorbing compounds in soils and Uusitalo, R. & Aura, E. Relationships between P pools in runoff versus soil test P Table 3. Table of results, with standard errors and confidence intervals (CI), for the linear relationships between the concentration of soil P Ac and P forms in runoff (see Fig. 3).The units are: mg l -1 soil for P Ac , mg l -1 runoff water for DRP, and mg kg -1 eroded sediment for AER-PP/TSS and BD-PP/TSS.The estimates associated with the headings with a suffix 50_ are for the whole studied material (n = 15; P Ac scale up to about 50), whereas those with the suffix 25_ are estimates after omitting the two soils with P Ac concentrations of about 40 and 46 mg l -1 (n = 13; P Ac scale up to about 25).It seems that prediction of the sediment content of AER-extractable P and BD-P from soil P Ac could also be possible.However, due to the limited number of observations and the diversity of the properties of the soils studied here (e.g.clay content, pH, Al/Fe ratio; Table 1), the estimates presented in Table 3 have a high degree of uncertainty.As example, the slope values of the P Ac versus AER-PP/TSS and BD-PP/TSS relationhips significantly differed from value zero when the whole material of 15 soils was utilized, but not when the two soils with the highest P Ac were omitted (Table 3).
To compare our rainfall simulation data on PP losses to those of field studies, the estimates of mean slope and intercept for the whole P Ac range (50_P forms) were utilized to predict concentra-Vol.14 (2005): 335-345.tions of potentially bioavailable P forms in runoff sediment at four field sites that have been intensively monitored (Table 4; observed values retrieved from Uusitalo 2004).Predictions using the mean values of Table 3 overestimated the AER-PP/ TSS in runoff by 14-50% at three field sites with P Ac values 4-9 mg l -1 and underestimated this ratio by 22% at one site (Aurajoki) where P Ac was highest (11-23 mg l -1 ).When using the mean estimates for slope and intercept, BD-PP/TSS was underestimated at all of the four fields, by 21-55%, and the more the higher was soil P Ac .Even if the upper limit-values of the 95% confidence interval for the slope and intercept estimates were used in these calculations, the predicted BD-PP/TSS would have been smaller than the observed mean concentrations.It is noted that the simulated runoff contained more suspended matter than field runoff samples contain, and high concentration of soil matter probably contributed to a relatively low P recoveries in BD extractions of simulated runoff (see Uusitalo and Turtola 2003).

Fig. 1 .
Fig. 1.Map of Finland showing sampling locations of the 15 soils used in rainfall simulations.Soil characteristics are given inTable 1.

Fig. 2 .
Fig. 2. Relationship between the degree of P saturation in soils (DPS M3 ) and ammonium acetate-extractable P (P Ac ); the parallel dotted lines indicate the 95% confidence band of the fitted curve.The agronomic interpretation of P Ac concentrations for clayey soils (with 1.75-3.5% total C) at the upper end of the scale is inserted, the limits being indicated by the horizontal lines.

Fig. 3 .
Fig. 3.The concentrations of dissolved molybdate-reactive P (DRP) in runoff water (uppermost figure), and desorbable PP (AER-extractable PP; figure in the middle) and redox-labile PP (solubilized at the negative redox potential of BD extraction; lowermost figure) in a kilogram of runoff sediment, as a function of the degree of soil P saturation (DPS M3 ).The dotted lines indicate 95% confidence bands of the (solid) fitted lines.For details on runoff chemistry, see Table 2.
Fig. 4. The concentrations of dissolved molybdate-reactive P (DRP) in runoff water (uppermost figure), and desorbable PP (AER-extractable PP; figure in the middle) and redox-labile PP (solubilized at the negative redox potential of BD extraction; lowermost figure) in a kilogram of runoff sediment, plotted against ammonium acetate-extractable soil P (P Ac ).The dotted lines indicate their 95% confidence bands of the (solid) fitted lines.The estimates of the slopes and intercepts of these lines are given in Table 3.

Table 2 .
Concentrations of dissolved molybdate-reactive P, anion exchange resin-extractable P, redox-labile P, total P, and total suspended solids in the runoff samples obtained from rainfall simulations.Source soil numbering as in Table1.

Table 4 .
Observed mean concentrations of potentially bioavailable PP in runoff from four clayey fields of southwest Finland where runoff analyses have included determinations of AER-P and BD-P (data retrieved from Uusitalo 2004), and predicted mean concentrations for these soils, calculated using the average values for slopes and intercepts listed in Table3(50_-values) and the soil test P (P Ac ) values given here in parentheses.