A flow-through reactor to assess potential phosphate release from agricultural soils

Geoderma 219–220 (2014) 125–135

Contents lists available at ScienceDirect

Geoderma journal homepage: www.elsevier.com/locate/geoderma

A flow-through reactor to assess potential phosphate release from agricultural soils Emmanuel Frossard a,⁎, Paolo Demaria a, Sokrat Sinaj a,b, Michael Schärer a,c a b c

Institute of Agricultural Sciences, ETH Zurich, 8315 Lindau, Switzerland Agroscope research station Changins-Wädenswil ACW, Route de Duillier 50, Case postale 1012, 1260 Nyon, Switzerland Federal Office for the Environment, Water Division, 3003 Berne, Switzerland

a r t i c l e

i n f o

Article history: Received 24 July 2013 Received in revised form 7 December 2013 Accepted 11 December 2013 Available online xxxx Keywords: Desorption Flow-through reactor Isotopic exchange Kinetics Phosphate Soil

a b s t r a c t Controlling phosphate (P) release from agricultural soils to water while maintaining optimal plant growth conditions remain a major challenge for the development of sustainable agricultural systems. To achieve this, it is important to have a proper knowledge of the amount of soil P that can be mobilized by water and of the kinetics of P release. We evaluated the ability of a flow-through reactor in which 33P labeled soils can be inserted and leached continuously with deionized water, to assess P release. The experiment was conducted on five grassland soils presenting a large range in P availability. The availability of P in these soils was further modified by submitting them to 0 to 3 plant growth cycles with Italian ryegrass (Lolium multiflorum) with three levels of P added (0, 20 and 40 mg P kg soil−1 ). The P input–output balance, water and oxalate extractable P, the degree of P saturation of the soil and the amount of isotopically exchangeable P (E value) were assessed in all samples. A subset of these soil samples was labeled with 33P, introduced in a flow-through reactor and the release of P and 33P measured over 14 days. The cumulated amount of P released after 14 days was strongly correlated to the amount of oxalate extractable P, isotopically exchangeable P (E value), and water extractable P. The P release kinetics was modeled with a 2 pools model with each pool following first order kinetics. Plants were able to take up P from both pools. Assuming that the leached P had the same isotopic composition as the pool of soil P it came from it became possible to quantify the amount of isotopically exchangeable remaining in the soil which was called the D value. D decreased during the three first days of the flow-through experiment and then increased linearly with time reaching a maximum after 14 days. This maximum remained lower than the oxalate extractable P. Processes contributing to this increase were isotopic exchange and possibly also some organic P mineralization. The D value was strongly linearly correlated to E values measured after different exchange times, but for a given exchange time, the D value was lower than the E value, whereas equality could have been expected. This difference was related to the high rate of 33P export from the soil at the beginning of the flow-through experiment. The D value was also strongly correlated to the oxalate and water extractable P. In conclusion, we suggest that the use of the flow-through reactor yields relevant information on the amount of P that can be leached from a given soil, and that the D value delivers information on the amount of isotopically exchangeable P remaining in the soil and therefore which could still be leached if sufficient time would be given. © 2013 Elsevier B.V. All rights reserved.

1. Introduction Excessive use of phosphate (P) is a problem in many intensively managed agro-ecosystems, especially in those presenting a high livestock density (Sutton et al., 2013). Inputs of P in amounts that exceed plant needs lead to P losses to water bodies and to their eutrophication. Haygarth et al. (2005) provide a framework to analyze P losses from agro-ecosystems in which they consider the potential amount of soil P which can be mobilized, the mobilization of P by water, and the P transfer to water bodies. Some works attempted to decrease the transfer of

⁎ Corresponding author. Tel.: +41 52 354 91 40; fax: +41 52 354 91 19. E-mail address: [email protected] (E. Frossard). 0016-7061/$ – see front matter © 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.geoderma.2013.12.015

P to water bodies, e.g. by installing buffer strips between fields and water bodies. However, the efficacy of such measures has been reported to be variable (Noij et al., 2013). The option of decreasing P mobility by increasing soil sorption capacity for P has also been explored, for instance by amending the soil with sorbents rich in Fe or Al (Chardon et al., 2012; Groenenberg et al., 2013; McDowell and Nash, 2012). Although promising results have been obtained, the efficiency of such amendments and especially of Fe-rich amendments might not hold on the long-term because of the cycles of reduction/oxidation decreasing the efficiency of Fe oxides as sorbent (Schärer et al., 2007, 2009). In any case, the amount of P that can be mobilized has also to be decreased. This can be achieved by decreasing or stopping P inputs and/or by increasing P outputs (Dodd et al., 2012; Koopmans et al., 2004a; van der Salm et al., 2009). These approaches are also promising but depending

126

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

on plant growth and initial available P stocks, it may take decades to decrease P availability to a level that is low enough to hinder losses (Dodd et al., 2012). It is important, before implementing a strategy aiming at decreasing P losses, to correctly estimate the amount of soil P which can be released to water and the rate at which it can be released. Kinetics of P release have been studied in batch experiments in which soils were extracted by P-free aqueous solutions in the presence or not of a sink for P (Freese et al., 1995; Hosseinpur and Pashamokhtari, 2008; Lookman et al., 1995; McDowell and Sharpley, 2003; van Rotterdam et al., 2009). These works showed that P release is non-linear and that in P-rich soils, desorption does not reach a plateau even after a very long desorption time. Lookman et al. (1995) extracted soil P using an Fe-hydroxide inserted in a dialysis membrane and modeled the kinetics of P release by considering two pools, each one being described by a first order rate equation. Other authors proposed different models to fit P release kinetics (Hosseinpur and Pashamokhtari, 2008; Koopmans et al., 2001; McDowell and Sharpley, 2003). Although many authors assume that in acid sandy soils the oxalate extractable P can be considered as fully extractable on the long term (Koopmans et al., 2004a; Lookman et al., 1995), they do not provide direct evidence of the mobility of P remaining on the soil's solid phase. Other approaches, based on the use of P radio-isotopes, can identify a fraction of soil P which is exchangeable, i.e. which can reach the soil solution at a given point in time, without having to extract the entire pool of exchangeable P (Di et al., 1997; Larsen, 1967). Isotopically exchangeable P is measured in soil/water suspensions yielding the socalled E value or in soil/plant systems yielding the so called L value (Di et al., 1997; Larsen, 1967). The amount of soil P that is isotopically exchangeable is a function of time and is the main source of P for most crop plants (Fardeau, 1996). A criticism which can be made to experiments conducted in batch i.e. to most approaches measuring P desorption kinetics and to measuring the E value, is that they are conducted in suspensions with a high solution to soil ratio which are regularly shaken. This can lead to aggregate dispersion, particle abrasion and to the exposition of new surfaces during the experiment (Koopmans et al., 2004b; Randriamanantsoa et al., 2013). This in turn can increase the rate of reaction (e.g. of P desorption or P exchange) between the soil and solution (Koopmans et al., 2004b; McDowell et al., 2001; Sinaj et al., 1997). A way to minimize these modifications would be to conduct P release experiments in a flow-through reactor as the one designed by Freese et al. (1999), as its reactor can accommodate soil at a soil to solution ratio close to 1 and as it can be leached continuously with deionized water. Furthermore, using soil labeled with radioactive P in such a flow-through reactor experiment could provide, in addition to the amount of P effectively leached, information on the amount of isotopically exchangeable P remaining in the soil and therefore which could still be leached if sufficient time would be given. The amount of exchangeable P remaining in the soil could be calculated from the isotopic composition (33P/31P) of the leached P, assuming that the leached P would stem from a pool of soil P having the same isotopic composition. The rationale behind this assumption is that the P present in the soil solution is in constant isotopic exchange with the P present on the solid phase. This hypothesis has been made earlier for the calculation of the isotopically exchangeable P as a source of plant available P by Larsen (1967). This type of approach has to our knowledge not yet been evaluated. The objectives of this study were: i) to assess the isotopic composition of P released from 33P labeled soil samples inserted in a flowthrough reactor and leached continuously with deionized water; ii) to model the amounts of P released from soils; iii) to assess the amount of soil P being in isotopic equilibrium with the leached P (which will be called thereafter the D value) and iv) to compare the amount of P released from the soil and the D value with indices of soil P availability measured in batch experiments and from plant growth experiments.

2. Materials and methods 2.1. The soil samples We sampled soils in 5 permanent grasslands to a depth of 10 cm in early January 2006 in the lake Baldegg watershed, canton of Lucerne, Switzerland. This area supports an intensive livestock production and the excessive P application rates on grasslands have been a major source of P for the lake during the last decades (Kanton Luzern, 2001). Liebisch et al. (2013) have shown that in this watershed intensively managed grasslands can still receive more than 110 kg P ha-1 year-1. Soils were sieved at 5 mm and then oven-dried at 40 °C for 7 days. Stones, roots and other plant residues were discarded during sieving. Dried subsamples of about 400 g were further sieved at 2 mm and stored for laboratory analyses (P release kinetics, oxalate extractable P, isotopically exchangeable P and other soil characteristics). The rest was kept for the pot experiments described hereafter. Sampling location and selected soil properties are listed in Table 1. 2.2. Pot experiments Three sequential growth cycles (8 weeks for each cycle) were carried out on each soil using Italian ryegrass (Lolium multiflorum, cultivar Onyx). For each cycle, pots were filled with 500 g (dry weight) of 5 mm sieved soil and sown with 1 g seed kg soil−1 . For each of the 5 soils, 3 P fertilization levels were considered: no P added (0P), addition of 20 mg P kg soil−1 (P20) and 40 mg P kg− soil−1 (P40). Phosphate was added at the beginning of each growth cycle as KH2PO4. The soil samples are designated in the text with their abbreviation, P fertilization rate and growth cycle. For instance “A1 start” designates the original A1 sample, whereas “A1 0P 3 cycle” designates the A1 soil which has been submitted to 3 cycles of plant growth without any P inputs. Nitrogen and potassium were added to the soil prior to sowing and after each cut as NH 4 NO 3 (100 mg N kg soil −1 prior sowing, 100 after the 1st cut, 122 after the 2nd cut and 112 after the 3rd cut) and as K 2 SO 4 (200 mg K kg soil −1 prior sowing, 83 after the 1st cut, 100 after the 2nd cut and 97 after the 3rd cut). Other nutrients were added prior sowing as follows: 34 mg Ca kg soil −1 as CaCl2.2H2O; 16 mg Mg kg soil−1 as MgCl2·6H2O; 2 mg Cu kg soil−1 as CuSO4·5H20; 2 mg Mn kg soil−1 as MnSO4·H20; 1 mg Zn kg soil−1 as ZnSO4·7H20; 1 mg B kg soil−1 as H3BO3; 0.1 mg Mo kg soil−1 as (NH4)6Mo7O24·4H20. The pot experiments were carried out in a greenhouse during the year 2006: the 1st cycle from mid-January to mid-March, the 2nd cycle from the beginning of May to the end of June, and the 3rd cycle from the end of July to the end of September. The pots were placed in the greenhouse in a completely randomized block design. An automatic watering system controlled by a Gardena C1060-profi electronic timer (Gardena, Switzerland) was set up delivering up to 80 ml of deionized water (Osmose water II) daily per pot. The grass was cut four times, every 2 weeks, at about 2 cm above the soil surface, dried at 40 °C, weighed and stored. After the 4th cut, the soils were recovered from the pots, sieved at 5 mm after removal of roots and oven-dried at 40 °C. Soil samples from the various replicates of each soil and treatment were pooled. A fraction (about 400 g) of the pooled sample was sieved at 2 mm and stored for further analyses (P release kinetics, oxalate extractable P and isotopically exchangeable P). The remaining amount of soil was used for the following growth cycle. The number of replicates for each soil/treatment combination was 24, 10 and 4 giving a total number of pots of 360, 150 and 60 for the 1st, 2nd and 3rd growth cycle, respectively. 2.3. Isotope exchange kinetics Isotopic exchange kinetics (IEK) experiments were conducted in triplicate on all soil samples (before plant growth, and after each plant

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

127

Table 1 Selected properties of the 0–10 cm horizon of the studied grassland soils. Soil

A1 C1 C3 C5 E5

Sampling location

Rain Gelfingen Gelfingen Gelfingen Kleinwangen

pH

Olsen-P

Total-P

(H2O)

mg P kg−1

g P kg−1

6.0 6.4 7.1 6.9 6.5

49.3 97.4 188 107 19.6

1.18 1.74 3.36 2.02 0.69

Organic P

0.63 0.83 1.14 1.01 0.53

growth cycle for the three P fertilization levels). The IEK experiment has been described in the literature (Fardeau, 1996). We give here only an outline of the method. Ten grams of air-dried soil were weighed in 250 ml plastic bottles and equilibrated with 99 ml of de-ionized H2O for 16 h at 22 °C on a rotating shaker. Before the beginning of an IEK, the bottles were placed on a magnetic stirring plate and stirred at 300 rpm to maintain a homogeneous suspension. At time t = 0, 1 ml containing 0.1–1 MBq ml−1 of carrier-free 33P ions (as H33 3 PO4 in HCl, specific activity at delivery N3.6 TBq 33P mg−131P, Hartmann Analytic, GmbH) was injected in the soil suspensions and about 5 ml samples were taken with a polyethylene syringe 1, 10, 30 and 60 min after 33P addition. The suspensions were rapidly filtered through a Sartorius 0.2 μm acetate cellulose membrane filter and 33P was measured by scintillation counting in the filtrates. The concentration of orthophosphate in the solution (CP) was measured after filtration at 0.2 μm with the malachite green colorimetric method (Ohno and Zibilske, 1991) after the 60 minute sampling.

Feox

3.25 3.96 2.80 2.22 2.36

Alox

0.91 1.47 1.07 1.76 1.18

CEC

Corg

cmol kg−1

g kg−1

21.5 27.4 29.2 31.2 22.3

66 63 88 108 56

Clay

Silt

Sand

100 334 258 322 300

556 341 284 375 352

344 325 458 303 348

filters clogging by fine soil particles. The flow rate of de-ionized water through the reactor was 0.18 ml min−1. The system ran continuously for 14 days. Eluates were collected daily, weighted and an aliquot of 20 ml was filtered at 0.2 μm with a Sartorius 0.2 μm acetate cellulose membrane filter before measuring 33P by scintillation counting and 31 P with the malachite green method (Ohno and Zibilske, 1991). The amount of leached 31P and 33 P and the total amount of added 33P were used to calculate the amount of soil P being in isotopic equilibrium with the leached P (the D value).

2.4. Flow-through reactor design and experiment The P release kinetics were studied by carrying out flow-through reactor experiments in which 33P-labeled soils were continuously leached with de-ionized water for 14 days. Preliminary experiments showed that the use of 2 mMCaCl2 caused a strong decrease of the leachate pH at the beginning of the flow-through experiment (Demaria, 2004). As this pH decrease could have caused the dissolution of soil P species we preferred to use de-ionized water as eluent. These experiments were carried out on 26 soil samples: those taken before ryegrass growth, those taken after the 1st and 3rd growth cycles from the 0P and P40 treatments and in addition on the E5 sample that had been fertilized with 40 mg P kg−1 and collected after the 2nd growth cycle (E5 40P 2 cycle). We selected these soil samples as they provided a wide range in P availability. Three replicates were made yielding a total of 78 flow-through experiments. Soil samples were labeled with carrier-free 33P ions and incubated at about 60% of soil water holding capacity for 14 days, at 22 °C, 50% of atmospheric humidity and in absence of light. The amount of added 33P was adapted in order to obtain a measurable 33P concentration in the leachates throughout the 14 days of the flow-through reactor experiment. The added amounts ranged between 277 and 731 MBq kg soil−1 . The flow-through reactor employed in this study was designed by Freese et al. (1999) and further developed by Demaria (2004; Fig. 1). The reactor column was a polymethylmethacrylate cylinder with an internal diameter of 25 mm and, from bottom to top, consisted of: a ringshaped silicon gasket; one 1.6 μm Whatman GF/A glass microfiber filter; about 4 g of acid washed silica sand (Siegfried AG, Switzerland); 10 g (dry weight) of 33P-labeled soil sample; another layer of 6 g of silica sand; one 0.7 μm Whatman GF/F glass microfiber filter. The columns were connected with Tygon tubes (Fischer Scientific AG, Switzerland, internal diameter of 2.06 mm) to a 24 channels peristaltic pump (Ismatec, IPC-N—ISM 939). The eluates were collected into 250 ml conical flasks. An upward flow was chosen in order to reduce the effects of

Fig. 1. Scheme of the flow-through reactor (see text for a detailed description). a) Filter holder bottom piece; b) silicon ring-shaped gasket; c) 1.6 μm Whatman GF/A glass microfiber filter; d) acid-washed silica sand; e) soil (10 g dry weight); f) 0.7 μm Whatman GF/F glass microfiber filter; g) filter holder top-piece.

128

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

2.5. Other analyses 2.5.1. Soil analyses The five start soil samples were analyzed before plant growth for pH (H2O), organic carbon, texture and cation exchange capacity following the reference methods of the Swiss research stations for agriculture (FAL et al., 1996). In addition we measured the amount of soil P extractable using the Olsen method (Olsen et al., 1954), the amount of organic P using the method proposed by Saunders and Williams (1955) and the total P content of the soils. Total P was estimated by extracting 1 g of soil with 50 ml of 0.5 M H2SO4 for 16 h, after ignition of the sample at 550 °C for 3 h. Organic P was calculated by subtracting the amount of P extracted from 1 g of non-ignited soil sample with 50 ml of 0.5 M H 2 SO 4 for 16 h from the total P (Saunders and Williams, 1955). The extracts were filtered with a Sartorius 0.2 μm acetate cellulose membrane filter before the measurement of the P concentration with the malachite green method (Ohno and Zibilske, 1991). Oxalate extractable Fe, Al and P were measured on all soil samples before the pot experiment and after each growth cycle and for each P fertilization regime. We used the method of McKeague and Day (1966) for the extraction oxalate extractable Fe, Al and P and their concentrations were measured on an Inductively Coupled Plasma Mass Spectrometer (ICP/MS). The analyses were conducted on single samples with only a few replicates. The coefficient of variation of these replicates was in average lower than 2%. The oxalate extractable Al, Fe and P results were used to estimate the degree of P saturation of the soil (DSP) considering a α coefficient of 0.5 as in Schoumans and Groenendijk (2000). We are aware that α varies from soil to soil (van der Zee and van Riemsdijk, 1988) and the value of 0.5 is used here as a first approximation. 2.5.2. Plant analyses For the 1st and the 2nd growth cycles, 4 replicates per treatment were randomly chosen and the shoots were analyzed for total P. For the 3rd growth cycle all replicates were analyzed as this last pot experiment had only 4 replicates. Dried shoots were ground and 100 mg were incinerated at 550 °C for 3 h. The ashes were dissolved in 2 ml of 22 M HNO3 and diluted with de-ionized water up to 50 ml. The extract was then filtered with a Sartorius 0.2 μm acetate cellulose membrane filter and stored at 4 °C before measuring P concentration by ICP/MS. 2.6. Data analysis and statistics 2.6.1. Phosphorus uptake and balances from the pot experiments The P uptake by plant shoot is reported in a cumulative manner. P uptake in shoots from the 1st growth cycle is the sum of uptake from the 4 cuts. The amount of P exported in a cut is calculated as the product between the concentration of P and the dry matter produced. P uptake at the end of the 2nd growth cycle is equal to the P uptake during the 1st growth cycle plus the P uptake during the 2nd growth cycle. Finally P uptake at the end of the 3rd growth cycle is the sum of P taken during the 1st, the 2nd and the 3rd growth cycles. The root biomass and the concentration of P in roots were not measured but estimated at the end of each growth cycle using results from another series of pot experiments carried out with English ryegrass (Lolium perenne) under similar conditions (soil volume, pot size, seeds amount, environmental conditions) by Gallet et al. (2003a). We considered a root biomass of 4 g dry matter kg soil−1 and per cycle for all soils and treatments, except for the 0P treatment of the low-P soil (E5) for which we considered a root biomass of only 2 g kg soil−1 as observed by Gallet et al. (2003a) in an unfertilized low-P soil. Finally we considered that the P concentration of the roots was similar to that of the aerial parts measured at the end of each cycle. Using these shoot and root uptake data, considering that no P losses had occurred during the

pot experiments by leaching, and knowing the amounts of P added to the soil before each growth cycle, we calculated an input–output P balance for each treatment. 2.6.2. Isotope exchange kinetics (E value) The decrease with time of the 33P added in solution can be described by the following function:
 1 −n rt r ¼ m  t þ mn þ ∞; R R

ð1Þ

where rt and r∞ (MBq) are the radioactivity remaining in solution after t minutes of exchange and after an infinite time of exchange, respectively; R (MBq) is the initially added radioactivity; t is the time (in minutes) elapsed after the radioactivity addition and m and n are soil specific parameters calculated from a non-linear regression between rt / R and t. The r∞ / R value is estimated as the ratio of the water soluble P expressed in mg P kg−1 to the oxalate extractable P also expressed in mg P kg− 1. The amount of water soluble P is calculated as 10 × CP with CP expressed in mg P L−1 and 10 being the solution to soil ratio (100 ml:10 g). The amount of soil isotopically exchangeable P (Et, expressed in mg P kg soil−1 ) was calculated assuming that (i) 33P and 31 P ions have the same fate in the system and (ii) that at any given time t the specific activity (SA = 33 P/ 31 P) of the phosphate ions in soil solution (rt / 10 × CP) equals the specific activity of the pool of isotopically exchangeable P ions in the whole system (R / Et): r t =ð10  C P Þ ¼ R=Et ;

ð2Þ

which, solved for Et, gives: Et ¼ 10  C P  ðR=r t Þ:

ð3Þ

We calculated for this paper the amounts of P isotopically exchangeable within 1 min (E 1min , mg P kg soil −1 ), 60 min (E 60min , mg P kg soil−1 ) and 14 days (E14days, mg P kg soil−1 ). 2.6.3. Modeling P desorption We assumed as in Lookman et al. (1995) that our data could be described by the sum of two pools P that would desorb following first order kinetics. The amounts of desorbed P with time were therefore fitted to the following equation:

 −k t −k t þ Q 2  1−e 2 ; P dest ¼ Q 1  1−e 1

ð4Þ

where Pdest is the cumulated amount of P desorbed (mg P kg soil−1 ) after time t (hours); Q1 is the amount of rapidly desorbable P (mg P kg soil−1 ); k1 is the rate of desorption (hour−1) of P from Q1; Q2 is the amount of slowly desorbable P (mg P kg soil−1 ) and k2 is the rate of desorption (hour−1) of P from Q2. We assume that the maximum amount of P that can potentially be desorbed is the amount of oxalate extractable P (Poxmg P kg soil−1 ) as suggested in Lookman et al. (1995) and therefore when time is infinite: P ox ¼ Q 1 þ Q 2 :

ð5Þ

The model was calculated for the mean values of the 3 replicates of each treatment from the flow-through reactor, as we have a single Pox value per treatment. 2.6.4. Calculation of the D Value In analogy to works reporting the L value, which is the amount of isotopically exchangeable P measured in a plant growing on a soil labeled with radioactive P (Larsen, 1967), we introduce here the D value (D for desorption) which is the amount of P remaining in the soil at a

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

given time having the same isotopic composition as the P collected in the leachate. The D value is calculated as follows for each sampling time: Dn ¼

n −1 X

R−

i¼1

! r Di =

r Dn ; P Dn

ð6Þ

where: D n is the D value at the n th sampling step; R is the total amount of radioactivity added to the soil (MBq kg soil−1 ) and r Dn (MBq kg soil−1 ) and PDn (mg kg soil−1 ) are, respectively, the amounts of 33P and 31P released from the soil at the nth sampling step. The sum n−1

129

2.6.5. Statistics We report the mean and standard deviation for all measurements (except for Pox, DSP and the P balance). Linear and non-linear regressions were carried out using StatGraphics. Goodness of fit was evaluated with the coefficient of determination (r 2) and the standard error of the estimate. Significance of the regressions is given at the 95% confidence level. 3. Results and discussion

∑ r Di is the cumulated amount of 33P released between the first and

3.1. Changes in soil P availability following plant growth cycles and P inputs

the (n − 1)th step. All radioactivity measurements were calculated back to the date of radioactivity input and expressed in percent of the introduced radioactivity.

The removal of P by ryegrass shoots from the soils submitted to various growth cycles and to different P fertilization regimes, the estimated input–output P balance from these treatments and indices of

i¼1

Table 2 Phosphorus uptake by shoots of Italian rye grass after 1, 2 or 3 growth cycles fertilized with 0, 20 or 40 mg P kg soil−1 and estimated input–output P balance for each treatment; oxalate extractable P (Pox), degree of saturation with P (DSP) and the amount of P exchangeable within 1 min (E1min) measured on the initial soil and after each growth cycle. The mean values and, when possible, the standard deviation are presented. Soil/treatment

P uptake shoots Average

Input–output P Balance

Pox

E1min Average

mg kg−1 A1 Start A1 0P 1cycle A1 0P 2 cycle A1 0P 3 cycle A1 20P 1 cycle A1 20P 2 cycle A1 20P 3cycle A1 40P 1 cycle A1 40P 2 cycle A1 40P 3 cycle C1 Start C1 0P 1 cycle C1 0P 2 cycle C1 0P 3 cycle C1 20P 1 cycle C1 20P 2 cycle C1 20P 3 cycle C1 40P 1 cycle C1 40P 2 cycle C1 40P 3 cycle C3 Start C3 0P 1 cycle C3 0P 2 cycle C3 0P 3 cycle C3 20P 1 cycle C3 20P 2 cycle C3 20P 3 cycle C3 40P 1 cycle C3 40P 2 cycle C3 40P 3 cycle C5 Start C5 0P 1cycle C5 0P 2 cycle C5 0P 3 cycle C5 20P 1 cycle C5 20P 2cycle C5 20P 3 cycle C5 40P 1 cycle C5 40P 2 cycle C5 40P 3 cycle E5 Start E5 0P 1 cycle E5 0P 2 cycle E5 0P 3 cycle E5 20P 1 cycle E5 20P 2 cycle E5 20P 3 cycle E5 40P 1cycle E5 40P 2 cycle E5 40P 3 cycle

DSP

SD

40.1 88.4 118.3 55.3 108.3 150.7 67.2 128.2 177.5

4.5 8.7 8.8 7.3 7.5 5.8 7.7 3.8 5.2

−57.7 −120 −164 −58.2 −108 −148 −56 −95.2 −125

36.9 101.1 155.4 44.2 114.7 170.7 50.1 121.4 186.3

4.5 3.4 5.0 0.9 5.5 4.7 1.6 2.0 3.3

−56.1 −139 −218 −46.7 −116 −179 −34.2 −87.6 −139

44.7 107.5 147.2 53.6 122.4 168.8 47.5 113.6 161.4

5.4 8.9 10.1 5.3 4.3 4.6 2.5 4.0 3.8

−66.5 −146 −205 −58 −126 −173 −30.7 −75.4 −105

37.4 101.1 143.8 35.6 83.8 131.3 37.9 88.8 137.6

1.0 6.4 6.5 3.7 3.9 7.1 6.2 6.3 8.1

−56.9 −139 −204 −36 −80.5 −130 −19.2 −47.7 −79

8.7 24.9 33.5 12.8 36.1 56.3 18.4 51 87.4

0.3 0.7 1.0 1.7 2.2 2.3 1.0 2.2 1.0

−12.4 −31.6 −43.4 −1 −13.4 −25 10.7 4.6 −8

597 605 560 550 602 569 546 562 537 539 846 727 694 647 775 699 693 763 712 686 1535 1369 1361 1392 1507 1346 1343 1441 1476 1427 900 824 787 735 880 869 877 905 858 868 222 216 214 207 236 225 221 258 243 241

25.7 25.0 23.1 22.0 25.1 23.8 22.4 23.0 22.2 21.3 27.9 26.7 25.1 23.7 27.8 25.6 25.0 28.7 26.0 25.3 70.9 67.0 63.7 61.9 71.5 64.5 61.6 72.1 68.9 67.5 40.1 36.6 35.6 32.3 39.5 38.9 39.4 40.2 37.7 38.4 11.2 12.2 11.9 11.3 12.4 12.1 12.0 14.5 13.0 13.4

SD mg kg−1

% 13.9 9.40 3.63 2.21 9.42 3.23 2.91 7.93 3.62 3.12 24.5 18.8 10.4 10.5 17.5 12.6 11.2 21.2 16.3 14.1 49.7 27.7 10.9 13.5 18.2 18.5 14.9 21.6 24.1 19.6 28.9 23.0 9.66 10.3 19.3 18.2 13.1 20.1 21.3 15.5 5.06 2.20 1.44 1.51 3.99 1.70 2.49 5.31 4.57 3.96

0.46 0.50 1.06 0.43 0.75 0.10 0.15 0.12 0.30 0.14 4.64 1.90 1.1 1.97 0.11 1.28 0.36 1.25 0.71 0.85 1.24 0.69 0.98 0.58 2.07 1.63 0.61 0.71 2.20 0.99 2.65 0.17 0.92 0.39 0.51 1.43 1.00 1.10 2.26 2.06 0.35 0.15 0.11 0.04 0.15 0.10 0.22 0.47 0.19 0.24

130

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

P availability measured on all soils are reported in Table 2. There was no effect of P inputs on dry matter production of ryegrass except in the soil E5 during the 1st and the 3rd cycles and in the soil C5 during the 3rd cycle where P inputs increased shoot production (data not shown). The responsiveness of E5 was related to its low P availability (Tables 1 and 2). These results are in agreement with the works of Gallet et al. (2003b) and of Morel et al. (1992) showing that the input of P on soils having a E1min larger than 5 mg P kg soil−1 or an Olsen P content larger than 15 mg P kg soil−1 does not increase biomass production. The responsiveness of C5 soil on the contrary could not be explained. In all soils, P inputs led to an increased P uptake in shoots except for the soils C3 and C5 which showed variable results (Table 2). Phosphorus uptake after the 3 growth cycles was higher in all C1, C3, C5 soils and in the A1 samples fertilized with P, it was intermediate in the A1 soil that was not fertilized with P and was lower in the E5 samples. Thus P uptake was higher in the soils that contained the largest amounts of available P and lower in the soils that contained the lowest amounts of available P (Tables 1 and 2). The different growth cycles, in the presence or not of P inputs, resulted in most of the cases in negative P input–output balances reaching in the treatment C1 0P 3 cycle a value of − 218 mg P kg soil−1 . Positive balances were observed in only 2 treatments, E5 40P 1 cycle and E5 40P 2 cycle, and they were lower than 11 mg P kg soil−1 . Increasing P depletion was paralleled by decreasing available P. The decrease in labile P (E1min) was stronger in proportion of its initial value than the decrease of the more bound P (Pox) as observed earlier by van der Salm et al. (2009). The P balance was for each soil linearly correlated to the changes in P isotopically exchangeable within 1 min (ΔE1min)

and in oxalate extractable P (ΔPox) which were calculated by subtracting the P measured after each growth cycle from the P measured on the non-cultivated start soils (Fig. 2a and b). Fig. 2a shows that the relation between the P balance and ΔE1min could be fitted for each soil by a statistically significant linear correlation. The slopes of these equations were similar. However the ordinates at the origin (showing the effect of a very small P depletion on E1min decrease) varied from −1.1 for E5 to −23.5 mg P kg soil−1 for C3. The absolute values of these ordinates at the origin were significantly related by a power function to the degree of P saturation (DSP) of the non-cultivated soils (r2 = 0.92, n = 5, relation not shown). This shows that ΔE1min induced by a very small negative P balance will be much stronger for soils that have a high DSP, while ΔE1min will be smaller in soils with a low DSP. The relationships between the P balance and the changes in oxalate extractable P were also linear (Fig. 2b). Only one of these correlations was not significant, the one relating ΔPox to the P balance of the soil C3 (p = 0.08, n = 9), while the other correlations were significant. The relations observed between the changes in oxalate extractable P and the P balance suggest that oxalate extractable P is the “ultimate” source of available P for ryegrass in these neutral loamy soils confirming the findings of Lookman et al. (1995) and Koopmans et al. (2004a) for acidic sandy soils.

3.2. P release from soils: measurements and modeling The concentration of P recovered in the leachate at each sampling time and the cumulated release of P with time are shown in Fig. 3 for the soil C3. The highest P concentrations in the leachates were observed

10

E1min (mg P kg-1)

0 -10 -20 -30

A1: y = 0.067 x - 1.95; r2 = 0.82 C1: y = 0.057 x - 3.35; r 2 = 0.82 C3: y = 0.068 x - 23.5; r 2 = 0.55 C5: y = 0.071 x - 5.88; r 2 = 0.77 E5: y = 0.073 x - 1.07; r2 = 0.77

-40 -50 -250

-200

-150

-100

-50

0

50

P concentration in leachate (mg L-1)

a) a)

3.0 C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle

2.5 2.0 1.5 1.0 0.5 0.0 0

50

100

input-output P balance (mg P kg-1)

b)

b)

250

300

350

400

300

350

400

300 C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle

250

-50 -100 -150

2

A1: y = 0.43 x + 10.2; r = 0.47 C1: y = 0.58 x - 69.7; r2 = 0.85 C3 relation not significant C5: y = 0.73 x + 9.95; r2 = 0.66 E5: y = 0.86 x + 18.4; r2 = 0.83

Pdest (mg P kg-1)

Pox (mg P kg-1)

0

-250 -250

200

Time (hours)

50

-200

150

200 150 100 50 0

-200

-150

-100

-50

0

50

input-output P balance (mg P kg-1) Fig. 2. Relations between the P input–output balances caused by 1 to 3 growth cycles of Italian ryegrass submitted to different P fertilization regimes on 5 soils and a) changes in P isotopically exchangeable within 1 min (ΔE1min = E1min in a given treatment − E1min of the start soil), and b) changes in oxalate extractable P (ΔPox = Pox in a given treatment − Pox of the start soil).

0

50

100

150

200

250

Time (hours) Fig. 3. a) Concentration of P in the leachate at each sampling time and b) cumulated amount of released P from the soil C3 which had been submitted to 0 to 3 growth cycles with Italian ryegrass and different P fertilization regimes, measured in flow-through reactor experiments over a period of 14 days. The average values are shown surrounded by twice the standard deviation.

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

at the beginning of the experiment. After a while the concentration reached a minimum around 0.3 mg L−1 which remained constant till the end of the experiment. The cumulated P release increased with time, first rapidly and then slowly. A plateau could not be reached after 14 days of experiment. We observed similar trends with the other soils albeit with lower P concentration in the leachates (results not shown). These results are in full agreement with those presented in earlier publications on soil P release kinetics (Freese et al., 1995; Hosseinpur and Pashamokhtari, 2008; Koopmans et al., 2004a; Lookman et al., 1995; McDowell and Sharpley, 2003; van Rotterdam et al., 2009). The cumulated amount of P desorbed after 14 days (P des14days ) reached 10% of P ox in the E5 start soil, 13% of P ox in the A1 start soils and varied between 16 and 17% of P ox in the C1, 2 and 3 start soils. The total amount of P desorbed after 14 days (Table 3) was highly significantly linearly correlated to CP, E1min, E60min, E14days and Pox (r2 N 0.85, n = 26, graphs not shown). Relationships were observed between the input–output P balance and the changes in desorbable P after 14 days (ΔPdes14days). Two groups of soils can be distinguished in Fig. 4a, the C3 soil where a given negative P balance has a very strong effect on the depletion of desorbable P (this relation is not statistically significant because of the low number of points), and the other soils which fall in a group of points that can be described by a single significant linear regression. We explain the different behavior of soil C3 by its high phosphate saturation degree as discussed above. The two-pools model proposed by Lookman et al. (1995) fitted our P release kinetic data very well. The coefficient r2 was in all cases equal or higher than 0.99. The amount of P that is rapidly releasable (Q1) varied from 2.8 mg P kg−1 in E5 0P 3 cycle to 91.6 mg P kg−1 in C3 start (Table 3). The amount of P present in Q1 varied from 1.4% of Pox in E5 0P 3cycle to 6.8% of Pox in A1 start. The amount of P in Q1 was strongly linearly correlated to CP, E1min and E60min (r2 N 0.9, n = 26, graphs not shown) and less with E14days and Pox (r2 N 0.6, n = 26, graphs not shown). While k1 could not be explained by any of the variables

131

reported in this work, k2 was significantly positively linearly correlated to the logarithm of CP, E1min, E60min and E14days (r2 N 0.5, n = 26; graphs not shown). As for the cumulated amount of P released after 14 days relations were observed between the input–output P balances created in the pot experiments and the changes in Q1 (Fig. 4b). The results from the C3 soils deviated here again from the others (Fig. 4b). Except for the few soils showing a positive P balance and for a C3 soil, the absolute value of ΔQ1 was always much lower than the input–output balance showing that ryegrass had been able to take up P both from both Q1 and Q2 (Fig. 4b). Whereas the Pox, the DSP and the k1 values of our soils were similar to those reported by Lookman et al. (1995), our Q1 values were smaller for a given P ox . We report a Q 1 of 3 mmol P kg soil −1 for a Pox of 49 mmol kg soil−1 , while Lookman et al. (1995) reported Q1 values from 8.5 to 9.4 mmol kg soil−1 for Pox values of 45 to 51 mmol P kg soil−1 . Furthermore the k2 values we obtained were lower by one to two orders of magnitudes compared to Lookman et al. (1995). The lower Q1 and k2 values observed in our study might be due to the system we used, as we had no other sink for P than the P-free solution flowing through the reactor, or/and to the different soil properties as our soils were less acidic and had a lower sand content than those studied by Lookman et al. (1995). The low solution to soil ratio, the absence of stirring in our system and the low percolation rate also limited the transfer of P from soil aggregates to the solution as discussed in Koopmans et al. (2004b). Finally, a strong linear positive correlation was observed between P des14days and the organic P content measured in all soil samples (r2 = 0.77, n = 26, data and graph not shown). Whereas substantial mineralization rates of organic P can be observed in grassland soils (Bünemann et al., 2012; Simpson et al., 2012), we hypothesize that mineralization was not the main source of leached P in our experiment. Indeed, the introduction of Pox and soil organic P content in a multiple regression, showed a significant contribution of Pox (p = 0.001) in

Table 3 Cumulated amount of P released after 14 days (Pdes 14days) during the flow-through reactor experiments, and parameters of the 2 pools kinetic model (Pdest = Q1 ∗ (1 − e−k1t) + (Pox − Q1) ∗ (1 − e−k2t) in soils submitted to 0 to 3 growth cycles with Italian ryegrass and to different P fertilization regimes. Average and standard deviation are shown for Pdes 14days, while for Q1, k1, and k2 the estimated value is shown together with its lower level (LL 95%) and its higher level (HL 95%) for an interval of confidence of 95%.The standard error of the estimate is also shown for the model. The coefficient r2 was in all cases higher or equal to 0.99. Soil/treatment

Pdes 14days Average

Q1 SD

Estimate

LL95%

k1 HL95%

Estimate

LL95%

mg P kg soil−1 A1 Start A1 0P 1 cycle A1 0P 3 cycle A1 40P 1 cycle A1 40P 3 cycle C1 Start C1 0P 1 cycle C1 0P 3 cycle C1 40P 1 cycle C1 40P 3 cycle C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle TC5 Start C5 0P 1 cycle C5 0P 3 cycle C5 40P 1 cycle C5 40P 3 cycle E5 Start E5 0P 1 cycle E5 0P 3 cycle E5 40P 1 cycle E5 40P 2 cycle E5 40P 3 cycle

82.2 65.7 26.3 50.7 32.8 148 127 82.7 116 104 257 201 149 163 167 151 140 97.7 126 117 21.3 14.3 7.18 26.8 27.2 24.6

0.82 0.97 0.93 1.38 0.56 0.53 5.09 2.90 0.98 1.28 13.5 1.95 1.36 3.42 2.17 4.43 2.94 2.74 1.55 2.05 0.65 0.91 0.34 0.38 0.74 0.75

40.4 28.2 10.7 22.5 13.3 44.6 34.8 24.5 36.6 31.3 91.6 54.1 37.2 39.0 42.1 51.1 40.2 20.6 36.1 35.4 8.3 7.1 2.9 11.7 9.0 9.48

k2 HL95%

Estimate

SE Estimate

LL95%

HL95%

2.24E−04 1.92E−04 8.33E−05 1.53E−04 1.10E−04 4.03E−04 4.22E−04 2.92E−04 3.38E−04 3.40E−04 3.50E−04 3.44E−04 2.49E−04 2.71E−04 2.77E−04 3.63E−04 4.02E−04 3.40E−04 3.17E−04 2.99E−04 1.86E−04 8.86E−05 4.93E−05 1.84E−04 2.38E−04 1.98E−04

2.46E−04 2.18E−04 9.05E−05 1.71E−04 1.16E−04 4.35E−04 4.53E−04 3.14E−04 3.54E−04 3.64E−04 3.87E−04 3.66E−04 2.65E−04 2.87E−04 2.92E−04 3.98E−04 4.27E−04 3.49E−04 3.41E−04 3.22E−04 1.92E−04 1.11E−04 7.27E−05 1.96E−04 2.54E−04 2.09E−04

hour−1 38.9 26.4 10.2 21.4 12.9 41.4 32.3 21.8 35.2 29.4 84.9 50.5 34.2 36.3 39.5 47.3 37.7 19.9 33.5 32.9 8.2 6.5 2.3 11.4 8.5 9.14

41.9 29.9 11.1 23.7 13.7 47.8 37.3 25.2 37.9 33.3 98.3 57.8 40.2 41.8 44.7 54.8 42.6 21.3 37.7 37.8 8.5 7.7 3.5 12.1 9.5 9.83

3.43E−02 3.45E−02 2.84E−02 3.51E−02 2.33E−02 2.53E−02 3.43E−02 2.63E−02 3.75E−02 2.58E−02 2.58E−02 2.20E−02 1.94E−02 2.84E−02 2.30E−02 2.61E−02 2.89E−02 3.87E−02 2.74E−02 2.57E−02 3.23E−02 3.14E−02 2.70E−02 4.01E−02 1.91E−02 1.96E−02

3.10E−02 2.87E−02 2.54E−02 3.02E−02 2.19E−02 2.17E−02 2.79E−02 2.22E−02 3.37E−02 2.24E−02 2.19E−02 1.93E−02 1.68E−02 2.39E−02 2.03E−02 2.21E−02 2.50E−02 3.49E−02 2.31E−02 2.24E−02 3.08E−01 2.52E−02 1.43E−02 3.67E−02 1.73E−02 1.84E−02

3.75E−02 4.00E−02 3.13E−02 4.00E−02 2.47E−02 2.89E−02 4.07E−02 3.03E−02 4.14E−02 2.91E−02 2.97E−02 2.47E−02 2.20E−02 3.30E−02 2.57E−02 3.00E−02 3.29E−02 4.26E−02 3.16E−02 2.92E−02 3.38E−01 3.77E−02 3.97E−02 4.35E−02 2.09E−02 2.07E−02

2.35E−04 2.08E−04 8.69E−05 1.62E−04 1.13E−04 4.19E−04 4.37E−04 3.03E−04 3.46E−04 3.52E−04 3.67E−04 3.55E−04 2.57E−04 2.79E−04 2.85E−04 3.80E−04 4.15E−04 3.44E−04 3.29E−04 3.11E−04 1.89E−04 9.98E−05 6.10E−05 1.90E−04 2.46E−04 2.04E−04

0.578 0.69 0.166 0.471 0.108 0.895 0.956 0.513 0.565 0.573 1.95 0.872 0.621 0.906 0.669 1.094 0.797 0.305 0.814 0.708 0.059 0.213 0.195 0.15 0.104 0.072

132

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

20

Pdes14days (mg P kg-1)

0

A1, C1, C5, E5; y = 0.29 - 4.98; r2 = 0.84 C3 relationshipp not significant

-20 -40 -60 -80 -100 -120 -250

-200

-150

-100

-50

0

50

input-output P balance (mg P kg-1)

b)

10

Q1 (mg P kg-1)

0

A1, C1, C5, E5: y = 0.12 x - 3.30; r2 = 0.71 C3 relation not significant

-10 -20 -30

The amount of P in the soil column having at a given time the same isotopic composition as the leached P (the D value) is shown in Fig. 6 for the different treatments of the soil C3. Similar trends are observed in the other soils. At the beginning of the flow-through experiments, the D value decreased, reaching a minimum after about 3 days, and then increased till 14 days. The D values calculated for 3 days (D3days) and 14 days (D14days) are shown in the Table 4. The increase in D over these 11 days varied from 1.8 mg P kg soil−1 in A1 40P 3 cycle, to 89 mg P kg soil−1 in C3 40P 1 cycle. Actually in all soils and treatments, except the treatments A1 40P 3 cycle and E5 0P 3 cycle, the relation between D and time from 3 to 14 days was linear and statistically significant (relations not shown). The D value calculated for 14 days remained lower than the amount of oxalate extractable P. The D value measured at 14 days ranged from 17% of Pox in A1 0P 3 cycle to 35% of Pox in C1 0P 1 cycle. Finally, as for Q1 and the total amount of P desorbed after 14 days, changes in D3days and in D14days were related to the P balance observed in the pot experiments (graphs not shown). Highly significant correlations were observed between D3days and E1min, E60min, E14days and Pox (Fig. 7) and CP (graph not shown). The E14days values extrapolated from the short term isotopic exchange kinetics conducted in batch were about 3 times higher than D3days, whereas they should have been similar as D3days was measured after 17 days of isotopic exchange (14 days of preliminary incubation + 3 days of

-40 -50 -60 -250

a) -200

-150

-100

-50

0

50

input-output P balance (mg P kg-1)

33P

Fig. 4. Relations between the P input–output balances caused by 1 to 3 growth cycles of Italian ryegrass submitted to different P fertilization regimes on 5 soils and a) changes in the cumulated amount of P released after 14 days (ΔPdes14days = Pdes14days in a given treatment − Pdes14days of the start soil) and b) changes in the quantity of P present in the pool of rapidly releasable P (ΔQ1 = Q1 in a given treatment − Q1 of the start soil). The empty circles represent the C3 soils, the full circles the other soils. The equations are shown for all soils except the C3 soil where the regression is not significant.

20

released per sampling (% total introduced radioactivity, R)

a)

3.3. Release of radioactive P from soils and the D value The radioactivity released to the solution at each sampling time and the cumulated amounts of radioactivity released with time are presented for the different treatments of the C3 soil in Fig. 5. The amount of radioactivity released during the first 24 h was very high (10% to 15% of the introduced radioactivity). Afterwards the amount of radioactivity released decreased steadily reaching 1.2 to 1.5% of the added radioactivity at the last sampling points. As for the non-radioactive P, a desorption plateau was not reached within the 2 weeks of experiment. The other soils presented similar trends. The proportion of radioactivity recovered after 14 days from all soils is shown in the Table 4. It ranged between 12% for E5 0P 3 cycle and 51% for C5 40P 1cycle. It was higher than the proportion of P released from the column expressed in per cent of the oxalate extractable P. We observed a significant relationship with the form of a function Y = a + b / X between the DSP (X) and the amount of radioactive P recovered after 14 days (Y) (r2 = 0.83, n = 26, graph not shown) suggesting that at low DSP the added 33P remained more strongly sorbed than at higher DSP.

16 14 12 10 8 6 4 2 0 0

50

100

150

200

250

300

350

400

Time (hours)

b) 60

Cumulated amount of 33P released (% total introduced radioactivity, R)

explaining Pdes14days while the contribution of soil organic P was not significant (p = 0.55). Furthermore the flow-through reactor experiment operates under water saturated conditions, which are not optimal for organic matter mineralization. Yet, in order to fully exclude organic P mineralization, it would be necessary to repeat this experiment under sterile conditions e.g. by using γ irradiated soils.

C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle

18

50 40 30 20 C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle

10 0 0

50

100

150

200

250

300

350

400

Time (hours) Fig. 5. a) Recovery of 33P in the leachate at each sampling time (in % of he introduced radioactivity R) and b) cumulated recovery of 33P (in % of he introduced radioactivity R), from the soil C3 which had been submitted to 0 to 3 growth cycles with Italian ryegrass and different P fertilization regimes, measured from flow-through reactor experiments over a period of 14 days. The average values are shown surrounded by twice the standard deviation.

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

133

Table 4 Cumulated amounts of radioactive P (33P) released from the flow-through reactor experiments after 14 days of leaching, D values calculated after 3 days (D3days) and after 14 days of experiment (D14days) in soils submitted to 0 to 3 growth cycles with Italian ryegrass and to different P fertilization regimes. Average and standard deviation are shown for each result. Cumulated 33P release

Soil/treatment

Average % of added A1 Start A1 0P 1 cycle A1 0P 3 cycle A1 40P 1 cycle A1 40P 3cycle C1 Start C1 0P 1cycle C1 0P 3 cycle C1 40P 1 cycle C1 40P 3 cycle C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle C5 Start C5 0P 1 cycle C5 0P 3 cycle C5 40P 1 cycle C5 40P 3 cycle E5 Start E5 0P 1 cycle E5 0P 3 cycle E5 40P 1 cycle E5 40P 2 cycle E5 40P 3 cycle

D3days SD

Average

42.9 42.2 27.4 41.5 29.7 45.8 44.6 38.3 43.5 40.5 49.6 51.1 47.4 52.5 48.9 48.1 50.3 49.4 51.3 47.8 27.3 24.5 12.5 33.2 29.6 31.2

Average

SD

164 125 90.7 103 94.2 271 254 202 240 234 422 351 268 287 283 267 238 176 230 216 74.6 68.2 68.5 75.6 80.8 68.3

10.4 5.6 2.5 2.6 6.6 19.2 5.8 7.6 8.8 2.7 54.2 9.1 2.0 19.1 8.9 4.1 43.1 9.6 7.4 15.4 2.6 9.9 4.8 6.7 4.9 3.8

−1

33

P

mg P kg 0.98 1.62 0.52 0.73 1.18 1.41 1.89 1.30 1.19 0.31 4.54 0.80 0.77 1.18 0.28 0.95 4.69 2.85 0.98 0.30 1.71 1.77 0.44 1.08 0.65 1.05

146 119 81.1 91.0 92.4 243 206 166 195 192 369 269 225 198 238 229 191 137 161 175 67.5 50.1 51.4 64.1 76.4 62.8

flow-through experiment). This high E value could be due to a faster rate of exchange and diffusion in the stirred soil/solution suspension, but this is not the sole explanation as L values measured in soil/plant systems for the same exchange time than E values have been reported to be similar for plants using only isotopically exchangeable P (Morel and Plenchette, 1994). We suggest that this discrepancy between E14days and D3days is explained by the rapid and significant release of radioactivity from the flow-through reactor at the beginning of the experiment. As can be seen from Eq. (6) a strong release of radioactivity

500 450

D Value (mg P kg-1)

D14days SD

400 350 300

4.2 4.3 2.7 2.4 4.6 4.1 3.6 6.0 8.2 3.7 26.5 5.2 3.9 7.7 7.0 7.9 25.5 7.4 0.9 1.2 3.5 1.1 3.2 2.9 3.7 1.9

in the leachate will lead to a strong decrease in the numerator and to a low D value. Fardeau (1996) states that isotopically exchangeable P should be measured in systems at steady-state for P. This is often the case for plant experiments as plants remove only a small fraction of the added radioactivity with time, but this is not the case in the present flow-through reactor from which we removed too much radioactivity within a short time. This hinders a direct comparison between E and D values. The subsequent increase in D between 3 and 14 days might be due to the mineralization of soil organic P (Bünemann et al., 2007) and/or to isotopic exchange proceeding with time (Fardeau, 1996; Larsen, 1967). Bünemann et al. (2012) and Oehl et al. (2004) reported gross mineralization rates of 1 to 8 mg P kg soil−1 and day− 1. These increases are in the order of magnitude of the daily increase in D values observed in our soils between 3 and 14 days. However, as discussed above for Pdes14days we believe that organic P mineralization is not an important process and that the increase in D is largely due to simple continuing isotopic exchange with time. Furthermore we expect that D will increase towards Pox as time will increase. 4. Conclusions

250 200 150 0

50

100

150

200

250

300

350

400

Time (hours) C3 Start C3 0P 1 cycle C3 0P 3 cycle C3 40P 1 cycle C3 40P 3 cycle

Fig. 6. Changes in the D value in the soil C3 after 0 to 3 growth cycles with Italian ryegrass and submitted to different P fertilization regimes over a period of 14 days. The average values are shown surrounded by twice the standard deviation.

The flow-through reactor first designed by Freese et al. (1999) and further developed by Demaria (2004) provided relevant information on the amount of P that can be mobilized from a soil without having to disperse and stir it. The amounts of P released after 14 days were strongly correlated to the amounts of oxalate extractable P, isotopically exchangeable P (E value), water extractable P and to the P saturation degree. The changes in released P were related to the input–output P balance. The two-pools model proposed by Lookman et al. (1995) described very well the kinetics of P release from soils. Ryegrass was able to deplete P from both pools. Labeling the soil inserted in the reactor with 33P allowed quantifying a fraction of isotopically exchangeable P remaining in the reactor (the D value). The D value was not a constant but after an initial decrease, it

134

E. Frossard et al. / Geoderma 219–220 (2014) 125–135

a) 140

E1min and E60min (mg P kg-1)

E1min; y = 0.13 x - 5.87; r2 = 0.88 E60min; y = 0.37 x -12.9; r2 = 0.95

120 100 80 60 40 20 0 0

50

100

150

200

250

300

350

400

300

350

400

D3days (mg P kg-1)

b) 1800 Pox: y = 4.54 x + 27.4; r2 = 0.78

-1 E14days and Pox (mg P kg )

1600

E14days: y = 3.20 x - 42.4; r2 = 0.84

1400 1200 1000 800 600 400 200 0 0

50

100

150

200

250 -1

D3days (mg P kg ) Fig. 7. Relationships between the D value measured after 3 days (D3days) of flow-through experiment and a) the amount of P that is isotopically exchangeable within 1 and 60 min (E1min, E60min), and b) the amount of P extractable with oxalate (Pox) and isotopically exchangeable within 14 days (E14days) in soils that have been submitted to 0 to 3 growth cycles with Italian ryegrass and to different P fertilization regimes.

increased with time. Factors contributing to this increase were continuing isotopic exchange and possibly some organic P mineralization. As isotopic exchange provides information on the rate of release of P from the solid phase of the soil to the solution (Fardeau, 1996), and since D values were strongly correlated to oxalate extractable P, water soluble P and E value, we suggest that the D value gives relevant information on the potential amount of P that can still be leached or become available to plant. Acknowledgments We thank the Swiss Federal Office of the Environment (contract Nr. Wasser GS 06.0X, Rubrik-Nr. 810.3189.049) and the Swiss National Science Foundation (project no. 31-50528.97) for funding Paolo Demaria, Thomas Flura for help in the laboratory, and Dirk Freese for early discussions on the flow-through reactor. We also thank the two reviewers who helped to improve this paper. References Bünemann, E.K., Marschner, P., McNeill, A.M., McLaughlin, M.J., 2007. Measuring rates of gross and net mineralisation of organic phosphorus in soils. Soil Biol. Biochem. 39, 900–913.

Bünemann, E.K., Oberson, A., Liebisch, F., Keller, F., Annaheim, K.E., Huguenin-Elie, O., Frossard, E., 2012. Rapid microbial phosphorus immobilization dominates gross phosphorus fluxes in a grassland soil with low inorganic phosphorus availability. Soil Biol. Biochem. 51, 84–95. Chardon, W.J., Groenenberg, J.E., Temminghoff, E.J.M., Koopmans, G.F., 2012. Use of reactive materials to bind phosphorus. J. Environ. Qual. 41, 636–646. Demaria, P., 2004. Factors controlling phosphate exchangeability and release in agricultural soils. (Doctoral thesis) Diss ETH Nr. 15864.Swiss Federal Institute of Technology, Zurich (available at http://dx.doi.org/10.3929/ethz-a-005004872). Di, H.J., Condron, L.M., Frossard, E., 1997. Isotope techniques to study phosphorus cycling in agricultural and forest soils: a review. Biol. Fertil. Soils 27, 1–12. Dodd, R.J., McDowell, R.W., Condron, L.M., 2012. Predicting the changes in environmentally and agronomically significant phosphorus forms following the cessation of phosphorus fertilizer applications to grassland. Soil Use Manag. 28, 135–147. FAL, RAC, FAW, 1996. Méthodes de références des stations fédérales de recherches agronomiques. Analyse de terre et du substrat pour le conseil de fumure, vol. 1. Agroscope, Zurich-Rechenholz. Fardeau, J.C., 1996. Dynamics of phosphate in soils. An isotopic outlook. Fertil. Res. 45, 91–100. Freese, D., Lookman, R., Merckx, R., van Riemsdijk, W.H., 1995. New method for assessment of long term phosphate desorption from soils. Soil Sci. Soc. Am. J. 59, 1295–1300. Freese, D., Weidler, P.G., Grolimund, D., Sticher, H., 1999. A flow-through reactor with an infinite sink for monitoring desorption processes. J. Environ. Qual. 28, 537–543. Gallet, A., Flisch, R., Ryser, J.P., Nösberger, J., Frossard, E., Sinaj, S., 2003a. Uptake of residual phosphate and freshly applied diammonium phosphate by Lolium perenne and Trifolium repens. J. Plant Nutr. Soil Sci. 166, 556–567. Gallet, A., Flisch, R., Ryser, J.P., Frossard, E., Sinaj, S., 2003b. Effect of phosphate fertilization on crop yield and soil phosphorus status. J. Plant Nutr. Soil Sci. 166, 568–578. Groenenberg, J.E., Chardon, W.J., Koopmans, G.F., 2013. Reducing phosphorus loading of surface water using iron-coated sand. J. Environ. Qual. 42, 250–259. Haygarth, P.M., Condron, L.M., Heathwaite, A.L., Turner, B.L., Harris, G.P., 2005. The phosphorus transfer continuum: linking source to impact with an interdisciplinary and multi-scaled approach. Sci. Total Environ. 344, 5–14. Hosseinpur, A., Pashamokhtari, H., 2008. Impact of treated sewage sludge application on phosphorus release kinetics in some calcareous soils. Environ. Geol. 55, 1015–1021. Kanton Luzern, 2001. Baldeggersee. Auswertung der Zuflussuntersuchungen 1995 bis 1999. Amt für Umweltschutz, Luzern (Switzerland). Koopmans, G.F., van der Zeeuw, M.E., Chardon, W.J., Dolfing, J., 2001. Selective extraction of labile phosphorus using dialysis membrane tubes filled with hydrous iron hydroxide. Soil Sci. 166, 475–483. Koopmans, G.F., Chardon, W.J., Ehlert, P.A.I., Dolfing, J., Suurs, R.A.A., Oenema, O., van Riemsdijk, W.H., 2004a. Phosphorus availability for plant uptake in a phosphorusenriched noncalcareous sandy soil. J. Environ. Qual. 33, 965–975. Koopmans, G.F., Chardon, W.J., de Willigen, P., van Riemsdijk, W.H., 2004b. Phosphorus desorption dynamics in soil and the link to a dynamic concept of bioavailability. J. Environ. Qual. 33, 1393–1402. Larsen, E., 1967. Soil phosphorus. Adv. Agron. 19, 151–210. Liebisch, F., Bünemann, E.K., Huguenin-Elie, O., Jeangros, B., Frossard, E., Oberson, A., 2013. Plant phosphorus nutrition indicators evaluated in agricultural grasslands managed at different intensities. Eur. J. Agron. 44, 67–77. Lookman, R., Freese, D., Merckx, R., Vlassak, K., van Riemsdijk, W.H., 1995. Long term kinetics of phosphate release from soil. Environ. Sci. Technol. 29, 1569–1575. McDowell, R.W., Nash, D., 2012. A review of the cost-effectiveness and suitability of mitigation strategies to prevent phosphorus loss from dairy farms in New Zealand and Australia. J. Environ. Qual. 41, 680–693. McDowell, R.W., Sharpley, A.N., 2003. Phosphorus solubility and release kinetics as a function of soil test P concentration. Geoderma 112, 143–154. McDowell, R.W., Sinaj, S., Sharpley, A.N., Frossard, E., 2001. The use of isotopic exchange kinetics to determine P availability in overland flow and subsurface drainage waters. Soil Sci. 166, 365–373. McKeague, J., Day, J.H., 1966. Dithionite and oxalate extractable Fe and Al as aids in differentiating various classes of soils. Can. J. Soil Sci. 46, 13–22. Morel, C., Plenchette, C., 1994. Is the isotopically exchangeable phosphate of a loamy soil the plant available P? Plant Soil 158, 287–297. Morel, C., Plenchette, C., Fardeau, J.C., 1992. La fertilisation phosphatée raisonnée de la culture du blé. Agronomie 12, 565–579. Noij, I.G.A.M., Heinen, M., Heesmans, H.I.M., Thissen, J.T.N.M., Groenendijk, P., 2013. Effectiveness of buffer strips without added fertilizer to reduce phosphorus loads from flat fields to surface waters. Soil Use Manag. 29, 162–174. Oehl, F., Frossard, E., Fliessbach, A., Dubois, D., Oberson, A., 2004. Basal phosphorus mineralisation in soils under different farming systems. Soil Biol. Biochem. 36, 667–675. Ohno, T., Zibilske, L.M., 1991. Determination of low concentrations of phosphorus in soil extracts using malachite green. Soil Sci. Soc. Am. J. 55, 892–895. Olsen, S.R., Cole, C.V., Watanabe, F.S., Dean, L.A., 1954. Estimation of available phosphorus in soils by extraction with sodium bicarbonate. USDA Circ. 939. Randriamanantsoa, L., Morel, C., Rabeharisoa, L., Douzet, J.M., Jansa, J., Frossard, E., 2013. Can the isotopic exchange kinetic method be used in soils with a very low water extractable phosphate content and a high sorbing capacity for phosphate ions? Geoderma 200–201, 120–129. Saunders, W.M.H., Williams, E.G., 1955. Observations on the determination of total organic phosphorus in soils. J. Soil Sci. 6, 254–267. Schärer, M., Stamm, C., Vollmer, T., Frossard, E., Oberson, A., Flühler, H., Sinaj, S., 2007. Reducing phosphorus losses from over-fertilized grassland soils proves difficult in the short term. Soil Use Manag. 23, 154–164.

E. Frossard et al. / Geoderma 219–220 (2014) 125–135 Schärer, M., De Grave, E., Semalulu, O., Sinaj, S., Vandenberghe, R.E., Frossard, E., 2009. Effect of redox conditions on phosphate exchangeability and iron forms in a soil amended with ferrous iron. Eur. J. Soil Sci. 60, 386–397. Schoumans, O.F., Groenendijk, P., 2000. Modeling soil phosphorus levels and phosphorus leaching from agricultural land in the Netherlands. J. Environ. Qual. 29, 111–116. Simpson, M., McLenaghen, R.D., Chirino-Valle, I., Condron, L.M., 2012. Effects of long-term grassland management on the chemical nature and bioavailability of soil phosphorus. Biol. Fertil. Soils 48, 607–611. Sinaj, S., Frossard, E., Fardeau, J.C., 1997. Isotopically exchangeable phosphate in size fractionated and unfractionated soils. Soil Sci. Soc. Am. J. 61, 1413–1417. Sutton, M.A., Bleeker, A., Howard, C.M., Bekunda, M., Grizzetti, B., de Vries, W., van Grinsven, H.J.M., Abrol, Y.P., Adhya, T.K., Billen, G., Davidson, E.A., Datta, A., Diaz, R., Erisman, J.W., Liu, X.J., Oenema, O., Palm, C., Raghuram, N., Reis, S., Scholz, R.W.,

135

Sims, T., Westhoek, H., Zhang, F.S., 2013. Our Nutrient World: The Challenge to Produce More Food and Energy with Less Pollution. Global Overview of Nutrient Management.Centre for Ecology and Hydrology, Edinburgh on behalf of the Global Partnership on Nutrient Management and the International Nitrogen Initiative (www.unep.org www.gpa.unep.org/gpnm.html). van der Salm, C., Chardon, W.J., Koopmans, G.F., van Middelkoop, J.C., Ehlert, P.A.I., 2009. Phytoextraction of phosphorus-enriched grassland soils. J. Environ. Qual. 38, 751–761. van der Zee, S.E.A.T.M., van Riemsdijk, W.H., 1988. Model for long term phosphate reaction kinetics in soil. J. Environ. Qual. 17, 35–41. van Rotterdam, A.M.D., Temminghoff, E.J.M., Schenkeveld, W.D.L., Hiemstra, T., van Riemsdijk, W.H., 2009. Phosphorus removal from soil using Fe oxide-impregnated paper: processes and applications. Geoderma 151, 282–289.