Re-defining and quantifying inorganic phosphate pools in the soil and water assessment tool

The soil and water assessment tool (SWAT), a large-scale hydrologic model, is used to estimate phosphate (P) loading in streams and water bodies. The labile, active, and stable P pools are currently used to represent P cycling in SWAT; however, these pools are conceptual without any chemical basis. The current structure allows SWAT to reasonably represent P cycling; however, restructuring and incorporation of recent research results should produce more accurate P simulations. This paper presents a redefined SWAT structure using four inorganic soil P pools (water soluble, labile, active, and stable) combined with soil extraction methods to determine initial values for each pool. The redefined structure was selected by examining the chemical characteristics of laboratory tests relative to soil P pools and analyzing the relationships between test results. Water soluble P, labile P and active P are determined using water, H3A, Mehlich 3 extraction, respectively. Stable P is determined with an acid digestion. Redefining SWAT’s inorganic soil P pools based on soil and extractant chemistry will improve the defensibility, credibility, and the accuracy of SWAT P routines in water resource planning, management, and decision making.


INTRODUCTION
Phosphate (P) is an essential nutrient for plant growth and is applied to agricultural and urban lands in inorganic (fertilizer) and organic forms (manure).Excessive application of P to the landscape increases P losses to surface waters, which may contribute to eutrophication of fresh water systems and coastal estuaries (Sharpley et al., 1994).Intensive cropping systems can produce soils with elevated nutrient contents, and the associated tillage increases sediment losses from these enriched fields and streams banks to lakes and streams.
While the application of inorganic P is relatively uniform on a regional basis on urban and agricultural lands, manure produced by animal feeding operations tends to be a regional issue.Animal feeding operations tend to cluster around existing animal support and processing infrastructures.Intensive manure applications in these areas often exceed the capacity of the soils to retain nutrients.Repeated applications of excess manure can increase P losses in runoff (Sharpley et al., 2001;Pierson et al., 2001) to stream and river waters, with significant consequences for downstream water quality.
Phosphate concentration in surface water may be tied *Corresponding author.E-mail: Rick. .
Abbreviations: SWAT, Soil and Water Assessment Tool; STP, soil test phosphorus; PAI, phosphorus availability index; NAPT, North American proficiency testing.
(a) (b) Figure 1.Soil inorganic phosphorus pools and processes that move P in and out of pools: a) as currently defined in SWAT (Neitsch et al. 2005), and b) as redfined.
directly to water quality degradation.Total P concentrations of 50 μg L -1 are sufficient to sustain chlorophyll concentrations of 10 μg L -1 , the lower boundary for a trophic state in temperate streams (Dodds et al., 1998) and a eutrophic state in lakes (Smith, 2003).Using the Redfield (1958) ratio P concentrations exceeding 65 μg L -1 are sufficient to support algal growth in many water bodies when nitrate-N concentrations exceed 10 mg L -1 . It has been shown that these concentrations of total P can be from ground water in intensive agricultural regions (Burkart et al., 2004).Dissolved P concentrations from compiled peer-reviewed nutrient export studies on agricultural lands averaged 1.38 mg L -1 ; however, the median concentration was 0.40 mg L -1 since most concentrations were well below the average value (Harmel et al., 2008).
Tools to estimate P runoff for assessment of conservation effects or land management planning range from simple export coefficients (Reckhow et al., 1980) to process based models such as the soil water assessment tool (SWAT) (Arnold et al., 1998) and PPM Plus (White et al., 2010).SWAT is a river basin or watershed scale model developed to predict the impact of land management practices on water, sediment, and agricultural chemical yields over time (Neitsch et al., 2005).SWAT predicts the impact of topography, soils, land use, management, and weather on water, sediment, nutrient (N and P), and agricultural chemical yields from the landscape at a variety of spatial scales (Gassman et al., 2007).SWAT simulates the fate and transport of water, sediment, nutrients, and pesticides from upland areas by routing these constituents through a channel network where they are subjected to various in stream processes.
SWAT has been extensively used to study land management impacts on water quantity and quality.For example, SWAT has been used in to evaluate precision feed management programs for controlling P losses and soil-P build-up at field and watershed scales (Ghebremichael et al., 2008), to quantify P sources throughout the Rock River basin in Wisconsin and quantify impacts from the application of best managements practices (Kirsch et al., 2002), and to evaluate best management practices related to dairy manure management (Santhi et al., 2001).SWAT also has been integrated into USEPA's modeling framework, Better Assessment Science Integrating Point and Nonpoint Sources (BASINS).
The current version of SWAT (SWAT, 2009) contains three inorganic soil P pools (labile, active, and stable) as shown in Figure 1a.SWAT contains a full P balance; P may be added to the soil by fertilizer, manure, or crop residues, and lost from the soil through plant uptake and runoff (sediment bound and soluble P).In addition, organic P forms transform into inorganic P forms through mineralization.In SWAT and as in natural soils, most of the inorganic P resides in the stable pool.In SWAT, inorganic P in the labile pool is assumed to be in rapid equilibrium with the active pool, while the active pool is in slow equilibrium with the stable pool.Users may define the initial labile P value for various soils.If the user does not input initial labile P, labile P in all soil layers is initially set to 5 mg P kg -1 .Active and stable inorganic P pools are not input by the user but are initially set using the following equations: (1) Where psp is the phosphorus availability index (PAI), which can be user input (default value is 0.4).
A PAI of 0.4 indicates that 40% of P added to the soil remains as labile P and 60% becomes active P (Vadas and White, 2010).For user input, the PAI may be determined experimentally using a 6 month long incubation study (Vadas and White, 2010) or estimated from soil texture, pH, and base saturation (Sharpley et al., 1984(Sharpley et al., , 1989)).Total P is defined in SWAT as the sum of the labile, active, and stable inorganic P values plus the fresh and humic organic P pools.
SWAT's soil P routines were originally developed for the EPIC model (Jones et al., 1984), where labile P is synonymous with solution P and was quantified through extraction with anion exchange resin (Sharpley et al., 1984).Vadas and White (2010) indicate that any P extraction method that is comparable to the anion exchange resin can be used to quantify solution P and that P extraction results from routine soil tests, such as Mehlich 3, Bray 1, or Olsen, are sometimes well related to anion exchange resin results.Therefore, it is possible to use routine soil test P data to initialize solution P. Since the default values for the active, stable, and labile P pools are arbitrary and because guidance is not provided on methods for estimating reasonable initial input values (Vadas and White, 2010), the initial P pool values may not reflect actual soil conditions.It is more accurate for users to input the initial labile P concentration (Vadas and White, 2010), and the same should be true for the active and stable P pools as well.Therefore, our research objectives were to develop a new structure for inorganic soil P pools in SWAT and to identify readily available laboratory methods to determine initial values for each pool.This structure redefines inorganic soil P pools from conceptual to chemically based and allows site-specific initial value determination; both of which will improve the accuracy and defensibility of SWAT P simulation.

MATERIALS AND METHODS
The methods used in this study consist of two major steps.First, soil samples were analyzed with numerous common soil extractants to determine soil P concentrations.Then, results from the various extractants were all compared to each other to examine potential correlation, and the chemical "appropriateness" for each extractant in terms of representing soil P pools was examined.

Soil testing
Sixty-two soils of varying type from the North American Proficiency Testing (NAPT) program were used to evaluate various soil test methods for determining initial values for soil P pools.NAPT is a voluntary proficiency testing program of the Soil Science Society of America for laboratories throughout the country.Soil samples are sent to laboratories quarterly for testing, and laboratory results are statistically analyzed.Samples are also available for sale for use in research and as laboratory standards.Analytical data from each quarter are available at www.naptprogram.org.Seven of the 62 soils used in this study were from 2003 (laboratory identification numbers 103, 105, 112, 113, 114, 117, and 119).The remaining samples were from 2005 (laboratory identification numbers 101-104 and 106-120), 2006 (laboratory identification numbers 101-114 and 116-120), and 2007 (laboratory identification numbers 101, 103-104, 106-107, and 109-120).Each soil sample received from the NAPT program were previously dried and ground to 2 mm.NAPT reports the following extractable P results: water; Olsen et al. (1954); Bray (1:7); Bray (1:10), Bray and Kurtzb (1945); AB-DTPA, Modified Morgan, Modified Kewlona, Strong Bray, Mehlich 1, and Mehlich 3 (Mehlich, 1984).
In our laboratory, the samples were extracted with a 10:1 extractant to soil ratio with H3A (Haney et al., 2010), shaken for 10 min at 180 rev/min, centrifuged at 3500 rpm, and filtered through Whatman 2V filter paper.Filtered extracts were then analyzed for PO4-P using a Rapid Flow Analyzer (Flow Solution IV, OI Analytical; College Station, TX).For quality control, twenty soil samples were selected randomly and extracted using Mehlich 3, Olsen, Bray (1:7), and water for P determination.The results were compared with the NAPT results for the same samples, and no significant differences were detected (data not shown).

Selecting extractants to represent inorganic P pools
The results from the various soil extractants were examined in terms of their chemical "appropriateness."In other words, extractants were examined based on their design intent and more importantly their chemical composition to determine whether they could realistically represent the inorganic soil P pools.Since water is the universal extractant and the key to natural rainfall, runoff, and leaching processes, the examination started with water.All possible combinations of soil P data from the ten extractants used in NAPT were compared to determine significant correlations (α = 0.05).Specifically, correlation between paired extractant results indicated that both extractants measured some portion of the same soil P pool or at least that a consistent relationship existed between the two.

Soil testing
Results of soil tests conducted by NAPT and by our laboratory are shown in Table 1.As shown, soils varied widely in terms of relevant physical and chemical properties, particularly pH, organic C, % clay, and extractable P; therefore, the redefined pool structure should be applicable for a majority of soil types.

Selecting extractants to represent inorganic P pools
When the correlation between soil P levels for the 10 extractants was examined, it was interesting that the correlations were much better for the commonly applied  and H3A).In fact, the correlation between each of these extractants was significant at α = 0.05 (Table 2).These results indirectly support the appropriateness and the common use of Mehlich 3, Bray 1, Olsen, and H3A extractants.

Water soluble pool
In redefining the inorganic soil P pools, it was important to start with water extraction (Figure 2) because dissolution and transport of soil P by rainfall, leaching, and runoff is a fundamental natural process and a fundamental modeling need in SWAT.As such, it is more appropriate to define water soluble P as its own pool, which is fed by the labile pool and diminished by runoff and leaching (Figure 1b), instead of simply including it as a component of labile P (Figure 1a).In addition, anion exchange can overestimate this value relative to water since loosely bound water extractable P is generally a very small fraction of the total P in soils or sediments (Kuo, 1996).

Labile pool
The labile pool, which includes the water soluble pool, is fed by fertilizer addition and mineralization of organic P and diminished by plant uptake, represents the readily plant available pool in SWAT (Figure 1b).The H3A extractant (Haney et al., 2010) was chosen to quantify the labile pool (Figure 2) because H3A extractable P was more strongly correlated with water extractable P than any of the other extractants (Figure 3, Table 2) and because of the chemical nature and design of H3A.For soil tests to accurately represent P availability, they should respond to soil characteristics in manner similar to that of plants (Kuo, 1996); therefore, H3A utilizes organic acids that mimic plant root exudates in the extraction of soil nutrients (Haney et al., 2010).H3A extracts P at or near soil pH which is important in simulating natural soil conditions since pH and P solubility are highly correlated (Golterman, 1998;Sharpley, 1993).H3A extractable P is comparable to resin exchangeable P upon which the original EPIC labile pool was based, yet is a much simpler laboratory analysis (unpublished results).In addition, H3A extractable P was strongly correlated with Olsen, Bray, and Mehlich 3 extractable P (Table 2), which are known to be highly correlated with resin-extractable P.

Active pool
Whereas the water soluble P pool (runoff, leaching) and labile P pool (plant available) are easily visualized, the active pool is a bit more nebulous (Figure 1b).The active pool is in essence a reserve P supply pool that is relatively stable and composed of slowly desorbable P (Vadas et al., 2006).Mehlich 3 was chosen to quantify the active pool (Figure 2) because H3A and Mehlich 3 extractable P are strongly correlated (r = 0.88, Table 2) and Mehlich 3 extracted more P than H3A (Figure 4).Mehlich 3 extracted more P than the other extractants studied (Table 1) because it dissolves iron (Fe-), aluminum (Al-) and calcium (Ca-) phosphates (Nelson et al., 1953) with its strongly acidic solution (buffered to pH 2.6).The Fe-, Al-, and Ca-phosphates dissolved by Mehlich 3 extraction are relatively insoluble P forms in the soil ( Ketterings et al., 2002), which are more representative of relatively slow reacting pools than labile or water soluble P pools.Mehlich 3 is poorly correlated with water extractable P; however, the relationship is significantly improved when water and H3A extractable P are summed (Figure 5), which further supports the redefined pool structure.

Stable pool
Similar to active P, stable P is more difficult to visualize than water soluble and labile P (Figure 1b).P in the stable pool is strongly adhered to mineral particles or occurs in low solubility particles; therefore, a small portion of P in the stable pool becomes available at an extremely slow rate.Quantifying this soil P pool is best accomplished by determining the total soil P through acid digestion (Figure 2).

Redefined inorganic soil p pool structure
The correlation between and increasing magnitude of water, H3A, and Mehlich 3 extractable P (Tables 1 and  2), along with the design intent and chemical basis of these extractants, provided support for a redefined inorganic soil P structure in SWAT.The correlation between water and H3A extractable P (r=0.84) and between H3A and Mehlich 3 extractable P (r=0.88)indicates that the extractants measured some portion of the same soil P pool or at least that consistent relationships exist between the pools.Although many fractionation processes exist for determining the inorganic P status and availability in soil, the alternatives are time consuming and not practical for most SWAT users; therefore, the soil test methods listed in Table 3 corresponding to each pool are recommended.
The soil P pools as shown in Figures 1b and 3 are subset of the next larger set (Table 3).For the soils analyzed in this dataset, the water soluble pool comprised 14% of the labile pool (H3A extractable) which comprised 47% of the active pool (Mehlich 3 extractable).Similarly, the active pool comprised 14% of the total P pool.These defined pools and relative sizes provide an initial basis for estimating the relative speed of transfer between the soil P pools.Qualitatively, when water soluble pool is diminished by runoff or leaching, it is fed by the labile P pool.Similarly, when the labile pool is diminished by plant uptake, it is rapidly fed by fertilizer addition and more slowly fed by mineralization and the active pool, which in turn is even more slowly replenished by the stable pool.
Incorporating this redefined inorganic soil P structure into SWAT will require a relatively simple modification of the code.The model is not dissimilar to the current approach.To fully realize an improved mineral soil P model, additional research will be needed to further refine transformation rates between the proposed pools.The transformation rates between these pools dictate how long it takes applied P to move from highly available pools (water, labile) to less available forms (active, stable).The definition of these pools by soil extractant allows these rate coefficients to be estimated from laboratory data.A soil sample can be spiked with mineral P and tested over a period of time to quantify P movement from pool to pool.With a sufficient sample population, pool to pool rate transformation coefficients can be estimated.Both the relative pool sizes derived from this research and transformation coefficients are needed to accurately simulate the fate of P fertilizer applied to soils.

Conclusions
Redefining the inorganic soil P pool structure in SWAT will allow users to rapidly and accurately estimate initial values for each pool with commonly applied soil test methods.The new soil P pool structure defines water soluble P as water extractable P, labile P (plant available) as H3A extractable, and active P (reserve supply pool) as Mehlich 3 extractable.The stable P pool, a majority of which is only very slowly transformed to active P, is determined by acid digestion.This redefined soil P pool structure based on commonly used soil test methods has several advantages.SWAT users will be able to rapidly and inexpensively determine initial values and relative magnitude of soil P pools from soil samples collected from watersheds of interest instead of using SWAT default ratios.This will produce more accurate soil P values, which will directly improve simulation of soluble P in runoff and other P processes, since calculations are directly dependent on soil P pool values.Further improvement in P simulation will also result from more accurate estimation of parameters (e.g.phosphate soil partitioning coefficient (Kd)) that are affected by soil P values.The improved simulations will be especially valuable in P threatened or impaired watersheds.Another benefit is the ability to better characterize P transformation rates between soil pools in quasiequilibrium, which is a critical aspect of the model (as important as the relative magnitudes of the pools themselves).With soil pools defined with available laboratory tests, the transformation rates can be evaluated under differing environmental conditions and pool imbalances due to fertilization and crop uptake.By testing soils over time, the dynamics between the pools can more readily be identified.Finally, the redefined inorganic soil P pool structure will improve the defensibility of model predictions, since soil P pools will be definable and reproducible and not simply conceptual.

Figure 2 .Figure 3 .
Figure 2. Redefined SWAT inorganic P pools presented with the soil test method used to quantify initial values for each pool.

Table 1 .
Descriptive statistics of relevant soil properties, particularly extractable P, for the 62 samples used in this study.

Table 3 .
Mean and standard deviation of soil P levels averaged across all soils.