Vadas & Sims: Predicting Phosphorus Desorption from Coastal Plain Soils

of agricultural P to water quality. In Florida, FHANTM 2.0 (Field Hydrologic and Nutrient Transport Model, Pollution of surface waters by P from agricultural areas is a water version 2.0; Fraisse and Campbell, 1997) was developed quality issue in Delaware. The FHANTM 2.0 computer model can help identify areas with a high potential for P loss, but the model’s to simulate water and P movement from individual fields representation of P desorption from soils to runoff waters needs reas part of an effort to reduce P loads to Lake Okeeevaluation. The equation, Pd K Po t W , has been proposed to chobee. The hydrology of FHANTM 2.0 is based on predict such P desorption, but equations originally proposed to predict DRAINMOD (Skaggs, 1980), and the nutrient compovalues for the constants K, , and from the ratio of soil clay content/ nents are based on GLEAMS (Leonard, 1987). Because soil organic C content may not be accurate for Delaware soils. ThereFlorida’s physical and hydrologic conditions of flat fore, we measured P desorption for 23 sandy Delaware soils for times of 5 to 180 min, water/soil ratios of 10 to 1000 L kg 1, and three initial fields, high water tables, and high P sandy soils are levels of soil desorbable P. Values for the constants K, , and were similar to those in Delaware, FHANTM 2.0 could potencalculated and related to soil properties. We found that K, , and tially be used in Delaware to simulate field-scale P exvalues were not well related to clay/OC, but were better related to port. However, several of FHANTM 2.0’s mathematical the ratio of oxalate-extractable Fe/OC content ( ) or the sum of representations of soil P processes were designed either oxalate extractable Fe and Al ( and K ). These results can be used for pesticide transformations in soils for GLEAMS or to help refine the FHANTM 2.0 model in predicting P loss from agricultural areas in Delaware and similar landscapes in the Midfor specific chemical and physical properties of Florida Atlantic Coastal Plain. soils. Therefore, to use FHANTM 2.0 in Delaware, its P components must be modified to more accurately represent the chemical and physical processes in DelaT nonpoint source pollution of surface waters ware soils. One such modification is the representation by P is an international environmental quality issue. of P desorption to runoff waters. Currently in FHANTM In Delaware, water quality in the Inland Bays national 2.0, the quantity of P in the topsoil available for runoff, estuary, which consists of the Rehoboth, Indian River, (Cav )p (mg kg 1 ), is calculated with the equation and Little Assawoman Bays, has been impaired by P to such an extent that the state must now comply with (Cav)p (CPLAB) exp [{ (Pr Q regulations of the Clean Water Act (United States, AWS)}/{(SSG) Kd (1 POR) 1967). As part of this compliance, total maximum daily loads (TMDL) have been established as the level of POR}], [1] pollution below which the Inland Bays will meet water quality standards. One goal of the TMDLs is a 70% Abbreviations: Alox, acid ammonium oxalate-extractable Al; AWS, amount of rainfall needed to saturate the topsoil layer in the reduction of nonpoint source P loads to the Inland Bays FHANTM 2.0 model; B, extraction coefficient used in the FHANTM (State of Delaware, 1995). Because agriculture has been 2.0 model; (Cav )p, quantity of P in the topsoil available for runoff used identified as a significant nonpoint source of P in the in the FHANTM 2.0 model; CPLAB, quantity of desorbable P in the Inland Bays watershed (Ritter, 1992), it is necessary to topsoil used in the FHANTM 2.0 model; (Cw )p, concentration of P in runoff used in the FHANTM 2.0 model; Feox, acid ammonium identify agricultural areas that have a high potential for oxalate-extractable Fe; FHANTM, Field Hydrologic and Nutrient P export. Transport Model; OC, organic C; OM, organic matter; UDSTP, UniField-scale nutrient transport models have been proversity of Delaware Soil Testing Program; K, empirical constant in posed as a means to characterize the environmental risk soil P desorption equation; Kd, partitioning coefficient used in the FHANTM 2.0 model; Pd, amount of P desorbed from the soil; Po, initial concentration of desorbable P in soil; POR, porosity used in P.A. Vadas, USDA, ARS, ANRI, AMBL, B-163F Rm. 5, BARCthe FHANTM 2.0 model; Pr, value for precipitation used in the East, 10300 Baltimore Ave., Beltsville, MD 20705; J.T. Sims. Dep. of FHANTM 2.0 model; Q, value for runoff used in the FHANTM 2.0 Plant and Soil Sciences, 152 Townsend Hall, University of Delaware, model; SSG, soil specific gravity used in the FHANTM 2.0 model; t, Newark, DE 19717. Received 11 Oct. 2000. *Corresponding author time of P desorption; TMDL, total maximum daily load; W, water/ ([email protected]). soil ratio during P desorption; , empirical constant in soil P desorption equation; , empirical constant in soil P desorption equation. Published in Soil Sci. Soc. Am. J. 66:623–631 (2002). 624 SOIL SCI. SOC. AM. J., VOL. 66, MARCH–APRIL 2002 where CPLAB (mg kg 1 ) is the quantity of labile P in constants can be predicted from known or easily estimated soil physical and chemical properties. Therefore, the topsoil as defined for the FHANTM 2.0 model, Pr (cm) is precipitation, Q (cm) is runoff, AWS (cm) is the Sharpley (1983) related K, , and to soil properties and found statistically significant relationships with the amount of rainfall necessary to fill the topsoil layer to saturation, SSG (g cm 3 ) is soil specific gravity, Kd is a ratio of soil Fe/OC and clay to OC for 43 soils collected from throughout the USA. However, the clay and Fe partitioning coefficient, and POR is soil porosity (cm cm 1 ). Given that the values of SSG, Kd, and POR are contents of these soils were much greater than those typically found in the sandy soils of Delaware’s Inland constant for a given soil and that the value of AWS is much less than Pr and Q, Eq. [1] mathematically Bays watershed and of the Mid-Atlantic Coastal Plain in general. Because the clay and Fe contents of soils demonstrates that (Cav )p is curvilinearly related to CPLAB as a function of the absolute value of the differhave a strong influence on P desorption phenomena, the relationships provided by Sharpley to predict K, , ence between Pr and Q. Essentially, as the difference between Pr and Q decreases, the amount of P available and and to subsequently predict P desorption may not accurately represent Delaware soils. Given these for runoff increases. The actual concentration of P in runoff, (Cw )p (mg L 1 ), is then calculated in FHANTM considerations, the objectives of this research were to determine if Sharpley’s relationships for predicting K, 2.0 based on partitioning (Kd ) and extraction (B) coefficients with the Eq. [2]. , and are applicable to Delaware soils; and, if not, to develop relationships for predicting K, , and from (Cw)p [(Cav)p B]/[1 (Kd) B] [2] known or easily measured properties of Delaware soils. Although the research presented here was conducted The partitioning coefficient, Kd, assumes that the equifor only Delaware soils, it should be applicable to Midlibrium relationship between (Cav )p and (Cw )p is linear Atlantic Coastal Plain soils with similar physical and and is a function of the Mg and OC content of the chemical characteristics. topsoil. The extraction coefficient, B, accounts for the fact that the concentration of P in runoff is typically less than the P concentration in the soil solution and is MATERIALS AND METHODS calculated based on the value of the soil’s Kd value. Soil Selection and Characterization Because Eq. [1] and [2] were originally designed to represent pesticide transformations in soils (Leonard et Twenty soil samples were obtained from the archives of al., 1987), and were calibrated in FHANTM 2.0 for the University of Delaware Soil Testing Program (UDSTP). Florida soils, they may not accurately represent desorpThese soil samples had been used in a previous study to assess tion of P from soils to runoff for conditions in Delaware. relationships between soil test P, soluble P, and P saturation (Pautler and Sims, 2000). Of Delaware’s three counties, two In many studies of P desorption, it is commonly obsoil samples were taken from the northernmost New Castle served that desorption reactions occur rapidly at first County, 12 were taken from the southernmost Sussex County, and then decrease as equilibrium is approached. It is and six were taken from Kent County. An additional three also observed that the quantity of P desorbed is largely soil samples were taken from samples collected from a Sussex a function of the time allowed for desorption and the County field site that was used in a previous study to assess water/soil ratio during desorption (Barrow, 1979). Such the impact of agricultural drainage on water quality (Sims et desorption data are typically best described by exponenal., 1998). These three soil samples were chosen because they tial or logarithmic equations, such as Elovich or Freundrepresent the artificially drained, high organic matter (OM) lich equations (Chien and Clayton, 1980; Kuo and Lotse, soils that will likely be subject to future simulations of P export 1974). Sharpley et al. (1981) proposed a simplified P using the modified FHANTM 2.0 model. The number of soil samples from each county was chosen to reflect the percentage desorption equation: of samples submitted to the UDSTP each year. Overall, the Pd KPot W [3] 23 soil samples were chosen to represent a range in clay, Fe and Al oxide, and OM content, all of which have a strong where Pd is the amount of P desorbed (mg kg 1 ) in influence on P desorption. The quantity of desorbable P initime, t (min), at a water/soil ratio W (mL g 1 ), Po is the tially present in the soil samples was a secondary considinitial amount of desorbable P present in the soil (mg eration. kg 1 ), and K, , and are empirical constants for a given All samples had been previously air-dried and ground to soil. Equation [3] represents a significant improvement pass a 2.0-mm sieve. We characterized the soils for pH (1:1 from the current P desorption equations in FHANTM soil/water ratio), soil test P (STP; Mehlich-1 extraction; 1:4 ratio of soil/0.05 M HCl 0.0125 M H2SO4; 5-min reaction time 2.0 (Eq. [1] and [2]) because it accounts for the nonlinear [Sims and Heckendorn, 1991]), particle size by the hydrometer characteristics of P desorption and the strong influence method (Bouyoucos, 1962), and acid ammonium oxalateof time and water/soil ratio. For any given runoff event, extractable Al and Fe (Alox and Feox; 1:40 ratio of soil/0.2 M either simulated or observed, the time, water/soil ratio, [NH4]2C2O4, 2-h reaction time in darkness [McKeague and and desorbable P parameters are either known, as is Day, 1996]). Soil OC was measured by the Walkley-Black wet the case with FHANTM 2.0, or can be easily measured oxidation procedure (Nelson and Sommers, 1982). The initial or calculated. Sharpley (1983) pointed out that the appliamount of desorbable P (P0, Eq. [3]) was measured with Fecation of Eq. [3] is limited if the values of the constants oxide impregnated filter strips (1:40 ratio of soil/0.01 M CaCl2 K, , and must be experimentally determined for a Fe-oxide coated filter paper strip; 16-h reaction time, followed given soil before P desorption can be predicted. Appliby desorption of P with 1 M H2SO4 [Chardon et al., 1996]). The P in the Mehlich-1 extraction and the Fe and Al in the cation of Eq. [3] is much broader if the values for the VADAS & SIMS: PREDICTING PHOSPHORUS DESORPTION FROM COASTAL PLAIN SOILS 625 oxalate extraction were measured by inductively coupled (1971). To do this, r values were converted to z values using statistical tables. The differences between z values were then plasma atomic emission spectroscopy (ICP-AES). The P in the filter strip procedure was measured by the molybdate tested for significance using a two-tailed t test. These comparisons were made only among equations for each constant , blue method of Murphy and Riley (1962) with absorbance measured at 882 nm. , and K, and not between constants. Phosphorus Desorption Experiments RESULTS AND DISCUSSION The desorbable P status of each soil was varied by adding Soil Characteristics 0, 95, and 190 mg P kg 1 soil as a solution of KH2PO4 (equivaThe soils used in this study were representative of lent to fertilizer application rates of 0, 50, and 100 kg P ha 1 to a 4-cm soil depth). Both the P-amended and unamended many soils in the Coastal Plain and Piedmont regions soils were incubated at 25 C for 3 d prior to the P desorption of Delaware and other Mid-Atlantic states. In general, study. This incubation time was chosen to duplicate the experisoils were moderately acidic and low in OC, although ments of Sharpley et al. (1981), and because Sharpley and this varied slightly among counties (Table 1). Soils from Ahuja (1982) showed that longer incubation times did not New Castle County, which is in the Piedmont plateau, have a significant effect on the values of K, , and . The are generally finer-textured (silt loams) than the Coastal desorption of P from soils was investigated by shaking dupliPlain soils from Kent and Sussex Counties (sandy loams cate samples with distilled water at water/soil ratios of 10:1, and loamy sands). The dominant soil orders in New 100:1, and 1000:1 (mL g 1 ) on an end-to-end shaker at 25 C Castle and Kent Counties are Ultisols (80–90%) and for times of 5, 30, 60, and 180 min. After shaking, each solution Alfisols (9%). The soils of Sussex County are typically was filtered through Gelman 0.45m millipore filters(Gelman Sciences, Ann Arbor, MI). The P in the filtered solutions was very sandy with low OC, although some may have high measured colorimetrically on a Sequoia-Turner model 340 OC because of poor drainage. Dominant soil orders in spectrophotometer (Sequoia Turner Corp., Mountain View, Sussex County are Ultisols (57%) and Entisols (33%). CA) by the molybdate blue method of Murphy and Riley The average OC content of the 23 Delaware soils (1962) with absorbance measured at 882 nm. (1277 mmol kg 1 ) was comparable with the average OC content of Sharpley’s (1983) 43 soils (1083 mmol kg 1 ), Calculation of K, , and but the Delaware soils on average had less clay (86 g kg 1 ) than Sharpley’s soils (220 g kg 1 ). The average Given that Eq. [3] is valid and is a power equation, at any given combination of W and Po, the logarithm of Pd should Feox and Alox contents of the Delaware soils (Feox 8 be linearly related to the logarithm of t for each soil. The slope mmol kg 1 and Alox 32 mmol kg 1 ) were also much of this line is the value of for that soil at that combination of less than the extractable Fe and Al contents of W and Po. Similarly, at any given combination of t and Po, the Sharpley’s soils (Fe 98 mmol kg 1 and Al 208 mmol logarithm of Pd should be linearly related to the logarithm of kg 1 ). Sharpley used 1 M NH4Oac, pH 4.8, to extract Fe W for each soil. The slope of this line is the value of for and Al, whereas we used 0.2 M (NH4)2C2O4. Therefore, that soil at that combination of t and Po. In the final case, at some of the difference in the amount of Fe and Al any given combination of t and W, Pd should be linearly related content extracted from Sharpley’s and our Delaware to Po. The slope of this line is the value of K for that soil soils is likely because of the difference in the inherent at that combination of t and W. Throughout the desorption ability of acetate and oxalate to extract Fe and Al from experiments for each soil, nine values of from all combinations of W and Po, 12 values of from all combinations of t soil. However, the literature is sparse with research diand Po, and 12 values of K from all combinations of t and W rectly comparing the ability of acetate and oxalate to were calculated. Average values of , , and K were then extract soil Fe and Al. Myers et al. (1988) extracted Al calculated from these nine or 12 values. These average values from seven acid Ohio soils with both 1 M NH4Oac, pH of , , and K were then related to soil properties using a least4.8, and 0.2 M (NH4 )2C2O4 and found that the oxalate squares regession. The parameters of the resulting regression always extracted more Al than the acetate, by an averequations and their correlation coefficients were analyzed for age greater amount of nearly nine times. The same is significance by an ANOVA procedure. These statistical analylikely true for soil Fe, but we could not find any sources ses were performed within Microsoft EXCEL spreadsheets in the literature that directly compare the ability of (Microsoft, Inc., Redmond, WA). Correlation coefficients of oxalate and acetate to extract Fe from soil. Indirectly, the regression equations were compared for statistically significant differences as described by Snedecor and Cochran Kraske et al. (1989) extracted Fe from six New England Table 1. Selected soil physical and chemical properties of 23 Delaware soils used for P desorption studies. Property All soils New Castle soils Kent soils Sussex soils