Compatibility of the ammonium lactate (AL) and sodium bicarbonate (Olsen) methods for determining available phosphorus in Swedish soils

Abstract

Plant-available P was first extracted in soils 114 years ago and a number of different analytical methods have since been developed, but for good reasons none of these methods has been adopted as a standard for all soils. With increasing cooperation within research, there is a need to harmonise the interpretation of analytical data for fertiliser recommendations, research, and environmental control. This paper evaluates the compatibility of the Swedish standard ammonium lactate (P_AL) method and the widely used Olson's sodium bicarbonate (P_Ols) method in 82 topsoil samples taken from Swedish long-term soil fertility field trials. The P_AL-values were usually larger than P_Ols, with a mean P_AL/P_Ols quotient of 2.30±1.04 (0.67–6.58). The P_AL and P_Ols means and ranges were 126±89 (5–360) and 55.1±33 (3.1–122.8) mg P kg⁻¹ dry soil, respectively. Stepwise multiple linear regression analysis was used to evaluate the relationship between P_AL and P_Ols and how this relationship was affected by clay content, pH, and soil organic carbon content (SOC). After statistical transformation, it emerged that the square root of clay content (17.4%±13.82, range 1–54.4) and pH (6.45±0.54, range 5.5–7.7) significantly affected the relationship at partial R ²-values of 2 and 12%, respectively, while ln(SOC) (2.54%±1.21, range 1–6.03) did not, apparently due the narrow range. The regressions of predicted vs. measured values explained 95 and 94% of the variation in P_AL and P_Ols, respectively. The mean deviation of predicted compared with measured values was 21.3 and 8.3 mg P kg⁻¹ dry soil for P_AL and P_Ols, respectively, corresponding to 20 or 19% of the measured values. We conclude that a data set consisting of P_AL-values can be converted into P_Ols-values and vice versa with reasonably high accuracy when accounting for clay content and pH.

Keywords:

• Clay
• organic
• carbon
• pH
• prediction regressions

Introduction

There is a general consensus that the complexity of soil phosphorus (P) forms of different solubility coupled with diverse soil mineralogy and properties render this element one of the most difficult nutrients to extract accurately in forms available to plants. The diversity of soil P disposition explains the different approaches for its extraction (e.g., Williams et al., 1967; Olsen & Sommer, 1982; Turner et al., 2005). The first attempt to extract plant-available P from soils was probably made by Dyer (1894) using a 2% CH₃COOH solution to obtain what he termed ‘plant mineral food’. Since then, a number of different methods have been developed and the search for new methods or modifications of existing method continues. However, none of these methods has been adopted as a standard for all soils.

As research cooperation is widening, there is also need to harmonise the interpretation of different soil P tests for fertiliser recommendations, research, and environmental control. For example, this need was highlighted in the EU COST Action 832 (EU COST832, 2000), which noted the need for a common action plan based on P test data for management of agricultural P to minimise eutrophication in EU member states. The questions arising here are whether soil P tests are compatible and whether analytical data obtained using one method can be translated accurately into those of another method. Theoretically, it is not possible to establish an exact relationship between analytical datasets obtained using different methods. However, the general relationship between the methods can be determined and can enable comparison and interpretation of results collected in different countries.

The ammonium lactate (AL) solution (0.1 M NH₄-lactate+0.4 M HOAc, adjusted to pH 3.75) was developed by Egnér et al. (1960) to extract plant-available P in soils. The method is currently standard in some European countries including Sweden, Norway, Belgium, Portugal, the former USSR Baltic states, and some Balkan states, though shaking time defers. In some countries soil samples are shaken for two hours, while in others the shaking time is 90 minutes. The P in AL extracts are hereafter designated P_AL. The AL solution also provides an estimate of plant-available K and Mg content in the soil. However, in addition to accessing available P (Karlsson & Jansson, 1959; Semb, 1984), AL also dissolves insoluble Ca-P compounds, such as apatite (Semb, 1984) and Ca₃(PO₄)₂ in limed and calcareous soils (Hahlin & Eriksson, 1981; Fernandes et al., 1999; Ivarsson & Gustafsson, 2001) and soils fertilised with phosphate rock (da Silva & Van Raij, 1999). Evidently, P_AL is an overestimation of available P in such soils.

The reason why the P_AL method remains the official method in Sweden is that cultivated soils have been regularly analysed using this method and there is wealth of data on P_AL, pH, clay, and soil organic carbon (SOC) contents from the mid-1950s until now. Moreover, Swedish advisory authorities and farmers are conversant with P_AL and can interpret it for soil fertility assessment and management. In addition, P_AL is a valuable tool in research, teaching, and soil classification (Kirchmann et al., 1999).

The weakly alkaline sodium bicarbonate (SB) solution (0.5 M NaHCO₃, pH 8.5) proposed by Olsen et al. (1954) is widely used to complement established national standard methods in the USA and several European countries. The P extracted is hereafter termed P_Ols. This method was originally designed to measure plant-available P content in alkaline soils, but has also predicted the P requirement for rice grown in submerged soils (Dobermann et al., 1996) as well as some acid soils (Bationo et al., 1991). In Denmark P_AL is still included in routine analysis, but about two decades ago P_Ols became the standard method. Similarly, in Norway, where P_AL is still the official method, analysis of P_Ols is recommended for sedimentary clay soils with a high pH (Riley & Steenberg, 1985). In Croatia, where P_AL is widely used in soil fertility assessment, P_Ols is also attracting interest (Lončarič et al., 2006). Both P_AL and P_Ols have been used as tools for risk assessment of P-leaching losses (Sharpley & Rekolainen, 1997; Hesketh & Brookes, 2000; Bechmann et al., 2003; Börling, 2003; Djodjic et al., 2004).

This study was therefore undertaken to determine and evaluate the compatibility of P_AL and P_Ols and to develop a model that can used to convert their data sets into each other. Because of high analytical costs, only those two methods were compared.

Materials and methods

Sites and soil analysis

Eighty-two topsoil (0–20 cm) samples were taken from Swedish long-term fertility trials located in 14 provinces (Table I). The samples were air-dried, crushed and passed through a sieve (≤2 mm). The pH of duplicate samples was measured in a 1:5 suspension (1 g dry soil suspended in 5 mL deionised water) on a pH meter. Two replicates were analysed for P_AL (1 g shaken with 20 mL AL solution for 90 min) and another two replicates were analysed for P_Ols (1 g shaken with 20 mL SB solution for 30 min). Both AL- and SB-extracts were filtered through Whatman no. 42 paper. Phosphorus concentrations in the extracts were measured calorimetrically on a spectrophotometer. For P_AL, the stannous chloride–molybdate procedure (Egnér et al., 1960) was used and absorbance was measured at 882 nm. For P_Ols, the ascorbic acid–molybdate procedure was used and absorbance was measured at 882 nm (Murphy & Riley, 1962). Soil organic carbon was determined by dry combustion and infrared gas analysis on LECO CNS-2000 equipment. Samples from one calcareous soil were first treated with HCl (0.5 M) to remove carbonates. Clay was determined as described by Day (1965). Details of the sites, soil properties, soil classification, and analytical procedures have been published elsewhere (Kirchmann et al., 1999; Börling et al., 2001, 2004).

Table I. Sites and treatments (n=82 soil samples).

		Treatment
Site	Province	A	B	C	D	E	F
Umeå	AC	X		X		X
Fors	C	X	X		X		X
Kungsängen	C	X	X		X		X
Örsundsbro	C	X				X
Högåsa	E	X	X		X		X
Klostergården	E	X	X		X		X
Ingelstad	G	X		X		X
Valstad	H	X		X		X
Kristianstad	L	X		X		X
Ekebo	M	X	X		X		X
Fjärdingslöv	M	X	X		X		X
Lomma	M	X		X		X
Orup	M	X	X		X		X
S. Ugglarp	M	X	X		X		X
Örja	M	X	X		X		X
Laholm	N	X		X		X
Dingle	O	X		X		X
Hunnebostrand	O	X		X		X
Vargön	P	X		X		X
Bjertorp	R	X	X		X		X
Skara	R	X		X		X
Östervåla	U	X		X		X
Viksjöfors	X	X		X		X
Undrom	Y	X		X		X
A=0 kg P ha^−1 y^−1; B=replacement; C=15 kg P ha^−1 y^−1; D=replacement+15 kg P ha^−1 y^−1; E=30 kg P ha^−1 y^−1; F=replacement+30 kg P ha^−1 y^−1.

Statistical analysis

Stepwise multiple linear regression analysis was used (SAS Institute, 1999–2000) to evaluate the relationship between P_AL and P_Ols and how this relationship is affected by the measured independent variables pH and contents of SOC and clay. Only variables significant at the 5% probability level were included in the resulting regression models. However, where it was deemed necessary, statistical transformation was performed to fulfil the requirements of normal distributions according to the tests (procedure univariate) implemented in the statistical software. For model evaluation, we regressed measured on predicted data according to Piñeiro et al. (2008) and estimated the root mean squared deviation (RMSD) representing the mean deviation of predicted values with respect to those measured (Kobayashi & Salam, 2000) by means of Equation (1):

1

Results

Selected soil properties

Of the 82 soils investigated, 51 had a pH range 6.0–6.9, 15 had a pH range 5.5–5.9, and 12 had pH range 7.0–7.7 (Table II). Mean and range of soil pH was 6.45±0.54 (5.5–7.7), representing the pH range commonly found in Swedish cultivated mineral soils. One soil had pH 7.7 and contained about 3.5% CaCO₃ (Börling et al., 2004). Clay and SOC contents also varied widely in the soils. The majority of Swedish cultivated soils are in P_AL Classes IV and V (Mattsson & Carlgren, 2001; Mattsson, 2002). In this present study, the soils generally were in Classes III–V (Table III).

Table II. Available P status and selected soil properties in the 82 soils.

Variable	Mean±SD	Minimum	Maximum
Soil pH (H2O)	6.45±0.54	5.5	7.7
SOC (%)	2.54±1.21	1	6.03
Clay (%)	17.4±13.8	1	54.4
PAL (mg P kg^−1)	126±89	5	360
POls (mg P kg^−1)	55.1±33	3.1	122.8
PAL/POls quotient	2.31±1.04	0.67	6.58

Table III. Soil classification into Classes I–V according to the standard Swedish P_AL test.

Class	P (mg kg^−1)	Soils
I	≤19	11
II	20–40	8
III	41–80	14
IV	81–160	24
V	≥161	25

The P_AL-values were usually larger than P_Ols, with a mean P_AL/P_Ols quotient of 2.30±1.04 (range 0.67–6.58). Relationships among soil properties and P tests based on Pearson's linear correlation coefficient (r) are presented in Table IV. The P tests were significantly and positively correlated with each other, negatively with clay content, positively with SOC, and weakly with pH.

Table IV. Correlations (Pearson's r-values) between soil properties and soil P tests.

Variable	PAL	POls	Clay	SOC
POls	0.8595 ***
Clay	−0.4146 **	−0.4292 **
SOC	0.4725 **	0.516 **	−0.3875**
Soil pH	0.1279 ^ns	−0.247 *	0.4222**	−0.2142 ^ns
ns=not significant, p>0.05; *^, **^, and ***=significant at p<0.05, p<0.01, and p<0.001, respectively.

Discussion

Relationship between P_AL and P_Ols

Of the soil properties determined, pH and clay could be used in prediction regression equations. Rather surprisingly, the pH-values weakly correlated with the two P tests (Table IV). When untransformed data were used to predict the P tests, the results strongly diverged from the corresponding analytical values, especially when soil content of P_AL was ≤40 mg kg⁻¹, and predicted P_Ols also markedly diverged from analytical values when P_AL content was above about 250 mg kg⁻¹. Among the three transformed independent variables tested, i.e. ln(SOC) and the square root of clay content and pH, only the latter two significantly affected the relationship between P_AL and P_Ols, with partial R ²-values of 2 and 12%, respectively. The resulting regression models [Equations (2) and (3)] accounted for between 95 and 94% of the total variation in the dataset. After raising both sides to the second power, the resulting equations become:

2

3

Plots of measured P_AL against predicted P_AL are depicted in Figure 1, while plots of measured P_Ols against predicted P_Ols are depicted in Figure 2. The mean deviation of predicted compared with measured values was 21.3 and 8.3 mg P kg⁻¹ dry soil for P_AL and P_Ols, respectively, corresponding to 20 or 19% of the measured values. Hence, a data set consisting of P_AL-values can be converted into P_Ols-values and vice versa with reasonably high accuracy when accounting for clay content and pH.

Figure 1.  Regression lines of measured vs. predicted P_AL-values (mg kg⁻¹ dry soil).

Figure 2.  Regression lines of measured vs. predicted P_Ols-values (mg kg⁻¹ dry soil).

The strong relationship (Table IV) obtained between P_AL and P_Ols suggests that these two P tests accessed similar soil P. Strong relationships between the methods have also been reported elsewhere (Börling, 2003; Lončarič et al., 2006). However, the AL and SB solutions are different in composition, strength, and buffering. The AL solution is a strong solution buffered at pH 3.75, whereas the SB solution is a weak alkaline solution buffered at pH 8.5. Evidently, the two methods differ in the mechanisms by which they extract soil P. Owing to its acidic pH, AL can also hydrolyse P in insoluble A1-humic-P substances in soils (McDowell & Sharpley, 2001) and dissolves insoluble Ca-P compounds (Semb, 1984; Fernandes et al., 1999; da Silva et al., 1999). Moreover, the AL solution can liberate P from the Al-P and Fe-P compounds by chelation of the metals, and P can also be exchanged by the anion (COO)⁻ present in the solution. Similarly, SB solution can displace P from the exchange sites by anion-exchange reactions promoted by the HCO₃ ⁻, CO₃ ²⁻, and OH⁻ anions present in the SB solution. Probably by virtue of its slightly alkaline reaction (pH 8.5), the SB solution can also hydrolyse Al(OH)₃, producing [Al(OH)₄]⁻ anion thus: [Al(OH)₃]^o+HOH ↔ [Al (OH)₄]⁻+H⁺. The [Al(OH)₄]⁻ anion can displace (H₂PO₄)⁻. Moreover, P can be released by hydrolysis of Al- and Fe-P thus: AlPO₄+3HOH ↔ Al(OH)₃+H₃PO₄. Mobilisation of P via hydrolysis reactions could probably be important in tropical soils rich in Al and Fe oxides. Indeed, SB has predicted the P requirement for rice grown in some tropical acid soils (Bationo et al., 1991). These differences in the P stocks extracted by the two P tests are reflected in Figure 1 & 2. In Figure 1, most of the points generally fell along the regression line, with some predicted P_AL tending to be either underestimated or overestimated. Similar trends were noted for P_Ols (Figure 2), where some points clustered below or above the regression line, R ²=0.94 compared with R ²=0.95 for the P_AL regression line. According to the regression models, the results from these two P-extraction methods are directly comparable. We conclude that a data set consisting of P_AL-values can be converted into P_Ols-values and vice versa with reasonably high accuracy when accounting for clay content and pH.

Acknowledgements

Financial support was provided by the Swedish Board of Soil Fertility and Plant Nutrition, Royal Academy of Forestry and Agriculture (KSLA).