Comparative investigation and inter-calibration of different soil P tests

Vergleichende Untersuchung und Interkalibrierung verschiedener P-Extraktionsmethoden

Material and Methods

Besides extractable P, “so-called total” P was determined by aqua regia digestion. Furthermore, C_total, soil pH and Al, Ca, Cu, Fe and Zn dissolved by the Westerhoff extract (1954/1955) were analysed (Tab. 2).

The results of the extraction methods and other chemical soil parameters are listed in Tab. 4. The wide range indicates the heterogeneity of the samples used in this study. Furthermore, significant differences in the extraction force of the different methods were found. The extraction force decreased in the order: P_AR > P_AL, P_M3 ≥ P_AAAc-EDTA, P_DL ≥ P_CAL ≥ P_Olsen, P_AAAc ≥ P_W (Fig. 1). As expected, AR extracted by far the highest amounts of P, while water extracted the lowest. It is known that extractants which are more acid or alkaline than the soil solution will also extract P forms with a low plant-availability (Self-Davis et al., 2000). In contrast, the water extraction maintains the soil pH within one unit of its original value. Furthermore, it is expected to simulate the release of P to run-off or leaching by water more accurately than stronger chemical extractants and can thus be used for agronomic purposes from an environmental perspective (Moore et al., 1998). Accordingly, a close correlation of water with DRP (= dissolved reactive P)-concentrations in run-off from agricultural land has been identified (Pote et al., 1996). Furthermore, it could also be observed in the present study that the addition of EDTA to the AAAc-extract increases the amount of extracted P significantly. This confirms the findings of Lakanen and Erviö (1971) who also observed a higher extraction force of AAAc-EDTA. The purpose of the application of EDTA is the complexation of phosphate binding cations such Ca, Al or Fe and a number of trace metals (e.g. Cu) in order to prevent their re-precipitation with phosphates. Otherwise, this re-precipitation might occur during the extraction process as a secondary reaction (Gallet, 2001; Cottenie et al., 1979).

Fig. 1. Extractable P (mean) (mg/kg) determined by 9 dif­ferent methods (significant differences between groups were determined by Tukey post-hoc test (p < 0.05) and are denoted by different letters).

The results shown in Tab. 4 and Fig. 1 are in good agreement with the findings of Neyroud and Lischer (2003), who compared the extraction force of 16 different extraction procedures. The amount of extracted P in their study decreased in the order^**: P_total > (P_oxal.) > P_AL > P_M3 (> P_Bray) > P_AAAC-EDTA, P_DL, P_CAL > P_Olsen (> P_{Paper Strip}) > P_AAAc, (P_Morgan) > P_W, (P_CO2¸ P_CaCl2).

The differences in extraction force between the various methods can be mainly attributed to different extraction mechanisms which base on individual active components in the extract. The three strongest methods, AL, M3 and AAAc-EDTA, all contain one or more strong chelating agents such as lactate, EDTA, NH₃F and acetate (the latter, however, according to Eriksson (2009) displaying only mild chelating properties), which are able to release P from Al- and Fe-P-compounds. In addition, AL and M3 have a low pH (< 4), leading to the hydrolysis of P in insoluble Al-humic-P substances and the dissolution of sparingly soluble Ca-P compounds (Ottabong et al., 2009). In contrast, the two methods with medium strength, DL and CAL, both contain only one chelating agent, lactate in combination with Ca as active cation. Ca²⁺ may precipitate some of the extracted P in the extraction solution and thus will lead to lower P-values compared to those extractants containing NH₄⁺ as active cation (like AL) (Eriksson, 2009). The weaker extractants, Olsen and water, only hydrolyse those Fe-/Al-P compounds which are easily soluble. In addition, the alkaline Olsen-extract releases P from Ca-oxides and some of the P found in Ca-phosphates (Olsen et al., 1954). Due to its alkaline pH, Olsen may show an advantage for soils rich in organic P such as peat soils, as it accesses the organic P pool more aggressively than acid extractants (Ottabong et al., 2009). As only very few of the soil samples investigated in this study showed an OM-content higher than 20%, this finding can neither be confirmed nor disproved here. The extraction force of AAAc is mainly based on its content of acetic acid. Therefore it can also be included into the group of weak P extractants. Besides the active components in the extract, factors such as soil: solution ratio as well as duration and power of shaking the samples influence the extraction force of the different methods (Eriksson, 2009).

The differing extraction forces of the various methods are useful for different purposes and also ask for different interpretation strategies for each method (Neyroud and Lischer, 2003). Weak extraction methods tend to reflect immediately available P and may be useful for immediate fertilisation recommendations. In contrast, the stronger extractants are rather related to P species which are more strongly bound and will become available on a medium to long-term basis, and thus essential for planning fertilisation on a mid to long-term basis. Regarding the assessment of the actual risk of P loss and environmental pollution (eutrophication of ground and surface waters), the weak extraction methods appear to be more suitable to the authors of this study.

Of course, soil characteristics also play an important role in the performance of the different extraction methods. Thus, key factors influencing the type of P binding such as soil pH, organic carbon, and contents of Fe_WH, Al_WH, Ca_WH, Zn_WH and Cu_WH were also included into the regression equations.

The main soil parameters for the soil samples examined per country are presented in Tab. 5.

The average pH of the soil samples from Estonia was 6.3. Worth mentioning were the significantly elevated Ca_WH-concentrations and comparatively low Cu_WH and Zn_WH concentrations in these soils.

The parameters of the Finnish soil samples differed significantly from those of the three other countries; the soils were most acidic (pH 5.0), the C-contents were by far the highest and the results for Cu_WH, Zn_WH, Fe_WH and Al_WH were elevated, as well. The average pH of the soil samples from Germany was 6.2 (slightly acidic) and Cu_WH- and Zn_WH concentrations were elevated in comparison to the Estonian soil samples. This tendency was also observed for the Polish soils.

Tab. 6 shows the average amounts of P extracted by different extraction methods. Among all soil samples, both, the “so-called total” P-content and extracted P were significantly higher in the samples from Poland. In contrast, the mean total P content in the Finnish soils was also elevated, but the share of extractable P in these soils was on a very low level for all methods. Since the pH of the Finnish soils was moderately acid and the Fe_WH- and Al_WH-concentrations were distinctively elevated (Tab. 5), it can be assumed that the formation of insoluble Fe_WH- and Al_WH-complexes was responsible for an increased P-fixation and resulted in the low share of extractable P. The average amount of extractable P in the Estonian soils was also very low and comparable to the Finnish soils. With view to high Ca_WH-concentrations this might be explained by the precipitation of sparingly soluble Ca-P-compounds (Tab. 5). The amounts of extractable P in the German soils ranged between those of Finland/Estonia and Poland.

At present, different standard methods are used among different countries to assess and to interpret the amount of available P in agricultural soils (see Tab. 1). Exemplary, the standard P extraction methods and the respective classification systems for Estonia, Germany, Poland and Sweden are listed in Tab. 7.

Differing reference methods and classification systems exist in each country which consequently results in a divergent assessment of the P supply of agricultural soils (Fig. 2). While the German classification tends to group sites into lower (deficient) classes, they are assessed as being sufficiently or even excessively supplied with P if one of the other national procedures is applied. These findings underline the necessity for a harmonisation of P-extraction and classification standards.

Fig. 2. Percentage distribution of P-classes obtained by applying different national standard extraction me­thods and the corresponding classification system (see also Tab. 7) (n = 183).

Satisfactory results (R² > 80%) were already achieved with simple regression equations for the calculation of P_CAL from P_AL, P_M3, P_AAAc-EDTA and P_water, which also showed very strong bivariate correlations (r > 0.9) between each two respective extraction methods (Tab. 8).

Despite different extraction mechanisms, these methods seem to be easily convertible into each other. This is also confirmed with the simple regression equations calcu­lated for each of them (Tables 10, 13, 15, 17). A multiple regression equation revealed satisfactory results for the calculation of P_CAL from P_DL. However, for P_Olsen and P_AAAc no coefficient of determination < 80% could be obtained by a multiple regression equation (Tab. 9).

The bivariate correlations (r) between P_Olsen and the other P-tests applied in this study were highly significant, but not as strong as those between P_CAL and other P-tests (0.512 ≤ r ≤ 0.810). This can be attributed to the alkaline nature of the Olsen-extract, a unique feature among the extractants tested in this study. Accordingly, different soil reactions can be expected, rendering the direct interchangeability with acidic extracts less likely. Consequently, the generation of simple regression equations for transferring Olsen-data into data obtained by acid extracts displayed comparably low coefficients of determination, ranging between R² = 25.8 and 65.5%. The inclusion of additional soil parameters into multiple regressions did not improve the outcome; coefficients of determination still remained below R² = 80%, which is considered too low to apply these equations for a reliable transfer of P-Olsen into another soil test (data not shown here).

P_AL showed highly significant and very strong (r > 0.9) bivariate correlations with P_CAL, P_M3, P_AAAc-EDTA and P_water. In this case, simple regression equations with only the corresponding P-test as regressor proved to be fully viable for estimating P_AL with a satisfactory level of determination (R > 82%) (Tab. 10).

For the relation P_AL-P_DL, the multiple regression yielded a satisfactory level of determination (85.2%) which was not the case for the relations P_AL-P_Olsen and P_AL-P_AAAc (Tab. 11). As indicated before, the reason for this can be most likely attributed to the chemical composition of the extractants (AL – acid/Olsen – alkaline). The main difference between AL and AAAc is assumed to lie in the higher chelating power of AL, which contains lactate as a chelating agent, while AAAc contains acetate, having only mild chelating properties and serving mainly as a pH-reducing component. Furthermore, the ammonium cation in the AL extract results in a high desorption of P (Eriksson, 2009).

Bivariate correlations (r) for DL with other extracts were strong and highly significant, but remained below 0.9 (data not shown here). Consequently, the coefficients of deter­mination for the simple regression equations were below 80%, i.e. estimates based on these equations will not be precise enough (data not shown here). Even with multiple regression equations, only levels of determination below 80% were achieved for P_DL estimated from P_Olsen, P_M3, P_water or P_AAAc. Reliable estimates (R² > 80%) could only be calculated for P_CAL, P_AL or P_AAAc-EDTA from P_DL data when additional soil parameters were included (Tab. 12).

As mentioned earlier, M3, water and AAAc-EDTA were easily convertible into each other as well as into CAL or AL-values by simple regression equations (see Tab. 13, 15 and 17). If further soil parameters are available, in some cases even higher coefficients of determination were achieved by using multiple regression equations (data not shown here).

The application of multiple regression equations did not lead to satisfying results for the conversion of M3 or AAAc-EDTA into Olsen, DL and AAAc (Tab. 14 and 16).

The application of multiple regression equations leads to a satisfying coefficient of determination for AAAc-EDTA-DL. For Olsen and AAAc no coefficient < 80% could be achieved (Tab. 18).

The weakest bivariate correlations were displayed by AAAc when related to other extraction methods. Hence, neither simple nor multiple regression equations could be generated for a reliable conversion of AAAc-data into data for other methods with a sufficiently high coefficient of determination (data not shown here).

For a first evaluation and validation of the models, the application of the best regression equations for each pair of methods to the soil samples they were derived from, and then regressing measured data on predicted data is suitable (Ottabong et al., 2009). In this study, a first model evaluation showed promising results with view to the validity of the derived regression equations (data not shown here). The respective coefficients of determination (of transformation regression equation and regression lines for measured vs. predicted values) were almost identical, indicating that both regressions are based on the same basic data set (data not shown here). Thus, an additional data set containing samples which were not used for calculating the regression equations will be used for a statistically sound validation.

Fig. 3. Regression equations and coefficients of determination for measured (y) vs. predicted (x) P-values (mg/kg air-dried soil) using an independent data set for validation.

Coefficients of determination for the regressions between predicted and measured values ranged from 47 up to 97% (Tab. 19). Out of the 26 equations tested, 8 produced predictions with a coefficient of determination R² < 80%, i.e. a transformation of STP using these equations would not yield a statistically satisfactory result.

Footnotes: