Repeated, grid-based forest soil inventories such as the
National Forest Soil Inventory of Germany (NFSI) aim, among other things, at
detecting changes in soil properties and plant nutrition. In these types of
inventories, the only information on soil phosphorus (P) is commonly the
total P content. However, total P content in mineral soils of forests is
usually not a meaningful variable with respect to predicting the availability of P to trees.
Here we tested a modified sequential P extraction according to Hedley (1982) to
determine the distribution of different plant-available P fractions in soil
samples (at depths of 0–5 and 10–30 cm) from 146 NFSI sites, encompassing a wide
variety of soil conditions. In addition, we analyzed relationships between
these P fractions and common soil properties such as pH, texture, and
soil organic carbon content (SOC). The total P content among our samples ranged from
approximately 60 to 2800 mg kg
Insufficient or even critical phosphorus (P) nutrition in forest trees has
been repeatedly observed in Europe over the last few decades (e.g., Jonard et
al., 2015). A large proportion of forest trees that were examined within
the framework of the Second National Forest Soil Inventory in Germany (NFSI
I-II) showed insufficient P nutrition (20 % of all
It has been shown that the total P content in mineral forest soils is not a
significant predictor of tree nutritional status, expressed as foliar P
content, in different tree species (Ilg et al., 2009). This indicates that
most soil P is unavailable or not directly plant available. Correlations between
total soil P and foliar P contents have been observed in a few studies, e.g.,
for
Other approaches that comprise the quantification of a number of different P
fractions in mineral soils have been successfully employed
to describe potential sources of P uptake by trees in forest soils. One analytical approach
that allows for the partitioning of total P into fractions
of different (plant) availability in mineral soils is the Hedley fractionation method (Cross
and Schlesinger, 1995; Hedley et al., 1982; Tiessen and Moir, 2008). The
original method (Hedley et al., 1982), which was modified by Tiessen and
Moir (2008), provides a total of seven inorganic and four organic P
fractions. These P fractions are often grouped into pools of distinct plant
availability: a labile, fast cycling pool (labile P), which is considered to
supply the short-term P demand of plants; a slow cycling pool (moderately labile
P), which can be converted into labile P forms; and a pool of occluded P
(stable P), which is assumed to hardly contribute to plant nutrition in the
short-term (Guo and Yost, 1998; Stevenson and Cole, 1999; Johnson et al.,
2003). There are a number of studies that have examined changes in P stocks
in forest ecosystems using the Hedley fractionation method. Some of them have
followed the development of P fractions over time to gain information on the
relevance of these P fractions for tree nutrition during ecosystem
development (Richter et al., 2006; De Schrijver et al., 2012). In other
studies, the influences of different forest
management systems on the distribution of P fractions in soils were investigated (Alt et al.,
2011). These were case studies at single sites or only at few different
sites and thus had only a limited population of inference (the population to
which the results from the sample can be extrapolated) (Binkley and
Menyailo, 2005). To our knowledge, no studies have currently addressed the
distribution of Hedley P fractions and P pools in forest soils on the basis
of large-scale inventories. Thus, there is little information on how
different soil variables such as pH value, C and N content, or soil texture,
which have been found to influence P availability (Alt et al., 2011;
Franzluebbers et al., 1996; Prescott et al., 1992; Silver et al., 2000;
Stevenson and Cole, 1999; Thirukkumaran and Parkinson, 2000; Turner et al.,
2007), affect the distribution of different P fractions across a variety of
forest soil types. Therefore, we determined the Hedley P fractions in
mineral soil samples from 145 sites of the National Forest Soil Inventory of
Germany, covering a wide range of P contents and many different soil parent
materials (Niederberger et al., 2015). With this study we
addressed the following questions:
How do commonly measured soil properties such as pH value, soil organic
carbon content, and soil texture influence the distribution of Hedley P
fractions, representing different P pools believed to have different
bioavailability? Are the foliar P contents of trees related to specific Hedley fractions, the respective Hedley P pools of soil P, or other soil variables?
For the purpose of our study, we used archived soil samples from the National Forest Soil Inventories of Germany (NFSI I and II) including samples from the state of Baden-Württemberg that originated from the first NFSI I in 1990, and samples from the states of Hesse, Lower Saxony, and Saxony-Anhalt that originated from the second NFSI II in 2006. In total, 285 archived soil samples from 147 sites (Fig. 1) were included in this study. In most cases, two depths, 0–5 and 10–30 cm, were analyzed for each site. At a depth of 0–5 cm, 145 samples were analyzed, and at a depth of 10–30 cm, 140 samples were available. The selected samples represented a wide range of total P contents (Table 1). Additional soil and site data as well as foliar element concentrations were provided by the Thünen Institute, the Forest Research Institute of Baden-Württemberg, and the Northwest German Forest Research Institute. Here, we used pH value, C and N content (Table 1) and soil type, forest floor type, soil texture, and dominant tree species as predictor variables. Sampling approaches and analytical methods used in the NFSI have been described in detail by Wolff and Riek (1996).
Variation in selected variables for soil samples from the German Forest Soil Inventory.
To analyze the effects of soil variables on distribution of P across
fractions of different plant availability, we classified the samples by
their variation in soil properties. To classify them by pH value, we used
the buffer ranges suggested by Ulrich (1981). In relation to the SOC content of
mineral soil, samples were grouped into classes of “low” (< 1.2 % SOC), “medium” (1.2–2.8 % SOC), “high” (2.8–5.6 % SOC), and
“very high” (> 5.6 % SOC) content following the German soil
assessment protocol for forest soils (Ad-hoc-Arbeitsgruppe Boden, 2005).
Furthermore, we grouped samples into the broad texture classes sand (
Locations of the 147 sample sites from the German Forest Soil
Inventory (NFSI) dataset; B represents Baden-Württemberg (
For 118 NFSI sites, the foliar P contents of the main tree species (
The P fractionation was carried out using the Hedley method (Hedley et al., 1982) modified by Tiessen and Moir (2008). For this purpose, 0.5 g of soil was repeatedly extracted using different extractants with increasing chemical strength (Fig. 2). A detailed description of the fractionation procedure used in this study is provided in Niederberger et al. (2015, 2016).
Sequential P fractionation schema according to Hedley modified by Tiessen and Moir (2008). Gray boxes indicate fractions with organic and inorganic P forms, and the dashed lines separate the Hedley P pools with different availability, following Niederberger et al. (2015).
The NFSI provides the total P contents for individual soil samples, which are determined in different labs employing different analytical methods. In a preliminary study, we found considerable differences between this total P (of the NFSI) and the sum of our Hedley fractions. Therefore, we carried out an additional, independent analysis to determine total P. For comparability with the Hedley fractionation procedure, we used nitric acid digestion to quantify this additional measure of total P. It should be kept in mind though that this analytical procedure might underestimate the real total P content in mineral soils (BMEL, 2014; Schwartz and Kölbel, 1992). This measure of total P content also served as a control to verify the recovery rate of the sum of Hedley P fractions (see also Niederberger et al., 2015).
In a first step, data were analyzed using descriptive statistics. As most
of our soil and site data were not normally distributed (Shapiro–Wilk test,
To compare variables between soil depths, we used the paired nonparametric Wilcoxon test. To compare variables among different soil classes within the same soil depth, for instance classes of different SOC content, we used the nonparametric Mann–Whitney U test.
In addition, we used linear regression models to explain Hedley P pools and P fractions with the predictor variables' depth, pH, SOC, and soil type. For modeling soil P content, we applied a log transformation to the Hedley P fractions and P pools. Transformations of the major soil types (sand, loam, silt, and clay) into grain size distribution (expressed as sand content) led to considerable model improvements. To transform soil type information in this way, we used mean values of sand content found in specific soil types based on the German soil assessment protocol for forest soils (Ad-hoc-Arbeitsgruppe Boden, 2005). To assess further model improvements we separated our sample set into calcareous and noncalcareous soils, as we expect strong differences in processes affecting P availability (Prietzel et al., 2014). Despite the large number of sites included in our survey, there were only 8 out of 143 with pH > 6.5. As this number is too low to develop robust statistical models for this collection of samples, we could only test whether model results for the group of soils with pH < 6.5 (noncalcareous soils) changed when calcareous soils were excluded. However, we found only minor improvements as well as some minor deterioration of model quality. Hence, for all further analyses, calcareous soils were not separated from the other soil samples.
Furthermore, we used linear regression models to test whether the foliar P
contents of
The total P content calculated as the sum of all Hedley P fractions ranged
from 58 to nearly 2800 mg kg
The sum of stable P fractions (P HCl
Box plots of Hedley P fractions and P pools for all analyzed
mineral soil samples separated by depth (all values in mg kg
Organic P forms were the largest single fractions within labile and
moderately labile Hedley P pools at both depths. At a depth of 0–5 cm, organic P
forms contributed even more than the sum of the inorganic forms to these two
Hedley P pools (Fig. 3). At a depth of 0–5 cm, contents of all labile P fractions
were significantly higher than at a depth of 10–30 cm. While organic P in the
moderately labile pool (P
Hedley P pools and total P (mean values), grouped by pH classes
and soil depths. Lowercase letters indicate significant differences between
pH classes within Hedley P pools and per depth (nonparametric
Mann–Whitney U test,
Total P decreased with increasing soil acidity (Fig. 4). This decrease was mainly attributable to a significant decline in the pool of stable Hedley P fractions, which decreased both in absolute as well as relative terms. The portion of stable P dropped from 48.7 % to 16.0 % of total P at a depth of 0–5 cm and from 56.8 % to 26.0 % at a depth of 10–30 cm (Fig. 4). In contrast, the labile P pool in surface soils (0–5 cm) increased significantly in absolute as well as relative terms with increasing acidity. At a depth of 10–30 cm, absolute quantities of labile P remained relatively constant, whereas its relative share of total P increased with increasing acidity (Fig. 4). The moderately labile Hedley P pool showed comparatively small differences between pH classes at both depths; only at a depth of 10–30 cm was this pool significantly smaller in the most acidic soils when compared to other pH classes.
Considering individual P fractions, we found an
increase of P content with decreasing pH value at a depth of 0–5 cm for all labile P fractions (Table S2),
but this was only significant for the P
Total P as well as P in all Hedley fractions increased strongly and significantly with increasing SOC content in mineral soil at both depths (Fig. 5). In contrast, the relative proportions of Hedley P pools showed no or only minor changes with increasing SOC at both depths (Fig. S2, Tables S4 and S5).
Hedley P pools and total P (mean values), grouped by SOC content
(in percent; %) and depth. “
Total P and in particular the stable Hedley P pool increased with decreasing particle size from sand to clay. The lowest contents of total P and P in the stable and moderately stable fractions were found in sandy soils at both depths (Fig. 6). In particular, the 1M HCl soluble P fractions showed extremely low P contents in sandy soils (Table S6). Accordingly, the highest proportion of labile P was also found in sandy soils: 40 % at a depth of 0–5 cm and 35 % at a depth of 10–30 cm. The proportion of labile P decreased with decreasing grain size. At a depth of 10–30 cm, the increase in total P with increasingly finer soil texture was mainly caused by stable P forms.
Hedley P pools and total P (mean values) grouped by soil texture
and depth. “
The best linear regression models were found between soil variables and organic P fractions in labile and moderately labile and stable Hedley P pools, whereas model quality for the inorganic P forms were considerably lower (Table 2). Sand content was a negative predictor in all cases, whereas SOC was always a positive predictor. In contrast, the pH value was a negative predictor for labile P, in particular for labile organic P, but positive for stable P forms.
Adjusted
For the labile Hedley P pool and the organic P fractions, the SOC had a stronger influence than the sand content, whereas the reverse was true for stable Hedley P pools. For moderately labile P, both predictors were similarly important, except for the 1M HCl soluble P fraction, where SOC content was not a significant predictor.
Models with a goodness of fit above 0.4 (Table 2) were only obtained for
organic P fractions (P
In samples from a depth of 10 to 30 cm (Table S7), we found generally lower model qualities but comparable patterns to those observed at a depth of 0–5 cm.
Total soil P and Hedley P pools with different availability varied
considerably between NFSI plots dominated by different tree species (Fig. 7). Under
Box plots of Hedley P pools and the sum of all Hedley fractions in
mineral soils of analyzed NFSI plots for sites dominated by the three main
tree species. Different letters indicate significant differences among soils sampled under different tree species for the respective P pool as well as the Hedley P sum in mineral soil (nonparametric
Mann–Whitney U test,
The quality of the linear regression models for foliar P content varied
considerably amongst the tree species examined. For
However, the vast majority of the relatively low number of needle samples of
Model quality and standardized, significant regression coefficients of linear regression models to explain foliar P content using Hedley P pools with different availability and other soil variables determined at soil depths of 0–5 and 10–30 cm.
Multiple linear regression models to predict foliar P content in the three tree species with soil variables and P pools from depths of 0–5 and 10–30 cm achieved a moderate quality when applied across the three species (Fig. 8). However, they were not suited to predict foliar P contents when datasets were considered for each of these species separately.
Results of multiple linear regressions to predict foliar P content
in
Our results show that soil properties, like acidity, SOC content, soil texture, and depth, have an important influence on the quantity and distribution of plant-available P in forest soils. However, to date, there have been very few studies that have investigated this issue (Augusto et al., 2017; Buckingham et al., 2010; Shang et al., 1992; Zederer and Talkner, 2018).
In our analysis there was no single soil variable that was consistently the
best predictor of the different Hedley pools or fractions. However, there
were consistent patterns such as the fact that the SOC content always had a positive influence and
sand always had a negative influence. In contrast, the influence of pH, which
was never a stronger predictor than either SOC or sand, could be positive or
negative. Labile P
In the following discussion, we first address the general assumptions
regarding the influence of these variables on the P distribution in mineral
soils and relate them to our results. Secondly, we discuss the results of
the regression models that were used to examine the relationship between
Hedley P pools or fractions and the foliar P content of
The vertical distribution of P in the soil profile is dominated by
biological turnover processes. Plant P uptake and transformation of
inorganic into organic P in combination with microbial activity lead to an
enrichment in organic and labile P and, in turn, also in total P in the topsoil
and forest floor layers (Jobbágy and Jackson, 2001). This could be clearly
demonstrated by our results which showed significantly higher amounts of
organic P forms and labile Hedley P in the upper soil layer.
Furthermore, our findings of a significant decrease in organic P fractions (P
Our results are also in agreement with the observation of a relative
accumulation of more stable inorganic P forms in deeper soil layers by P
fixation, for instance as secondary P minerals or in clay minerals
(Buckingham et al., 2010; Vitousek et al., 2010). This increase in
moderately labile and stable P
Solubility and fixation of soil P is strongly affected by soil acidity (pH) (Hinsinger, 2001; Shang et al., 1992; Stevenson and Cole, 1999). The optimum availability of P to plants typically occurs around pH 6.5. At a pH value below 6, P is fixed as Fe or Al phosphates or adsorbed to oxide surfaces (Shang et al., 1992; Stevenson and Cole, 1999) and above a pH value of 7, P is fixed in the form of Ca phosphates (Stevenson and Cole, 1999). It has been shown that there are considerable changes in the relative importance of P fractions with pH, even if there is only a minor influence of pH on total P (Turner and Blackwell, 2013). For example, phosphonates were only found in acidic soils, and the amount of P in DNA increased with increasing soil acidity (Turner and Blackwell, 2013). However, specific organic P forms can unfortunately not be picked up by the Hedley fractionation; thus, we could not analyze these pH-related shifts.
In accordance with other studies, the highest portion of labile P was found in the most acidic soil samples (Alt et al., 2011; Turner and Blackwell, 2013). The decrease in labile organic P with increasing pH might be caused by its enhanced mineralization in soils with a more favorable acid–base status for microbes (Stewart and Tiessen, 1987). In contrast to Turner and Blackwell (2013), we observed a positive influence of pH on total P contents at both soil depths. This disagreement between the studies may be attributed to the much larger and more variable dataset used in our study.
The contrasting effects of pH (negative on labile P and positive on stable P forms) might be explained by the different processes that influence P availability and fixation in mineral soils at different pH levels (Hinsinger, 2001; De Schrijver et al., 2012). The negative impact of increasing pH on labile P might be caused by the enhanced decomposition of organic matter by microorganisms at higher pH levels and increased mixing of organic matter with the mineral soil matrix (Paré and Bernier, 1989; Scheffer and Schachtschabel, 2010). Likewise, the positive effect of pH and higher portions of clay minerals on stable P may be explained by the increased fixation of P by clay minerals (Sugihara et al., 2012) and the occurrence of primary and or secondary P-containing minerals in soils at higher pH levels (Hinsinger, 2001).
As organic P forms can account for more than 50 % of total P in
mineral soils (Fox et al., 2011; Stevenson and Cole, 1999; Turner, 2008), we
assumed a strong influence of SOC on the distribution of P in mineral soils.
In agreement with expectations, total P and all P fractions increased with
increasing SOC content, whereas the relative contribution to total P of P in
pools of different availability remained stable (Fig. 5). Like other
studies, we found a strong positive correlation between SOC and total
organic P content (
The rather constant relative proportions of stable, moderately labile, and labile Hedley P pools across soils with different SOC contents (Fig. 5) and the close relationship between SOC and P (Johnson et al., 2003) suggest that these pools may be predicted by SOC content.
The significantly lower P in all Hedley fractions in sandy soils when
compared with other soil texture types may be related to lower organic matter
content, fewer possible fixation opportunities, e.g., at clay minerals, and
higher acidity in these soils. Strong negative correlations between sand
content and organic P and strong positive correlations with inorganic P were
found in agricultural soils, whereas silt content had the opposite effect
(O'Halloran et al., 1985). These were explained by sorption of phosphate
monoesters at clay minerals. Likewise, we interpret the negative influence
of sand content on total P content in all significant models of our study as
the decreasing amount of surfaces to which P could be adsorbed or fixed.
As
Nutrient content in leaves and needles are subjected to strong spatial and temporal (inter- and intra-annual) variability (Netzer et al., 2017; Rennenberg and Herschbach, 2013). Therefore, statistical models to explain the foliage nutrient content based on soil properties should be based on a large number of replicate foliage samples to capture the spatial heterogeneity as well as the temporal variation (Wehrmann, 1959), which may be caused by climatic conditions or mast years. Unfortunately, the NFSI only provides data on foliage nutrient content for one point in time, and based on a small sample from three trees per plot (Wellbrock et al., 2016). Therefore, the low quality of our statistical models may be partially caused by the sampling design of the NFSI, and could be improved by collecting foliage samples over several consecutive years (Zederer and Talkner, 2018).
The mean values of Hedley P pools of different availability in forests soils
stocked with different main trees species (Fig. 7) indicated considerable
differences in P availability. While for
However, some earlier studies indicated that there was a correlation, albeit not
strong, between total P content in soil and tree nutrition for
Foliar P contents in
Our best multiple linear regression models for foliar P in
The model results for
The results of the study of Magnhabati et al. (2018) are remarkable because
total soil P was as good a predictor of the P nutrient status in
Using the Hedley fractionation approach, we assessed the distribution of soil P forms of different availability in a range of German forest soils and analyzed relationships between these fractions and selected soil properties. Although our dataset was not representative for German forest sites, it clearly showed that approximately half of the soil P is contained in moderately labile fractions, whereas stable and labile fractions contribute to approximately one-quarter of the total P in the upper mineral soil. With increasing depth, the labile Hedley soil P pool declines in favor of the stable Hedley pool. Common soil properties such as pH, SOC, and soil texture may be used to predict certain Hedley P pools in large forest soil inventories. However, additional soil and site variables should be considered to improve these models.
Despite high quantities of P in labile fractions in mineral soils, which
greatly exceed the annual uptake demand of trees, correlations between these
fractions and foliar P contents of
Data are available upon request.
The supplement related to this article is available online at:
JN was responsible for the sample acquisition, analysis, and interpretation of data, he made substantial contributions to the conception and design of the study, and is the lead author of the paper. JB conceived the idea for this study. MK and JB contributed to the design of the study as well as to data interpretation and writing the paper.
The authors declare that they have no conflict of interest.
We wish to thank Renate Nitschke and Konstantin Winschu for technical support, and Klaus von Wilpert from the the Forest Research Institute Baden-Württemberg and Jan Evers from the Northwest German Forest Research Station for providing the NFSI samples. We also wish to thank the editors and the three anonymous referees, who helped to improve the paper.
This research has been supported by a grant to Jürgen Bauhus, provided by the Federal Ministry of Food, Agriculture and Consumer Protection (BMELV) administered through the Thünen Institute, Eberswalde, Germany.
This paper was edited by Stuart Grandy and reviewed by three anonymous referees.