Soil phosphorus status and P nutrition strategies of European beech forests on carbonate compared to silicate parent material

Sustainable forest management requires understanding of ecosystem phosphorus (P) cycling. Lang et al. (2017) [Biogeochemistry, https://doi.org/10.1007/s10533-017-0375-0] introduced the concept of P-acquiring vs. P-recycling nutrition strategies for European beech (Fagus sylvatica L.) forests on silicate parent material, and demonstrated a change from P-acquiring to P-recycling nutrition from P-rich to P-poor sites. The present study extends this silicate rock-based assessment to forest sites with soils formed from carbonate bedrock. For all sites, it presents a large set of general soil and bedrock chemistry data. It thoroughly describes the soil P status and generates a comprehensive concept on forest ecosystem P nutrition covering the majority of Central European forest soils. For this purpose, an Ecosystem P Nutrition Index (ENIP) was developed, which enabled the comparison of forest P nutrition strategies at the carbonate sites in our study among each other and also with those of the silicate sites investigated by Lang et al. (2017). The P status of forest soils on carbonate substrates was characterized by low soil P stocks and a large fraction of organic Ca-bound P (probably largely Ca phytate) during early stages of pedogenesis. Soil P stocks, particularly those in the mineral soil and of inorganic P forms, including Al- and Fe-bound P, became more abundant with progressing pedogenesis and accumulation of carbonate rock dissolution residue. Phosphorus-rich impure, silicate-enriched carbonate bedrock promoted the accumulation of dissolution residue and supported larger soil P stocks, mainly bound to Fe and Al minerals. In carbonate-derived soils, only low P amounts were bioavailable during early stages of pedogenesis, and, similar to P-poor silicate sites, P nutrition of beech forests depended on tight (re)cycling of P bound in forest floor soil organic matter (SOM). In contrast to P-poor silicate sites, where the ecosystem P nutrition strategy is direct biotic recycling of SOM-bound organic P, recycling during early stages of pedogenesis on carbonate substrates also involves the dissolution of stable Ca-Porg precipitates formed from phosphate released during SOM decomposition. In contrast to silicate sites, progressing pedogenesis and accumulation of P-enriched carbonate bedrock dissolution residue at the carbonate sites promote again P-acquiring mechanisms for ecosystem P nutrition. Supplementary Information The online version contains supplementary material available at 10.1007/s10533-021-00884-7.


Detailed description of methods: Soil sampling, XANES spectroscopy, statistics
Soil sampling using the Quantitative Pit approach For quantitative pit establishment, a square wooden frame with an interior lateral length of 50 cm was prepared and registered optical targets for photogrammetric analysis were fixed to the upper side of the frame. The frame was fixed as reference plane to the soil surface using steel pins. The entire organic layer was cut off alongside the inner edge of the frame with a knife. The individual humus layers (Oi, Oe, Oa) were sampled separately by hand.
Roots crossing different layers were cut off at the respective layer boundaries and removed to prevent mixing of material from different layers. The mineral soil sampling followed layers representing single diagnostic horizons.
If the thickness of a diagnostic horizon exceeded 5 cm for A horizons and 10 cm for B horizons, a new sample layer was started. The procedure was repeated until a maximum sampling depth of 1 m or the consolidated bedrock was reached. We placed all soil material (including rocks and roots) in containers, transported it to the laboratory, and air-dried it at 40°C. Then, we manually separated all soil samples into the following fractions: fine earth (<2 mm), gravel (2-20 mm), stones (>20 mm), coarse roots (>2 mm), fine roots (<2 mm), other soil constituents (e.g., wood, seedlings). After weighing, all fractions were stored dry, cold, and in the dark for further analysis. For soil analysis, we used dried fine earth material. The volume quantification of the different soil layers was carried out based on photogrammetry (Haas et al. 2016). Details concerning the stand representativeness of the QP approach have been discussed earlier (Lang et al. 2017).
Following the protocol of Werner and Prietzel (2015), we conducted LCF from -14 to +46 eV with respect to E0.
The lower energy used for baseline correction varied between -43 and -30 eV (1 eV step), the upper energy between -19 and -9 eV (0.5 eV step) with respect to E0. The lower energy used for normalization varied between +34 and +40 eV (0.5 eV step), the upper energy between +50 and +65 eV (1 eV step) with respect to E0. All spectra were fitted automatically, using baseline-corrected, normalized standard spectra as predictor compounds, allowing for up to six P species being included in the fits. Phosphorus speciation shares <5% of total P were excluded from the result list and LCF was repeated without the respective standard. To improve the accuracy and precision of the LCF results, we used the averages of the five "best" results, i.e. with the smallest R factors as LCF results following the suggestion of Schmieder et al. (2020). Moreover, (i) spectra of P adsorbed to Al or Fe minerals with different degrees of mineral order (e.g. ferrihydrite vs. goethite), different Al mineral types (e.g. gibbsite, Al-saturated montmorillonite, amorphous Al(OH)3), as well as (ii) spectra of inorganic and organic P adsorbed to the same pedogenic mineral are very similar (Prietzel et al. 2016;Gustafsson et al. 2020). Because they therefore are hard to quantify specifically by LCF (Gustafsson et al. 2020), we combined the XANES P speciation data with the results of our wet-chemical determination of organic and inorganic P in the respective samples to yield our final P speciation results. In detail, in topsoil (O, Ah) horizons with pH values <6.5, which were dominated by organic P according to the results yielded with the Saunders and Williams (1955) method, Ca-bound P was assumed to be organic. Inorganic P in these horizons was assumed to be Al-bound, and in some horizons where also Fe-bound P was identified, also Fe-bound. In C and BwC horizons with pH values >7, Ca-bound P was assumed to be organic as well as inorganic, apatite P (Table S1).

References:
Gustafsson JP, Braun S, Tuyishime JRM et al (2020)  (1) P stocks: The level of total P stocks of 90-340 g P m -2 was at the level of the P-poor soils on silicate parent materials ( Figure 2 in main manuscript). At present, no inventory information based on a large number of samples is available that might allow for comparing P stocks of forest soils on carbonate with those on silicate bedrock on global scale. On regional scale, an inventory of 56 forest soils in the 70,000 km² German State of Bavaria (Schubert 2002) showed that six of the ten P-poorest soils had carbonate parent material; the other four P-poor profiles were formed from quartz-rich sand or sandstone. The lower P status of carbonate-than of silicate-derived soils is well in line with the lower P content of most carbonate parent materials, particularly those of high purity, compared to most silicate parent materials (Porder and Ramanchandran 2013). Second, due to intensive proton buffering of carbonates, chemical weathering of carbonate parent material proceeds much slower than silicate weathering.
Together with the low P content of most carbonate bedrock types, slow weathering results in low lithogenic P input as well as low soil accumulation rates of P-retaining sesquioxides and clay minerals.
In forest soils on silicate parent material, P stocks were mainly dependent on the general soil nutrient status, including the stocks of total and exchangeable Ca, Mg, and K, which in turn are strongly linked to parent material type (basalt > gneiss > quartz-rich sediments; Lang et al. 2017 (2) Soil P speciation: In the carbonate-derived soils, a larger portion of total P (66-90%, on average 77%) was Porg than in the soils on silicate bedrock (35-52%, on average 43%) studied by Lang et al. (2017). This was probably partly an effect of impeded enzymatic cleavage of Ca-Porg precipitates, as well as of advanced microbial immobilization and cycling of P in microbial residues, which may comprise a majority of SOM constituents (Liang With progressing carbonate dissolution and pedogenesis, P-retaining weathering residues consisting of sesquioxides and clay minerals initially present in the carbonate parent material accumulate, and therefore, the amount of Pinorg in soils with carbonate parent material as well as its contribution to total soil P increase with ). Precipitation of IHP with Ca 2+ restrains P cycling, and probably is a key factor for the low supply of bioavailable P in carbonate soils at early stages of pedogenesis. Yet, the XANES results also indicated gradual dissolution of Ca-IHP precipitates and accumulation of Al-or Fe-bound Porg with progressing pedogenesis, e.g. in the topsoil of BAE, which has a pH of 5.4. Not only monoester-P, but also diester-P contents were elevated in the carbonate soils: Topsoil contents of both monoester-and diester-P were at the level of the P-richest silicate soil BBR, whereas topsoil Pinorg contents were markedly smaller. This indicates that diester-P compounds are stabilized and accumulated (Bünemann et al. 2008) in Ca-rich topsoil horizons of carbonate soils. However, diester-P/monoester-P ratios were strongly decreased in the carbonate soils compared to the silicate soils and particularly compared to the P-poor silicate soils CON and LUE This can be attributed to the fact that on average phosphodiesterase activities were five times higher in the carbonate than in the silicate soils (Table 5 in main   manuscript, Table S6), whereas phosphomonoesterase activities were only elevated by <40%, resulting in markedly decreased phosphodiesterase/phosphomonoesterase ratios in the carbonate compared to the silicate soils.
The monoester P accumulation in the carbonate soils thus can be explained by (i) pronounced formation of stable Ca inositol phosphates (i.e., monoesters), (ii) low monoesterase activities, and (iii) the circumstance that enzymatic cleavage of Ca-bound inositol phosphate results in mobilization of the organic inositol moiety than the phosphate group, because binding of inositol phosphate to Ca 2+ is mediated by the phosphate group. This requires an additional step of Ca-phosphate bond destruction to render the remaining phosphate molecule plant-or ecosystemavailable, further hampering the utilization of Ca-bound inositol phosphate by plants and soil microorganisms. In summary, our results generally support hypothesis (1) that temperate forest soils formed from carbonate bedrock differ from those formed from silicate bedrock regarding P stocks and P speciation: They are characterized by smaller total P stocks, predominance of Ca-bound organic P. The circumstance, that temperate forest soils on carbonate bedrock with progressing pedogenesis (e.g. BAE) become more similar to silicate soils with respect to their P status, does not really refute hypothesis (1), because old soils, such as BAE, are rare in temperate regions (e.g. of Europe and North America), where old soils except for few small spots at most places have been removed by (peri)glacial processes and/or covered with autochthonous material (e.g. Loess) during the Pleiszocene, resulting in a re-start of pedogenesis about 12,000 years ago and often of conversion into mixed-parent material soils. Hypothesis (1) thus may be rephrased as follows: Typical temperate forest soils formed from carbonate bedrock in Central Europe are dissimilar to soils formed from silicate bedrock regarding P stocks and P speciation.
(3) Plant and ecosystem P availability: Stocks of plant-available oPO4 in the carbonate soils were at the level of the P-poorest silicate soils CON and LUE, or even smaller in the MAN N profiles. Additionally, stocks of labile or moderately labile Hedley P fractions were at the level of the P-poorest silicate soils CON and LUE, or even smaller in the MAN N profiles. Low beech foliage P contents (<1.3 mg g -1 ; except for MAN S; Table 1 in main manuscript) also indicate poor ecosystem P availability at all carbonate sites. Moreover, for sites MAN N1, TUT NE, and SCH, where contents of microbial-bound C (Cmic), N (Nmic), and P (Pmic) have been quantified, Cmic/Nmic ratios (5-7) were considerably smaller compared to those in the silicate-bedrock soils (10-16;

Calculation of theoretically expected P accumulation in the Bw horizons of the Cambisols Mangfallgebirge (MAN) S1 and Bärenthal (BAE) by carbonate weathering
Theoretically expected vs. measured P contents in Bw horizons of MAN S1 and BAE At sites MAN S1 and BAE the bedrock has a Ca/Mg carbonate (dolomite) content of 950 mg g -1 and a P content of 0.15 mg g -1 (Table S1). Thus, rock weathering, assuming complete dissolution of its dolomite fraction and subsequent removal of the carbonate dissolution products Ca 2+ , (Mg 2+ ; MAN) and HCO3with the soil seepage water, and assuming that the non-carbonate weathering products remain in the soil as "carbonate dissolution residue", results in a remaining carbonate solution residue mass of 5 g from originally 100 g bedrock, or of 0.05 g per 1 g of original dolomite bedrock, corresponding to a reduction in soil volume by a factor of 20. The expected P content of the carbonate dissolution residues in soils MAN S1 and BAE, assuming absence of P loss during carbonate weathering, is thus concentrated by a maximum of up to a factor of 20, thus amounting to 20 * 0.15 mg P g -1 = 3 mg P g -1 . This is 7.5 times as much as the measured P content of the Bw horizon (0.4 mg g -1 ) in profile MAN S1, and 6 times as much as the measured P content of the Bw horizon (0.5 mg g -1 ) in profile BAE, i.e. there was considerable P uptake and loss from these horizons.

Theoretically expected vs. measured fine earth P stocks in Bw horizons of MAN S1 and BAE
At sites MAN S1 and BAE the parent material has a Ca carbonate (calcite) content of 950 mg g -1 and a P content of 0.15 mg g -1 (Table S1). The densities of dolomite (bedrock of MAN S1) and calcite (parent material of BAE) are 2.9 g cm -3 and 2.7 g cm -3 , respectively. The density of accessory silicates (5% of total rock mass at both sites) is 2.75 g cm -3 and thus similar to carbonate densities. Total thicknesses of the Bk/Bw horizons of profiles MAN S1 and BAE are 15 cm and 20 (BwAh) +16 (Bw) = 36 cm, respectively (Table 1). Fine earth bulk densities, corrected for stone content, in the Bk/Bw horizons of MAN S1 and BAE are 1.87 and 1.1 g cm -3 (Stahr and Böcker 2014), respectively.
At site MAN S1, dissolution of 1 m carbonate rock, using the data reported above would result in a Bk horizon consisting of carbonate dissolution residue with a thickness of (100-95) * 2.9/1.87 = 5 * 2.9/1.87 cm = 7.75 cm.
Thus, to form a Bk horizon with a thickness of 15 cm (which is present at MAN S1), a column of 15/7.75 * 1 m = 1.93 m dolomite rock must have been weathered. This is a conservative estimate, assuming no loss of carbonate dissolution residue in colloidal or dissolved form with the seepage water. At a dolomite density of 2.9 g cm -3 , this corresponds (slope effects on horizontal surface area projection being neglected) to a carbonate rock mass of 2.9 [g cm -3 ] * 193 [cm] [1 cm -2 ] per cm 2 soil surface = 560 g cm -2 = 5,600 kg m -2 . The amount of P contained in that rock mass (at a P content of 0.15 mg g -1 ) is 0.15 * 10 -3 * 5,600 kg m -2 = 0.84 kg P m -2 or 8,400 kg P ha -1 . However, the total soil P stock of MAN S1 is only 4,000 kg ha -1 , and the P stock in the subsoil (Bk) is even only about 1000 kg ha -1 (Figure 2). Thus, if one reasonably assumes that the P content of the rock whose residues have produced the existing Bw horizon is similar to the P content of the rock analyzed in our study, the majority of the P amount released from the parent rock during carbonate dissolution must have been lost during pedogenesis leading to the formation of the Bk horizon at MAN S1. Probable reasons for that loss are discussed below.
At site BAE, dissolution of 1 m carbonate rock, using the data reported above would result in a Bw horizon consisting of carbonate dissolution residue with a thickness of (100-95) * 2.7/1.1 cm = 5 * 2.7/1. ] per cm 2 soil surface = 792 g cm -2 = 7,920 kg m -2 . The amount of P contained in that rock mass (at a P content of 0.15 mg g -1 ) is 0.15 * 10 -3 * 7,920 kg m -2 = 11.9 kg P m -2 or 11,900 kg P ha -1 . However, the total soil P stock of BAE is only 3,200 kg ha -1 , and the P stock in the subsoil (Bw) is even only about 2,400 kg ha -1 . (Figure 2). Thus, assuming that the P content of the parent rock whose residues have produced the existing Bw horizons is similar to the P content of the rock analyzed in our study, >70% of the P amount released from the parent rock during carbonate dissolution must have been lost during the pedogenesis that lead to the formation of the Bv horizons at BAE.

The Ecosystem Phosphorus Nutrition Index -Application, limitations, perspectives
The ENIP approach seems to be useful for a comprehensive evaluation and ranking of temperate beech forest sites on silicate and carbonate rock regarding their ecosystem P nutrition strategy, and even allowed the development of a new conceptual model ( Figure 10). Yet, it must be emphasized that the ENIP values calculated for our study sites are interval-scaled rather than ratio-scaledthe scale was defined arbitrarily (even though based on good reasons) by the end-members BBR and LUE which had been identified as "most P acquiring" and "most P recycling" silicate sites, respectively, by Lang et al. (2017). Application of that scale to the carbonate sites in our study generally yielded reasonable results. Yet, ENIPs down to -7.1 for site MAN S2 considerably extended the original scale (ranging from -1 to +1) in the negative (i.e. P-recycling) direction. As mentioned before, data on the P-recycling indicator N6 were only available for two of the eight carbonate sites, and data on the P-acquiring indicators N2 and N3 were only available for four carbonate sites (Table 7). These data deficits theoretically may have confounded the ENIPs calculated for the respective sites. We therefore repeated our ENIP calculations for all sites, omitting N6 (Table S7), or N2, N3, and N6 (Table S8). The ENIPs calculated without inclusion of N6 were very similar to those calculated with inclusion of N6: Without inclusion, profiles MAN N1 and N2 both had ENIPs of -2.0, and an ENIP for TUT NE of -0.7 instead of -0.5 indicated a somewhat larger influence of P-recycling ecosystem nutrition (Table S7). Also, the ENIPs calculated without inclusion of N2, N3, and N6 (Table S8)  of that site to a dominating P acquiring nutrition strategy (Table 7) instead of more likely coexistence of P-acquiring and P-recycling strategies. Nevertheless, site BAE with its oldest soil according to our different calculations always was the least P-recycling and most P-acquiring forest of all carbonate sites. In this context, it must be pointed out that the ENIP concept in its present state provides a ranking tool rather than a quantitative assessment of ecosystem P nutrition. Thus, it surely cannot be claimed that MAN S2 is a seven times more P-recycling ecosystem than LUE.
The ENIP concept itself as well as the variables it is currently based on surely represent the first stage of a longer development process. Concept and variables used have to be checked rigorously for additional soils on silicate and carbonate parent material in future studies, and improved, if necessary. Moreover, the applicability of the concept for other tree species and climate regions has to be tested. Yet, we suggest that our concept to integrate a large set of variables describing ecosystem P nutrition (acquisition, recycling) strategies into one well-defined ENIP for comparing forest ecosystems with respect to their dominating P nutrition strategy is a promising tool for forest ecology and biogeochemistry.   Table S2: Content of different P species in the investigated soils. Porg and Pinorg determined according to Saunders and Williams (1955). Percentages of Ca-bound P, Al-bound P, and Fe-bound P determined by synchrotron-based XANES spectroscopy. Error of XANES speciation results <10% (Werner and Prietzel, 2017).         Table S5: Contents of microbial C (Cmic), N (Nmic), and P (Pmic), Cmic/Pmic and Cmic/Nmic mass ratios as well as mass ratios of organic C (Corg) over organic P (Porg) and organic N (Norg) in the Ah1 horizons of the soils on silicate parent material described in Lang et al. (2017). Note that Pmic in the soils with silicate parent material in contrast to that in the soils on carbonate parent material has been quantified using the gaseous ("CFE") instead of liquid fumigation ("Resin") method. According to Bergkemper et al. (2016) the amount of Pmic yielded from a given soil with the CFE method is two times that yielded with the resin method from the same soil. Pmic values of the silicate soils printed in italic fonts in brackets were calculated from the respective Pmic value obtained by the CFE method divided by the factor 2 to enable a comparison between silicate and carbonate sites.     Table 3. ND: Not determined. *Normalization to the interval Bad Brückenau -Lüss as described in Method Section. Dolostone and limestone soils are ordered according to their stage of pedogenesis. **N6 is not included in ENIP calculation because data are only available for two of eight carbonate sites.      Figure S1: Citric acid-extractable (plant-available) soil P stocks at temperate beech forest sites on silicate parent material; data from Lang et al. (2017). Shown are absolute and relative contributions of organic and inorganic P. For a detailed description of sites, please read caption of Fig. 2