1.1 Background

Phosphorus (P) is an essential nutrient for all living organisms as it is part of many biochemical compounds like DNA and ATP. Available P levels in the soil are, however, often limiting crop production and farmers need to apply P fertilizer. The primary source for P fertilizer is rock phosphate, which is widely distributed throughout the world both geographically and geologically with USA, China, Morocco and Western Sahara and Russia accounting for 72% of the world total (Zapata and Roy 2004). In addition, farmers in many developing countries cannot always afford P fertilizer, but the soils in those countries are often amongst those with the lowest P availability. Besides the positive impacts of P fertilizers increasing crop productivity, P can cause negative impacts on the environment when applied in excess. Algal blooms in aquatic systems due to large input of P are still an issue in many aquatic systems including the Great Lakes in North America (Paytan et al. 2017). We therefore need to better understand P cycling in the environment. Radioisotopes and stable isotopes are a useful tool to investigate nutrient cycling in the environment (Di et al. 1997; Hogberg 1997; Barbour 2007). In regard to P, researchers relied for a long time mainly on the two artificial radioisotopes of P, 32P and 33P, since P only has one stable P isotope (31P) and several radioisotopes (from 26P to 30P and from 32P to 38P). The radioisotope approach is often used to study the uptake of P from different phosphate rocks and fertilizers by plants (Zapata and Roy 2004; Frossard et al. 2011; IAEA 2013, 2016a, b; Nanzer et al. 2014; Nguyen et al. 2017; Scrase et al. 2020; Wolff et al. 2020), the effects of organic manure and inorganic fertilizer application on soil P cycling (Ma et al. 2020) or the residence time of P in different P pools in soils or plants (Mimura et al. 1996; Helfenstein et al. 2018). It is a useful tool to study P cycling, but its drawbacks are the short half-life of the two radioisotopes (14.3 days for 32P and 25.4 days for 33P) and possible health risks, which has led to the restrictions of applying them in the field, and the need for special safety equipment and national regulations to monitor radiation. Obtaining the licence to work with radioisotopes of P that emit high beta (β) radiation will require operational radiation protection and safety procedures, as well as adequate training of personnel to handle these nuclides safely (IAEA 2016a, b). There is therefore a need for a stable isotope approach to investigate P cycling in the environment.

In recent years, various studies have indicated that the analysis of the stable isotopic composition of oxygen (O) bound to P (δ18OP) to better understand P cycling in the environment, has become a promising tracer (proxy) to investigate soil P transformation and to trace the source of P from the soil to water bodies and the environment. The two main underlying facts for using this method to study P cycling are (1) P is mainly bound to O in the environment and (2) the bond between P and O is stable under earth surface conditions and in the absence of biological activity (Winter et al. 1940). The two ways of altering the δ18OP signature are by (a) sorting between heavier and lighter isotopologues or (b) cleaving the P–O bond.

1.1.1 Sorting of Different Isotopologues

The sorting of heavier and lighter isotopologues could happen due to biological P uptake or inorganic processes, like the preferential sorption onto iron oxides (Jaisi et al. 2010). There are only a few studies which investigated the effect of P uptake by biota. Blake et al. (1997) report a fractionation factor of around −3‰ for Escherichia coli. More recently Ferrera et al. (2015) investigated the effect of P uptake by different coral species on the δ18OP. Depending on the coral species, fractionation factors varied between −18.80 and +0.91‰, with the majority of the investigated species having negative fractionation factors. If an uptake effect, however, is visible in environmental systems, is questionable as a significant amount of P (>50%) must be removed from a P pool without being replenished in order to observe an effect on δ18OP values. This seems rather questionable in case of P in the soil solution since it is usually constantly replenished. Another kind of sorting can occur due to inorganic processes like sorption/desorption of P onto iron or aluminium oxides and precipitation/dissolution of P minerals (Jaisi et al. 2010). These processes play an important role in the environmental P cycle (Arai and Sparks 2007), but fractionation factors associated with inorganic processes seem to be rather small and are still debated (Jaisi et al. 2010; Melby et al. 2013).

1.1.2 Cleaving the P–O Bond

Biological activity like the hydrolysis of organic P by enzymes, can cleave the P–O bond and thus lead to an exchange of O between phosphate and water (Blake 2005). This leads to a change of the initial δ18OP value (Table 1.1).

Table 1.1 Examples for the effect of different enzymes on δ18OP values
Full size table

Inorganic pyrophosphatase (PPase), an ubiquitous enzyme catalyses the hydrolysis of pyrophosphate (= diphosphate) into phosphate (Cohn 1958; Blake et al. 1997) and leads to a temperature-dependent equilibrium between O in water and in phosphate. It can be calculated by the equation of Chang and Blake (2015), as rearranged by Pistocchi et al. (2017):

$${\text{EQ}}\delta^{18} {\text{O}}_{{\text{P}}} = - 0.17 \cdot {\text{T}} + 26.5 + \delta^{18} {\text{O}}_{{\text{w}}}$$
(1.1)

where EQδ18OP is the δ18OP value of phosphate at equilibrium with O in the ambient water in ‰,

T is the ambient temperature in °C and δ18Ow is the isotopic composition of O in water in ‰.

What makes the inorganic PPase so important for the δ18OP is that O exchange can not only occur during the hydrolysis of pyrophosphate but also by locking a phosphate molecule into the active site of the enzyme (Cohn 1958; Blake et al. 1997). When the inorganic PPase is involved, the initial δ18OP signature is erased completely (see Table 1.1).

Hydrolytic enzymes like acid and alkaline phosphatases only lead to the exchange of 1–2 oxygen atoms and are temperature-independent (Liang and Blake 2006a; von Sperber et al. 2014, 2015). The associated fractionation factors vary greatly, from +20 to −30‰, and consequently the δ18OP of the released phosphate varies depending on the involved enzymes. Some fractionation factors like the one of phytase and acid and alkaline phosphatase are substrate dependent (von Sperber et al. 2015; Bai et al. 2020).

1.2 From Past to Present

Early 18O studies worked with 18O-enriched samples rather than natural abundance (Winter et al. 1940; Cohn and Hu 1978; Larsen et al. 1989) and were designed to understand the mechanism of the studied enzymes. Natural abundance studies only came later and were not that common, most likely due to the challenges of purifying extracts for δ18OP analysis. At the beginning, bismuth phosphate (BiPO4) was used as analyte (Tudge 1960), however, BiPO4 is highly hydroscopic, making it more difficult to handle and store (Vennemann et al. 2002). Nowadays, Ag3PO4 is used as analyte (Firsching 1961; Crowson et al. 1991), which is easier to handle and faster to precipitate. Advancements have also been made in the analysis of Ag3PO4 (Vennemann et al. 2002). The analysis with a high temperature conversion elemental analyser (TC/EA) coupled to an isotope ratio mass spectrometer (IRMS) is now the most common method.

The origins of the 18O method lie in chemical, biochemical and paleotemperature studies (Tamburini et al. 2014). In the early 2000s, Blake et al. (2001) postulated that the δ18OP method could be used to study biological P cycling. From then onwards, the δ18OP method became more widespread and scientists used it in a diverse range of environmental samples, including lake and marine sediments, salt and fresh water, soils, mineral and organic fertilizers, rocks, dust and plants (Tamburini et al. 2014). Most naturally occurring δ18OP values, reported so far, range between 10 and 30‰, with the exception of igneous rocks, which range between −0.8 and +12‰ (Smith et al. (2021)). For a more detailed overview of δ18OP values in the environment, the reader is referred to Bauke (2020) for δ18OP values in soils; Smith et al. (2021) for δ18OP values from rocks; for aquatic ecosystems, see for example Davies et al. (2014) and Gooddy et al. (2015). General overviews of δ18OP values can, for example, be found in Tamburini et al. (2014) and Jaisi & Blake (2014). Figure 1.1 shows examples for δ18OP values of potential P inputs like animal faeces and mineral P fertilizers into soils and aquatic systems (see Annex  Table 1.2 for references).

Fig. 1.1
figure 1

Examples for δ18OP values of potential phosphorus (P) inputs into soils and aquatic systems. Vegetation Pi = inorganic P in vegetation; vegetation Porg = organic P in vegetation; org. fertilizer = organic fertilizer; min. fertilizer = mineral fertilizer; WWTP = wastewater treatment plant

Full size image

The δ18OP method was also considered ideal to proof life on other planets, for example, on Mars (Greenwood et al. 2003). To be able to analyse the δ18OP from such a diverse range of substrates, several obstacles had to be overcome. The main obstacles for the δ18OP method are low P concentrations and/or high concentrations of interfering compounds like organic matter (Tamburini et al. 2010; Goldhammer et al. 2011b; Nisbeth et al. 2019). Therefore, modern protocols of the δ18OP method involve several purification steps before the final precipitation of silver phosphate (Ag3PO4) which is then used to determine the δ18OP via TC/EA-IRMS. McLaughlin et al. (2004) adapted the method for seawater samples adding a MAGIC step (precipitation of Mg (OH)2) and purifying the samples using cerium phosphate. Tamburini et al. (2010) applied this protocol to soil extracts but found that the Ag3PO4 could be still contaminated with organic matter. It was therefore suggested to use precipitation of ammonium phosphomolybdate and magnesium ammonium phosphate to purify extracts (Kolodny et al. 1983), adding an additional purification step (DAX-resin) at the beginning of the protocol.

Nowadays, the δ18OP method is more and more used to investigate P cycling in the environment, including sediments (Jaisi and Blake 2010; Goldhammer et al. 2011a; Pistocchi et al. 2017; Liu et al. 2019), soils (Amelung et al. 2015; Gross and Angert 2015; Helfenstein et al. 2018; Pistocchi et al. 2020; Siegenthaler et al. 2020; Pfahler et al. 2020) and plants/algae (Pfahler et al. 2013; Mellett et al. 2018). The focus in case of aquatic systems is often on source apportionment (McLaughlin et al. 2006a; Gross et al. 2013; Granger et al. 2017). Other studies characterise the δ18OP of different P inputs into the environment like mineral fertilizers and farm-yard manure (Granger et al. 2018) or investigate the effect of enzymes on δ18OP values (Blake et al. 1998; Liang and Blake 2006b; von Sperber et al. 2014).

1.3 Examples of δ18OP Studies

1.4 Sediments

Goldhammer et al. (2011a) analysed dissolved P extracted from sediment pore water, using marine sediments from the Benguela upwelling system of the coast of Namibia in Africa. They found dissolved P δ18OP values in and out of equilibrium, ranging between 12.8 and 26.6‰, and attributed these values to different P uptake strategies of microorganisms at low and high inorganic P availability.

Investigating river sediments from the Redon River in France, Pistocchi et al. (2017) showed that the δ18OP can potentially be used to trace P sources and study P cycling in river sediments. They also adapted the method by Tamburini et al. (2010) successfully to river sediments. Liu et al. (2019) explored a pre-treatment method for the δ18OP of different P fractions in sediments. They also concluded that the δ18OP can be a promising tool to trace P and study P cycling in freshwater sediments.

1.5 Soils

With the δ18OP approach, Tamburini et al. (2012) investigated the importance of microorganisms for soil P cycling. They extracted available, microbial, vegetation and mineral P and analysed the corresponding δ18OP values from soils taken along a soil chronosequence at the Damma glacier fore field in Switzerland. Regardless of the contribution of vegetation, mineral or P released by organic P hydrolysis, available P δ18OP values were always close to microbial P δ18OP values. This showed, for the first time under field conditions, the importance of microbial P cycling for the available P.

Recently, Bi et al. (2018) combined the δ18OP method with measuring the abundance of genes related to the P cycle like phoX (acid phosphatase) and phoD (alkaline phosphatase D) to investigate P cycling in agricultural soil from the Fengqiu State Key Experimental Station for Ecological Agriculture (Henan, China). Like Tamburini et al. (2012), they measured acid phosphatase activity and additionally also alkaline phosphatase, phosphodiesterase and dehydrogenase activity. They showed the importance of microbial P cycling in agricultural soil with δ18OP values of the more labile P pools (water, sodium bicarbonate and sodium hydroxide extractable inorganic P) tending towards equilibrium.

1.6 Salt and Fresh Water

After McLaughlin et al. (2004) successfully developed a method to analyse the δ18OP of phosphate in seawater samples, Elsbury et al. (2009) applied this method to investigate P cycling and inputs into Lake Erie. Along with water samples from different locations at Lake Erie, they also analysed water samples from seven rivers feeding into Lake Erie. The δ18OP values ranged between 10 and 17‰, with an expected equilibrium value of 14‰. The δ18OP values of the rivers were around 11‰ and could thus be one source of P, however, they did not find a P source with higher δ18OP values (Elsbury et al. 2009). They proposed that P released from sediments might cause the higher δ18OP values found in Lake Erie. Gooddy et al. (2016) used the δ18OP along with δ15N and δ18O of nitrate and δ15N of ammonium to investigate eutrophication. They took water samples from River Beult (Kent, UK) twice within six months and at seven sampling sites along the river. One of their conclusions was that abiotic processes in the river cause the P concentration changes as the isotopic values did not change dramatically along the river.

1.7 Dust

The main goal of the study by Gross et al. (2013) was to trace the input of P into Lake Kinneret in Israel through atmospheric deposition of dust. They analysed the δ18OP values of available P from soil and dust samples and found that in natural soils, the δ18OP values were between 17.4 and 18.2‰ and between 19.3 and 22.1‰ in agricultural soils. Resin P extracted from dust samples had δ18OP values of around 22‰, indicating that agricultural soils were the main source for dust in the region around Lake Kinneret. Following this study, Gross et al. (2016) found that P associated with Saharan dust can be traced to South America.

1.8 Mineral Fertilizers, Manures and Rocks

The δ18OP values of mineral fertilizers range from 6.4 to 25.9‰ (Fig. 1.1). This wide range is most likely caused by the bedrock material used to produce the mineral fertilizers (Gruau et al. 2005), which δ18OP values vary similarly to the ones of mineral fertilizers (Smith et al. 2021). Phosphorus leached from fresh cattle faeces can be an important P input into aquatic systems, but is sometimes hard to quantify (Bond et al. 2014). Granger et al. (2018) investigated how variable faeces δ18OP values are. They collected faeces samples from seven different cows, grazing on different pastures in Devon (UK) and differing in gender, race and age. Values ranged between 13.2 and 15.3‰ without correlation with any of the variables (pastures, gender, race and age), but they were within equilibrium range calculated with the δ18OP of groundwater and cattle body temperature (Granger et al. 2018).

In one of the early δ18OP studies, Mizota et al. (1992) determined the δ18OP of a range of lithogenic material and soil samples from the Great Rift Valley in Africa and Java (Indonesia). Volcanic ashes had δ18OP values between 5.3 and 6.2‰, apatite from hydrothermal deposits had δ18OP values between 2.4 and 12.2‰, and apatite from carbonatites δ18OP values between 0.2 and 10‰. The soil samples were extracted sequentially, first with 2.5% acetic acid and then with 1M ammonium fluoride (NH4F). Their soil δ18OP values ranged between 10.1 and 20.6‰ (acetic acid-extractable P) and between 12.7 and 24.8‰ (NH4F-extractable P). Their conclusion was that available soil P δ18OP values are due to biogenic and volcanic inputs.

1.9 Plants

Inorganic P in plants tends to have higher δ18OP values compared to the soil due to the 18O-enriched plant water compared to the soil water (Pfahler et al. 2013). However, as Pfahler et al. (2017) revealed, P limitation lowers inorganic P δ18OP values in soybean leaves. Using 33P along with measurement of the δ18OP, the authors could also calculate P fluxes to and from different plant parts. Qin et al. (2018) used the δ18OP method to investigate P uptake by maize plants. They found that the δ18OP value of the P applied to the soil and taken up via the roots directly was found in maize shoots, whereas the δ18OP value had changed if P was taken up via arbuscular mycorrhizal fungi (Qin et al. 2018).