Rothamsted Repository Download

Understanding the phosphate oxygen isotope ( d 18 O-PO 4 ) composition of bedrock phosphate sources is becoming ever more important, especially in areas of soil research which use this isotope signature as a proxy for biological cycling of phosphorus (P). For many of these studies, obtaining a sample of the source bedrock or applied mineral fertiliser for isotope analysis is impossible; meaning there is now a demand for a comprehensive characterisation of global bedrock d 18 O-PO 4 to support this work. Here we compile d 18 O-PO 4 data from a wide range of global bedrocks, including 56 new values produced as part of this study and a comprehensive overview of those within the previously existing literature. We present d 18 O-PO 4 data from the range of major phosphatic lithologies alongside as much metadata for the samples as could be gathered. Much of the data comes from bedrocks of marine sedimentary origin (< 1 Ma = > +22 % , > 540 Ma = » +12 % ), but we also present data from bedrocks associated with guano (range: +19.5 to +15 % ) and igneous deposits (range: +12 to –0.8 % ), both of which have distinct d 18 O-PO 4 signatures due to their formation mechanisms. We show that where repeat measure-ments of the same formation have been undertaken, regardless of method or exact sample location, there is an average within formation error of – 1.25 % . This is important, as is constitutes a reasonable level of uncertainty for phosphorus cycling studies which need to estimate bedrock d 18 O-PO 4 composition based on the literature. In combination, this data set presents 284 d 18 O-PO 4 values from 56 countries; a comprehensive starting point for researchers interested in understanding bedrock end member d 18 O-PO 4 .


Introduction
Bedrock phosphate (PO 4 ) is the major source of mineral phosphorus (P) within most soil systems, either as the natural weathered parent material underlying the soil or in the form of an applied mineral fertiliser (Ruttenberg, 2003). The nature and amount of P in the bedrock, as well as the bedrock material itself, varies (Van Kauwenbergh, 2010), influencing soil P transformations during pedogenesis on a range of scales (Walker and Syers, 1976). It is therefore critically important to understand this endmember to the soil P system, not least for studies using the oxygen isotope signature of PO 4 (δ 18 O-PO 4 ) to ascertain the extent and mechanisms of P uptake and cycling within the soil-plant system (Bauke, 2021).
Bedrock phosphate (PO 4 ) deposits can be loosely grouped into three major categories based upon their formation mechanism; these are (1) marine sedimentary, (2) guano and (3) igneous. A fourth category, including bedrocks derived from the weathered remnants of the former categories, could also be classified Orris, 2002a, 2002b;Szilas, 2002). (1) Marine sedimentary PO 4 is derived from the upwelling of P rich organic ocean sediments to the continental shelf and the release of P during the bacterial breakdown of these compounds (Goldhammer et al., 2011). The subsequent precipitation and reworking of P minerals, such as francolite (carbonate fluorapatite), forms highly enriched phosphorite beds, containing pellets and nodules of francolite (Hiatt and Budd, 2001;Shemesh et al., 1988). (2) Phosphorus rich guano deposits are comparatively small accumulations of excreta associated with large colonies of birds or bats. The breakdown of fresh guano by microorganisms produces a number of highly soluble P precipitates (Szilas, 2002). These deposits are vulnerable to secondary weathering either leaving surface deposits or leaching down into underlying bedrock and precipitating P rich carbonate apatites (Ayliffe et al., 1992;Szilas, 2002). (3) Igneous P rich deposits are normally found as sheets of intrusive alkaline crystalline rock associated with 2021 British Geological Survey UKRI. Journal of Plant Nutrition and Soil Science published by Wiley-VCH GmbH www.plant-soil.com carbonatites. These form during the crystallisation of high temperature calcic alkaline silicate melts (Pufahl and Groat, 2017;Szilas, 2002). Each of these different formation mechanisms leaves an oxygen isotope fingerprint on the PO 4 molecule, related to inorganic and organic processes involved during the formation of the P mineral phase.
The use of phosphate oxygen isotope (δ 18 O-PO 4 ) signatures in environmental research is rapidly developing as a robust solution for environmental tracing of the P cycle (Bauke, 2021). Much recent work has focused on P uptake within soils and subsequent nutrient transfer to the biosphere, where P is one of the fundamental elements for life (Davies et al., 2014;Tamburini et al., 2014;George et al., 2017;Pistocchi et al., 2018), but also a major environmental pollutant especially when it enters waterways, acting to reduce natural P limitation and enhance eutrophication effects . To fully understand P cycling and pollution dynamics it is important to have a reliable, natural, non-radiogenic tracer for the modern P cycle. Phosphorus only has one stable isotope, 31 P, rendering traditional stable isotope tracing impossible and leaving our understanding of environmental P cycling far behind that of carbon and nitrogen . In recent years, a number of researchers have exploited the fact that most P in the environment is bound to oxygen (O). Looking at the δ 18 O of different P compounds and/or pools can help elucidate the primary origin, and secondary physical and biological processes acting upon this molecule Tamburini et al., 2012). In this way the δ 18 O signature of different P pools has been used to try and assess the extent, rate and controls over P cycling between these pools and uptake/release from biota (Helfenstein et al., 2018), as well as to try and trace PO 4 sources and potential pollutants, including agricultural fertilisers and wastewater discharges (Gooddy et al., 2016;Granger et al., 2017Granger et al., , 2018. The δ 18 O-PO 4 method relies upon the fact that P is mainly associated to O in the environment and that the P-O bond is normally very stable in the absence of biological activity (O'Neil et al., 2003). The δ 18 O-PO 4 can be altered due to a sorting of heavier or lighter isotopologues. Sorting can occur during abiotic processes, however, the effects of these abiotic processes on the δ 18 O-PO 4 are still under discussion and appear to cause minimal alteration of the δ 18 O-PO 4 signature Jaisi, 2013;Li and Jaisi, 2015). Biological uptake of P can also result in an isotopic effect due to a preference of microorganisms for the lighter isotopologue, so far, mainly negative fractionation factors have been reported. Phosphorus uptake by the bacteria Escherichia coli, for example, has a fractionation of about -3% . Whilst isotope effects associated with P uptake do lead to small changes in the δ 18 O-PO 4 , more pronounced shifts in δ 18 O-PO 4 are observed due to enzyme cleavage of the P-O bond . The enzymes, which cleave the P-O bond, can be divided into two groups: phosphomono-and diesterases which lead to an exchange of 1-2 O atoms with water from the surrounding environment, and the intracellular enzyme inorganic pyrophosphatase, which leads to the complete exchange of all four O atoms in phosphate . Phosphomono-and -diester-ases catalyse the hydrolysis of organic P into PO 4 and the δ 18 O-PO 4 of the released PO 4 is typically depleted in 18 O. These enzymes include RNase, 5 -nucleotidase, DNase, alkaline and acid phosphatase and phytases, with fractionation varying between -30 and +20% Blake, 2009, 2006;Von Sperber et al., 2014). Intracellular enzymatic hydrolysis by inorganic pyrophosphatase (Cohn, 1957) leads to equilibrium fractionation defined by Eq. (1) (Pistocchi et al., 2017), rearranged from Chang and Blake (2015): where (1)]. But to be able to draw any concrete conclusions regarding P cycling we must also know the δ 18 O-PO 4 value of the other endmembers of the system including importantly the δ 18 O-PO 4 of inorganic and organic P sources (George et al., 2017;Granger et al., 2018, Bauke;2021). The inorganic P may in natural systems comprise simply local bedrock P, but in agricultural settings this may include anthropogenically applied, exogenous, sources in the form of fertilisers. These inorganic fertilisers are sourced globally, from high P containing geological deposits. International scale fertiliser operations often target larger marine sedimentary and igneous deposits, while high P guano deposits are more commonly utilised locally, due to economies of scale (Ayliffe et al., 1992;Mizota et al., 1992;Szilas, 2002;. It is important that studies undertaking source apportionment obtain samples of those endmembers, however, this is often not possible. In these cases, only knowledge of the geology of the underlying bedrock or applied inorganic (and organic) P may be available. For this reason, a wide-reaching understanding of the δ 18 O-PO 4 signature of a range of phosphatic lithologies of different ages, locations and formation histories would prove extremely useful when trying to establish a representative endmember value for uptake into the P cycle.
Here we present the most comprehensive archive of δ 18 O-PO 4 data to date, collected both from existing studies and our own rock archive. We present as much formation, age, mineralogy and location information as could be obtained for the three major phosphatic rock types (Tabs. S1-3). We use these data to highlight trends within the data set, which will help with correct identification of endmember δ 18 O-PO 4 values over a wide range of lithologies and locations.

Materials
This archive of δ 18 O-PO 4 data from a range of global phosphatic lithologies is derived from data collated from numerous papers as well as 56 new samples analysed as part of this project. These additional samples were kindly provided from the collections of Don Appleton at the British Geological Survey (BGS), John McArthur (University College London), and the Rothamsted Research (Harpenden, UK) sample archive. Metadata for old samples and often those within the literature was at times sparse. Where possible, we used the available location information from the original papers and supplemented this using the comprehensive global phosphate data set of Orris, 2002a, 2002b). This archive allows us to provide approximate location, age and mineralogy for many of the samples (Tabs. S1-3). These additional data should prove invaluable for researchers interested in understanding geographic distribution of the δ 18 O-PO 4 values. It should be noted that the grid references provided for sample locations are often based on our best knowledge, not a recorded sampling location, and these location data may relate to a local town or region where the exact sampling location is unknown. For some of the oldest references very minimal metadata were supplied; in these cases we have presented only the approximate location (country), age and δ 18 O-PO 4 values as presented in the original articles.

Methods
Over the last decades the preparation and analysis of samples for δ 18 O-PO 4 has changed due to improved preparation methods and a switch from conversion to BiPO 4 and subsequent fluorination to the production of silver phosphate (Ag 3 PO 4 ) and high temperature elemental analysis (Vennemann et al., 2002). These changes have been designed to both reduce the level of organic P available for hydrolysis (Tamburini et al., 2010) and enable the tracing of any hydrolysis through an isotope labelling technique. These changes have resulted in the robust extraction protocol used below, eliminating organic P, though a series of precipitation and dissolution stages and reducing NO 3 -O isotope exchange by using hydrochloric rather than nitric acid extractions. There have also been changes in the reported value for NBS120c, which has been used to standardise much of the δ 18 O-PO 4 data, especially in the marine palaeoclimate community (Pucé at et al., 2010). It is possible that this change in analytical methodology has led to a change in the recorded δ 18 O-PO 4 ( +2.2%), and this should be considered when comparing data across methods (Pucé at et al., 2010). However, rather than attempting to apply a correction across our data set, we have reported the raw published data and the reference from which this data was obtained (Tabs. S1-3).
For comprehensive analytical methodologies we direct the reader to the original articles and to Vennemann et al. (2002) for a review of the methods, and Pucé at et al. (2010) for discussion of potential isotope offsets between methods.
Below, we briefly describe the analytical procedure followed for the purification and analysis of samples which were part of this study. Samples were purified in two separate labs, Rothamsted Research, UK, and at the National Environmental Iso-tope Facility (NEIF), British Geological Survey (BGS), following the same procedure which broadly follows Tamburini et al. (2010). All isotope analysis was undertaken at the NEIF facility BGS, UK.

Purification and production of Ag 3 PO 4
Bedrock samples were hand ground and subsampled. Approximately 0.5 g of material were added to 50 mL of 1 M HCl and shaken at room temperature for 16 h before filtering through a Whatman filter (grade GF/F) to remove any undissolved residue. The supernatant was retained for concentration and isotope analysis. From here we followed a modified Tamburini et al. (2010) method, but omitted the DAX-8 resin step as organic contamination was minimal in rock samples (Granger et al., 2018). Briefly, dissolved PO 4 was first precipitated as ammonium phospho-molybdate (APM), ensuring a pH of 1 with the addition of 1 mL of concentrated sulphuric acid. The APM was filtered and then dissolved in NH 4 -citrate solution before PO 4 was precipitated as magnesium ammonium phosphate (MAP). The MAP precipitate was filtered and dissolved in 0.5 M HNO 3 and shaken with cation resin (AG 50 8X). The final precipitation stage was achieved by adding 7 mL of Ag-ammine solution to precipitate Ag 3 PO 4 . Precipitation was left to occur slowly at 40°C for at least 16 h, the precipitate was filtered before being oven-dried at 50°C.

Mass spectrometry
The analysis of Ag 3 PO 4 was undertaken by weighing approximately 300 μg of Ag 3 PO 4 into a silver capsule, in triplicate. The sample was then converted to carbon monoxide (CO) using a thermal conversion elemental analyser (TC-EA, Ther-moFinnigan, Germany) at 1,400°C in the presence of glassy carbon. The product CO mixed with a helium carrier gas and was analysed on a Delta + XL mass spectrometer (Thermo-Finnigan, Germany). All analysis was undertaken at the BGS. The δ 18 O-PO 4 values were calculated by comparison to an internal Ag 3 PO 4 standard, ALFA-1 (ALFA-1 = δ 18 O VSMOW value of +14.2%). In the absence of an international Ag 3 PO 4 reference material, the value for ALFA-1 is derived by comparison to the Ag 3 PO 4 standard 'B2207' (Elemental Microanalysis Ltd., England), which has been measured in an inter-laboratory comparison study to have a δ 18 O value of +21.7% versus VSMOW. Triplicate repeats had a typical precision σ 0.4%; furthermore, sample purity was assessed by determining the CO yield compared with the yield of Ag 3 PO 4 standards, and rejecting samples where this differed by 10% (Tamburini et al., 2010).

Global distribution of samples
Within the presented data set we have samples from many of the major global PO 4 reserves including several sites which were/are responsible for the supply of bedrock PO 4 for commercial purposes. We also compile the currently available δ 18 O-PO 4 dataset for sedimentary, igneous and guano derived bedrocks produced for purely scientific interest, creating a highly diverse, data set (Tabs. S1-3). Figure 1 displays the sample locations and global distribution of the bedrocks in this study, for which good quality metadata are provided (sampling location). In total we include 283 samples from 56 countries in the supplementary data tables.
Whilst this data set is a comprehensive overview of available δ 18 O-PO 4 data, it only covers a relatively small number of the global phosphatic lithologies detailed in the Orris (2002a, 2002b) mineral reserves list, which exceeds 1600 sites, if you include those derived from or related to guano deposits. This highlights a major dearth of available isotope data in comparison to the number of PO 4 reserves deemed significant enough to make the mineral reserve list. More work is needed by the community to try and develop the isotope data set. This will allow a more thorough assessment of the relationships between deposit formation mechanism, preservation history, location and isotope value and critically will enable a more reasoned assessment of potential endmember values for the soil δ 18 O-PO 4 system.
For ease of discussion, we break down the data set into the three common formation mechanisms for phosphatic lithologies rather than geographic context.

Sedimentary bedrock isotope values
The largest set of δ 18 O-PO 4 data comes from sedimentary bedrock sources with 179 samples compiled from existing literature and 32 new values from this study (Fig. 2). The data produced as part of this study include samples from 14 different countries (Tab. S1). These samples span 535 million years from the youngest, associated with the Varswater Orebody in South Africa Lower (Pliocene, 5 million years), to a sample from the Indian Lower Tal Formation (Proterozoic-Cambrian, 540 million years). The δ 18 O-PO 4 values range from +24.8% (Tanzania) to +10.1% (India). Our samples follow a distinctive trend with younger samples having higher isotope values than older ones (Fig. 2, Tab. S1). When we include sedimentary samples from the literature, this data set spans the widest geological range of all the rock types. Five extremely old samples presented in Blake et al. (2010) originate from the Onverwacht Group (South Africa) and are of Archean age (approx. 3.5 billion years), but the majority of the sedimentary data set spans from the Lower Proterozoic (approx. 2000 million years) to modern Holocene age sediments (Tab. S1, Fig. 2). All samples have, where possible, been classified by their geology and formation name in Tab. S1, following either the information with the original samples or based on the mineral reserves list Orris, 2002a, 2002b). More modern sediments have characteristically higher isotope values (+16 to +25%) than those more ancient deposits (+10 to +17%) (Fig. 2). There are a few exceptions to this rule with lower isotope values being reported for modern (100-0.1 Ma) sediments. These are mostly represented in one of the oldest archives, that of Longinelli and Nuti (1968), an archive which Pucé at et al. (2010) suggests should be corrected when comparing to modern analytical methods. A shift of +2.2% would align this data set far more closely with other similar age marine sedimentary values.
Several of the older data sets, including that of Shemesh et al. (1988), were originally compiled to test the application of δ 18 O-PO 4 to precise ocean palaeothermometry. This dataset built on the conclusions of Longinelli and Nuti (1968) and Shemesh et al. (1983), who found a trend of decreasing δ 18 O-PO 4 with increasing age (Fig. 2). Shemesh et al. (1983) suggest that the general trend for lower δ 18 O-PO 4 values with age is driven by an equilibrium between original formation 2021 British Geological Survey UKRI. Journal of Plant Nutrition and Soil Science published by Wiley-VCH GmbH www.plant-soil.com   (Colman et al., 2005).
However, Shemesh et al. (1988) do caveat that some bedrocks exhibit clear isotopic disequilibrium or secondary alteration, leading to calculated formation temperatures far in excess of the natural surface conditions (> 100°C). They attribute these findings to the secondary breakdown of the bedrock mineral by biota, with microbially driven O exchange (via an unidentified mechanism), influencing the original δ 18 O-PO 4 value and rendering the palaeo-thermometry technique void for some samples and locations (Shemesh et al., 1988). Goldhammer et al. (2011) show that under modern settings actively depositing marine P can be out of equilibrium with surrounding waters, this disequilibrium is not a secondary, post deposition, mechanism but controlled strongly by varying microbial uptake strategies within modern deposition settings. Studies have also shown that there is the potential for wide variations in δ 18 O-PO 4 value over a single geological formation related to that formations geomorphology. Hiatt and Budd's (2001) study of the Upper Permian Phosphoria Formation in the USA highlights the starkly different depositional environments of a single phosphorite formation, and whilst the δ 18 O-PO 4 value proved useful in characterising these for-mation environments the authors describe a range of δ 18 O-PO 4 values from +13.7 to +20.2%.
The clear heterogeneity across formation and the potential for disequilibrium at the point of formation means that trying to use the sediment age to attribute a δ 18 O-PO 4 to sedimentary bedrocks is unwise. Whilst general trends can be identified (Fig. 2), it is critically important researchers aim characterise a local bedrock endmember δ 18 O-PO 4 value before applying the δ 18 O-PO 4 methodology for its application as a tracer of the soil P system.

Guano deposit isotope values
Twenty eight of the deposits described here, including those from Ayliffe et al. (1992) and new samples (our archive) from islands in the Indian and Pacific Oceans, are classified as 'bird guano' derived (Tab. S2). This classification is descriptive of a depositional environment where there has potentially been chemical reactions between bird guano and carbonate during the development of these phosphorites, including guano derived P into the mineral phase (Aharon and Veeh, 1984). Guano deposits can roughly be split into two categories: (1) those younger (< 200,000 years) surface deposits recently formed locally from guano leaching, including phosphatic calcarenites which consist of carbonate sand and conglomerate cements with apatite, or (2)   leaching of guano derived P (often following the dissolution of category one phosphate minerals) into the underlying bedrock has occurred, followed by secondary precipitation (Ayliffe et al., 1992;Szilas, 2002). Where any one deposit may lie in this categorisation is often complex to interpret and needs site specific study, and dating of these deposits is known to be complex (Ayliffe et al., 1992). It should be noted that, whilst we subdivide the guano data set into modern and ancient, these bird guano derived bedrocks are all relatively modern (most < 1 Ma) in comparison to sedimentary and igneous rocks.
Here we present 13 new guano derived deposits in this study (Tab. S2). All these deposits are, to the best of our knowledge, from young formations (Holocene; 11,600 -0 years BP) and have a range of δ 18 O-PO 4 +14.7 to +19.6%. All the other modern samples listed here are from Ayliffe et al. (1992) and have δ 18 O-PO 4 values ranging between +20.1 and +22.8% similar to, or with slightly higher isotope values (1-2%) than the modern bird guano collected at the same sites (+19.8 to +23.1%) and associated with their formation (Ayliffe et al., 1992). Ayliffe et al. (1992) suggest that the slight (if not significant) difference between the modern guano and the local mineral precipitates is a result of an isotopic fractionation associated with the transformation of bird guano to apatite. This transformation is associated with microbial uptake, enzyme breakdown and cycling of P, which leaves the mineral precipitate δ 18 O-PO 4 in near temperature dependent Eδ 18 O-PO 4 with evaporative surface/soil waters [Eq. (1)]. More ancient island phosphorites, such as those found in emergent atolls, appear to have δ 18 O-PO 4 values that are lower (+14.7 to +19.1%) than the values of their assumed source guano material or local surface minerals (Ayliffe et al., 1992). The most convincing explanation for this lower isotope value being preserved in ancient apatite is the within karst enzymatic breakdown of guano derived P, leading to Eδ 18 O-PO 4 with meteoric waters within the karst. These waters have a lower δ 18 O value than that of the evaporatively enriched soils, leading to a measurable change in δ 18 O-PO 4 in these deposits in comparison to either a bird guano or primary mineral source.
We also provide data from six cave guano deposits associated with bat colonies (Tab. S2). One of our sample set is a cave limestone (Java, Indonesia) derived from bat guano (+17.2%), this sample is thought to be older (66 Ma) than the rest of the guano data set although dating is complex. The only comprehensive work we could find on this type of deposit is that of Chang et al. (2010) who describe two types of apatite deposits in South Korean caves. The first deposit type described by Chang et al. (2010) was an autogenic apatite formed within sediments directly below large bat guano deposits (U2-sed; Tab. S2). These apatite formations have similar δ 18 O-PO 4 values (+21.9%) to that of the original guano material (+21.4%), suggesting that these deposits were not significantly metabolised by microorganisms prior to deposition, potentially due to the high concentration of P available from the large guano deposits and the direct formation of apatite from leaching guano. The second type of deposit, hydroxylapatite crusts, are isotopically much lighter (+14.6 to +15.6%; Tab. S2) than the original guano material, attributed to the cycling of P by microorganisms and enzyme mediated O exchange, the same as the mechanism suggested by Ayliffe et al. (1992). Based on cave water d 18 O H 2 O and annual mean temperature, Chang et al. (2010) suggest that these crusts are at, or very near, expected Eδ 18 O-PO 4 , suggesting intracellular P cycling by pyrophosphatase. They hypothesise that the difference in formation mechanism between these crusts and apatites formed directly below guano deposits is due to the lower availability/limitation of P in the crusts as these are formed away from the concentrated guano deposits from bat urea. Where guano derived P is limited, this is all cycled by microbial communities, leading to the distinctive Eδ 18 O-PO 4 end member, 6% lower than the original guano .
Due to the conditions of their formation the δ 18 O-PO 4 of mineral deposits associated with both bird and bat guano samples are often influenced by microbiological cycling of P and the subsequent deposition of inorganic P. It is likely that the high prevalence of microbial biota in guano is responsible for the production of enzymes capable of facilitating the rapid breakdown of the P-O bond and the re-equilibrium with surrounding water . This abundance of microbes and the high availability of organic P in the guano clearly promotes the deposition of inorganic P rich substrates, which have undergone significant isotope exchange or complete equilibrium with surrounding waters, re-setting the original guano δ 18 O-PO 4 signature. Therefore, those wanting to estimate an end member value for a guano derived phosphorite (which is often a local or regional source of fertiliser), understanding the age, depositional history and setting of deposition is key.

Igneous bedrock isotope values
Samples of igneous origin (39 total) have some of the most distinctive δ 18 O-PO 4 values in the dataset, irrespective of age. These formations have a range much lower than the majority of sedimentary and guano deposits. Low δ 18 O-PO 4 values are a result of these rocks high temperature formation mechanism, where higher temperatures have been shown to result in a negative isotope shift (Lecuyer et al., 1999). Samples from our archive include the lowest recorded bedrock δ 18 O-PO 4 value reported in any of the literature, at -0.8%. This sample was from the Tundulu complex in Malawi and is of Late Jurassic-Early Cretaceous age (Tab. S3), our archive also includes several samples from Panama, Uganda, and from Kola mine in Russia. These new data span the full range of δ 18 O-PO 4 , previously reported within the literature, -0.8 to +12.1% (Fig. 3, Tab. S3). However, until this study, the best single combined archive of igneous rock δ 18 O-PO 4 values was provided in Mizota et al. (1992). Their study compiled the first clear archive of igneous values to act as a reference to their work focussed on understanding the relative contribution of P sources within soils of volcanic origin in Java, Indonesia and the African Rift Valley. These reference samples were collected globally and included volcanic ash, carbonatite and hydrothermal geodes (Tab. S3). Tamburini  (1)]. As igneous bedrock values are much lower than those of guano and sedimentary deposits (Fig. 3), this divergence is more prominent and P cycling more apparent when it occurs. Shen et al. (2020) highlight that fractionations associated with DNA and RNA hydrolysis (-25 and -5%, respectively), alkaline and acid phosphatases (approx. -25%), and 5'-nucleotidase hydrolysis (-10%) all cause a negative shift in isotope values (Liang and Blake, 2006), whilst inorganic pyrophosphatase will cause a positive shift based on equilibrium at most modern temperature and d 18 O H 2 O combinations. Therefore, highly positive isotope values (+20 to +25%), which have been significantly shifted away from an igneous bedrock endmember can only reasonably be attributed to the microbial cycling of P via inorganic pyrophosphatase (Shen et al., 2020).

Inter study comparison
This combined data set facilitates the opportunity to assess the homogeneity of published isotope data at geological structures, where several samples have been independently collected and analysed. For some locations we have samples from the same geological structure but that have been collected, stored and prepared at different times and in different places, as well as being analysed by different researchers, following different methodologies (Tabs. S1-3). Comparisons across these data sets give a far more robust idea of the expected heterogeneity across geological structures and methods. Based on the 10 sites where cross study comparison was possible (Tab. 1), an average 1stdv of 1.25% appears to be representative with a maximum recorded error at any one site of 3.78% (Minjingu, n = 4). We therefore suggest that workers who wish to attribute an end member value based on this data set or secondary literature, consider this level of error as a reasonable starting point, where direct analysis of δ 18 O-PO 4 endmembers is not possible.
These data consider all possible forms of error, from heterogeneity across the geological structure, sampling inconstancies, purification and analytical error. It should be highlighted that much of this error could be associated with heterogeneity within the geological structure as demonstrated by (Hiatt and Budd, 2001) at the Phosphoria formation, USA. The wide range of isotope values presented in their study relate to differing depositional environments across this large geological deposit and are therefore to be expected. The other site with significant differences in δ 18 O-PO 4 is the Minjingu guano deposit in Tanzania. These samples were included in Tab. 1 to highlight the potential for isotopic heterogeneity in guano derived formations, possibly due both to variable formation conditions over 1000's of years and the unquantified extent of isotope exchange during leaching. Most of the formations, however, are either less geologically massive than the Phosphoria or less complex in formation history than Minjingu. Where these have been sampled over multiple studies, we generally observe a reduced heterogeneity (Tab. 1).  Considering that many of these studies follow divergent sampling, purification and analysis protocols, such a level of homogeneity is promising and suggests that unless a site has distinct heterogeneity associated with the original depositional environment that using the isotope value of existing literature with appropriate error consideration can accurately reflect the δ 18 O-PO 4 of that structure. This gives significant confidence to studies using previously published isotope data to constrain endmembers to their δ 18 O-PO 4 system.

Conclusion
Based on a number of new bedrock δ 18 O-PO 4 isotope analyses and a comprehensive compilation of existing literature we identify the following trends in bedrock δ 18 O-PO 4 signature both over time and between formation mechanisms.
. This information will prove invaluable to those wishing to understand bedrock δ 18 O-PO 4 as an endmember to their soil P system. Whilst this is now the most comprehensive archive of geological δ 18 O-PO 4 data and we have been able to compile data from 285 samples representing many globally important reserves, this is still a sparse dataset in comparison with the number of major P deposits found worldwide Orris, 2002a, 2002b). Importantly, the range of geological bedrock δ 18 O-PO 4 values is diverse (+24.8 to -0.8%) meaning that studies may decide to focus their research in areas which will provide a bedrock δ 18 O-PO 4 starting point for maximum potential change due to P cycling. When aiming to characterise the bedrock δ 18 O-PO 4 for soil P cycling investigations we strongly suggest the following working procedure: (1) undertake a primary characterisation of endmember δ 18 O-PO 4 at the study site; (2) where this is not possible, but data from the same geological formation has been published, use this, incorporating a working error of 1.25% associated with within formation heterogeneity; (3) if no data are available for the specific geological formation of interest, consider its formation mechanism and age, this should provide a robust starting point for δ 18 O-PO 4 estimation. Following this procedure should enable the most accurate method for endmember estimation. Whilst this data set is a good start, much further work is still required by the community to characterise the δ 18 O-PO 4 values of major global P reserves and the effect of converting mineral P to fertilisers through chemical treatment, which may impart a hitherto unknown isotopic effect on the source P mineral.