When considering the effects of waterlogging on plant δ³⁴S, we refer to the partitioning of different amounts of sulfur in organic and inorganic compounds in the soil and their bioavailability to plants (the plant-available fraction), the isotopic composition of these different compounds (δ³⁴S), and microbial sulfur cycling processes that may incur isotopic fractionations affecting the isotopic composition of the compounds. Around 95% of sulfur in soil is in organic form (Kilham, 1994), derived from biomass of plant, microbial, and animal origin. However, plants are not able to utilize this organic sulfur directly. It is broken down slowly and is only very gradually mineralized and oxidized to sulfate by soil microbes (Chaudhary et al., 2023), with negligible change in δ³⁴S (e.g., Norman et al., 2002), to be replenished by organic sulfur in fresh biomass deposits. Breakdown of organic matter is suppressed if the soil is waterlogged (Dušek et al., 2020). Water also limits oxygen diffusion into waterlogged and flooded soils and sediments, which are often formed on poorly permeable lithologies. Under water, there is decreasing oxygen diffusion with depth and distance from the air-water surface interface. As remaining oxygen is used up by aerobic microbial respirers and plant roots, a steep vertical gradient of oxygen depletion develops, stimulating the activity of different anaerobic microbial communities and progressive electron acceptor depletion (Kilham, 1994). In the anoxic zone, microbial sulfate reducing activity is stimulated where sulfate and organic carbon are available. Within these redox gradients the boundary between well-oxidized conditions near the surface and anoxic conditions at depth can move up or down, for example, in response to water level fluctuations and bioturbation. This can result in dynamic switching between reducing and oxidizing conditions, stimulating rapid microbial cycling of sulfate and sulfide from the soil inorganic sulfur pool through sulfate reduction and sulfide oxidation.

Sulfate reduction in anoxic conditions is associated with a varied and often large isotopic fractionation (c. 3‰ to 68‰, Sim et al., 2011), which results in the production of sulfide with lower δ³⁴S values than those of the original sulfate. The isotope effect associated with microbial sulfate reduction occurs via a series of fractionating, reversible enzymatic steps within the sulfate reducing organism. The magnitude of the sulfate reduction isotope effect in the environment is one of the main controls on wetland plant δ³⁴S. Large isotope effects from wetland sulfur cycling have been reported, for example from peat wetland environments (11‰ to 51‰, Price and Casagrande, 1991), brackish marshland (c. 33‰, Stribling et al., 1998), and saltmarshes (25‰, Carlson and Forrest, 1982; 50‰, Peterson et al., 1986), although further studies are needed, especially from freshwater environments. At near neutral pH levels, sulfide produced from sulfate reduction can accumulate in the soil as hydrogen sulfide (H₂S). Persistent large isotopic offsets between sulfate in water and sulfide in sediments are brought about via the formation of sulfide precipitates in the sediment which prevent the upward diffusion of dissolved sulfide to the surface where oxygen would enable its oxidation to sulfate. Most sulfide precipitates are formed with iron (Fe²⁺), as iron sulfide (FeS) and pyrite (FeS₂), thus iron availability is key to the formation of sulfide precipitates and maintaining the large isotopic offsets found between sulfate in water and sulfide in sediments.

Factors causing variation in the isotope effect of sulfate reduction are not fully understood, and include the ratio of sulfate to organic carbon (a larger effect is found when the sulfate to carbon ratio is high), and non-linear effects from differences in sulfate concentration, temperature, microbial strain, the type of organic carbon, the cell specific sulfate reduction rate, and iron and ammonium availability (Sim et al., 2012, 2023). High groundwater levels, river flooding, heavy rainfall, poor soil permeability and impermeable lithology can cause soil saturation, which stimulates sulfate reduction. Soil waterlogging coupled with sulfate reduction may be a temporary or periodic phenomenon, for example in floodplains and estuaries, or a more permanent phenomenon, for example in marshes and fens. A wide diversity of wetlands are prime sulfate reducing environments, with low δ³⁴S sulfide in soil and sediments.

Sulfur is an essential nutrient for plants. It is taken up via sulfate transporter chemicals in roots, and used to make different sulfur containing compounds including the amino acids cysteine and methionine, the protective antioxidant glutathione, as well as many essential metabolites such as vitamins, enzymes, and compounds used for plant defense and stress response (Hill et al., 2022). If plants take up excess sulfate, it can be stored in cell vacuoles until it is needed, and if sulfate uptake is restricted due to low availability sulfur may be recycled within the plant from glucosinolates (secondary plant metabolites) (Maruyama-Nakashita, 2017).

Sulfate reduction results in sulfide with lowered δ³⁴S. However, most plants are not sulfide tolerant and cannot take up sulfide directly. In order for a lowered δ³⁴S from sulfide to be incorporated into plant biomass, sulfide must first be oxidized to sulfate. Here we use the term “lowered plant δ³⁴S” as a relative term to refer to plant δ³⁴S values that are lower than those expected from plants grown in non-waterlogged or flooded soil with the same geological parent material. For plants which are not sulfide tolerant, sulfide oxidation to sulfate can occur after waterlogged or flooded soil drains, allowing the ingress of air to oxygenate the soil, for example on seasonally draining floodplains. This enables the oxidizing microbial community to become active resulting in sulfide oxidation to sulfate, utilizing sulfide produced during the preceding period of submergence. Soil sulfide oxidation usually causes negligible isotopic fractionation (Canfield, 2001; Zerkle et al., 2009) so the newly oxidized sulfate retains the lowered δ³⁴S from sulfide. This effect is enhanced by differences in the solubility of sulfate and sulfide. Sulfate is highly soluble and some soil sulfate is removed in water as it drains from temporarily waterlogged soil, as a function of soil water retention on draining, which varies with soil type. Sulfide precipitates such as iron sulfide and pyrite are insoluble and remain in the soil when the water drains away, after which they can be oxidized back to sulfate. The relative proportions of plant-available sulfate from soil oxidized sulfide to remaining sulfate in soil water after draining is referred to as the “soil sulfate source ratio” hereafter.

Some plants, primarily those adapted to wetland ecologies (e.g., wetland plants including the crops rice and taro), are sulfide tolerant. They have adapted to root zone (rhizosphere) waterlogging and can oxidize sulfide to sulfate in the root zone (Lamers et al., 2013; Simkin et al., 2013). This occurs through the development of a barrier along the root length to the tip which protects the plant from absorbing toxic sulfide along the root shaft, but allows leakage of oxygen into the rhizosphere from the root tip (Abiko and Miyasaka, 2020; de la Cruz Jiménez et al., 2021; Nishiuchi et al., 2012; Pan et al., 2021). The oxygen seeping from the root tip into the waterlogged rhizosphere (Ando et al., 1983), leads to localized root zone oxygenation and the development of a symbiotic bacterial community of sulfide oxidizers. This supports in situ microbial oxidation of sulfide with lowered δ³⁴S, forming sulfate which can be then taken up and utilized by the plant. Internal sulfide oxidation by symbiotic bacterial colonies within the roots may also occur (Lamers et al., 2013). Wetland plants can also develop adventitious roots which form laterally at the soil/water interface, allowing uptake of oxygen and sulfate from overlying water (Parent et al., 2008; Steffens and Rasmussen, 2016).

A critical factor determining wetland plant δ³⁴S is the relative amounts of sulfur taken up via these two pathways: the uptake ratio of sulfide oxidized in the root zone, to direct uptake of sulfate from overlying water; hereafter referred to as the “plant uptake ratio.” There are limited published data on wetland plant uptake ratios. The few studies that are available show saltmarsh plant uptake ratios ranging from 1.0: 0.0 to 0.6: 0.4 (Carlson and Forrest, 1982; Cornwell et al., 1995; Fry et al., 1982, 2023; Peterson et al., 1986; Price and Casagrande, 1991; Stribling and Cornwell, 1997; Stribling et al., 1998). In these sulfate rich environments sulfate δ³⁴S in overlying water is as high as 21‰, but plant δ³⁴S ranges from −14‰ to 13‰ due to root zone sulfide oxygenation. Although freshwater wetland plant uptake ratio studies are lacking, the range of δ³⁴S from freshwater wetland plants reported recently is similar in magnitude to that of saltmarsh plants, although with lower absolute values (−31‰ to −1‰, Evans et al., 2023; Lamb et al., 2023).

Interestingly, adventitious roots (but not rhizosphere sulfide oxidation), can be induced by waterlogging stress in non-wetland species including the crop plants oats, barley, wheat, maize, and millet (Li et al., 2023; Matsuura et al., 2022; Pan et al., 2021; Yamauchi et al., 2013). This stress response may have enabled early cultivation of crop plants on floodplains and lowlands which would have been at risk of periodic inundation during the growing season.

Plants living in wetland systems with flow-through hydrology (including natural wetland plants and the crops rice and taro) are sulfide adapted. The primary source of water and additional sulfate for many natural wetland environments is continuous lateral and/or vertical shallow groundwater flow (Winter, 1999) which supplies an ongoing input of sulfate derived from geological weathering and atmospheric sources for use by microbial sulfate reducers (Figure 2A). In such flow-through systems, sulfate which is reduced to sulfide in anoxic sediments is replenished with sulfate in the continuously flowing shallow groundwater. Such natural groundwater fed wetlands can be permanently waterlogged if shallow groundwater levels are stable, or temporarily waterlogged if shallow groundwater levels recede, and include riparian wetlands, marshes and fens. A similar hydrology existed where engineered flow-through systems were built, supplied with flowing water channeled from nearby sources such as springs and streams.

Referring to the partitioning of ³²S and ³⁴S within different reservoirs of sulfur during isotopic fractionation, flow-through wetlands maintain a near constant sulfate δ³⁴S and sulfate concentration in overlying water because the continuous replenishment of sulfate from flowing water erases the effect of preferential removal of ³²S by sulfate reduction (Figure 2A). ³²S accumulates in sedimentary sulfide resulting in a lowered sulfide δ³⁴S, which is passed on to plants via rhizosphere oxygenation and sulfide oxidation, resulting in plants with lowered δ³⁴S. The difference between sulfate δ³⁴S in the overlying water and sulfide δ³⁴S in the sediment is stable if the magnitude of the isotope effect of sulfate reduction is stable (Högberg, 1997; Kendall and Caldwell, 1998; Mariotti et al., 1981), and is the first control on the lowered δ³⁴S of the wetland plants. The magnitude of the isotope effect has been characterized for a limited number of wetlands (e.g., 11‰ to 51‰ Carlson and Forrest, 1982; Peterson et al., 1986; Price and Casagrande, 1991; Stribling et al., 1998). If minor fluctuations in groundwater sulfate δ³⁴S occur as a result of changes in the relative proportions of sulfate source contributions, such as after recharge from rainfall or during drought, this could result in fluctuations in sulfate and sulfide δ³⁴S values. As plant uptake ratios are not well known (see Section 2) and the sulfate reduction isotope effect is dependent on many factors (see Section 3.1), we use ranges from the literature, modeling plant uptake ratios of 0.6:0.4 to 0.8:0.2 and isotope effects of −10‰ to −40‰ (Figure 3A and Supplementary Information). Our models illustrate the effect of sulfate reduction as changes in plant δ³⁴S (denoted by Δ³⁴S) in different systems relative to baseline “initial sulfate,” driven by the geology, and are based on the simplifying assumption that there is no pre-existing sulfate or sulfide in the system. In the real world initial sulfate δ³⁴S values could vary significantly depending on the underlying geology, which can encompass values from −20‰ to 30‰ (Seal, 2006).

Most plants growing in persistent standing water hydrology (including natural wetland plants and the crops rice and taro) are also sulfide adapted. Standing water systems receive water containing sulfate from rainfall and include seasonally and ephemerally waterlogged landscapes in low lying areas, and on poorly draining soil. In engineered standing water systems, pulses of water may have been diverted from local streams, lakes, or seasonal floodwaters, and retained by field edge structures, while poor soil permeability may have been enhanced by soil puddling (tillage under flooded conditions to reduce permeability), to prevent vertical drainage (Figure 2B). In these systems sulfate in overlying water slowly percolates into anoxic sediments to support microbial sulfate reduction. In the overlying water, sulfate δ³⁴S and sulfate concentration undergo dynamic changes as sulfate reduction progresses, which in turn affect the δ³⁴S of sulfide produced. In standing water systems which receive a single input of water, sulfate is progressively transferred from the overlying water to sulfide in the sediments where it reacts with iron to form insoluble precipitates. This results in an increase in ³⁴S in the remaining sulfate in overlying water and in ³²S in reduced sulfide in sediments as sulfate reduction proceeds. In this system, sulfate concentration in the overlying water decreases over time and the δ³⁴S of sulfate rises well above the δ³⁴S of the initial sulfate, while the δ³⁴S of sulfide in the sediment rises gradually as more sulfide is sequestered, reaching a maximum δ³⁴S sulfide value the same as that of initial sulfate. This point is reached when all sulfate from the single input has been reduced to sulfide (Högberg, 1997; Kendall and Caldwell, 1998; Mariotti et al., 1981) (Supplementary Information). Due to the fact that sulfide δ³⁴S rises as sulfate reduction proceeds, there is a smaller difference between the initial sulfate δ³⁴S and that of sulfide than the difference seen from flow-through systems (Figure 3B).

In some standing water systems, periodically repeated inputs of water supply fresh sulfate. Although the δ³⁴S of sulfate and sulfide respond dynamically during sulfate reduction (Högberg, 1997; Kendall and Caldwell, 1998; Mariotti et al., 1981), the periodic inputs of sulfate buffer these effects, resulting in a similar effect to that seen from flow through systems. We also modeled plant δ³⁴S with standing water hydrology and periodic inputs of “initial” sulfate in water, using sulfate and sulfide δ³⁴S over time. The same plant uptake ratios and isotope effects are used as in the flow-through system. However, the range of resulting plant δ³⁴S values is slightly smaller (Supplementary Information). Note that the plant uptake ratio is particularly sensitive in the standing water scenario as, if a large proportion of sulfate is taken up from overlying water after significant sulfate reduction, plant δ³⁴S values could be raised.

Standing water hydrology and flow-through hydrology can occur on a continuum within the same wetland, often controlled by the seasonality of water inputs. For example, flow-through wetland systems may behave like standing water systems at their margins when water flow is reduced (e.g., Peterson et al., 1986), if they have poorly permeable sediments which restrict water movement, and with sediment depth (e.g., Carlson and Forrest, 1982; Price and Casagrande, 1991). In addition, many standing water systems receive sporadic inputs of water and sulfate from precipitation, flood events, high tides, or temporarily high groundwater levels (Supplementary Information).

Plants grown after drainage of flooded soil conditions, (such plants include wild grasses and the crops barley, wheat, sorghum, and barnyard millet), are not able to oxidize sulfide but can tolerate transient flooded conditions. These plants have to take up their sulfur directly as sulfate from the environment. Many flow-through and standing water wetlands are temporarily waterlogged rather than permanently inundated. When flooded soil drains, the water which submerged the soil is mostly lost. This can occur via gradual drainage when flood waters recede and shallow groundwater levels drop, or, in engineered systems, as a result of intentional release (Figure 2C). Due to the high solubility of sulfate, when overlying water drains, much of the sulfate is lost via transport in drainage water. When drainage occurs, oxygen enters the soil triggering sulfide oxidation, creating low δ³⁴S sulfate from the sulfide produced during waterlogging and stored in precipitates formed with iron in anoxic sediments (Figure 2C). Plants growing immediately after soils are drained incorporate sulfate primarily from sulfide which has been oxidized in the freshly aerobic soil. Where temporary wetlands are repeatedly flooded and drained, a cumulative effect can occur if conditions promote substantial sulfide oxidation each time the wetland is drained (for example if soils have high porosity, readily allowing oxygen to enter previously waterlogged pore space). Some wetlands, once drained, remain so permanently, whether as a result of natural processes (e.g., alterations in river courses), environmental change (e.g., lowered shallow groundwater levels as a results of decreases in recharging precipitation), or alteration in land use (e.g., marshland drained to create pasture or rice paddy fields drained to grow different crops). Permanent drainage enables oxygen to remain in the soil, which, as the soil is no longer waterlogged, no longer supports sulfate reduction. Within this drained soil, low δ³⁴S sulfide (produced and stored in the soil before it was drained), is gradually oxidized to sulfate and is not replenished. This results in gradually decreasing inputs of low δ³⁴S into the plant available sulfate pool from sulfide reserves near the soil surface; a process which may take many years (e.g., Chung et al., 2017). We have modeled δ³⁴S for plants grown on drained soils after a flow-through hydrology (Figure 3C). We have estimated the proportion of sulfate lost in water on draining by considering soil water content under saturation and plant-available water on draining, for three soil types with different drainage; sandy, loam and clay soils. The different porosity of these soil types means they hold different proportions of water when saturated, and on draining, these soil properties affect how much of this water is available to plants (Kumar Rai et al., 2017; Marshall and Holmes, 1996) (Supplementary Information). Based on these estimates, we use soil sulfate source ratios (the relative proportions of plant-available sulfate from soil oxidized sulfide to remaining sulfate in soil water after draining), of 0.7:0.3, 0.8:0.2 to 0.9:0.1 for loam, clay and sandy soils, respectively. We use the same isotope effects as in the previous models. This system results in lowered plant δ³⁴S. If before draining, the system had a standing water hydrology (rather than flow-through modeled here), we expect δ³⁴S values of plants grown in drained soil to be lowered, but not lowered to the extent of plants growing in a drained flow-through system, reflecting the smaller signal expected from standing water systems in general (5.2). In landscapes that are repeatedly flooded and drained, we expect δ³⁴S of plants grown in drained soil to be lowered by a similar amount as that expected from flow-through systems after draining.

Plants grown in simultaneously flooded and drained systems (including the crops barley, wheat, squash, maize, beans) are not able to oxidize sulfide, but can tolerate partially saturated soil conditions. Such conditions occur through minor variations in elevation in waterlogged and flooded natural landscapes, and in some engineered wetlands (e.g., ridge and furrow, raised fields, floating fields) (Figure 2D). The hydrology in these systems may be standing water or flow-through. Such systems simultaneously support sulfate reduction in the waterlogged soil at depth, producing low δ³⁴S sulfide values, and sulfate oxidation in the soil above the water, producing low δ³⁴S sulfate (Figure 2D). However, plants grown on unsaturated soil above a high-water table are expected to incorporate sulfate primarily from roots reaching the upper level of saturated soil and the overlying water directly. The sulfate δ³⁴S of these sources is controlled by the flow-through or standing water status of the water. We have modeled the plant δ³⁴S with standing water hydrology over a range of soil and water sulfate source ratios of 0.4:0.6 to 0.2:0.8, at 50% reduction of sulfate to sulfide (but as discussed above the extent of reduction is dynamic). In simultaneously flooded and drained systems with standing water hydrology, plant δ³⁴S could be raised above the initial sulfate δ³⁴S, as illustrated in Figure 3D. However, if such systems have flow-through hydrology we would expect plant δ³⁴S to be slightly below the initial sulfate δ³⁴S of the water.

These model results suggest that through analysis of archaeological crop δ³⁴S we may be able to differentiate different water management practices, many of which played a critical role in subsistence agriculture and its development and intensification. In ancient societies, the ability to access, control and distribute water effectively often determined the success or failure of agricultural production, and increasingly sophisticated techniques often accompanied agricultural intensification (Fuller, 2020). Some of the earliest arable farmers harnessed the natural environment through “least-effort” cultivation on floodplains, as these areas offered fertile soils enriched by seasonal flooding, along with a reliable supply of water (Ma et al., 2016). Later, arable farmers developed infrastructure to manage water supply either to enable cultivation on land that required irrigation, or to utilize waterlogged or frequently flooded land through drainage (Langlie, 2018; Zhuang et al., 2025). This utilization or management of water is argued to have enabled early farmers not only to continue producing food in unpredictable climates but also to create agricultural surpluses that are thought to have underpinned the development of complex societies (Zhuang et al., 2014). While broad patterns of natural and managed water use in prehistoric agriculture are well-documented, the study of past water management regimes primarily relies on indirect evidence such as traces of irrigation infrastructure and weed ecology (Petrie and Bates, 2017; Zhuang et al., 2025). These methods often fall short in accurately reconstructing past water conditions, restricting interpretations to a coarse resolution. Analysis of δ³⁴S values of archaeological crop remains may have the potential to elucidate past regimes of water management for cultivation and to determine the hydrological conditions under which agricultural production was taking place (Stevens et al., 2022), particularly if dryland bioavailable sulfur δ³⁴S baselines derived from local geological and marine sulfur can be well-characterized.

In addition to supporting early agricultural development and intensification, the human manipulation of natural wet and waterlogged environments and the creation of engineered wetlands for crop production are likely to have resulted in environmental harms to the immediate environment and beyond. Such detrimental effects could have included biodiversity loss, the disruption of sediment cycling, and increases in greenhouse gas emissions (Erickson, 1992; Fuller et al., 2011; Marston and Scott Branting, 2016; Robinson and Lambrick, 1984). By studying archaeological crop δ³⁴S we may be able to further our understanding of the timing and extent of these practices to clarify the likely extent of the environmental damage associated with managing water and wetlands for crop production.

The effects of past water management techniques on archaeological crop δ³⁴S depends on the hydrology of the technique involved, and any waterlogging adaptation of the crop grown (Table 1). Here we suggest which model scenario (Section 4.) would best represent a particular past water management technique for specific crops. Rice and taro are probably the only truly flood adapted crops able to oxidize sulfide in the rhizosphere and to thrive when partially submerged (Abiko and Miyasaka, 2020; Nishiuchi et al., 2012). In the natural environment some wild rice and taro species would have grown in permanent wetlands which were likely to have included flow-through and standing water hydrology (Fullagar et al., 2006; Müller et al., 2010; Zheng et al., 2009; Zong et al., 2007) (Figures 3A, B). Under paddy cultivation, flow-through systems and standing water systems were likely used to grow wet rice and taro (Kay et al., 2019; Lee et al., 2014), depending on water source availability, with paddy drainage before harvest. Where such seasonal drainage occurred this would have removed much of the sulfate in overlying water and allowed oxidation of sulfide in the soil, which would then have become available for uptake by catch crops, as in the case of the rice-wheat system (Bates et al., 2017) (Figures 3A, B, followed by Figure 3C). Irrigation techniques which involved temporarily submerging soil, for example by engineered trapping of seasonal flood waters, would have suited cultivation of crops with some waterlogging tolerance such as dry rice, sorghum, and barnyard millet (Bates et al., 2017; Bhinda et al., 2023; Kay et al., 2019; Khalifa and Eltahir, 2023; Dal Corso et al., 2022; Neumann, 2018 (Figure 3C).

Similarly, crops grown after natural drainage of flood waters would have been cultivated in environments with seasonally high groundwater levels such as riverbanks, floodplains, lake margins, river valley lowlands, and ridge and furrow on low lying, poorly draining soil, and were likely to be able to tolerate temporary flooding and high groundwater levels in the root zone (Brechbühl et al., 2024; Hamerow et al., 2020; Kay et al., 2019; Kingwell-Banham, 2019; Motuzaite-Matuzeviciute et al., 2012; Ortloff, 2009). Where simultaneous flooding and drainage was engineered, cultivation may also have taken place at elevations above the reach of high groundwater levels, and crop plants may not have required flood tolerance (Figure 3D). Raised fields, floating fields, and sunken gardens, found in the Americas and parts of Africa (Blatrix et al., 2022; Comptour et al., 2018; Erickson, 1987; Frederick, 2007; Ortloff, 2009), would have enabled a steady supply of water to reach plant roots while protecting crops from waterlogging. It is possible, therefore, that in cases where dryland environmental δ³⁴S baselines can be established, and in combination with other contextual information, analysis of δ³⁴S in archaeological crop remains could be useful to differentiate between different wetland utilization strategies, as represented in our four modeled scenarios (Figures 3A–D).

Where changes in δ³⁴S values are found in charred archaeological grains from the same site across different phases, this could suggest different water management practices or environmental change through time, especially if this can be supported by information from weed ecology or weed phytoliths. Since its domestication, rice has been cultivated using wet (flooded) and rainfed (dryland) methods. Switches between cultivation methods, and expansion into higher yielding wet rice cultivation have been linked to changes in climate in Iron Age North-eastern Thailand and the Lower Yangtze, China between 7000 and 6000 BP (Castillo et al., 2018; Patalano et al., 2015). These studies could be further elucidated by investigating variations in δ³⁴S of charred archaeological rice from these sites which may be associated with changes in cultivation method, and to ascertain whether signals correspond in time to changes seen in climate proxies. We would expect significantly lower δ³⁴S values from rice cultivated using wet techniques with flow-through hydrology than from rice grown using the rainfed method.

Sulfur in human and animal bone and dentine collagen is derived through diet. At the base of the food chain, sulfur, an essential nutrient for plants, is taken up as sulfate from soil. Once within the plant, sulfur is used to make amino acids cysteine and methionine, and many other compounds (Hill et al., 2022; Li et al., 2022). Sulfur is passed up the food chain with minimal trophic enrichment when plants are eaten, so plant δ³⁴S is broadly conserved during the transfer of sulfur from plants to animals (Cavallaro et al., 2022; Krajcarz et al., 2019; Nehlich, 2015; Raoult et al., 2024; Tcherkez and Tea, 2013). Collagen in bones and dentine primarily consists of type 1 collagen, and its sulfur content arises exclusively from methionine (Anné et al., 2019). Therefore, most sulfur preserved in archaeological collagen is derived from the essential amino acid methionine, which is only synthesized by plants. With respect to the transfer of δ³⁴S from plants to animals, it is not known whether the overall (bulk) plant δ³⁴S is close to, or is significantly different from plant methionine δ³⁴S. It has been suggested that plant methionine δ³⁴S could differ significantly from the δ³⁴S of bulk plant material. This is because methionine is produced via a series of reactions which may cause isotopic fractionations in both directions, possibly resulting in an overall isotopic offset (Tcherkez and Tea, 2013). Clarification would require studies using compound-specific measurements of plant methionine δ³⁴S in comparison to bulk plant δ³⁴S from the same samples. However, until such studies are published, useful information can be found in two controlled animal feeding studies, which compared faunal bone collagen δ³⁴S against plant material feed with characterized bulk plant δ³⁴S (Tanz and Schmidt, 2010; Webb et al., 2017). These studies found a very small isotopic offset in animal bone collagen δ³⁴S (which is primarily from methionine) relative to the bulk δ³⁴S of plant-based feed (−0.5‰, Tanz and Schmidt, 2010; −1.5± 0.8‰; Webb et al., 2017). This suggests bulk plant δ³⁴S is similar to plant methionine δ³⁴S, and that bulk plant δ³⁴S can be used to represent “base of the food chain” δ³⁴S values in archaeological studies of bone and tooth collagen within a small margin of error. In this way, herbivore bone collagen δ³⁴S can be seen as an integrator of environmental signals carried in plant δ³⁴S as plant δ³⁴S reflects the environment in which the plant grew. See Tanz and Schmidt (2010), and Tcherkez and Tea (2013), for further details on plant sulfur metabolism and isotopic fractionation, Anné et al. (2019) for discussion of collagen synthesis and sulfur preservation in archaeological bones and Devignes et al. (2022), for detailed information on the role of amino acids including methionine in bones.

Some studies also use δ³⁴S of archaeological keratin from antlers, horns, hooves and hair. However, sulfur in keratin is mainly in the form of the non-essential amino acid cysteine which can be synthesized by mammals from various dietary sources of sulfur, such that keratin may be a less direct tracer of wetland environments than collagen δ³⁴S. Despite this, evidence for a relatively minor effect from isotopic fractionation during sulfur uptake into keratin is also suggested by controlled feeding studies and wild animal tracking (Kabalika et al., 2023; Richards et al., 2003).

The incorporation of a wetland δ³⁴S signal into bone and dentine collagen has the potential both to complicate and enhance the interpretation of archaeological human and faunal isotope data. In addition, wetland δ³⁴S signals may overlap geologically driven δ³⁴S baselines, particularly where geological δ³⁴S reflects past large-scale microbial sulfate reduction. The incorporation of a wetland δ³⁴S signal by mammals via diet is relevant for studies of hunter gathers, mobile pastoralists, and sedentary agricultural communities. For example, significant wild wetland resources use by hunter gatherers could potentially be seen in human collagen δ³⁴S (Rand et al., 2021). Similarly, where a significant proportion of human dietary methionine intake originated from crops cultivated on wetlands and from livestock which grazed on or were foddered from the wetlands, a wetland δ³⁴S signal from plants will be reflected in the consumer's collagen δ³⁴S. This could be used to indicate the cultivation of wetlands by early agricultural communities.

Variations in the magnitude of a wetland signal in human and animal collagen δ³⁴S could originate from differences in the proportion of dietary protein from a waterlogged or recently drained environment over time. For example, this could occur with different farming strategies for floodplain use including floodplain grazing, seasonal fodder using floodplain grown plants, and flood recession arable farming (Figures 4A–C).

With well-characterized dryland baseline δ³⁴S, serial analysis of δ³⁴S in dentine collagen from archaeological herbivore teeth could provide high-resolution insights into ancient farming practices, including the movement of animals across regions with different geologies and the use of resources from wetlands or recently drained areas within landscapes. Dentine protein, which comprises c. 90% collagen (Goldberg et al., 2004), records a time series of isotopic inputs from diet which is unchanged after deposition. Dentine growth bands form in hypsodont herbivore teeth every few months (Zazzo et al., 2006), potentially recording seasonal signals in dentine collagen. Thus, incremental δ³⁴S analysis of hypsodont tooth dentine collagen, as has been conducted for δ¹³C and δ¹⁵N (Díez-Canseco and Tornero, 2024), could be used to explore different floodplain utilization scenarios (e.g., Figures 4A–C), perhaps with seasonal climate signals constrained by incremental analysis of δ¹⁸O in enamel bioapatite adjusted for differences in timing and mineralization processes between dentine and enamel (Balasse, 2002).