INTRODUCTION AND BACKGROUND
The 14C released by nuclear weapons tests during the 1950s and early 1960s indirectly created a diagnostic “bomb curve” of enhanced 14C levels in the environment. This rapid increase in 14C values, and subsequent decrease via the uptake of 14CO2 by the biosphere and oceans, can be used to determine the age of short-lived materials for forensic applications, and offers insight into changes within the global carbon cycle. In the same timeframe, increasing greenhouse gas (GHG) emissions from fossil fuel burning continued to alter the 13CO2 and 12CO2 content in the atmosphere, with a resulting relative dilution of the 14C content. Both these anthropogenic episodes influence the natural atmospheric 14C activity (“fraction modern” F14C) and δ13C values, as first predicted by Suess in the 1950s (Suess Reference Suess1953).
The natural F14C value is nominally set at a value of 1, almost reaching a value of 2 in 1963 when the nuclear weapons test ban treaty came into action and is presently approaching the “nominal” F14C value of 1 due to uptake by the biosphere and ocean, and increased fossil fuel emissions. While the widespread decline in F14C values below 1 will make the use of the diagnostic bomb curve problematic in the next few years, estimates of the rate and extent of decline below 1 are variable (Sierra Reference Sierra2018).
The fluctuations in atmospheric 14C concentrations and ancillary δ13C values have been monitored for many decades through direct atmospheric monitoring of CO2 (Levin and Hesshaimer Reference Levin and Hesshaimer2000; Manning et al. Reference Manning, Lowe, Melhuish, Sparks, Wallace, Brenninkmeijer and McGill1990; Nydal and Lövseth 1996; Hua et al. Reference Hua, Barbetti and Rakowski2013; Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022), or using annual F14C records, with studies based on measurements of single tree rings (Reimer et al. Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Buck, Cheng, Edwards and Friedrich2013, Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020), grain with a single, known year of growth (Hüls et al. Reference Hüls, Börner and Hamann2021) or wines and spirits to infer atmospheric F14C levels (Baxter Reference Baxter1969; PhD thesis).
Additional studies have confirmed an increase in CO2 concentration and a temporal shift in δ13C values in atmospheric measurements over the industrial period, monitored and identified in the Scripps CO2 Program since the late 1950s (https://scrippsco2.ucsd.edu/). δ13C measurements on air extracted from Antarctic ice core sections also confirm recent increasing CO2 concentration and a decline in δ13C (Etheridge et al. Reference Etheridge, Steele, Langenfelds, Francey, Barnola and Morgan1996; Francey et al. Reference Francey, Allison, Etheridge, Trudinger, Enting, Leuenberger, Langenfelds, Michel and Steele1999). The compiled records of carbon isotopes in atmospheric CO2 produced by Graven et al. (Reference Graven, Allison, Etheridge, Hammer, Keeling, Levin, Meijer, Rubino, Tans, Trudinger, Vaughn and White2017) reveal evidence of temporal changes in δ13C within the atmosphere.
F14C and δ13C data obtained from tree-ring studies have confirmed a decline in δ13C values observed over the industrial period (Freyer Reference Freyer1979; Stuiver et al. Reference Stuiver, Reimer and Braziunas1998; McCarroll et al. Reference McCarroll, Gagen, Loader, Robertson, Anchukaitis, Los, Young, Jalkanen, Kirchhefer and Waterhouse2009; Graven et al. Reference Graven, Allison, Etheridge, Hammer, Keeling, Levin, Meijer, Rubino, Tans, Trudinger, Vaughn and White2017; Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020) only partly explained by changes in the isotopic ratio of CO2, and correction procedures have been derived to estimate δ13C values that would have been obtained under pre-industrial conditions (McCarroll et al. Reference McCarroll, Gagen, Loader, Robertson, Anchukaitis, Los, Young, Jalkanen, Kirchhefer and Waterhouse2009).
Radiocarbon analysis of local annual plants has been used to infer atmospheric radiocarbon concentration in cities globally, providing an estimate of CO2 data during a growing season (Sierra Reference Sierra2018; Hüls et al. Reference Hüls, Börner and Hamann2021). Single year grain samples have been studied at SUERC for many years, forming part of the routine Quality Assurance procedures used in the SUERC Radiocarbon Laboratory. Many of these single year grain samples have also been used in International Radiocarbon Intercomparison studies over the past 30 years and reliable consensus values for the F14C, and data for δ13C, are available (Baxter Reference Baxter1990; Scott et al. Reference Scott, Cook, Naysmith, Bryant and O’Donnell2007; Scott et al. Reference Scott, Cook and Naysmith2010a; Scott et al. Reference Scott, Cook and Naysmith2010b; Scott et al. Reference Scott, Naysmith and Cook2016). The evaluation of our internal quality assurance data collected over many years indicated there was a potential change in δ13C values with time. Additional analysis of single year derived ethanol confirmed that there was a drift in δ13C values with time (Dunbar et al. Reference Dunbar, Cook, Murdoch, Xu and Fabel2018, Cook et al. Reference Cook, Dunbar, Tripney and Fabel2020). Conveniently, these observed changes in δ13C values in the late 20th century can be used to help identify which side of the diagnostic radiocarbon bomb curve (Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022) a sample may derive from i.e., the upslope or the downslope (Dunbar et al. Reference Dunbar, Cook, Murdoch, Xu and Fabel2018; Cook et al. Reference Cook, Dunbar, Tripney and Fabel2020).
Presented in this paper are radiocarbon F14C and δ13C values of duplicate grain samples from two adjacent sites deriving from 1852 until 2020. The new barley grain data are presented together with data from archive barley mash samples, some of which formed part of the International Radiocarbon Intercomparison studies over the past 30 years. This detailed dataset will be used to monitor F14C levels and determine whether a temporal trend in δ13C values exists within one geographical area. The additional F14C measurements from these barley grain samples add to the F14C data available over the pre-and post-bomb timeline, while detailed statistical analysis of the δ13C values will verify and quantify if any temporal shift in δ13C exists. As F14C values approach and decline below 1 it will be problematic to establish if biological materials are pre- or post-bomb. However, if there is a measurable shift in δ13C values within environmental compartments, e.g., biota, as F14C decline below 1, this shift in δ13C can be used in conjunction with the measured F14C values to extend the use of the bomb peak. Here the question asked is, if δ13C values of a single crop species were measured over a large time frame can we estimate the change in δ13C values occurring from the same geographical location, with a restricted land use?
METHODS
Biota of known date and provenance, beginning in the pre-nuclear era, are ideal for observing changing F14C and δ13C values within a single geographical region over the last 100-150 years. Rothamsted Research (Hertfordshire, UK) is home to the oldest continuing agricultural field experiments in the world, and is located 35 km north of London, UK (Latitude 51° 48’ 34.44” N; Longitude 0° 21’ 22.76” W), covering about 330 ha. Seven of these Long-Term Experiments (LTEs) continue today, including the Hoosfield Spring Barley which is an archive of annual grain samples from different agricultural experiments dating back to 1852. Permission was granted by Rothamsted Research to obtain archived spring barley grain samples from every 2–3 years between 1852–2020. Paired samples of barley grain grown in the same year were taken from two different treatment plots with a relatively constant history of fertiliser or manure inputs.
The bulk barley mash samples are archived Quality Assurance materials regularly measured in the SUERC laboratory. This processed barley mash material was obtained from various distilleries across Scotland and provides an ideal, independently sourced material for comparison, although the precise geographical origin of the barley is unknown.
SAMPLING
Barley (Hordeum vulgare L.) is a member of the grass family and is a major cereal grain cultivated in the UK, with an average spring growing season from March to September. Annual records from the “Yields of the Long-Term Experiments” are available on the electronic Rothamsted Documents Archive: e-RAdoc (https://www.era.rothamsted.ac.uk/eradoc/books/2) and confirm that the spring barley was sown in late March and harvested in late August/early September (± 1–2 weeks). The Hoosfield Spring Barley grain samples were grown in two different treatment plots, Plot 72 (FYMN0 – plot 72) and Plot 42 (N1PK-plot-42). Both plots have a constant history of fertiliser or manure inputs and are located within 20m of each other. The plots are similar in soil composition and drainage, and are subject to comparable environmental factors, such as rainfall, temperature, prevailing wind direction, and other atmospheric considerations.
The barley grain was harvested, dried, and stored in sealed glass bottles since their year of collection. For this study, 2-3 grains were selected from each plot from the year specified. In total 112 barley grain samples were sampled from the archive. The years of crop and barley varieties are detailed in Table 1.
Rothamsted Long-term Experiment: Hoosfield Spring barley varieties 1852–2021.
Notes: fallowed in 1912, 1933, 1943 and 1967 to control weeds. Short–strawed cultivars introduced in 1970.
Sample Pretreatment
Each grain was acid rinsed for 1–2 seconds with 0.1M HCl to remove residual grain and soil detritus, rinsed with deionised water, and dried overnight at 40°C (Dunbar et al. Reference Dunbar, Cook, Naysmith, Tripney and Xu2016). The barley grain remained intact. Half a single grain (each grain was halved down the centre midline, typically 20–30 mg) was weighed into a clean quartz insert for combustion (Vandeputte et al. Reference Vandeputte, Moens and Dams1996). The barley mash samples were acid rinsed with 0.1M HCl, rinsed with deionised water, and dried overnight at 40°C. Typically, 15 mg of barley mash was weighed into a clean quartz insert for combustion (Vandeputte et al. Reference Vandeputte, Moens and Dams1996). Replicate quality assurance standards generated from combustion of a single year of barley mash from the Glengoyne distillery, Sample A in the Third International Radiocarbon Intercomparison (TIRI) (Naysmith et al. Reference Naysmith, Cook, Freeman, Scott, Anderson, Xu, Dunbar, Muir, Dougans, Wilcken, Schnabel, Russell, Ascough and Maden2010) were also prepared. The resulting CO2 was cryogenically isolated and 3 mL subsamples were converted to graphite using the zinc and iron reduction method described by Slota et al. (Reference Slota, Jull, Linick and Toolin1987).
Radiocarbon Analysis
The F14C measurements were undertaken using a National Electrostatics Corporation (NEC) 5MV tandem accelerator mass spectrometer using a 134-position MC-SNICS source for running samples (Dunbar et al. Reference Dunbar, Cook, Naysmith, Tripney and Xu2016). Supplemental quality assurance standards were included in the measurement batch and the data listed in Table 2. The measured F14C values for the Rothamsted archive barley grain for Plot 72 and Plot 42 are listed in Table 3.
F14C for TIRI A and δ13C for Quality Assurance Samples used at SUERC RCL.
Rothamsted Long-term Experiment: Hoosfield Spring barley varieties F14C and δ13C values, 1852–2020.
* F14C value 1852–1938 IntCal20, Reimer et al. (Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020), 1941–2020 (average for March to September calculated), Bomb21NH1, Hua et al. (Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022).
Stable Isotope Measurements
δ13C analyses were performed on subsamples of CO2 using a VG SIRA II IRMS. Sample values are compared with those of working standard reference gases of known isotopic composition, produced from international reference materials NBS19 and IAEA-CO-1. The measurement results are expressed using the δ notation (Craig Reference Craig1957) as per mil deviations from the VPDB standard, with 1σ precision of ±0.1‰. CO2 aliquots from the primary and secondary 14C standards were also measured (oxalic acid II primary standard, humic acid secondary standard, and a barley mash or a background secondary standard) detailed in Dunbar et al. (Reference Dunbar, Cook, Naysmith, Tripney and Xu2016). These values are used for offline normalization of sample 14C/13C ratios. Supplementary δ13C values for anthracite and kerosine standards, used in the SUERC RCL, were measured to establish mean δ13C values for comparison (Table 2). The measured δ13C values for the Rothamsted archive barley grain and barley mash data are listed in Table 3 and Table 4, respectively.
Glasgow International Radiocarbon Intercomparison Studies barley mash sample F14C consensus values and SUERC RCL measured δ13C values.
Statistical Models
For the investigation of the temporal trend, a dog-leg (or changepoint) linear regression model was developed. This model (either with known or unknown changepoint) was compared using ANOVA to a simple, single linear regression (Julious Reference Julious2001). The model is defined in the equations below.
$${\delta ^{13}}{{\rm{C}}_i} = {\rm{f}}({{\rm{t}}_{\rm{i}}}) = {\alpha _1} + {\beta _1}{{\rm{t}}_{\rm{i}}}\,{\rm{for}}\,{{\rm{t}}_0} \le {{\rm{t}}_{\rm{i}}} \le {\rm{d}}$$
$${\delta ^{13}}{{\rm{C}}_i} = {\rm{f}}({{\rm{t}}_{\rm{i}}}) = {\alpha _2} + {\beta _2}{{\rm{t}}_{\rm{i}}}\,{\rm{for}}\,{\rm{d}} \le {{\rm{t}}_{\rm{i}}} \le {{\rm{t}}_1}$$
and subject to α1 + β1d = α2 + β2d
where d is the changepoint, and t is year.
RESULTS AND DISCUSSION
F14C
Presented in Table 3 are the radiocarbon F14C and δ13C values in duplicate grain samples from two adjacent sites at Rothamsted, deriving from 1852 until 2020. The annual growing seasons for the barley grain samples are detailed in Table 3 and range from March through to September (https://www.era.rothamsted.ac.uk/eradoc/books/2). Also included, for comparison in Table 3 are F14C values, for 1852 to 1938 from IntCal20 (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020) and for 1941 to 2020 (average March to September data estimated) from Northern Hemisphere Zone 1 data (Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022).
The CO2 generated and graphitised from each grain sample is representative of carbon assimilated within the plant in this growth season timeframe. The uptake of carbon and growth of the grain throughout a season is highly complex, with many processes from photosynthesis, carbon accumulation, and subsequent redistribution of carbon within the cereal. For comparison of the yearly data presented here, it is assumed that carbon assimilation occurs within a similar time frame each year.
Comparing the F14C values from the barley grain at Rothamsted sites Plot 72 and Plot 42, presented in Figure 1 and Table 3, with the F14C values for the pre-bomb timeline 1852 to 1940 from the IntCal20 calibration curve (Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020), suggests there is good general agreement between the Rothamsted barley grain data and the measured atmospheric F14C data sets. F14C values of barley grain from both Rothamsted plots reached their lowest values of 0.9654 ± 0.0029 in 1950, and 0.9586 ± 0.0029 in 1947, respectively, both are slightly lower than the estimated average for the NH Zone 1 for the period to September. These slightly lower values observed at Rothamsted in 1947 and 1950 may be the result of a few local or regional factors, possibly due to additional coal burning in the immediate post war period. The steady decline of F14C values immediately prior to the onset of nuclear weapons testing corresponds with the observed continuous decline of atmospheric δ13C data on a global scale, detailed in Table 3 and shown in Figure 1.
F14C values for barley grain samples at Rothamsted Long-term Experiment: Hoosfield Spring barley varieties 1852–2020 and Barley Mash (data from IntCal20, Reimer et al. Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020 [red diamond] and Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022 [red circle] are shown alongside). (Please see online version for color figures.)
The barley grain F14C values from 1941 to 2016, spanning the bomb peak period, were compared with the estimated average F14C values for the growing season (average for March to September) from NH Zone 1 data, as listed in Table 3 (Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022). From Figure 1, the data between the paired barley grain begins to deviate in 1960, but notably higher F14C values of 1.9126 ± 0.0056 and 1.9258 ± 0.0057 are identified in the barley grains from both Rothamsted plots in 1963, compared with a growing season estimate of F14C 1.6566 ± 0.1845 from the NH Zone 1 atmospheric data. Figure 1 illustrates that Rothamsted grain F14C data is consistently higher than the global average measured in following years, 1966 until 1983, suggesting that the influence of enhanced 14CO2 derived from the weapons testing was observed in the Rothamsted area, reaching higher 14C levels faster, and remaining elevated for longer. It must be considered that variation exists around the NH Zone 1 data, and local influences are a contributing factor. In the years since 1983 the Rothamsted data remain slightly higher than the atmospheric NH Zone 1 data. The most recent year sampled with grain and NH Zone 1 data for comparison is 2016, with Rothamsted data remaining slightly higher than the NH Zone 1 value, with paired values of 1.0236 ± 0.003 and 1.0291 ± 0.0003 compared with the average for March to September F14C value of 1.0195 ± 0.0014 (Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022). Recent grain harvested in 2020 had measured F14C values of 1.0065 ± 0.0024 from Plot 72 and 1.0118 ± 0.0024 from Plot 42, both approaching the “nominal” F14C value of 1 (Figure 1), however no additional data was available for comparison.
Additionally, the supplemental barley mash F14C data from the Intercomparison study samples collected over the last 30 years (Table 4) have decreased at a similar rate to the atmospheric levels (Hua et al. Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022) and the paired barley grain, with recent F14C levels remaining slightly above the “nominal” value of 1. It should be noted however that the barley mash differs from barley grain in that it is a processed material with the carbohydrate and protein fractions removed, i.e., the mash material is largely residual “husk”. The geographical location where the source barley was grown is confidential, so little comment can be made concerning the local 14CO2.
δ13C
The δ13C values for the barley grain plotted in Figure 2 suggest that, before 1960, values are relatively constant with a mean δ13C = −22.3 ± 1.2‰, with a shift to an average value of −26.8 ± 0.98‰ post-2000. The δ13C values start decreasing in both plots in the early 2000s (Figure 2). It must be noted that a small systematic offset of 0.268 ± 0.12‰ is observed between the annual paired δ13C measurements on the barley grain from each site. This offset between Plot 72 and Plot 42 remains unexplained as the sites are <20m apart and have undergone similar treatments within the timescale.
δ13C values for barley grain samples at Rothamsted Long-term Experiment: Hoosfield Spring barley varieties 1852–2020.
The original TIRI-A barley mash sample was collected in 1990 and has a mean δ13C value of –26.5‰ (n = 10), decreasing to –30.6‰ (n = 20), in the 2015 sample. A shift in δ13C values is also evident in the Barley Mash data from the replicate measurements (Table 4, Figure 2).
A basic linear regression (purple curve) with a segmented (dog-leg) regression (blue curve) fitted to the measured δ13C data (Figure 3) shows that there is a clear breakpoint at 1995, when the rate of change in δ13C has a negative slope, pre-1995, of 1.4‰ per century, and similarly, a negative slope, post-1995, of 3.6‰ per century. Measurement of “corrected” δ13C values from tree ring data obtained from different sites in Northern Europe spanning the 19th and 20th centuries by McCarroll et al. (Reference McCarroll, Gagen, Loader, Robertson, Anchukaitis, Los, Young, Jalkanen, Kirchhefer and Waterhouse2009) show variation in δ13C decline, as do the compiled studies from different location and sample types by Graven et al. (Reference Graven, Allison, Etheridge, Hammer, Keeling, Levin, Meijer, Rubino, Tans, Trudinger, Vaughn and White2017). These studies highlight that the rate of decline varies with sample type, location and the timeline applied. The reasons for this apparent change in δ13C pre- and post-1995 at this location remains highly complex and additional data is required. Although Rothamsted Research is the location of a weather station; CO2 data are not available over the timeline investigated. As a result, no corrections were applied to the δ13C values of the barley grain to account for any temporal trend in atmospheric δ13CO2 values—the appropriate correction is being considered.
Linear and Segmented Regression of δ13C values for barley grain samples at Rothamsted Long-term Experiment: Hoosfield Spring barley varieties 1852–2020.
Ongoing investigations include considering whether there are local sources of fossil carbon. According to recent data (Ritchie et al. Reference Ritchie, Rosado and Roser2023), CO2 emissions from UK fossil fuel sources have been declining since 1995.
DISCUSSION
The global carbon cycle is made more complex in the industrial period by the introduction of human sources interacting with natural atmospheric CO2 processes. There are many other external factors that influence the δ13CO2 but the most significant to be considered in the recent industrial period is the use of petroleum and natural gas. Studies simulating the F14C bomb peak levels, with and without the effect of fossil fuels, have shown the potential impact of fossil fuels on CO2 concentration and composition (Graven et al. Reference Graven, Keeling and Rogelj2020). The average δ13C value of anthracite is –23.3‰ (n = 11) whereas petroleum-based kerosine is –30.5‰ (n = 11) (Table 2 and Figure 2). While the δ13C of natural gas is source dependant, a δ13C value of –44.2‰ has been identified in the natural gas fuel supply of an urban area (Pugliese et al. Reference Pugliese, Murphy, Vogel and Worthy2017). Although anthracite burning has been common in the UK since the industrial revolution, the use of petroleum and natural gas has significantly increased in the last few decades. To some extent the impact of the use of anthracite will not be overly apparent in the δ13C value of the barley grain samples, as the values are too similar to those observed in archaeological barley grain with an average δ13C of –23.8‰ (n = 100). As Figure 2 illustrates, petroleum-based products have a distinctly different δ13C value. There are many factors that might influence δ13CO2, however using both F14C and δ13C measurements together may permit triangulation and identification of organic materials of recent origin. Anomalous δ13C values of –25.9‰ and –26.4‰ were observed for both Plot 72 and Plot 42 in 2020 inconsistent with the observed trend. The obvious assumption is an association with the temporary alteration in daily CO2 emissions during the COVID-19 compulsory confinement. Barley grains for 2021 and 2022 were not available at the time of analysis. Establishing the δ13C and F14C values from barley grain grown annually since 2020 will determine if this δ13C re-bound is indeed an anomaly, and if δ13C values will again continue to demonstrate more negative values.
CONCLUSIONS
This paper presents F14C and δ13C values in paired barley grain samples from 1852 to 2020 in one geographical location, using archive material from Rothamsted Research. The F14C values of barley grain and bulk barley mash intercomparison samples show a similar trend to the values from Reimer et al. (Reference Reimer, Austin, Bard, Bayliss, Blackwell, Ramsey, Butzin, Cheng, Edwards and Friedrich2020) and Hua et al. (Reference Hua, Turnbull, Santos, Rakowski, Ancapichún, De Pol-Holz and Turney2022), demonstrating that F14C values at the Rothamsted site unsurprisingly follow the global atmospheric trend, but are offset in that the paired barley grain from Rothamsted F14C levels increase at a faster rate, and remain high over a longer period. Analysis of the δ13C measurements from the barley grain and bulk barley mash samples reveal a temporal shift with statistical analysis suggesting a clear breakpoint occurring in 1995 where the rate of change in δ13C alters, in that δ13C, pre-1995, is declining at 1.4‰ per century, while δ13C, post-1995, is declining at 3.6‰ per century. The reason for this shift in the rate of decline is complex, but it is noted that the values are becoming more similar to those from petroleum and natural gas resources, the use of which continues to increase globally, but many external causes could be influential.
The F14C values of the barley grain samples will decline below the nominal F14C value of 1 within 2–3 years, however the quantifiable shift in δ13C values within the last 30 years presents a potentially useful indicator to identify pre- and post-bomb peak F14C data. Further analysis and data are required to establish if this trend is replicated in other C3 and C4 biota, for example δ13C values may be readily available from other temporal studies of similar plant materials, making it possible that widespread δ13C alterations could be used with the F14C values to extend the use of the bomb peak for forensic, ecological, and environmental applications.
ACKNOWLEDGMENTS
Rothamsted Research is a world-leading, non-profit research centre that focuses on strategic agricultural science to the benefit of farmers and society worldwide. Rothamsted Research is the longest-running agricultural research institution in the world. The authors would like to thank Dr Andy MacDonald and Holly Addis for their assistance gaining permission to obtain the samples from the archive, and for their help during the sampling process and Dr Andy Gregory for comments.
We thank the Lawes Agricultural Trust and Rothamsted Research for data from the e-RA database. The Rothamsted Long-term Experiments National Bioscience Research Infrastructure (RLTE-NBRI) is supported by the UK BBSRC (Biotechnology and Biological Sciences Research Council, BBS/E/C/000J0300, 2017-2023) and the Lawes Agricultural Trust. We are grateful to the referees for their constructive input reviewing the manuscript.
Open access
