ATMOSPHERIC RADIOCARBON FOR THE PERIOD 1910–2021 RECORDED BY ANNUAL PLANTS

ABSTRACT We present a timeseries of 14CO2 for the period 1910–2021 recorded by annual plants collected in the southwestern United States, centered near Flagstaff, Arizona. This timeseries is dominated by five commonly occurring annual plant species in the region, which is considered broadly representative of the southern Colorado Plateau. Most samples (1910–2015) were previously archived herbarium specimens, with additional samples harvested from field experiments in 2015–2021. We used this novel timeseries to develop a smoothed local record with uncertainties for “bomb spike” 14C dating of recent terrestrial organic matter. Our results highlight the potential importance of local records, as we document a delayed arrival of the 1963–1964 bomb spike peak, lower values in the 1980s, and elevated values in the last decade in comparison to the most current Northern Hemisphere Zone 2 record. It is impossible to retroactively collect atmospheric samples, but archived annual plants serve as faithful scribes: samples from herbaria around the Earth may be an under-utilized resource to improve understanding of the modern carbon cycle.


INTRODUCTION
Bomb radiocarbon ( 14 C) was produced in the 1950-1960s from atmospheric thermonuclear weapons testing primarily in the Northern Hemisphere (Hesshaimer et al. 1994). This period is increasingly viewed as a near-universal marker of the beginning of the Anthropocene (Turney et al. 2018). Since peaking in the early 1960s, tropospheric Δ 14 C has declined over time due to exchange with both the terrestrial (Trumbore 2000) and ocean (Druffel and Suess 1983;Broecker et al. 1985) reservoirs, and also the burning of 14 C-free fossil fuels (Hesshaimer and Levin 2000). Fortuitously, bomb 14 C provides a unique way to "age" recent (less than ∼60 years old) organic matter within 1-3-year resolution by accelerator mass spectrometry (AMS). Thus, quantifying the incorporation of bomb 14 C into different pools or tissues allows for estimating residence times and sources of carbon, and is a powerful tracer to study the modern global carbon cycle (Hua and Barbetti 2004).
Tropospheric records of bomb 14 C are based on atmospheric CO 2 captured by alkaline solution and flasks from land (e.g., Levin and Kromer 1997;Turnbull et al. 2017), aircraft (e.g., Telegadas 1971), and tree-ring records (e.g., Stuiver and Quay 1981;Yamada et al. 2005) from a small but increasing number of locations on Earth (Hua et al. 2022). These records show differences in the magnitude and timing of the bomb spike across the northern and southern hemispheres due to location and size of bomb detonation, atmospheric transport, and mixing times (Hesshaimer and Levin 2000). With growing interest in studying the global carbon cycle, and increased accessibility of measurements of 14 C by AMS, sampling locations added in the past two decades have led to better spatial and temporal representation (Levin et al. 2022). Hua et al. (2013Hua et al. ( , 2022 compiled bomb 14 C records to develop a set of synthetic datasets that accounted for atmospheric transport and mixing. These calibration curves are specific to five latitudinal zones, and offer improved regional dating accuracy, but the potential for local deviations from these zonal curves has not been fully characterized. The objective of this work was to develop a century-long bomb 14 C record for the southern Colorado Plateau region of the southwestern United States. The Colorado Plateau is a remote area spanning parts of Arizona, Utah, Colorado, and New Mexico. It is characterized by low population density, high elevation, and a generally arid environment. This region has a bimodal precipitation pattern, with winter snow and rain (November-April) and summer monsoon rainfall (July-September). We took advantage of annual plants as unique samplers of atmospheric 14 CO 2 to construct this record. Annual plants can have certain advantages over traditional flask sampling and tree-ring records. First, annual plants complete a lifecycle in less than one year (usually one season). In contrast to longlived plants like trees, annual plants do not have nonstructural carbon stored from previous years that can be used to grow biomass (e.g., tree rings; Carbone et al. 2013;McDonald et al. 2019) in subsequent years. Thus, with the exception of the initial seed from which it is grown, all carbon in an annual plant is produced from atmospheric CO 2 assimilated within the same year or season. Second, annual plants sample the atmosphere through photosynthesis over many days to months, integrating the atmospheric 14 CO 2 signal over longer periods than flask sampling (minutes to hours). Finally, annual plants are common and widely distributed, and include many crops and non-native weedy species that can be found across ecosystems and therefore are often present in herbaria collections.
The value of annual plants as a proxy for atmospheric 14 CO 2 has been known for many years (Godwin 1969). Annual plants have been used to develop short term (< 10 years) site-specific background atmospheric 14 CO 2 records to accurately date recent terrestrial organic matter when anthropogenic fossil fuel emissions may cause localized lower 14 CO 2 relative to the northern hemispheric average Richardson et al. 2013;Furze et al. 2018Furze et al. , 2020. Creatively, annual plants have been collected across large spatial scales to map and quantify contributions of fossil fuel derived CO 2 to the atmosphere in a given year (Hsueh et al. 2007;Riley et al. 2008;Pataki 2010, 2012). Most recently, Hüls et al. (2021) created a 75-yr 14 CO 2 record from annual plants (agricultural wheat seed archives) documenting bomb 14 C as well as the fossil fuel contributions over the past four decades.
In recent decades, archived herbarium specimens have increasingly been used to study the impact of global change on plants (Meineke et al. 2018;Lang et al. 2019). Specific examples include early studies investigating the effects of rising atmospheric CO 2 on both stomatal density (Woodward 1987) and leaf isotopic composition (δ 13 C; Peñuelas and AzcónBieto 1992), as well as the effects of increasing temperature on both phenology (Willis et al. 2017) and herbivory (Meineke et al. 2019). We are not aware of herbarium records having been used previously to develop a long-term record of 14 CO 2 in the atmosphere.
Here, we present the application of an herbarium collection of annual plants to develop a smoothed annually resolved record of bomb spike 14 C, from 1910 to 2021. We describe the 14 C timeseries derived from analysis of 100 individual annual plant samples, and compare these samples to existing western U.S. records, as well as the most current calibration curves for the region. We then use smoothing techniques to develop a synthetic, annual-resolution (summertime values) curve with uncertainty for local dating of terrestrial organic matter. Finally, we discuss the potential to use annual plants, including leveraging of herbaria collections, to complement existing records and further improve understanding of local-to-regional variation in tropospheric 14 CO 2 .

Annual Plant Samples
Archived annual plant specimens were sampled from the Deaver Herbarium (ASC) at Northern Arizona University in Flagstaff, Arizona, USA (Thiers 2022). We chose herbarium specimens based on annual plants species that had the best representation and abundance during the period 1950-2016. Herbarium specimens in order of abundance include Xanthisma gracile, Townsendia annua, Plantago argyraea, Erigeron divergens, Bromus rubens, and Bromus rigidus. We prioritized specimens from Coconino and Yavapai counties, which include the southern Colorado Plateau and the adjacent Arizona transition zone of the Mogollon Rim. From each specimen, ∼10 mg of leaf, flower, and/or inflorescence material was removed with tweezers, weighed, and placed in a glass vial. We attempted to sample different regions of each specimen, both basal and distal, to ensure that sampling was representative of the atmosphere during the entire period of growth. We were careful to avoid areas of the plant that had been attached with glue or tape to the specimen mounting paper. Figure 1 shows an example of a Xanthisma gracile specimen from 1964 that was sampled for 14 C.
Additional annual plants were collected by the authors in the Flagstaff area from 2015-2021. These include Bromus tectorum, Lupinus kingii, Ambrosia acanthicarpa, and Solanum lycopersicum. Plants were harvested at the end of the summer growing season (August and September). After oven-drying at 60°C, leaves were homogenized with mortar and pestle. No chemical pretreatment or washing of plant material was conducted on herbarium specimens or field samples. Potential carbon contamination by dust or human oils was assumed to be minimal in comparison to the carbon in the sample.

C Analyses
All annual plant samples were prepared for 14 C analysis in 2021 at the Arizona Climate and Ecosystem (ACE) Isotope Laboratory at Northern Arizona University. For each sample, approximately 2.5 mg of dry organic matter was weighed into a tin capsule and converted to graphite using the Automated Graphitization Equipment (AGE 3, Ionplus, Switzerland). The 14 C content of the graphite was measured using accelerator mass spectrometry (AMS) on a Mini Carbon Dating System (MICADAS, Ionplus, Switzerland). The data (decay corrected Δ 14 C) are reported in per mil (‰) following standard methods (equation 3.19) summarized in Trumbore et al. (2016). Instrument error is reported for all Δ 14 C data; for most samples, it was approximately 1-2‰.

Data Analyses
Annual plant Δ 14 C values were compared to the most current synthetic records for the Northern Hemisphere zone 2 from Hua et al. (2022) referred to as NHZ2 summer and NHZ2 monthly from here on. From 1950 to 1972, the NHZ2 summer is a compilation of Figure 1 Example of a Deaver Herbarium annual plant specimen (Xanthisma gracile) that was harvested in 1964 at the peak of bomb spike in Flagstaff, Arizona, USA. samples from atmospheric CO 2 captured by alkaline solution (in Spain, Israel, and Senegal) and tree rings (Oregon, Arizona, Mexico, Japan, and South Korea) from clean-air sites. From 1973 to 2019, Hua et al. (2022) does not distinguish different zones for the Northern Hemisphere record and synthesizes many more samples and locations across the Northern Hemisphere. The NHZ2 monthly is derived from similar records as the NHZ2 summer with additional curve fitting and smoothing. We compared our data against the NHZ2 monthly record, with the difference (commonly reported as ΔΔ 14 C) calculated as (annual plant Δ 14 C) -(NHZ2 Δ 14 C), using the NHZ2 value for the month in which the annual plant was harvested. To account for the potential integration of 14 C in annual plant biomass as the plant grows, the difference between annual plant Δ 14 C and the mean NHZ2 value of the previous 1, 2, and 5 months was also calculated, representing integration times of 2, 3, and 6 months, respectively. Total error for ΔΔ 14 C was combined in quadrature from the NHZ2 monthly dataset 1σ uncertainty, and the annual plant AMS instrument error.
To develop an annual resolution 14 C smoothed record applicable for the southern Colorado Plateau centered near Flagstaff from 1911-2021 (nicknamed RITA, Radiocarbon In Terrestrial Annuals), we used loess smoothing (PROC LOESS in SAS OnDemand for Academics, https://welcome.oda.sas.com/; SAS Institute Inc., Cary NC, USA) to fit a nonparametric local regression surface. We used the original date of collection for all annual plants, and because our dataset was lacking any samples collected between the spring of 1952 and the summer of 1959 we used 1950-1959 data (annual summertime means) from NHZ2 as a secondary constraint. We weighted our observations as the reciprocal of the squared analytical uncertainty (average 2‰), while we weighted NHZ2 summer values using the reported 1σ uncertainty (average 6‰, with a range from 2‰ to 11‰. We then compared the resulting RITA curve (Supplemental Table S1) against the NHZ2 summer curve, as well as the 1850-2015 curve presented by Graven et al. (2017). Uncertainty estimates (1σ) for the RITA curve were calculated from the LOESS regression residuals, and hence these can be interpreted as the expected range within which an individual new measurement might fall, conditional on the data and our regression model.

Annual Plant Sample Characteristics
All 100 annual plant samples (Table 1)  ). We estimate the average lifespan, or atmospheric 14 C sampling/integration period, of the plants before being harvested was 1-3 months, and at most 6 months.
Annual plant Δ 14 C separated by genera are plotted against the NHZ2 summer record (Figure 3a-f). Annual plant Δ 14 C ranged from -44‰ in 1951 to 797‰ in 1964. In comparison to the NHZ2 record, no measurable bias in Δ 14 C was detected in the  The pre-bomb period with samples between 1910 and 1952 shows a strong decline with a slope of -0.6‰ per year (r 2 = 0.7 p<0.001; Figure 3b). This trend is similar to that observed previously (Stuiver and Quay 1981). The decline in the Δ 14 CO 2 of the atmosphere is called the Suess effect (Keeling 1979) following work by Hans Suess (Suess 1955;Revelle and Suess 1957) and is caused by the addition of 14 C-free CO 2 to the atmosphere from anthropogenic burning of fossil fuels. However, the annual plant sample Δ 14 C values are lower (-8±2‰, mean ± 1SE, n = 11) than the NHZ2 record, indicating a higher local anthropogenic background which coincides with major timber and railroad industries centered in Flagstaff (Reid 2014).  ; ‰ ± instrument error). (C-F) Zoomed-in plots of same data shown in (A) for specific years; y-axis plots differ across plots. Error is much smaller than the size of the symbol.

Atmospheric 14 C for 1910-2021 367
Bomb Spike 14 C Differences between the NHZ2 record and the annual plants occurs with the rise and peak of the bomb spike in the 1960s. Figure 4 shows the difference in Δ 14 C between the annual plant samples and the NHZ2 monthly record where values below zero pre-1964 indicate the annual plant values were lower than the NHZ2 monthly record and values above zero post-1964 indicate annual plant values were higher than the record. This suggests a delayed rise (1962)(1963) and fall (1964)(1965)(1966) in atmospheric Δ 14 C in comparison to the NHZ2 monthly record. We explored whether some of this difference in timing could be due to different integration time (or growing time) of the plants. Increasing the integration time improved the agreement of the records, however even with a 6-month integration time (maximum estimated for these plant species, and likely not most representative) there is still a delay in peak of the bomb spike in comparison to the NHZ2 records.
The annual plants have elevated Δ 14 C in comparison to the NHZ2 records since 2015, differing from the summer values by as much as 4‰ (3±1‰, mean ± 1SE, n = 8), and only reaching zero in 2021, one to two years later than NHZ2 (Figure 3f). Finally, there is a noticeable flattening of the curve in 2020 and 2021, attributed to reduced fossil fuel emissions during the COVID-19 pandemic (Liu et al. 2020(Liu et al. ). 1960(Liu et al. 1961(Liu et al. 1962(Liu et al. 1963(Liu et al. 1964(Liu et al. 1965(Liu et al. 1966(Liu et al. 1967(Liu et al. 1968 Year -500

RITA Curve
Discrepancies between our annual plant samples and the most current synthetic record (see Figures 4 and 5) justify the need for more local records for accurate 14 C dating of terrestrial organic matter. While generally similar to annual-resolution summer atmospheric Δ 14 C records presented by Hua (NHZ2) and Graven et al. (2017), our smoothed RITA curve (Supplemental Table S1) is slightly but consistently lower (more negative Δ 14 C, by ≈6±2‰, mean ± 1SD) than the Graven curve over the period 1910-1949; the average RITA uncertainty over this period is 5‰. RITA does not rise as rapidly in the early 1960s as either NHZ2 or Graven, although RITA's peak value (800 ± 27‰, mean ± 1σ) in the summer of 1964 is intermediate between NHZ2 (784 ± 33‰) and Graven (836‰). In individual years between 1970 and 1985, deviations of up to ±15‰ between RITA and both NHZ2 and Graven are common. The RITA uncertainty during this period is 7‰ vs. NHZ2 of 9‰. Beginning in 1988, when RITA (at 158±6‰) is lower than either NHZ2 (172 ± 5‰) or Graven (175‰), the distance between all three curves progressively shrinks over the following two and a half decades. By about 2000, the difference between the three curves is reliably less than 5‰, which is comparable to the year-over-year decrease in Δ 14 C in all three curves, and similar in magnitude to the RITA uncertainty of 6‰. Intriguingly, since the summer of 2015 RITA has been somewhat higher than NHZ2, particularly in the 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 Year 100  Atmospheric 14 C for 1910-2021 369 most recent years. The strong 1-year lag autocorrelation (r = 0.84, over the period 1980-2019) of differences between RITA and NHZ2 shows that there are systematic discrepancies between our local record and NHZ2, which persist over time and cannot be attributed to random error.
NH Zone 2 In Figure 5, we compare the annual plant data and RITA record to the NHZ2 curves (annual and monthly), and existing bomb 14 C records of subannual tree rings of Sitka spruce from Washington (Grootes et al. 1989), the Sheridan Novitiate Oak (SNO; white oak) in Oregon (Cain et al. 2018), and atmospheric CO 2 captured by NaOH at China Lake, California (Berger et al. 1965(Berger et al. , 1966(Berger et al. , 1967(Berger et al. , 1968(Berger et al. , 1969(Berger et al. , 1987. The datasets are difficult to quantitatively compare due to differences in the timing of sample collections, but visually the annual plant record and RITA curve have a delayed rise and also a muted bomb peak in comparison to the other records. The annual plant data and RITA record are most similar to the NHZ2 curve, confirming the location of the Flagstaff region within NH Zone 2 along with the Oregon record, whereas the Washington and California records are believed to be in NH Zone 1 (Hua et al. 2022).

Unique Regional 14 C Record
Our annual plant record of 14 CO 2 , derived primarily from herbarium specimens, generally agrees with the regional synthetic record by Hua et al. (2022), but, surprisingly, our data show that there is some evidence for a more delayed arrival of the bomb spike in the southwestern U.S. than has been previously believed. With annual plants, we were able to identify independent herbarium specimens that differed in their active growing season, and spring versus summer phenologies, due to the steep elevation and climate gradient in Arizona. This sampling allowed for fine resolution independent 14 C measurements in October of 1962, March, May, and July of 1963 that recorded a delayed arrival of the rise in the bomb spike. Additional specimens in 1964-66 recorded a delay in the subsequent decline in the bomb spike. By broadening our search parameters to include a wider radius around Flagstaff, it may be possible to include samples from a larger number of sites, all of which could still be considered "regional," and thereby improve the temporal resolution of our record during this period when the atmospheric 14 CO 2 signal is extremely dynamic. This delay is most likely due to atmospheric circulation, where the polar and sub-tropical jet moved northward during this time period introducing air masses from the south with lower Δ 14 CO 2 values (Hua et al. 2022). Another explanation for the delay could be that the annual plants are not sampling the well-mixed atmosphere due to their proximity to the soil surface and are thus influenced by microbial decomposition and plant respiration sources, which would not yet have incorporated bomb carbon at this time. But, in the region we sampled, the vegetation canopy tends to be very open, and the near-surface air space is extremely well ventilated. Finally, we also note that Flagstaff falls ∼600-650 km between multiple testing sites in Nevada (upwind) and New Mexico (downwind), where low-yield atmospheric weapons testing took place as early as 1945, but mainly in the 1950s and early 1960s (Enting 1982), and we therefore cannot rule out these potential impacts on our localized record.
Our annual plant data also noticeably deviate from estimated tropospheric 14 CO 2 in the last decade. Elevated 14 C values could be due to cleaner air (i.e., less local fossil fuel contributions) due to the remoteness, as well as high elevation (>2000 m) in much of the region we sampled. Elevated 14 C values may also be the result of increased wildfires in the western U.S. (Zhuang et al. 2021) and localized biomass burning due to recent forest management efforts, which reintroduce bomb 14 C (Randerson et al. 2002;Schuur et al. 2003;Heckman et al. 2013) into the atmosphere during the summer growing season.
Accurate dating of recent terrestrial organic matter require that we take these regional to local scale deviations in the annual plant data into consideration. This is particularly crucial for dating faster cycling organic matter pools, like plant respired carbon and stored mobile plant carbon pools, where deviations of just 2-4‰ in the local background atmosphere can impact the attribution of current year carbon versus previous year's carbon .

Potential of Annual Plants as Widespread Samplers of Tropospheric 14 CO 2
Annual plants have several characteristics that make them appealing to use as samplers of CO 2 . These include: no carryover of nonstructural carbon pools from previous years, atmospheric integration times of weeks to months, and widespread abundance in both space (many are weeds or crops) and in time (due to herbaria collections and short lifespans). Our data additionally show that the genus of plant was not associated with any detectable bias in the measured 14 C, thus many species of annual plants may be available for this purpose. For terrestrial carbon cycling studies, annual plants record the 14 CO 2 that the ecosystem (plants and soil) experience, and thus may be more accurate for dating or attributing sources than "free" atmospheric records.
There are also disadvantages to annual plants as samplers of 14 CO 2 that lead to uncertainties that should be addressed. These include specimen curation and preparation that may introduce contamination to the 14 C measurement. However, the primary disadvantage we encountered in this analysis was uncertain sampling integration time. Most annual plants have short lifespans of 1-3 months, but up to 6 months; herbarium records indicate the date of collection but provide no information about when the plant germinated. An individual leaf could integrate carbon from the atmosphere over just weeks. Determining this integration time for individual plant types and tissues would be important for higher time resolution records. This integration time may depend on how much plant tissue can be sampled for 14 C, i.e., whether the whole plant is being sampled or just a few leaves. Alternatively, for certain applications, annual plants could be purposely grown from seeds (e.g., "iso-meters;" Körner et al. 2005;Carbone et al. 2016), and the observed period of growth used to estimate the atmospheric integration time more accurately. More detailed understanding of how the 14 C of the atmospheric is incorporated into different annual plant tissues of stems, leaves, flowers, seeds, and their nonstructural carbon, could better inform the use of herbaria data for new records. We also note that tree ring records may have much larger integration time uncertainty than annual plants, as tree nonstructural carbohydrate pools stored in bole tissue integrate years of photosynthetic activity Richardson et al. 2013).
We believe the ease of sampling and positive characteristics discussed above largely outweigh this time integration uncertainty and provide exciting potential for the use of annuals plants as widespread samplers of the past and future 14 CO 2 . Utilizing large numbers of herbarium collections that extend decades to centuries into the past (Lang et al. 2019) could allow for Atmospheric 14 C for 1910-2021 371 expansion to higher time resolution and greater spatial representation of 14 CO 2 records, and mapping of local-to-regional deviations from the hemispherical averages. Since AMS samples sizes can be very small, the amount of tissue collected should not present a problem for most herbarium specimens. Also, many herbaria recognize the value of allowing specimens to be subsampled for chemical and genomic analyses, as long as specimens are properly annotated. Finally, because many herbarium collections can be queried remotely online the time and effort required to identify potential specimens is, remarkably, quite minimal. Future sampling campaigns of annual plants could also include annual plants as recorders of the fossil fuel imprint on specific locations for carbon accounting purposes. Finally, we note that in addition to calls for increased high resolution flask sampling (Levin et al. 2022) annual plants could potentially complement information used to constrain Earth System Models (Graven et al. 2017) to understand global and regional scale exchange fluxes of the modern carbon cycle.

CONCLUSIONS
We used 100 annual plants that grew between 1910 and 2021 as a "proof of concept" to create a record of 14 CO 2 for the region near Flagstaff, Arizona, USA. This record is dominated by five commonly occurring annual plant species in the area, and most samples were previously archived herbarium specimens. We provide a localized synthetic record from which dating of recent terrestrial organic matter tissues and pools may be more accurate than synthetic global records. With increasing access to, and decreasing costs in AMS analyses, our results highlight the potential of planted and wild annual vegetation, as well as archived in herbarium collections, for increased time and spatial resolution of 14 C records.