Herbarium specimens reveal drivers of Arctic shrub growth

The Arctic biome is experiencing climate change at rates higher than anywhere else in the world (Jansen et al., 2020). Longer, warmer growing seasons have driven increases in plant height, biomass, and changes in distribution and abundance, including an expansion of shrub species (‘shrubification’), ultimately affecting vegetation community dynamics and carbon sequestration across the tundra (Myers-Smith et al., 2011, 2020; Bjorkman et al., 2018; Collins et al., 2021; Kemppinen et al., 2021). However, there is heterogeneity among sites and among species with some sites and some species show little or no response to warming (Elmendorf et al., 2012; Myers-Smith et al., 2015a; Bjorkman et al., 2020). Thus, understanding when and why some sites or species show dramatic responses to warming, while others less so, is critical to predicting the effects of future climate change on vegetation and associated ecosystem functions. However, existing time-series research, such as the long-term International Tundra Experiment (ITEX) established in 1990 (Molau & Mølgaard, 1996; Henry et al., 2022), is based on relatively short periods compared to the onset of anthropogenic global warming (AGW) 150–200 years ago (Franzke, 2014). Arctic shrubs provide a record of growth spanning many decades of environmental change. Radial growth (annual ring widths) has most commonly been used as an indicator of shrub performance over time, and to investigate climate impacts on vegetation processes, including climate reconstructions (Weijers et al., 2010; Buchwal et al., 2013), climate sensitivity (Myers-Smith et al., 2015a), and interactions with other drivers including herbivory (Vuorinen et al., 2022). Axial growth (annual stem growth increments), measured as the distance between winter bud scars or wintermarksepta, also reflects annual variation in growth (Myers-Smith et al., 2015a). While comparatively few studies have included stem increments, just like growth rings, stem increment measurements can also be used as climate proxies (Weijers et al., 2010) and to investigate changes in biomass and temperature sensitivities for Arctic shrub species (Hallinger et al., 2010; Le Moullec et al., 2019). Herbarium collections of Arctic plants have tremendous potential to be used as an extension of shrub growth chronologies. Herbarium specimens have been used to study the long-term impacts of climate change on plant function and structure (Panchen & Gorelick, 2017; Speed et al., 2022; Ahlstrand et al., 2023, 2025). While the interest in the application of herbarium specimens for studying global change has increased in recent years (Lang et al., 2019), the exploration of plant functional traits from herbarium specimens is still very much in its infancy (Heberling, 2022). Woody stems can be cut from mounted herbarium sheets and prepared to reveal tree ring chronologies, and historical specimens can greatly expand the timelines of shrub chronologies (Opała-Owczarek & Owczarek, 2023). However, this includes undesirable destructive sampling of historical records and rare specimens. Here, we demonstrate that digital images of herbarium specimens can be leveraged as a new resource to nondestructively measure plant growth in Arctic shrubs. World-wide digitisation efforts are increasingly making herbarium specimens and their associated biological and ecological data widely available, including hundreds of thousands of specimens collected from across the Arctic biome (GBIF, 2025). We investigate a novel derivation of trait data obtained directly from digital images of herbarium specimens to further our understanding of Arctic shrubification and global change: incremental annual stem length growth. Herbarium specimens of many woody plants uniquely document annual growth increments. Arctic shrub growth is relatively slow with very short annual increments; thus, multiple years of growth preceding the collection date can putatively be measured along the stem from a single specimen. For example, a specimen of Arctic willow (Salix arctica Pall.) collected in northern Greenland in 1908 could allow us to nondestructively measure annual growth increments dating back to 1899 (Fig. 1). Thus, a single herbarium specimen can provide a chronology of historical annual growth rather than a snapshot in time such as flowering time or leaf traits, providing us with repeated measurements in nature from the same individual and providing a better understanding of interannual effects of climate on Arctic plant growth and development. We use 482 Arctic specimens of four species in the genus Salix L. (the willows) collected in Greenland over 160 years to: (1) develop an approach for quantifying stem incremental growth chronologies from digitised herbarium specimens of Arctic shrubs and assess the feasibility and variability at the stem and individual level; (2) determine how Arctic shrub stem growth has changed over time; and (3) determine whether shrub growth is related to the summer temperature of the growing year based on shrub growth quantified from Greenlandic shrub willow specimens. We discuss collection biases in Arctic specimens that might impact measuring growth chronologies. The genus Salix was selected for this investigation based on the feasibility of detecting and measuring growth increments on stems directly from digital specimens. In addition, the genus is widespread, found in a variety of habitats across the Arctic, and several species have been extensively studied from various location across the Arctic (Gamm et al., 2018; Buchwal et al., 2020; Boyle et al., 2022; Prendin et al., 2022). We focused our study using digital specimens from the Natural History Museum of Denmark's Herbarium of the Vascular Plants of Greenland. Four species of differing growth habit, ecology and habitat preferences: Salix arctica Pall, Salix glauca L., Salix arctophila Cockerell, and Salix herbacea L. (Supporting Information Table S1) were selected for this investigation, all collected in western Greenland, broadly around the Disko Bay area in phytogeographic region W5 (Fredskild, 1996; Bay et al., 2017). We restricted our study to this region based on the large numbers and relatively continuous time series of specimens and the availability of climate data collected in the region dating to 1807 (John, 2021). Digital specimens are available on GBIF (Ahlstrand, 2023, 2025). A total of 482 digital herbarium sheets were assessed for their feasibility in detecting growth increments (Table S1). We excluded specimens where label data were obstructed (and therefore not legible), or for which growth increments were not detectable in digital images due to herbarium conservation or digitisation methods. We excluded specimens that were identified as hybrids based on the specimen labels, specimens that were lacking a year of collection, or specimens for which the most current growth increments were obstructed by leaves or mounting tape, and so on. Annual growth increments (cm) on stems were measured directly from digital herbarium specimens using ImageJ (Fig. 1) (Rueden et al., 2017). Rulers captured in the digital specimen images were used to calibrate the measurement tool in ImageJ. Only stems that exhibited obvious growth in the year of collection (determined by the presence of leaves, flowers, fruit, or opening buds) were selected. Stems with broken, browsed, unopened, or putatively dead buds were not selected as we were unable to confirm the year of growth. We aimed to measure multiple increment chronologies from three different stems on a single herbarium specimen; however, this was not always possible, as in many cases only one or two stems exhibited obvious or unobstructed growth increments. We avoided measuring increments on stems if roots were noted (in case of prostrate species). Each growth increment, beginning from the first full increment corresponding to the year before collection (since growth in the year of collection did not represent a whole growing season), was measured three times from septa to the next septa using the segmented line tool, which allowed for the measurement of curved stem increments, and the average value was used for our analyses. Digital images were annotated using the drawing tool to label stem numbers and increment numbers measured on each specimen. Digital data from specimen labels pertaining to the date of collection and location of sampling were used to refine our final dataset. A growth year was assigned to each annual growth increment. Historical climate data for Greenland dating to the 1780s were obtained from the Danish Meteorological Institute for the Weather Station number 4221 in Ilulissat in the Disko Bay region (latitude: 69°14′ N longitude: 51°04′ W; DMI, John, 2021). While these historical climate data lack some information in some years, we chose to use this over other climate data (i.e. CRU TS, Harris et al., 2020), as many of the specimens included lacked geographical coordinates. The historical data for this study area show that the average monthly July temperature has been steadily increasing at a rate of 0.1°C per decade (R2 = 0.20, F = 44.23, P < 0.001). Climate data were summarised to create the following climate variables: monthly temperatures and monthly precipitation for the months of January–September in the year of growth and for the months of October, November, and December, and annual mean temperature for the years preceding growth. Annual average temperature in the year of growth and in the year preceding growth were also included. Climatic data were coupled with the corresponding growth year for that increment, and the year preceding growth (growth year − 1) for the months of November, December, and for annual mean temperature. For example, a specimen collected in 1955 with four consecutive growth increments in the years preceding collection was coupled to climate data corresponding to the years 1954, 1953, 1952, and 1950 and with climate data from the preceding year in the case of the winter variables. Temperature anomalies were calculated by subtracting the baseline mean temperature for the period 1961–2010 from the mean annual temperature for a given year, both for the current growth year and for the year preceding growth. Positive anomalies meant that years were classed as warmer than average, whereas negative anomalies meant that the years were cooler than average. All analyses were conducted in R v.4.4.2 (R Core Team, 2023). We used Spearman rank coefficients to evaluate the relationship between the annual growth increments on the first and second stems on the same specimen and to explore the influence of climate on the growth of each Salix species. We then fitted generalised additive models (GAMs) separately for each Salix species to assess the effects of time and climate variables on annual growth increments using the R package mgcv (Pedersen et al., 2019). Annual growth increment was modelled as a function of either time (growth year) or climate predictors (temperature, precipitation, or temperature anomaly in the year preceding growth) using penalised regression splines. To account for variation among specimens (‘individuals’) and stems within specimens (on the same individual), we included random intercepts effects for specimen ID (NHMD_ID) and the nested stem ID (stem_No). This structure allowed us to control for repeated measures within individual specimens while accounting for additional variation at the stem level. Models were fitted using a Gamma distribution with a log-link function (family = Gamma (link = ‘log’)) as these parameters best fit our data and produced models with the lowest Akaike Information Criterion (AIC) values. Models were evaluated by inspecting standardised residuals, fitted values, and error distributions. A total of 482 digital herbarium sheets were assessed for their feasibility in detecting growth increments and included specimens collected between 1833 and 1993. Between 10% and 37% of the specimens we assessed met our criteria for growth increments (S. arctica 20%, S. glauca 10%, S. herbacea 35%, and S. arctophila 37%; Table S1). The number of growth increments ranged from one to nine per stem, and this varied by species, with S. arctophila and S. herbacea exhibiting greater numbers of increments on average per stem. The mean annual growth increments and variance in growth per species also differed between species, with S. herbacea consistently having the smallest incremental growth and lowest variance (mean = 0.36 cm yr−1, var = 0.07 cm) and S. glauca showing the largest growth (mean = 2.92 cm yr−1, var = 3.94 cm) (Table S1). Smaller, prostrate-growing willows were better candidates for this study. These consistently had fewer leaves obstructing stems, and the shorter annual growth increments meant that a greater number of annual increments could be measured from a single stem, compared to the upright willows and generally more leaf-dense specimens of S. glauca. However, specimen conservation practices such as the use of mounting tape and the positioning of herbarium labels were often what limited specimen use for this work. In addition, we observed many S. glauca and S. arctica specimens with broken or browsed branches, rendering these specimens unsuitable for measurement. We found a significant nonlinear effect of year on growth increment for S. herbacea (estimated degrees of freedom (edf) = 3.65, F = 3.72, P = 0.01). However, despite substantial variability in growth among years, no significant relationship between year and increment length was found for the other species, indicating that stem growth did not change over time (Fig. 2a; see GAM output in Tables S2–S4). Significant correlations were found between the annual growth increments on the first and second stems measured on the same specimens (Table S5; Fig. S1). Variance in growth was found to be greater in warmer than average years and in the period post-1950 for all species (i.e. growth may be more difficult to predict in warmer years and over time due to increased temperature variability in more recent years) (Fig. S2). Growth increments of S. glauca exhibited a significant nonlinear response to mean temperatures in April (edf = 8.05, F = 2.55, P = 0.02), July (edf = 2.65, F = 3.40, P = 0.021; Fig. 2b), and August (edf = 1.00, F = 6.37, P = 0.01), as well as temperature anomaly (edf = 1.00, F = 5.16, P = 0.026) and mean annual temperatures for the years preceding growth year (edf = 1.00, F = 10.35, P = 0.002). For S. herbacea, mean temperature in April (edf = 1.00, F = 4.77, P = 0.03) and June (edf = 5.11, F = 2.18, P = 0.04), temperature anomaly (edf = 1.00, F = 7.87, P = 0.005), and the mean temperatures in the months of November (edf = 1.86, F = 4.41, P = 0.01) and December (edf = 1.00, F = 4.39, P = 0.04) in the year preceding growth year all had significant effects on S. herbacea growth increments. Temperature anomaly in the year preceding growth was also found to be a significant factor in the growth of S. arctophila (edf = 8.26, F = 3.35, P = 0.001). No relationships were found between climate variables and growth for S. arctica based on our GAMs (Table S2); however, a significant correlation was found between July temperature and growth for this species based on Spearman rank correlation analyses (P = 0.03, Table S6). Our study demonstrates a novel use of digital herbarium specimens: extracting incremental stem measurements to understand interannual variation in the growth of Arctic shrubs. Across four Arctic willow species, we found distinct growth patterns and species-specific relationships with summer temperature in keeping with other studies that find that local conditions or microhabitat play an important role in explaining growth (Myers-Smith et al., 2015a; Panchen & Gorelick, 2016; Dobbert et al., 2022). The approach documented here advances our understanding of the climate sensitivity of shrub growth in the Arctic by quantifying incremental stem growth across many individuals of multiple species and sites, without having to do more costly and time-consuming field work, and extending chronologies back further than possible with other methods. This demonstrates the potential for further ecological and trait-based studies based on nondestructive assessments of digital herbarium collections and advances our understanding of Arctic vegetation responses to climate through the Anthropocene. Although all our samples were from the phytogeographic region around the Disko Bay area, the between-individual variation in growth was This may be due to habitat heterogeneity in Arctic tundra (Elmendorf et al., or to among substantial variation was found between of the same specimens (and this from species to other to growth such as or The variation between and individuals also summarised by Myers-Smith et al., the measurement of multiple per specimen and multiple individuals per and where The digitisation of herbarium specimens around the world greater and specimens across collections We found the growth of only one species, S. herbacea, to change over time; no changes were found in growth over time when accounting for the random which is with other studies (Myers-Smith et al., 2015a; Bjorkman et al., and may that the growth of some shrub species is or climate change (Franzke, Panchen et al., 2022). ecological such as habitat and growth habit, play a role in the species-specific climate sensitivities observed here (Table S1). such as et al., et al., 2017; et al., or of or et al., 2022), can the in annual growth we Salix glauca was the only species we found to have a relationship with mean July temperature. Salix is a upright willow that sites, and is associated with in the Arctic et al., The nonlinear relationship we found between growth and mean temperature in July for this willow is in keeping with et and Myers-Smith et where upright willows were found to be more to studies in and Greenland that summer temperature is a critical for Salix with S. glauca in exhibiting greater growth in warmer years et al., Prendin et al., 2022). However, while increasing July temperatures generally our show that growth to with temperature to before or indicating a potential This also patterns found in growth for the species in Greenland et al., et al., also found in the herbarium data show a relationship between summer temperatures and annual stem that warmer growing seasons consistently both and growth of S. glauca in Greenland and that and growth at temperatures Salix herbacea is a growth may be more by the of and temperatures in the as in our study with the species to temperatures in April and Radial growth for S. herbacea in that both temperature and precipitation the growth and that drivers of growth by sampling et al., 2022). In the et found that and growing seasons the performance of S. Salix arctica growth increments no response to variables in our which may be that the species is more in Greenland. Salix arctica is one of the most widely Arctic and can in a of may be more to change & However, in Arctic Greenland, growth of S. arctica was found to be to summer temperatures et al., 2019). Salix arctophila is a species found in tundra and and while studies on this are its growth could be more on changes than temperature (Myers-Smith et al., 2015a; et al., 2017; et al., 2022). Although there are some in the time series of Salix shrub growth here (Fig. the chronology within a single phytogeographic region better than many other Arctic due to collection biases et al., 2018; Panchen et al., 2019). We that our herbarium data lack a critical time period in of global change most recent specimens were collected in the However, the growth specimen data we obtained in this study the full of interannual climate variability and and between the years 1833 and (John, corresponding to our samples from Greenland, and this is critical for the effects of climate on growth. While these critical years of climate warming were not possible to from these the herbarium data do field which in the through the Our study as a of and for historical stem incremental growth from shrubs across the Arctic are to digital specimen images from across the Arctic biome on GBIF (GBIF, 2025). of stem growth and stem growth chronologies in analyses as well as growth to interannual is also potential for individual growth rates with specimen through The collection of shrub specimens in collections is to to to understanding the long-term of vegetation response to climate change. mounting can be to trait-based for example, specimens with herbarium sheets is better than tape, and tape are better than While and have been not included in shrub we are advances including ring in shrub et al., and such advances in could extracting data from the digitised herbarium specimens more in the future et al., has been used for quantifying traits across herbarium specimens including flowering and leaf area & 2023). models could be to stem growth increments, the potential for of growth from be Digital are for studying chronologies of plant growth. application can to the of new plant growth dynamics at greater and than other of Herbarium specimens and based studies by extending timelines and included to our understanding of how shrubs to climate change over the whole period of We like to the for their to the largest collection of herbarium specimens collected in Greenland and at the Natural History Museum of We also like to two with the measurement of growth increments. We that the specimens used in this were collected historical that do not current often without the or the of these specimens are and their and can be the the collected and and the of the and all to the of the and to the and of the Salix specimens used in this study are available on the Information at The numbers of the specimens used in the study and growth measurements can be found on at Fig. between growth increments on first and second stems and stem from stems on the same specimen for Salix species. Fig. in growth increments over time for Salix species. Table Salix species collected in the Disko Bay phytogeographic region of Greenland. Table of generalised additive models Table Variance for each fitted Table from Table Spearman rank correlation for annual growth increments on the first and second stems from the same specimen for Salix species. Table Spearman rank correlation analyses of annual growth increment and mean temperature in July corresponding to the year of growth. is not for the or of Information by the than be to the The is not for the or of information by the than be to the corresponding for the The with to in and in