Climate sensitivity of shrub growth across the tundra biome

,

Shrubs are woody perennial species that can live from decades to centuries.In seasonal climates, they form annual growth rings, allowing analysis of radial growth over time.Many shrub species are widely distributed across the tundra biome and are often dominant, owing to their canopy height, longevity and ability to outcompete low-growing plants.With wide geographic distributions and annual growth records, shrubs are ideally suited for quantifying tundra vegetation responses to climate warming.Assembled annual growth records from sites across the tundra biome provide a unique opportunity to test competing hypotheses of shrub responses to climate change over the past half-century.
Previous ecological monitoring and dendroecological studies have identified temperature, growing season length, summer precipitation and snow cover as important variables explaining spatial and interspecific variation in shrub growth 1,10,13,14,[16][17][18] .However, there is a lack of consensus regarding which climate variables best explain growth across all tundra ecosystems.We therefore do not know whether climate-growth relationships are consistent in direction, strength and magnitude among species and among sites where plant composition, climate trends and environmental parameters di er.At present, most large-scale vegetation models assume high climate sensitivity and a uniform growth response to warming among shrub species and populations 8,19 .These models predict pronounced positive climate feedbacks as a result of tundra vegetation change 5,8 .Yet, if shrub growth responses to climate are constrained, then changes in shrub dominance should vary regionally, and feedbacks across the tundra biome as a whole could be weaker than predicted at present.
We quantified the climate sensitivity of shrub growth-that is, the strength of relationship between annual growth and climate variables (including temperature and precipitation, specific calculations described below)-to test four hypotheses: (1) The greatest climate sensitivity of growth should occur at northern or high-elevation range edges if plant performance is more climate limited in the harsher growing conditions at range edges than in the centre of species distributions [20][21][22] .(2) Climate sensitivity of growth should be greatest in the centre of species distributions if populations growing under more stressful conditions at range edges have evolved conservative life history strategies limiting their ability to respond when conditions improve 23 .(3) Climate sensitivity of growth should vary along spatial gradients if the response of species to warming is limited by other factors, such as soil nutrients, soil moisture or biotic interactions 21 .Alternatively, (4) climate sensitivity of growth could be uniform across the tundra biome.
We synthesized published and unpublished time series of shrub growth across the tundra biome.Our data set extends beyond previous analyses by including sites across the circumpolar Arctic, comprising dwarf, low and tall canopy species, and encompassing 60 years of annual-resolution shrub growth.We used crossdated, radial and axial growth measurements spanning 1950-2010, collected at 37 sites, and for 25 shrub species in 8 genera.We analysed climategrowth relationships for 46 genus-by-site combinations using linear mixed models to estimate climate sensitivity, with 33 candidate climate models as predictors of shrub growth increments.All data were normalized before analysis and model terms included seasonal temperatures and precipitation as fixed e ects and year as a random e ect (see Supplementary Information).
We calculated four complementary indices of climate sensitivity from the mixed model analysis for each genus-by-site combination: (1) the di erence in Akaike information criterion (AIC) between the best climate model and a null model (1AIC), (2) the R2 for the best climate model, (3) the absolute value of the slope of the relationship between growth and summer temperature and (4) the proportion of individuals that had significant linear relationships between growth and summer temperature (the best predictor from the overall analysis).We assessed these indices of climate sensitivity across abiotic (wet day frequency, soil moisture, growing season length) and biotic gradients (distance to range edge and specieslevel maximum canopy height, see Supplementary Information).In Fig. 1, we report both 1AIC and model slopes to illustrate spatial variation in climate sensitivity (all indices reported in Supplementary Fig. 12).In Fig. 2 we report the percentage of models indicating climate (temperature or precipitation) sensitivity in the model comparison analysis; Fig. 3 shows relationships between all four climate-sensitivity indices across di erent gradients.
Climate-growth relationships were not uniform across the tundra biome (Fig. 1), contrasting with the common assumption used in Arctic vegetation models 19 .Overall climate sensitivity was high: 76% (35/46) genus-by-site combinations exhibited climatesensitive growth (Supplementary Table 5).Summer temperature variables best explained variation in shrub growth across the 46 genus-by-site combinations and 33 climate models (Fig. 2), with 46% (21/46) genus-by-site combinations showing positive growthsummer temperature relationships and 17% (8/46) showing negative relationships (Fig. 1 and Supplementary Table 5).Individuallevel climate sensitivity of growth varied considerably: 5-97% of individuals at each site and ⇠36% of all individuals showed significant summer temperature sensitivity (Supplementary Table 5).A moving window analysis demonstrated the relatively consistent  4).The shaded colouring indicates the percentage of models that were considered climate sensitive (either positive or negative) for each of the four categories of climate variables for each of the genus-by-site combinations with a 1AIC value of greater than 2 between the given climate model and the null model for all one-parameter models in the model comparison analysis.
climate sensitivity of shrub growth over time, despite the increase in sample size in recent years (Supplementary Fig. 13).
Climate sensitivity of shrub growth was highly heterogeneous across the tundra biome (Fig. 1).Climate sensitivity was greatest in the northwest Russian Arctic and northern Europe, and more heterogeneous among sites in North America (Fig. 1), where many sites exhibited weak relationships between growth and summer temperatures (Supplementary Table 5).Across gradients, climate sensitivity was greater in wetter sites relative to drier sites as indicated by the number of days with precipitation and satellitederived soil moisture (Fig. 3a,b).We found support for our first hypothesis: shrubs growing near their northern latitudinal or elevational range limits showed greater climate sensitivity, as did taller (>50 cm maximum canopy height) versus shorter species (<50 cm; Fig. 3c,d).Overall, shrub climate-growth relationships were not uniform across the tundra biome, but instead varied according to soil moisture, species canopy height and geographic position within the species ranges.
Our results highlight the importance of soil moisture as a driver of climate sensitivity of shrub growth.In tundra environments, soil moisture is influenced by several factors including rainfall during the summer, snow distribution, duration and melt, permafrost status, soil properties and topography, making it more challenging to quantify than climate variables 24 .We observed high climate sensitivity and positive climate-growth relationships at many sites with high soil moisture (Figs 1 and 3); however, eight sites exhibited negative summer temperature-growth relationships (Fig. 1) and some of these sites were located in areas with high soil moisture at the landscape scale (Supplementary Fig. 14).These negative relationships with summer temperatures could indicate drought limitation of growth in woody species, which can occur in both wet and dry landscapes 25 , although in sites with increasing soil moisture, standing water can also lead to reduced growth and shrub dieback 6 .
Previous studies have identified summer temperatures as an important driver of vegetation change 1,13,14,26 , but the role of soil moisture is less often examined.A recent synthesis of two decades of ecological monitoring (the International Tundra Experiment Network) showed that increased shrub abundance was most pronounced at sites that had experienced summer warming and in wetter versus drier sites 1 .In addition, landscape-level studies of shrub change in northern Alaska showed greater increases in wet floodplains relative to well-drained hill slopes 3,10 .Our study, using a new circumarctic dendroecological data set consisting of almost exclusively di erent sites from those in previous studies, also demonstrates broad geographic patterns in the climate sensitivity of shrub growth, with higher climate sensitivity at sites with higher soil moisture.Taken together these results suggest that, with continued warming 11 , potentially more variable precipitation 11 and uncertainty in the future soil moisture regime 11,24 , water availability or flooding could play an increasingly important role in limiting future shrub expansion.However, analyses of plant water availability in tundra ecosystems are limited by the lack of high-resolution soil moisture data 24 .
In our study, climate sensitivity of shrub growth was greatest at the northern or elevational range margins of individual species (Fig. 3).Climate sensitivity of shrub growth was thus greatest at the transition zone between tall and low shrub tundra (Fig. 1).The largest ecosystem transitions in shrub dominance could occur at these mid-Arctic latitudes, rather than at the northern limits of the tundra biome as a whole.The patterns of climate sensitivity of growth in tundra shrub species can be compared to patterns observed in treeline ecotones.Half of the latitudinal and elevational treelines studied so far have advanced poleward or upslope, often associated with warming 27 .Temperature sensitivity of tree growth has been found to be highest at the upper or northern-most margin of the forest-tundra transition zone 20,27 and moisture sensitivity to be highest at southern or lower range edges 28 .Our results suggest that for tundra shrubs, both temperature and soil moisture control growth at range edges, whereas further from the range edge other factors such as competition, facilitation, herbivory and disease 21 may be more important.Herbivore densities vary spatially and temporally across our study locations 12,29 , and this could be one of the factors explaining the variation in climate sensitivity.Relationships between the climatic and biotic factors influencing growth are probably complex and deserve greater study.
We find that climate sensitivity of growth is greater for tall shrubs, than for low-statured shrub species (Fig. 3b).This has important implications for Earth system models, as changes in tall shrub cover will contribute more markedly to ecosystem-climate feedbacks than changes in dwarf shrub cover 8 .Increases in canopy height and abundance of taller species relative to lower-stature shrub species was a major finding of two recent syntheses of plot-based ecological monitoring and passive warming experiments; however, these studies did not include taller alder and willow species 1,26 .Tall shrub species may more readily exploit favourable climate conditions, particularly at the transition zone from tall to low shrub tundra, by competing for limited light and nutrient resources 30 .In particular, in contrast to previous work that has not explicitly tested biogeographic patterns of climate sensitivity 1 , our analysis demonstrates that the climate sensitivity of both tall and dwarf shrub species was often greater towards colder range margins (Fig. 3a).This results in an overall pattern of high climate sensitivity at mid-latitudes, but also high climate sensitivity for some species growing in the High Arctic (Fig. 1).
In conclusion, climate sensitivity of shrub growth is generally high at sites across the tundra biome, which provides strong evidence for the attribution of tundra shrub increases to climate warming 4 .However, pronounced increases in shrub growth with warming are unlikely to occur in all regions, and the greatest shrub growth responses are instead likely to occur in the transition zone   between tall-and low-statured shrub tundra and where soil moisture is not limiting.A pressing research question is whether temperatureinduced increases in shrub growth will continue to occur at current or accelerated rates or whether factors such as water availability, herbivory, pathogen outbreaks, nutrient limitation or fire will play a greater role in limiting future tundra shrub expansion.Further experimental manipulations of temperature 26 , moisture regime, biotic interactions and atmospheric CO 2 concentration are necessary to predict shrub growth responses under future environmental scenarios.Improved soil moisture records 24 (resulting from, for example, ESA http://www.esa-soilmoisture-cci.organd NASA http://smap.jpl.nasa.gov)and other locally influenced climate and biological variables and expanded networks of in situ tundra vegetation observations 1 will further improve predictions.
Only with a combination of enhanced ecological monitoring, multifactorial experimentation and additional data synthesis can we make improved projections of vegetation feedbacks to future climate change.

Figure 1 |
Figure 1 | Climate sensitivity across the tundra biome.The size of the circle shows the strength of the summer temperature sensitivity as indicated by the 1AIC.The colour of the circles indicates the direction of the relationship with summer temperature variables.Locations with multiple circles indicate study sites where multiple species were sampled.The coloured regions indicate the bioclimatic zones of the Circumpolar Arctic Vegetation Map (http://www.geobotany.uaf.edu/cavm).

10 Figure 2 |
Figure 2 | Comparison of climate models.Summer temperature models were more frequently climate sensitive than other temperature or precipitation models in the model comparison analysis of 46 genus-by-site combinations and 33 climate models (Supplementary Table4).The shaded colouring indicates the percentage of models that were considered climate sensitive (either positive or negative) for each of the four categories of climate variables for each of the genus-by-site combinations with a 1AIC value of greater than 2 between the given climate model and the null model for all one-parameter models in the model comparison analysis.

Figure 3 |
Figure 3 | Climate sensitivity across gradients.a-d, Greater climate sensitivity was found for shrub species growing at sites with a greater number of wet days (a), higher soil moisture (b), closer to northern/elevational range limits (c) and for species with higher maximum canopy heights (d).e,f, Climate sensitivity varied among genera (e; Supplementary Table2) and between the two growth measures of stem increments and annual ring widths (f).The lines and associated p values indicate beta regression of the dierent climate-sensitivity metrics; the shaded areas indicate the 90th quantile of these regressions and the error bars (e,f) indicate the range of values.The distance to the range edge (c) is the distance between the sampling location and the northern or elevation range edge for each species converted to relative latitudes (see Supplementary Information).This gives an index of how far a sample population is located from the maximum extent of the distribution of that species either northward in the Arctic or upslope in alpine tundra.

2
Figure 3 | Climate sensitivity across gradients.a-d, Greater climate sensitivity was found for shrub species growing at sites with a greater number of wet days (a), higher soil moisture (b), closer to northern/elevational range limits (c) and for species with higher maximum canopy heights (d).e,f, Climate sensitivity varied among genera (e; Supplementary Table2) and between the two growth measures of stem increments and annual ring widths (f).The lines and associated p values indicate beta regression of the dierent climate-sensitivity metrics; the shaded areas indicate the 90th quantile of these regressions and the error bars (e,f) indicate the range of values.The distance to the range edge (c) is the distance between the sampling location and the northern or elevation range edge for each species converted to relative latitudes (see Supplementary Information).This gives an index of how far a sample population is located from the maximum extent of the distribution of that species either northward in the Arctic or upslope in alpine tundra.
Figure 3 | Climate sensitivity across gradients.a-d, Greater climate sensitivity was found for shrub species growing at sites with a greater number of wet days (a), higher soil moisture (b), closer to northern/elevational range limits (c) and for species with higher maximum canopy heights (d).e,f, Climate sensitivity varied among genera (e; Supplementary Table2) and between the two growth measures of stem increments and annual ring widths (f).The lines and associated p values indicate beta regression of the dierent climate-sensitivity metrics; the shaded areas indicate the 90th quantile of these regressions and the error bars (e,f) indicate the range of values.The distance to the range edge (c) is the distance between the sampling location and the northern or elevation range edge for each species converted to relative latitudes (see Supplementary Information).This gives an index of how far a sample population is located from the maximum extent of the distribution of that species either northward in the Arctic or upslope in alpine tundra.