Tree demographic and neighbourhood responses to regional environmental gradients of the northwestern United States

Studies of drivers shaping forest communities frequently include abiotic or biotic factors, while their interactive effects remain understudied. Here, we combined data on two prominent abiotic gradients, climatic moisture deficit and wildfire probability, along with tree functional neighbourhoods (i.e. trait differences of close neighbours) to assess variation in survival and growth of 56 tree species in the northwestern US. We asked two questions: (1) How does functional neighbourhood dissimilarity vary with environmental gradients? and (2) How do demographic rates of tree species in the northwestern United States vary with the interactive effects of environmental gradients and functional neighbourhood? We expected functional neighbourhoods to become more similar as environmental stress increased, due to a convergence of species towards an optimum stress tolerance strategy. We also predicted the interactive effects of abiotic and biotic factors on tree demography and high variation in species‐specific responses to these interactive effects due to divergent species life history strategies. Functional neighbourhoods defined by dissimilarities in stem conductivity, litter decomposition, resprouting ability and specific leaf area changed with climate, shifting to more diverse neighbourhoods as climatic moisture deficit and wildfire probability increased. Results supported interactive effects of the functional neighbourhood and climatic moisture deficit or wildfire probability on tree demography, but only when the identity of dominant species was considered. Species‐specific responses were highly variable in their direction and magnitude and often demonstrated opposing effects of climate and the functional neighbourhood and climatic moisture deficit on tree demography. Synthesis. Our findings show that climate and tree neighbourhood functional dissimilarity jointly impact tree demography; however, the effects are species‐specific. Results of this study highlight the need to consider the interactive effects of abiotic and biotic contexts and individual species responses to their environment to adequately understand tree persistence under current and future climate conditions.


| INTRODUC TI ON
Documenting how vegetation varies across large environmental gradients is foundational to understanding long-term persistence of forests.Given that wildfires are more frequent, extensive and severe than in recent history (Hessburg et al., 2005;Westerling, 2016) and that drought events occur more frequently and for more extended periods (Marlier et al., 2017), the persistence of forest vegetation under historically novel conditions is an open question.This is the case in the western US, where large-scale tree mortality events are becoming more frequent (Allen et al., 2010;Millar & Stephenson, 2015).The interactive effects of drought and wildfires can range from mild impacts eliciting local-scale background mortality, to severe impacts that can cause regional-scale die-off events over very short time periods (Allen et al., 2010).Despite strong evidence for links between environmental stress and tree demographic responses, these responses are still poorly understood.Likewise, concurrent plant-plant interactions remain understudied.Combining measures of climate along with a biotic context is essential to better understand variation of tree demographic responses across environmental gradients (Buechling et al., 2017;Carnwath & Nelson, 2016;Zambrano et al., 2017).
How species respond to abiotic conditions depends on the effects of close neighbours and their traits.The trait-based approach has provided a mechanistic understanding of tree demographic variation, through the consideration of morphological, physiological and phenological traits that determine species interactions (He et al., 2013) and responses to the environment (Lavorel & Garnier, 2002).Along with the trait-based approach, spatially explicit neighbourhood models have been used to determine the effects of plant-plant interactions on tree demography (Canham & Uriarte, 2006;Fortunel et al., 2018;Jourdan et al., 2020;Rubio et al., 2021;Zambrano et al., 2017Zambrano et al., , 2019)).Variation in tree performance can be explained by trait similarity and dissimilarity between a focal tree and its neighbouring trees (i.e.functional neighbourhood).Increased functional neighbourhood dissimilarity may reflect niche differentiation, with species avoiding strong interactions through limiting similarity (MacArthur & Levins, 1967) resulting in increased individual tree survival and growth.Alternatively, tree performance could increase with functional neighbourhood similarity if species share a dominant trait or traits that allow them to occupy a given habitat, minimizing fitness differences through environmental filtering (Woodward & Diament, 1991).However, species sensitivity to increased functional neighbourhood similarity might differ with environmental heterogeneity.For example, negative effects of a similar functional neighbourhood on tree performance can be exacerbated under stressful conditions (Zambrano et al., 2017), explaining tree demographic variation within species' ranges.
Differences in tree survival and growth rates will determine the persistence of an individual tree in a given location, which subsequently drives the turnover of species within communities or across the species' range (Ettinger et al., 2011).Thus, how individuals respond to their environment directly contributes to determining contractions and expansions of species within their geographic ranges.
More dominant tree species are therefore expected to respond differently to abiotic and biotic factors than rare species do, in ways that support their higher abundance across their ranges (DeMarche et al., 2018;Ehrlén & Morris, 2015).Environmental responses and species interactions could coincide to contribute to dominance in different ways.Therefore, to better predict how species ranges could shift under current and future environmental conditions, we first need to understand the drivers of individual variation in species-specific responses to the interactive effects of abiotic and biotic factors.
By combining information of the dual stress gradients of climatic moisture deficit and wildfire probability with drought-and fire-related trait data, we investigated tree demographic variation in the northwestern US.We asked two questions: (1) How does functional neighbourhood dissimilarity vary with environmental gradients?
As environmental stress increased, we hypothesized more similar functional neighbourhoods if species converge on a common set of traits associated with stress tolerance.Alternatively, we hypothesized more dissimilar functional neighbourhoods if species diverge in their responses to increased climatic moisture deficit and wildfire probability.(2) How do demographic rates of tree species in the northwestern United States vary with the interactive effects of environmental gradients and functional neighbourhood?We hypothesized that the interaction of the functional neighbourhood with climatic moisture deficit and wildfire probability would have significant effects on tree survival and growth.Furthermore, given the high variation in tree traits, biogeographic histories and relative abundances of northwestern US trees, we expected that dominant species would diverge in their responses to the interactive effects of abiotic and biotic factors, reflecting differences in their ecological strategies that allow them to co-occur in given habitats.
often applied prescribed fire to manage timber, game, berries and food and fibre products at landscape scales (Boyd, 1999;Walsh et al., 2015).We used national-scale tree data from the US Forest Service's Forest Inventory and Analysis (FIA) programme (Gray et al., 2012), which has been used extensively to assess the status and trends of forests, largely in relation to climate impacts on forest demography (Tinkham et al., 2018).We created a stem map of every tree above 12.7 cm diameter at breast height (DBH;McWilliams et al., 2012) in each of the four subplots of the FIA plot (see Appendix S1 for additional information on the FIA sampling).
Each subplot received a unique identification number, hereafter referred to as 'plot ID'.Plots were measured when a randomly located point in each 2,430-ha cell of a hexagonal grid covering the United States fell on forested land (USDA FS, 2021).We defined survival as a binary variable taking on the value 1 when a tree was alive at the second sampling event and 0 when the tree was dead from natural causes, as harvested trees were not considered.Specifically, a tree was defined as 'alive' in the presence of living parts (leaves, buds and cambium) at or above the point of diameter measurement.Any temporarily defoliated trees were considered 'alive' as well.We characterized tree growth as a relative growth rate, defined as the annualized difference in diameter stem growth (log 10 -transformed) between two sampling events, divided by diameter at the first sampling event.In total, we assessed individual trees from 37,739 unique plots in the FIA network.

| Environmental variation
Broad environmental variation exists in the northwestern US, which is physiographically divided by the north-south cordillera of the Cascade Mountain Range.The dominant environmental gradients in the region include regional variation in climatic moisture deficit and wildfire probability.The western part of the Cascades has historically fostered shorter fire return intervals and lower-severity fires than the eastern part, prior to fire suppression and fuel accumulation in recent decades (Everett et al., 2000).We considered climatic moisture deficit (CMD) as a proxy for long-term average drought.The index is based on evapotranspiration relative to precipitation, sourced for each FIA plot location as 30-year normals  from the ClimateNA model (Wang et al., 2016), which we log 10 transformed.We also modelled wildfire probabilities (Short et al., 2016), which were calculated by a Fire Simulation (FSim) system that simulates the development and growth of wildfires in modelled, hypothetical fire seasons.Results of the model are compared against historical fire statistics and patterns to assess accuracy.Specifically, we calculated the product of burn probabilities and conditional flame lengths for 'impactful' flame lengths (>1.2 m), and to avoid unrealistically extreme values, we calculated the average burn probability of all 30-m pixels in a 2000-m radius surrounding each plot location.We used these probabilities as a proxy for the long-term future influence of drier conditions on regime shifts, rather than direct predictions of future wildfire probability that would require a historical record.Values for wildfire probabilities were third-root transformed to allow for easier interpretation and discrimination of differences at the low end of the scale, where most values occurred.Lastly, we mapped climatic moisture deficit and wildfire probability for each FIA plot in the studied region and computed the Pearson's correlation coefficient between the two environmental gradients, to assess how they covary across the region.The Cascade Mountain Range causes a rain-shadow effect such that permissive, moist conditions prevail westward towards the Pacific Ocean, while more arid conditions prevail inland to the east (Figure 1).

| Trait information
We used the TRY plant trait database (Kattge et al., 2011) and the USDA PLANTS database (USDA-NRCS, 2020) to determine trait information for each of our target species.We selected eight traits which have strong a priori relationships with drought and fire mortality, and for which we could assign values to nearly all species (Appendix S2: Table S1).We used trait imputation methods based on phylogeny to fill occasional missing trait values, as described in Appendix S2.Of the eight traits we considered, two relate directly to tree-level hydraulic responses: resistance to xylem embolism (P 50 ) and stem hydraulic conductivity.We also included information on vegetative characteristics by considering specific leaf area (SLA) or the ratio of the amount of leaf area to a unit of leaf dry mass, bark thickness or the thickness of the bark layer to the cambium, and rooting depth that represents the minimum depth of soil required for good growth (USDA-NRCS, 2020).Lower SLA reduces leaf flammability (de Magalhães & Schwilk, 2012) and increases drought tolerance (Greenwood et al., 2017;O'Brien et al., 2017), while higher bark thickness reduces flammability (Frejaville et al., 2013), and higher rooting depth aids in the avoidance of both drought and wildfire, and increases water access (Brunner et al., 2015;Phillips et al., 2016).Litter decomposition rate was considered as a secondary characteristic related to fuel accumulation and flammability (Cornelissen et al., 2017), with higher decomposition rates leading to slower accumulation of flammable fuels.
Lastly, traits representing postfire persistence and regeneration included seed dry mass and vegetative resprouting ability, defined as resprouting potential following the removal of above-ground biomass (USDA-NRCS, 2020).Both traits are involved in species responses to fire and reflect a trade-off between vegetative growth and seed production (Bellingham & Sparrow, 2000).Resprouting is more frequently observed in species exposed to high-intensity fire (Vesk & Westoby, 2004), with high resource allocation towards resprouting at the expense of seed production (Bellingham & Sparrow, 2000).Comparatively, seed mass is an important factor in early seedling survival during establishment postfire (Moles & Westoby, 2004) and represents resource investment in the establishment of a future generation after disturbance.

| Principal component analysis of tree traits
We performed a principal component analysis (PCA) to determine axes of variation of the studied traits and to reduce collinearity.
We adjusted or rescaled trait values as appropriate to account for the use of multiple units in the database entries, and we performed transformations to linearize relationships for the following traits: root depth, SLA, bark thickness and seed dry mass (log); stem hydraulic conductivity (log 10 ); and litter decomposition (square-root).
We considered resprouting ability a binary numeric trait.We conducted the PCA using the prcomp function of the R 'stats' package (R Development Core Team, 2020) and created plots using the package 'ggplot2' (Wickham, 2016).See Table S2 in Appendix S1 for loadings for individual functional traits.

| The functional neighbourhood and its variation with abiotic predictors
To quantify the effect of neighbours' functional trait similarity, we calculated a neighbour crowding index of trait dissimilarity (NCIS) based on the FIA stem maps.We first determined a crowding index (CI), where the negative influence of neighbours varies as a direct function of the squared diameter of the neighbour (DBH j 2 ) and inverse function of the squared distance of the focal tree to close neighbours (d ij ) (Canham et al., 2004;Uriarte et al., 2004).This effect was summed over all neighbours j of the focal individual i: We Since functional neighbourhood similarity may vary with species richness and abundances, we used a null modelling approach to test whether the observed amount of functional dissimilarity among co-occurring heterospecific trees (NCIS) differs from a random expectation, following a previous approach (Zambrano et al., 2019(Zambrano et al., , 2020)).For the null approach, we randomized the names of species on the functional distance matrices 999 times and calculated a random neighbourhood dissimilarity index each time.The mean and standard deviation of the null distribution of 999 expected neighbourhood dissimilarity index was used to calculate standardized effect sizes (SES) for each focal tree: SES = (NCIS observed − mean null )/ SD null .Here, the SES indicated whether a focal tree differs with its neighbours more or less than expected at random, with more similar neighbours leading to negative SES values and dissimilar neighbours leading to positive SES values.We repeated this procedure for each of the 336,428 individual trees in the dataset and calculated SES values of functional dissimilarity for both the PC1 and PC2 axes for each individual tree, which will hereafter be referred to as the PC1 and PC2 functional neighbourhoods.
Next, we wanted to assess how functional neighbourhood dissimilarity and crowding varied across the environmental gradients.
To do this, we constructed simple linear models using the lm function in the 'stats' package (R Development Core Team, 2020) to regress the neighbourhood crowding index calculated for each plot (extreme values >99th percentile excluded) and SES values for the PC1 and PC2 functional neighbourhoods across the environmental gradients.

| Assessing tree demographic responses to interactive effects of abiotic and biotic predictors
To test for interactive effects of the climatic moisture deficit and wildfire probability gradients and functional neighbourhoods on tree demography, we modelled survival and growth separately as linear functions in mixed-effects models with random slopes and intercepts for each tree species: Here, 'fire' and 'moisture' are the wildfire probabilities and values of climatic moisture deficit we calculated for each FIA plot, respectively.
'Functional neighbourhood' is the SES values of PC1 functional neighbourhoods.We included only the SES values of the PC1 functional neighbourhood in the models, as the PC2 functional neighbourhood and the neighbourhood crowding index did not vary along the climate gradients.Functional neighbourhood SES values incorporate information of trait dissimilarity and the size and crowding of each heterospecific neighbour to each focal tree.We standardized fixed effects of the abiotic variables and functional neighbourhoods by subtracting the mean of each variable, from all individual trees censused, and dividing this value by the overall standard deviation of each variable.We included initial tree size at the first FIA census as a fixed effect to account for the impact of tree size/age on demographic rates and modelled it using a quadratic function following standardization to scale individual DBH measurements made during the initial census, relative to species-specific means and standard deviation.We included census year, census interval and plot ID as random effects.Census year was the year of the first re-census of each tree.We calculated census interval as the timespan between the first and second censuses, to account for differences in census frequency.We included plot ID to account for spatial autocorrelation in the models.We also included tree species with a random slope and intercept to allow for consideration of species-specific responses to the environment and for differences in the effect of the fixed effects on studied species, respectively.
We compiled separate datasets for the survival and growth models, with the survival models run across 307,136 individual adult trees of 56 species and growth models run across 270,577 individuals of 55 species (excluding individuals who did not survive to the second census event).Survival models used a binomial error distribution with a logit function, while growth models used a Gaussian error distribution with an identity function.We constructed the models using the 'lme4' package in R (Bates et al., 2015) and used a bootstrap by resampling for 999 iterations.We calculated mean estimates of the effects of tree size, the abiotic gradients, the functional neighbourhoods and their interaction on tree survival and growth across all tree species considered in the survival and growth models (56 and 55 species, respectively), along with 95% confidence intervals.
To test whether the interactive effects of abiotic and biotic factors were species-specific, we next determined demographic responses of the most dominant tree species censused in the northwestern US studied region in response to the interactive effects.To have a more direct estimation to assess species dominance, we calculated the Berger-Parker dominance index (Berger & Parker, 1970) for each of the studied species (Appendix S1: Table S1).The index measures the numerical importance of the most dominant species as follows: where N max is the number of individuals in the most dominant species, and N is the total number of individuals in the sample.We then used the 75th quantile to establish a cut-off point to determine the most dominant species, which yielded 15 tree species.For these species, we plotted effect sizes of the mean effect of each fixed effect on tree survival and growth with 95% confidence intervals.
To further investigate how tree demography varied with the interactive effects, we plotted mean survival and growth rates per We found no variation in the neighbourhood crowding along both moisture deficit (Figure 3a) and wildfire probability gradients (Figure 3b).However, we found that SES values of the functional neighbourhoods based on the PC1 axis of variation (PC1 functional neighbourhoods) increased with both climatic moisture deficit (R 2 = 0.18, p < 0.001) and wildfire probability (R 2 = 0.09, p < 0.001), indicating that neighbours became more functionally dissimilar as stress increases (Figure 3c the SES values of the PC1 functional neighbourhoods were included in the mixed-effects models as a fixed effect and will henceforth be referred to as simply 'functional neighbourhood'.

| Demographic responses to variation of the functional neighbourhood and abiotic predictors
Model results including all 56 species showed quadratic initial tree size had a significant effect on tree survival or growth (Figure 4), with a negative effect of tree size on survival, but a positive effect on growth.Contrary to expectation, we found no support for a significant interactive effect of abiotic and biotic predictors on tree demography when all species were considered together.The most variance in demographic responses was explained by random effects such as the FIA plot ID and tree species identity, while census year and interval explained very little variance (see Appendix S3: Table S1 for a summary of all model results).
When assessing the interactive effects of abiotic and biotic factors on the 15 most dominant censused species, we observed strong effects on all the species, along with high variation in the directionality and magnitude of effects on species-level tree survival and growth (Figure 5).
Overall, the interaction of moisture deficit with the functional neighbourhood had a positive effect on survival for 13 of the 15 species, and the interaction with wildfire probability had a positive effect on survival for eight species.Both interactions of the abiotic factors with the functional neighbourhood had a positive effect on growth for 10 of the 15 species.Demographic responses were further mixed in the magnitude of individual species responses, with species typically exhibiting a mix of strong and weak responses to the interactive effects.We found the strongest interactive effects on Thuja plicata, which exhibited relatively large negative survival and growth responses to the interaction of the functional neighbourhood with both abiotic factors.Furthermore, interactive effects on survival and growth, both positive and negative, were generally stronger for the functional neighbourhood and moisture deficit, than for the functional neighbourhood and wildfire probability.
Interestingly, the two most dominant tree species censused in the studied region, Pseudotsuga menziesii and Pinus ponderosa, exhibited countervailing demographic responses to the interactive effects of the abiotic factors and the functional neighbourhood (Figure 5).
Pseudotsuga menziesii showed a positive effect of the moisture deficit and neighbourhood interaction on survival, but a negative effect of the interaction with wildfire probability.Growth of P. menziesii responded positively to the interaction of the functional neighbourhood with both abiotic factors.Comparatively, Pinus ponderosa exhibited the same survival trends as P. menziesii, but for growth, P. ponderosa showed a negative effect of the interaction with the functional neighbourhood and moisture deficit and a positive effect of the interaction with wildfire probability on growth.
These differences in demographic responses were further apparent when we plotted mean survival and growth rates across the environmental gradients of moisture deficit and wildfire probability for each of the 15 most dominant censused species with the functional neighbourhood categorized according to their trait dissimilarity as 'less' , 'intermediate' Species scores on the first two trait axes obtained from the principal component analysis (PCA).The strength of correlation between each of eight traits and the two PC axes is shown by the relative length and direction of vector arrows.PC1 indicated a separation of species according to stem conductivity, litter decomposition, resprouting ability and SLA, while PC2 displayed separation based on P 50 , rooting depth, litter decomposition and SLA.Tree types exhibited a general separation along the PC1 axis.Conifers (green points) occupied lower positions on PC1 relative to angiosperm trees (blue points).See Appendix S1: Table S1 for species abbreviations.
and 'more' functionally dissimilar (Appendix S3: Figures S1-S4).Species exhibited stronger variation in survival than growth when looking at the interaction of the environmental gradients with the functional neighbourhood.For example, as moisture deficit increased, survival of Pinus contorta increased when functional neighbourhoods were more dissimilar (Appendix S3: Figure S1), consistent with the overarching trend observed across all species (Figure 3).In contrast, we found little variation in growth responses for species when considering functional neighbourhood dissimilarity along with the moisture deficit and wildfire probability gradients (Appendix S3: Figures S3 and S4).

| Functional neighbourhoods vary with environmental variation
Here, we described functional neighbourhoods based on well-known axes of trait variation.The first axis of variation included traits of stem conductivity, litter decomposition, resprouting ability and SLA, while the second axis included traits of P 50 , rooting depth, litter decomposition and SLA.These two axes of variation supported descriptions of axes of trait variation found in the literature.Along the PC1 axis, coniferous species were strongly skewed to the low values, while angiosperms (hardwood) were skewed to the high values.In assessing the traits associated with the first axis of variation, this spread is consistent with the trade-offs described in the plant economic spectra (Chave et al., 2009;Wright et al., 2004) that separates species according to conservative or acquisitive strategies.Furthermore, this division between groups is consistent with broad trends in variation in hydraulic traits (Anderegg et al., 2016), vegetative characteristics (Carnicer et al., 2013) and postfire regenerative capacity (Del Tredici, 2001).
Conservative species included coniferous species that are longer-lived, favour slower growth and resource acquisition (Maynard et al., 2022) and often display more conservative hydraulic strategies than acquisitive species (Reich, 2014;Yang et al., 2022).Coniferous species have lower stem conductivity and slow-decomposing litter with smaller, thicker leaves (Díaz et al., 2016).Conversely, acquisitive species included angiosperm species that typically favour faster growth and resource acquisition, faster-decomposing litter with large and thin Variance of functional neighbourhoods also increased as climatic moisture deficit increased, suggesting that a broader range of functional neighbourhood SES values occur at relatively higher stress.No significant trends were observed for the SES values for the functional neighbourhoods from the second PCA axis (e and f) across either abiotic gradient.
We expected more similar functional neighbourhoods if species converge on a common set of traits associated with increased moisture deficit and wildfire probability.We found that PC1 functional neighbourhoods tended to become more functionally dissimilar with increasing climatic moisture deficit and wildfire probability, with neighbourhood shifts from less dissimilar at low environmental stress, to more dissimilar under increased stressful conditions.This pattern can be attributed to multiple mechanisms which result in the same community structuring, such as environmental filtering or niche differentiation (Mayfield & Levine, 2010).Environmental filtering is predicted to lead to phenotypic clustering due to the selection of species capable of surviving and persisting in a given habitat (Kraft et al., 2015), while niche differentiation often results in high phenotypic disparity as species that are too similar are unlikely to co-occur (Chesson & Warner, 1981;MacArthur & Levins, 1967).Here, functional neighbourhoods became more dissimilar as both climatic moisture deficit and wildfire probability increased, likely the result of competing species occupying sufficiently dissimilar niche space to support species coexistence in stressful environments (MacArthur & Levins, 1967).Under conditions of increasing climatic moisture deficit, for example, variation in characteristics such as leaf size and rooting depth could allow species to coexist and reduce competition for limited resources.Decreasing leaf size could allow a tree to reduce water loss (Greenwood et al., 2017), and rooting at different depths could allow species to access alternate water resources in the soil to increase drought tolerance (Phillips et al., 2016).
The lack of a clear trend in PC2 functional neighbourhoods across the environmental gradients could suggest a scale dependency in the traits included in this study.PC1-associated traits are likely effective to describe regional-scale patterns, as significant relationships were observed at that resolution.Strategies for postfire regeneration are one example-while conifers regenerate almost exclusively from the seed bank, resprouting ability is most often observed in angiosperms (Del Tredici, 2001).Thus, variation in resprouting ability could reflect overall broader-scale patterns of tree species distributions and not fine-scale variation in local environment.By contrast, PC2-associated traits exhibited no significant variation at a regional scale, suggesting that a finer scale of within-species assessment may be necessary for traits associated with this axis.This is the case of P 50 , a measure of xylem pressure inducing 50% loss of hydraulic conductance, a trait associated Effects of initial tree size, abiotic variables, functional neighbourhood and environmentneighbourhood interaction on tree demographic rates for all studied species in the northwestern US.The impact of climatic moisture deficit (CMD) and its interaction with the functional neighbourhood (FN) on survival (a) and growth (c), and the impact of wildfire probability (fire), and its interaction with the functional neighbourhood on tree survival (b) and growth (d).Effect sizes are bootstrapped mean standardized coefficient estimates of the effect of the model fixed effects, bounded by 95% confidence intervals (CI).Filled points indicate a significant effect, and unfilled points indicate no significant effect when CI's overlap zero.Initial tree size was modelled with a quadratic function, and the CI is small enough that it is not visible around the mean.
with hydraulic safety that has been observed to vary at local scales in response to water stress (Condo & Reinhardt, 2019) and has been shown to display large intra-and interspecific variation (Anderegg, 2015).High plasticity in many hydraulic traits in response to fine-scale environmental and climatic variation (Stuart-Haëntjens et al., 2018) is expected to be essential for populations to persist long enough in their current ranges to adapt to future conditions (Leites & Garzón, 2023).Assessing the contribution of individual variation in plant traits to overall differences in demographic responses may be more important for some traits than for others (Reich, 2014) and has been observed to depend on environmental and spatial contexts as well (Auger & Shipley, 2013).

| Tree demographic variation is explained by the interactive effects of abiotic and biotic factors
With functional neighbourhoods varying along the climate gradients, we expected an interactive effect of both abiotic and biotic factors on tree demography.We found no support for this prediction when models included all 56 species.However, we found that initial tree size at the time of the first FIA census had a significant effect on survival and growth.This finding is consistent with previously described relationships of temperate tree demographic rates with tree size/age (Johnson & Abrams, 2009;Maringer et al., 2020) and environmental disturbance, with larger trees experiencing the most negative survival and growth impacts from droughts (Bennett et al., 2015).These trends are also partially concordant with the survival-growth trade-off, where the allocation of resources to one component is made at the expense of the other.Survival-growth trade-offs have been well documented in woody species (Bigler & Veblen, 2009;Negreiros et al., 2016;Russo et al., 2008;Umaña et al., 2022) and P 50 , rooting depth and SLA, traits associated with either PCA axis in this study, undergo regulation by trees to modify water acquisition and drought tolerance (Anderegg, 2015;Brunner et al., 2015;Bryukhanova & Fonti, 2013;Greenwood et al., 2017).
As droughts and wildfires continue to become more frequent and severe in the northwestern United States in the coming decades (Marlier et al., 2017;Westerling, 2016), predictions based on the well-known survival-growth trade-off may become more necessary to forecast how species will survive unprecedented climate conditions.
Our results supported the expectation for interactive effects of abiotic and biotic factors on tree demography when dominant species were considered independently.We found these interactive effects on tree survival and growth to vary greatly in both magnitude and direction.Differences in the magnitude of the interactive effects with moisture deficit and wildfire probability could be due to the additive effects of drought and wildfire in the northwestern US.Drought can increase the severity and intensity of wildfires (van Mantgem et al., 2018), while simultaneously increasing the vulnerability of trees to pathogens and pests such as bark beetles (Kolb, Fettig, et al., 2016).Interactions between disturbance factors are complex, and often result in disparate effects on tree demography in temperate forests (Benito-Garzón et al., 2013;Millar & Stephenson, 2015;van Mantgem et al., 2013;van Mantgem & Stephenson, 2007), as observed in this study.
Results also highlighted a variation in the direction of species responses to the interactive effects that is likely the result of divergent life history strategies.Species included in this study reflect a wellknown fast-slow plant economic spectrum (Reich, 2014) describing a trade-off between acquisition and conservation of resources that ultimately determine species ecological strategies (Maynard et al., 2022).Under water stress, plants with greater water acquisition and conservation should be selected against plants that acquire and use water rapidly and often experience high mortality in drier conditions (Montwé et al., 2015).While both Pseudotsuga menziesii and Pinus ponderosa display conservative strategies, the species possess divergent adapted responses to fire and water stress.Pseudotsuga menziesii is a mesic species with moderate fire resistance, whereas P.
ponderosa is highly fire resistant (Stevens et al., 2020), suggesting a trade-off associated with resistance and resilience to wildfire.Pinus ponderosa has greater fire resistance and faster juvenile growth, while P. menziesii produces cones at a younger age, likely increasing long-distance seed dispersal allowing colonization of new habitats after fires (Rodman et al., 2020).The similar magnitude and direction of both species' responses to the interactive effects with wildfire probability could suggest that their demographic responses do not vary greatly, even with disparate trade-offs of resilience and resistance to wildfire.
Furthermore, Pinus ponderosa experienced a strong positive effect on survival and a negative effect on growth to the interactive effects with moisture deficit, while P. menziesii experienced small positive responses to both interactions.These differences in the magnitude of responses to moisture deficit could be explained in part by strategies to control stomatal conductance that prioritizes minimizing water loss or maximizing growth in different ways.Under water stress, P. ponderosa is more drought tolerant than P. menziesii, and the species have been observed to use disparate divergent functional and structural regulation mechanisms improving stress tolerance, such as control of stomatal conductance to reduce water loss (Stout & Sala, 2002).Because P. ponderosa is more vulnerable to cavitation, particularly in the roots, it responds rapidly to reductions in soil water content by sharply reducing canopy and stomatal conductance, subsequently increasing drought tolerance and minimizing water loss often more effectively than P. menziesii (Kwon et al., 2018).Taken together, differences in regulation of stomatal conductance and fire resistance and resiliency could explain the observed variation in the magnitude and direction of Pseudotsuga menziesii and Pinus ponderosa demographic responses to moisture deficit and wildfire probability.
These adapted responses to water stress and fire are likely key to the ability of P. menziesii and P. ponderosa to persist and respond to the mosaic of abiotic and biotic stressors across their broad ranges.Notably, a recent study of tree demography in FIA plots in the western United States observed that both species are currently undergoing range expansions, compared with other dominant tree species, and are increasing in relative density in many forested stands (Stanke et al., 2021).Due to their broad ranges and complex biogeographic histories (Potter et al., 2015;Shinneman et al., 2016;Wei et al., 2011), populations of both species have high levels of genetic variation.This genetic variation, in combination with local acclimation (Reich et al., 2003), is the basis of intraspecific trait variation that has been observed in both species, such as in drought-related traits of SLA and branch P 50 (Condo & Reinhardt, 2019;Kolb, Grady, et al., 2016).The neighbourhood crowding index used in this study accounts for trait differences between heterospecific tree neighbours and thus does not incorporate plasticity or intraspecific trait variation.Previous studies have described an important role of genetic and phenotypic variation on tree responses to environmental stress by buffering the impact of changing environments on individual performance (Schaberg et al., 2008;Yang et al., 2021).
Furthermore, conspecific neighbourhood crowding has been described as an important force shaping forest communities (Ramage et al., 2017;Swenson et al., 2023;Zambrano et al., 2019) that needs to be accounted for to refine predictions of tree demographic responses to current and future climate changes.
In sum, our trait-based approach offers a direct investigation of tree-demography relationships, yielding strong support for the interactive effects of abiotic and biotic drivers on tree demography.The high variation observed in species-specific responses to the interactive effects of climate and the functional neighbourhood indicates that generalizing demographic trends to the community level would fail to represent important variation at the species level.Therefore, it is critical to quantify variation at multiple organization levels to adequately understand forest responses to future environmental changes.Furthermore, since the effects of the functional neighbourhoods on survival and growth varied widely across the measured environmental gradients, it is essential to include the extremes of these gradients when modelling demographic rates by expanding the geographic extent.As climate and disturbance regimes continue to shift in the northwestern US, atypical environmental conditions are set to become more common (Marlier et al., 2017).Therefore, future modelling approaches must incorporate these unprecedented conditions to better understand and predict future tree responses to changing conditions and assess the long-term persistence of forests experiencing unprecedented climate conditions.

AUTH O R CO NTR I B UTI O N S
This study was conceived by Jenny Zambrano.L. McKinley Nevins and Jenny Zambrano conducted the analyses, wrote the manuscript and performed revisions.

F
Map of northwestern US study area, and climatic moisture deficit and wildfire probability across the region.The red point and polygon indicate location of study area.Each point represents a USDA FIA Program plot, coloured by the intensity of (a) log 10 climatic moisture deficit (1.470-2.452mm year −1 ) or (b) wildfire probabilities (0.00-0.322).There is a significant positive relationship between the two gradients (r = 0.51, p < 0.001), and cool/purple colours indicate low climatic moisture deficit or wildfire probability, while warm/orange colours indicate high climatic moisture deficit or wildfire probability.Moist conditions and low wildfire probabilities occur to the north and west of the Cascade Mountains, and drier conditions with higher wildfire probability generally occur to the south and east.
then determined a trait dissimilarities (NCIS) index to determine the functional neighbourhood of a focal tree.The index includes neighbourhood crowding (CI) and trait differences among heterospecific neighbours in a plot, following Zambrano et al. (2017, 2020): where F s[i] and F s[j] are the values of the functional trait of interest for the species of the focal individual i and its heterospecific neighbour j, DBH j is the size (diameter at breast height) of neighbour j, and d ij is the distance between focal individual i and neighbour j.Increased neighbourhood crowding or trait dissimilarity may lead to more intense interactions between functionally similar individuals, where the effect of the focal individual i and neighbour j decreases with trait distance, measured by the absolute trait difference term |F s[i] − F s[j] |.We calculated separate functional neighbourhood indices for scores on the first and second axes of variation from the PCA separately, as PCA axes are mathematically independent and biologically interpreted as different axes of functional trait variation.

FIA|
plot for each of the 15 dominant tree species, across degrees of functional neighbourhood dissimilarity, and the moisture deficit and wildfire probability gradients.All values used were raw data, not predictions from the mixed-effects models.We categorized functional neighbourhood trait dissimilarity as 'less', 'intermediate' or 'more' dissimilar according to quantiles of SES values calculated for the PC1 functional neighbourhood.We calculated quantiles separately for the survival and growth datasets to account for differences in the surviving trees in each functional neighbourhood.We defined 'less dissimilar' neighbourhoods as those with SES values from the minimum to below the 25th quantile of values (−0.972 for survival, −0.973 for growth), neighbourhoods of 'intermediate dissimilarity' as those with values between the 25th and 75th quantiles, and 'more dissimilar' neighbourhoods as those with values above the 75th quantile (0.957 for survival and 0.954 for growth) to the maximum SES value.Variation in functional neighbourhood with climatic moisture deficit and wildfire probability Functional neighbourhoods were defined using the results of the first two principal components that explained 57.2% of the variance in species' measured traits (Figure2).PC1 described an axis of variation dependent on stem conductivity, litter decomposition, resprouting ability and SLA.Low PC1 values represented species with lower stem conductivity, slower litter decomposition, minimal ability to resprout and lower SLA values.High PC1 values represented species with higher stem conductivity, faster litter decomposition, greater ability to resprout and higher SLA values.Coniferous species tended to have lower PC1 values than angiosperm species (Figure 2).PC2 defined an axis of variation based on P 50 , rooting depth, litter decomposition and SLA.We found low PC2 values for species with more negative P 50 values, shallower rooting depth, slower litter decomposition and higher SLA values.High PC2 values represented species with less negative P 50 values, deeper rooting depth, faster litter decomposition and lower SLA.The PCA's upper left quadrant had subalpine and montane conifer species like Tsuga mertensiana, Larix lyallii, Pinus albicaulis and Pinus monticola, along with the broadly distributed generalist conifers Pseudotsuga menziesii and Tsuga heterophylla.The lower left quadrant featured many Californian conifers typical of subxeric or Mediterranean zones, such as Pinus jeffreyi, Pinus sabiniana, Pinus muricata, Calocedrus decurrens and Cupressus sargentii.The upper right quadrant depicted fast-growing and short-lived angiosperm trees like Prunus, Betula and Populus spp., whereas the lower right featured slower-growing and longer-lived angiosperms like Quercus spp., Chrysolepis chrysophylla and Aesculus californica.
,d).In contrast, SES values for functional neighbourhoods based on the second axis of trait variation (PC2 functional neighbourhoods, Figure 3e,f) did not vary significantly across either gradient.Because a significant relationship was observed between the SES values of the PC1 functional neighbourhoods and climate, only

F
Regressions of the SES values for the functional neighbourhoods from the two PCA axes and neighbourhood crowding index values on climatic moisture deficit and wildfire probability.Coloured lines represent linear regressions and 95% confidence intervals for mean values for each species in each FIA plot.R 2 values are the adjusted R 2 calculated for each linear regression.No significant relationships were observed for the neighbourhood crowding index (a and b) and the abiotic predictors.Significant relationships were observed between the SES values for the PC1 functional neighbourhoods and the abiotic predictors (c and d), demonstrating that as climatic moisture deficit and wildfire probability increase, the functional neighbourhood becomes more dissimilar (more positive SES values).
are observed as responses to water stress.When experiencing drought, plants may allocate more resources to traits related to water acquisition, thus increasing their potential for survival, at the expense of resource allocation to growth.Stem conductivity, F I G U R E 5 Effects of abiotic variables, functional neighbourhood and environment-neighbourhood interaction on demographic rates of 15 most dominant censused tree species in the northwestern US.The impact of climatic moisture deficit (CMD) and its interaction with the functional neighbourhood on survival (a) and growth (c), and the impact of wildfire probability, and its interaction with the functional neighbourhood on tree survival (b) and growth (d).Effect sizes are bootstrapped mean standardized coefficient estimates for each tree species, bounded by 95% confidence intervals (CI).Confidence intervals (CIs) around many individual points are small enough that they are not visible around each mean.Filled points indicate a significant effect, and unfilled points indicate no significant effect when CI's overlap zero.Species included were the 15 most dominant species censused in the studied region, according to the calculated Berger-Parker dominance index, and are organized from most to least dominant.