Radiation, Air Temperature, and Soil Water Availability Drive Tree Water Deficit Across Temporal Scales in Canada's Western Boreal Forest

Changes are projected for the boreal biome with complex and variable effects on forest vegetation including drought‐induced tree mortality and forest loss. With soil and atmospheric conditions governing drought intensity, specific drivers of trees water stress can be difficult to disentangle across temporal scales. We used wavelet analysis and causality detection to identify potential environmental controls (evapotranspiration, soil moisture, rainfall, vapor pressure deficit, air temperature and photosynthetically active radiation) on daily tree water deficit and on longer periods of tree dehydration in black spruce and tamarack. Daily tree water deficit was controlled by photosynthetically active radiation, vapor pressure deficit, and air temperature, causing greater stand evapotranspiration. Prolonged periods of tree water deficit (multi‐day) were regulated by photosynthetically active radiation and soil moisture. We provide empirical evidence that continued warming and drying will cause short‐term increases in black spruce and tamarack transpiration, but greater drought stress with reduced soil water availability.


Introduction
Water, carbon, and energy fluxes regulated by the boreal biome play a crucial role in global water and energy budgets and climate regulation (Bonan, 2008;Price et al., 2013).By 2100, climate change is expected to warm the boreal biome by 3-8°C resulting in water limitation and stress (Berner et al., 2005;Gauthier et al., 2015), with implications for boreal forest health (Allen et al., 2010;Peng et al., 2011) and resilience to disturbance (Baltzer et al., 2021;Whitman et al., 2019;Yang et al., 2022;Yi & Jackson, 2021).Without an increase in precipitation, forests in the boreal biome will experience more frequent and severe droughts over the next century (Wang et al., 2014;Zhang et al., 2019).Drought (i.e., water stress) is the duration and frequency of water limitation, and the intensity of water deficit (Cook et al., 2014;McDowell et al., 2008;Seidl et al., 2017).In boreal forests, like other systems, various controls contribute to drought stress in trees, including air temperature (Barber et al., 2000;Reich et al., 2022;Ruiz-Pérez & Vico, 2020), reduced rainfall, low soil water content (Hogg et al., 2005;Peng et al., 2011) and increased atmospheric water demand (Babst et al., 2019;Beck et al., 2011;Mirabel et al., 2022).These same controls contribute to increased evapotranspiration (ET) which in turn reduces water available for plant uptake, exacerbating drought stress in trees (Zhao et al., 2022).
Traditionally, dendrochronology has been used to explore the effect of drought on boreal tree species by examining annual growth increments.Studies have shown that tree growth is stimulated by greater water availability combined with warmer temperatures (D'Orangeville et al., 2016;Wang et al., 2023;Zhang et al., 2019) but limited by high atmospheric water demand, low water availability (Babst et al., 2019;Lloyd & Bunn, 2007;Mirabel et al., 2023;Sniderhan et al., 2021) and/or the timing and severity of drought stress (Mood et al., 2021).However, there are few studies assessing the impact of environmental controls on water stress in boreal trees at higher temporal resolution (from hours to days) (e.g., Balducci et al., 2019;Dulamsuren et al., 2023;Maillet et al., 2022).Thus, it remains unclear if the environmental controls of annual growth increment are also controlling water stress at smaller temporal scales.
Automatic stem dendrometers measure variation in tree radius, and are suitable for quantifying temporal dynamics of water stress in trees (De Swaef et al., 2015;Drew & Downes, 2009;Zweifel et al., 2000;Zweifel & Hasler, 2001).Stem shrinkage occurs as water is transported from phloem and roots to the canopy for transpiration via the xylem.Swelling typically occurs at night with water uptake and stem refilling (Steppe et al., 2015).Factors hindering stem refilling can lead to prolonged periods (days to weeks) of stem dehydration and drought stress in trees (Oberhuber, Hammerle, & Kofler, 2015;Oberhuber, Kofler, et al., 2015;Salomón et al., 2022;Schäfer et al., 2019;Zweifel et al., 2005).Therefore, the amplitude of the reversible shrinkage/swelling is a reliable measure of tree water deficit (TWD; Zweifel, 2016;Dietrich et al., 2018).Previous studies using automatic stem dendrometers have focused on the relationship between stem radial change (growth) and environmental controls (Dulamsuren et al., 2023;Maillet et al., 2022), with few exploring the drivers of TWD across multiple sites and temporal scales (but see Salomón et al., 2022).To our knowledge this study represents the first investigation into the driver of TWD in two widespread boreal species in Canada's western boreal forest.
Our goal is to quantify the environmental controls of tree water stress in black spruce (Picea mariana, (Mill.)B.S. P.) and tamarack (Larix laricina (Du Roi) K. Koch), from southern to northern treeline in Canada's western boreal forest during June-August by asking: Are the environmental controls of tree water deficit consistent across temporal scales in black spruce and tamarack?Identifying the controls of water stress in trees provides a more holistic understanding of tree health, resilience, and ecosystem dynamics in the face of rapid environmental changes.

Site Descriptions
Our work took place at five black spruce dominated forest stands spanning ∼2,000 km in Canada's western boreal biome (Figure 1a).From south to north, Old Black Spruce (OBS) is a permafrost-free site located ∼100 km northeast of Prince Albert, SK at the southern edge of the boreal treeline.Scotty Creek (SCC) is a subarctic boreal peatland complex characterized by sporadic discontinuous permafrost with forested permafrost plateaus interspersed with sparsely treed permafrost-free wetlands, located ∼50 km south of Fort Simpson, NT (Dearborn & Baltzer, 2021;Quinton et al., 2019).Baker Creek (BAC) is ∼15 km north of Yellowknife, NT with extensive discontinuous permafrost and forests located between bedrock outcrops (Morse et al., 2016).Similar to SCC, Smith Creek (SMC) is a subarctic boreal peatland complex characterized by extensive discontinuous permafrost (Natural Resources Canada, 2016), ∼15 km from Wrigley, NT.Havikpak Creek (HPC), is a woodland located in the forest-tundra ecotone ∼10 km south of Inuvik, NT where permafrost is continuous (Krogh et al., 2017;Table S1 in Supporting Information S1).

Stem Radius Change Measurements
We employed over-the-bark automatic circumference dendrometers (model DC-2 and 3, Ecomatik, Munich, Germany, Figure S1 in Supporting Information S1) to measure stem radius change (∆SR) in mature, healthy black spruce and tamarack trees.Black spruce ∆SR was measured at all five sites, while tamarack ∆SR was measured at OBS, SCC and BAC.Measurements of ∆SR represent irreversible radial growth and reversible shrinkage/swelling associated with hydration in living tissue (measured as TWD), internal tension in dead xylem, and stem thermal contraction/expansion (Steppe et al., 2015;Zweifel et al., 2021).To mitigate thermal contraction, freezing-induced bark degradation and hygroscopic swelling effects on ∆SR, measurements were confined to the peak growing season (June, July, August), and dead bark was removed (Oberhuber et al., 2020;Zweifel et al., 2021;Zweifel & Hasler, 2001).Dendrometers exhibited high thermal sensitivity (<0.2 μm/°C) and were not corrected for thermal stem expansion or the deformation of mature xylem under tension (Irvine & Grace, 1997).Thermal contraction of wood tissue due to colder air temperatures may have led to stem shrinking.
Data from up to four dendrometers were logged using HOBO data loggers (UX120-006M, Bourne, MA), and averaged at half-hourly intervals from 2018 to 2020.Measurements began between June 6-15, 2018 at SCC, BAC, SMC, and HPC.Gaps in the data ≤3 hr were filled using linear interpolation, while longer gaps remained unfilled.Species-averaged TWD was derived ΔSR by removing growth trends using the "Zero Growth Concept" (Zweifel et al., 2016, Figure S2 in Supporting Information S1).The Zero Growth concept posits that positive TWD values indicate dehydration and insufficient turgor pressure for cell division and enlargement, precluding growth, while TWD of zero signifies fully saturated stem tissue (Dietrich et al., 2018;Drew et al., 2011;Hinckley & Bruckerhoff, 1975;Lockhart, 1965;Zweifel et al., 2005Zweifel et al., , 2016)).Stems typically refill within a 24-hr cycle, with consecutive days of TWD above zero indicating stem dehydration, drought stress and growth suppression (e.g., Nehemy, Maillet et al., 2021;Nehemy, Benettin, et al., 2021;Zweifel et al., 2021).Full saturation (TWD = 0) was assumed at the beginning of the measurement period in June of each year due to large hydrological inputs of snowmelt in April-May across sites (Nehemy et al., 2022).

Eddy Covariance and Micrometeorological Measurements
Stand-level evapotranspiration (ET) and potential environmental controls of TWD including photosynthetically active radiation (PAR; measured as photosynthetic photon flux density; μmol m 2 s 1 ), air temperature (T air ; °C), vapor pressure deficit (VPD, derived from T air and relative humidity; kPa), and rain (mm) were measured continuously on micrometeorological towers located near the instrumented trees (OBS: Barr et al., 2012, Nehemy et al., 2023, SCC and HPC: Helbig, Pappas, & Sonnentag, 2016;Helbig, Wischnewski, et al., 2016, BAC: Spence & Hedstrom, 2018, SMC).Only peak radial-growing season (June, July, and August) measurements from 2018, 2019, and 2020 were used.Rain data at HPC were obtained from the Environment and Climate Change Canada station located at the Inuvik airport (<1 km from the site, ECCC, 2023).Near-surface soil moisture (θ; %), representing the primary rooting depth for black spruce and tamarack (Lieffers & Rothwell, 1987), was measured from the top 5-15 cm of the ground surface at each site, and calibrated using site-specific coefficients, or coefficients specific to the organic matter at the soil surface (e.g., Spence et al., 2020;Warren, 2015).

Analyses
We used wavelet coherence analysis to identify the correlation and temporal structures between paired time series of environmental drivers (ET, PAR, VPD, T air , rain, and θ) and the species-averaged TWD (Harmon et al., 2021;Hatala et al., 2012;Leštianska et al., 2020;Torrence & Compo, 1998).Hereafter, the predictors are denoted by X and TWD by Y.The wavelet coherence is computed by normalizing the continuous wavelet cross-spectrum (the product of the wavelet transform of X and Y) by the individual continuous wavelet power spectra (Torrence & Compo, 1998), producing a measure of correlation from 0 to 1 analogous to an R 2 (Yu & Lin, 2015).
Statistical significance of the wavelet coherence was estimated using Monte Carlo simulations (n = 1,000) with a significance level of α = 0.05 (Grinsted et al., 2004;Torrence & Compo, 1998).Analysis of species-specific wavelet coherence was done for each site in 2018, 2019, and 2020 (depending on data availability) using the Morlet wavelet implemented in the "WaveletComp" R package (Rosch & Schmidbauer, 2018).An ensemble was created for each species and site by time-averaging the continuous wavelet coherence and averaging across years.
Granger causality (or G-causality) analysis (Granger, 1969) was used to determine if the coherence identified in the analysis described above can represent causal relationships.G-causality, originally developed for economics and now used in many disciplines including geosciences (e.g., Detto et al., 2012;Hatala et al., 2012;Li et al., 2022;Salvucci et al., 2002;Wu et al., 2020;Zhang et al., 2022), exploits the temporal order of events and statistical prediction to infer cause and effect (Barraquand et al., 2021;Granger, 1988).A causal relationship exists between variables if time series X is better predicted when including the past values of time series Y than using the past values of X alone (or vice-versa; Granger, 1969).We applied the unconditional bivariate Gcausality to the spectrally transformed series to identify causal relationships as a function of periods ranging from 0.5 to 1.3 days and ≥1.3 to 14 days, herein referred to as daily and multi-day, respectively (Detto et al., 2012).Analysis was performed for each species, at each site in 2018, 2019 and 2020, where data were available.Confidence intervals were evaluated using a bootstrap threshold of α = 0.05 with the "grangers" R package (Farnè & Montanari, 2022).Data processing, statistical analysis and graphics were performed in R (version 3.6.3,R Core Team, 2020).

Tree Water Deficit and Environmental Conditions During the Study Period
The study period was generally cooler and wetter than climate normals between 1980 and 2010, with only SCC and BAC had warmer than average temperatures (in 2018 and 2020, respectively).Figure 1 illustrates the time series of tree water deficit and the associated environmental variables for OBS in 2020.A diel pattern typical for boreal trees during the peak growing season, with greater TWD near mid-day and lower TWD at night, was more pronounced during periods of water deficit (consecutive days without TWD returning to zero), compared to periods with no water deficit (periods with TWD returning to zero; Figures 1b and 1c).Rainfall was associated with an increase in θ and decreased TWD, T air , VPD, PAR and ET (Figures 1b-1h).The time series for other sites are presented in Figures S3-S7 in Supporting Information S1.

Periodicity in Tree Water Deficit and Coherence With Environmental Controls
The wavelet analysis revealed common and significant periodicities in tamarack and black spruce TWD across sites and years, with high wavelet power at daily and multi-day periods (Figure S8 in Supporting Information S1), that is typical for environmental and eco-physiological series (Stoy et al., 2009).Wavelet analysis best identifies oscillations with a period larger than the data collection time step, and shorter than the data collection period (Harmon et al., 2021;Torrence & Compo, 1998).Frequencies below 0.5 days had low wavelet power (Figure S8 in Supporting Information S1) and were statistically significant.Thus frequencies <0.5 days and beyond 2 weeks were not evaluated in the wavelet coherence and G-causality analysis.Wavelet coherence evaluated between TWD and the potential environmental controls was highest at daily periods and at multi-day periods (Figure 2 and Figures S7-S14 in Supporting Information S1).
Daily TWD showed highest coherence with PAR (not measured at BAC) followed by VPD, ET, and T air (Figures 2a-2h).At OBS, SCC, BAC and SMC, higher daily PAR, VPD, ET, and T air correlated with higher TWD.At HPC, greater TWD was also correlated with higher VPD, ET and T air but with lower PAR (Figure S15 in Supporting Information S1).High daily θ at SCC, BAC, SMC, and HPC was correlated with higher TWD however, the coherence was low (Figures 2a-2d, 2f, and 2g).By contrast, OBS had the lowest θ of all five sites but high coherence with daily θ TWD (Figures 2e and 2h).Rain had a weak coherence with daily TWD across all sites (Table S3, Figure S15 in Supporting Information S1).
At multi-day periods, wavelet coherence was strongest for PAR, followed by VPD, θ (not significant at HPC), rain (not significant at SCC and BAC), and ET (not significant at HPC or BAC; Figures 2a-2h).Higher PAR, VPD and ET were correlated with greater TWD while high θ and rainfall was correlated with a decrease in TWD.Air temperature had the lowest coherence with multi-day TWD, and was only significant at SCC (Figures 2a-2h, Table S3 and Figure S15 in Supporting Information S1).
From the G-causality analysis, we found differences in the interactions among the potential environmental controls at the daily and multi-day periods.Daily, T air and PAR controlled TWD (Figures 3b and 3d).Higher T air and PAR caused greater TWD in both species.At multi-day periods, PAR and T air remained important controls of TWD (Figures 3h and 3j), however the causal effect was diminished compared to the daily period.The causal relationship between TWD and VPD varied between time periods.VPD had a strong causal effect on TWD, at daily and multi-day periods, but the causal effect of TWD on VPD was stronger in multi-day periods (Figures 3c  and 3i).Tree water deficit also has a strong causal effect on ET, a relationship that was consistent across time scales (Figures 3a and 3g).Increasing TWD in both species, but especially tamarack, led to greater ET daily and at multi-day periods.Soil moisture had a consistent causal effect on TWD daily and during multi-day periods (Figures 3f and 3l), however, the coherence was low at the daily timescale (Figures 2a-2h).There was little indication that rain had a direct causal effect on TWD at either timescale (Figures 3e and 3k), despite rain having high coherence with TWD during multi-day periods.

Discussion
Our goal was to quantify the environmental controls that shape tree water stress in Canada's western boreal forests.Central to our investigation was to determine if daily TWD, and extended periods of stem dehydration were determined by the same environmental controls across a large latitudinal range and between two of the most common tree species.Our analysis revealed that the environmental controls of TWD in black spruce and tamarack varied across temporal scales.At the daily timescale PAR, T air , and VPD, exhibited the highest coherence and strongest causal influence on TWD.However, during multi-day periods, PAR and θ emerged with the highest coherence and strongest causal effect on TWD 3h and 31).
Photosynthetically active radiation, T air , and VPD influence transpiration in boreal (e.g., Oogathoo et al., 2020;Pappas et al., 2018;Patankar et al., 2015;Van Herk et al., 2011), temperate (e.g., Harrison et al., 2020) and tropical tree species (e.g., Brum et al., 2023;Ghimire et al., 2022).While PAR correlates with stem water storage in ponderosa pine ((Pinus ponderosa), Harmon et al., 2021), this relationship remains unstudied for boreal tree  Granger-causality (G-causality) between black spruce and tamarack tree water deficit (TWD) and evapotranspiration (ET; a and g), photosynthetically active radiation (PAR; b and h), vapor pressure deficit (VPD; c and i), air temperature (T air d and j), rain (e and k) and soil moisture (θ, f and l) at the daily and multi-day (1.3-14 days) periods.The first boxplot in each panel represents the G-causality of environmental variables on TWD (e.g., ET → TWD).The second boxplot in each panel represents the G-causality of TWD on environmental variables (e.g., TWD → ET).Individual points represent the causal effect of environmental controls on black spruce (circles) and tamarack (triangle) TWD or of the causal effect of TWD on environmental controls at each site.Dashed lines indicate the significance threshold (α = 0.05).Environmental controls have a causal effect when the G-causality of the environmental control on tree water deficit is higher than the G-causality of tree water deficit on the environmental control (and vice versa).
species.Higher PAR promotes stem shrinking by increasing water loss via transpiration when days are longer (i.e., larger window for photosynthetic uptake; Ahedor et al., 2018).Being higher in latitude, HPC had a lower potential for PAR, thus the relationship between TWD and PAR was weaker at this site.Air temperature and VPD increase with PAR, further enhancing transpiration and stem shrinking (Sack & Holbrook, 2006;Taiz & Zeiger, 2002;Zweifel & Hasler, 2001).Daily TWD peaked after evapotranspiration and transpiration (derived from sap flux density measurements at OBS, SCC, SMC and HPC between 2018 and 2021, Figure S17 in Supporting Information S1), indicating environmental controls of transpiration dominate daily TWD dynamics.Increasing TWD with elevated PAR, T air , and VPD had repercussions at the forest stand scale, driving an increase in ET at the daily period (Figure 3a).
The expected increases in peak growing season T air and VPD in the western boreal biome (Gauthier et al., 2015;Mirabel et al., 2023) will likely result in greater daily TWD and subsequently transpiration (Van Herk et al., 2011).This could enhance forest productivity, as seen with temperature-induced growth in areas of the boreal biome (Babst et al., 2019;D'Orangeville et al., 2016;Reich et al., 2018;Wang et al., 2023).However, soil water content could play a crucial role, limiting productivity by constraining water available for uptake, stem refilling, transpiration and growth.Throughout the study period, water availability (i.e., rainfall and θ) exhibited minimal daily cycles and remained consistently high compared to climate normals between 1980 and 2010.Consequently, the observed influence of rainfall and θ on TWD was likely constrained and remained above the species' wilting point at the daily time scale .
While soil moisture did not influence daily TWD, it significantly controlled TWD over multi-day periods, with lower θ strongly correlated with increasing TWD in black spruce and tamarack .This observation aligns with Maillet et al. (2022) who reported similar correlation between stem radius change of three boreal conifers and water availability (rainfall) at the weekly timescale.Trees rely on internal water reserves and uptake to support normal functioning, including transpiration (Čermák et al., 2007;Mészáros et al., 2011;Nehemy, Maillet et al., 2021;Nehemy, Benettin, et al., 2021;Steppe et al., 2006;Zweifel & Hasler, 2001).As TWD increases between rainfall events, trees increasingly rely on θ for transpiration and stem refilling.With narrow drought tolerance margins (Choat et al., 2012), declining summer precipitation and rising temperatures in western Canada by the end of the 21st century (Zhang et al., 2019), could heighten TWD, drought stress and growth reduction in boreal trees.The relationship between multi-day TWD and θ was strongest in the southern site (OBS) where θ was the lowest among the five sites, suggesting greater occurrence water deficit in the southern boreal forest compared to the northern region (Figures 2e and 2h and Figure S3 in Supporting Information S1).Additionally, areas with discontinuous permafrost may face drought stress associated permafrost thaw-induced active layer thickening and drying in the rooting zone of boreal trees (Helbig, Wischnewski, et al., 2016;Sniderhan et al., 2021;Zhang et al., 2008).
In the warmer and drier climate projected for western Canada, intermittent wet and warm periods may enhance growth in black spruce and tamarack, potentially increasing productivity (D'Orangeville et al., 2018;Wang et al., 2023).Some species, such as black spruce, have demonstrated low sensitivity of basal area increment to summer θ (Girardin et al., 2021), and prolonged stem dehydration has not consistently resulted in stress-induced growth reduction (Belien et al., 2014).Hence, boreal trees could exhibit resilience to anticipated drier and warmer conditions in the near future.
Species resilience may stem from specific physiological strategies that enable trees to maintain function and growth during water deficit (Bartlett et al., 2019).Despite black spruce and tamarack transpiration contributing minimally to ET (Perron et al., 2023;Warren et al., 2018), increasing TWD caused greater ET over multi-day periods (Figure 3g).This suggests that species maintain transpiration despite reduced soil water availability and increasing TWD, as supported by consistent daily sap flux density rates during decreased θ at OBS (Figure S18 in Supporting Information S1).Greater decoupling of ET from water deficit has been observed in moist sites where plants regulate stomata and hydraulic conductance, or access water deep in the soil to maintain transpiration during dry periods (Giardina et al., 2023).With shallow rooting zones, black spruce and tamarack likely did not access deeper soil water after the top 10-20 cm were depleted.Stomatal regulation to limit water loss during transpiration likely facilitated continued transpiration in both species, enabling sustained ET during multi-day periods with higher TWD (Dusenge et al., 2021;Giardina et al., 2023).
We anticipated that VPD would drive tree dehydration in black spruce and tamarack over multi-day periods of TWD.Vapor pressure deficit drives transpiration, impacting tree water stress, (Harmon et al., 2021;Nehemy, Maillet et al., 2021;Nehemy, Benettin, et al., 2021;Salomón et al., 2022;Xue et al., 2022), increasing TWD in larch species (Tian et al., 2018) and decreasing growth in spruce species (Mirabel et al., 2023).We found a significant coherence between VPD and TWD, with higher VPD consistently causing increased TWD in black spruce and tamarack across sites, highlighting the contribution of atmospheric dryness to multi-day periods of water stress.However, the relationship between TWD and VPD was found to be bidirectional and site-specific, with TWD exerting stronger causal effect on VPD (Figure 3i).The effect of TWD on VPD was most pronounced at OBS, the denser southern forest with lower θ and higher T air, ET, and transpiration compared to the sparsely forested sites further north (Figures S19 and S20 in Supporting Information S1).Boreal species, in response to high atmospheric water demands, have demonstrated greater coupling between stem water content and VPD, and sensitive stomatal control (Dusenge et al., 2021;Ewers et al., 2005).Thus, at OBS, black spruce and tamarack likely adjusted stomatal conductance before VPD peaked, resulting in changes in TWD preceding those in VPD.

Conclusions
This work determined that the environmental controls on tree water deficit (TWD) in black spruce and tamarack vary across temporal scales.PAR, T air , and VPD emerged as crucial controls of TWD at the daily scale, contributing to transpiration-driven stem shrinking.In contrast, PAR and water availability (θ) controlled multi-day TWD, where tree dehydration became pronounced.With warming and drying predicted for Canada's western boreal biome, trees will see greater daily TWD with benefits to productivity through increased transpiration and ET.However, with decreased frequency and intensity of rainfall events, trees will experience more numerous, longer, and more extreme periods of stem dehydration during dry periods, with potential consequences forest vulnerability to drought.This research underscores the challenges posed by warming and drying trends in boreal forests, highlighting the delicate balance between soil water availability, air temperature, and tree dehydration.

Figure 1 .
Figure1.Five sites (a) in Canada's boreal forest biome (shaded green) where black spruce (Picea mariana) and tamarack (Larix laricina) were sampled for tree water deficit: Old Black Spruce (OBS), Scotty Creek (SCC), Baker Creek (BAC), Smith Creek (SMC) and Havikpak Creek (HPC).Half-hourly measurements of species-averaged tree water deficit (TWD) in black spruce (b) and tamarack (c), and potential environmental controls at Old Black Spruce in 2020.Environmental controls include evapotranspiration (ET, d), photosynthetically active radiation (PAR, e), vapor pressure deficit (VPD, f), air temperature (T air , g), rain (h), and soil moisture (θ, h).Weekly average TWD, ET, PAR, VPD and T air are represented with gray lines (b-g).Breaks in the soil moisture data represent data that was periodically unavailable.The blue shading across panels represent large periodic rainfall events.

Figure 2 .
Figure 2. Wavelet coherence levels (0-1) between potential environmental controls and tree water deficit of black spruce and tamarack at Old Black Spruce (OBS), Scotty Creek (SCC), and Baker Creek (BAC), and black spruce at OBS, SCC, BAC, Smith Creek (SMC) and Havikpak Creek (HPC).Sites are listed from south (bottom) to north (top).Potential environmental controls include evapotranspiration, photosynthetically active radiation (PAR), vapor pressure deficit (VPD), air temperature, rain and soil moisture.Circles with dark shading indicate significant (α = 0.05) coherence between tree water deficit and environmental condition.

Figure 3 .
Figure 3. Across-siteGranger-causality (G-causality)  between black spruce and tamarack tree water deficit (TWD) and evapotranspiration (ET; a and g), photosynthetically active radiation (PAR; b and h), vapor pressure deficit (VPD; c and i), air temperature (T air d and j), rain (e and k) and soil moisture (θ, f and l) at the daily and multi-day (1.3-14 days) periods.The first boxplot in each panel represents the G-causality of environmental variables on TWD (e.g., ET → TWD).The second boxplot in each panel represents the G-causality of TWD on environmental variables (e.g., TWD → ET).Individual points represent the causal effect of environmental controls on black spruce (circles) and tamarack (triangle) TWD or of the causal effect of TWD on environmental controls at each site.Dashed lines indicate the significance threshold (α = 0.05).Environmental controls have a causal effect when the G-causality of the environmental control on tree water deficit is higher than the G-causality of tree water deficit on the environmental control (and vice versa).