Landscape position mediates drought vulnerability in California valley oak (Quercus lobata)

Future climate change will exacerbate drought stress in water-limited ecosystems. However, topography can alter the fine-scale climatic and hydrologic conditions that mediate plant response to meteorological drought. Here, with six new valley oak (Quercus lobata) tree-ring width chronologies, we assess how topography acts as a mediating factor on tree growth and drought sensitivity. Because valley oaks are known to be highly dependent on subsurface water, we predicted that trees growing in riparian sites would be less sensitive to precipitation variability due to greater access to groundwater. Trees were sampled in the Tehachapi Mountains of California across a landscape gradient of sites ranging from 375–1650 m elevation and across upland (55–69 m mean height above nearest drainage) and riparian (2–6 m mean height above the nearest drainage) hillslope positions. Interannual tree growth patterns and drought sensitivity varied substantially in association with hillslope position and elevation. Valley oak radial growth showed a consistently weaker response to precipitation at riparian sites. The influence of hillslope position on drought sensitivity varied with elevation, such that the riparian buffering effect was weakest at sites higher in the watershed and strengthened, progressively, toward the lower elevation sites with greater climatic water deficit. Upland tree growth exhibited a strong response to high-frequency interannual precipitation variability at the high elevation site, whereas trees at lower elevation upland sites responded more to low-frequency decadal trends in precipitation, possibly reflecting hydrogeological processes by which precipitation feeds groundwater lower in the watershed. Our results are consistent with groundwater-dependence of valley oak and indicate that riparian habitats are the most likely refugia for the species during prolonged drought.


Introduction
Hotter temperatures and drought, characteristic of climate change, are significant drivers of tree mortality in many ecosystems worldwide (Allen et al 2015, Hammond et al 2022).Predicting drought-induced tree mortality at regional to landscape scales remains a challenge, in part because localized environmental conditions mediate how plants experience broaderscale heating and drought.Topography is a primary control on processes of climate (e.g., orographic effects and lapse rates; Körner 2007) and hydrology (e.g.redistribution of soil water and groundwater; Fan 2015), which can create locally favorable microsites for plants (Dobrowski 2011).Topographically diverse landscapes will likely contain pockets of climate change refugia in many environments (Ashcroft 2010, Morelli et al 2016).
A growing body of evidence suggests that subsurface hydrologic processes play a key role in mediating tree mortality risk during drought (Dawson et al 2020, Sousa et al 2020, Chitra-Tarak et al 2021).Subsurface processes are critical in water-limited systems such as Mediterranean California, where trees rely on access to subsurface moisture to maintain transpiration throughout the dry season (Klos et al 2018).During the extreme 2012-2015 California drought, tree survivorship and mortality were linked to access to groundwater or deep subsurface water in remote sensing, ecohydrological, and field observations (Goulden and Bales 2019, McLaughlin et al 2020, Baldocchi et al 2021, Kibler et al 2021).Subsurface water availability is heavily influenced by topographic convergence and divergence at the local hillslope scale (Fan et al 2019).This redistribution of upslope water subsidies to downslope positions may be critical for buffering trees from anticipated future droughts in arid and semiarid environments (Hoylman et al 2018).Following related literature (e.g.Hoylman et al 2018, Martin et al 2018), in this paper we use the term 'hillslope position' to describe relative positions along the hill-to-valley hydrologic gradient (also referred to hereafter as upland and riparian positions).
Growth rings of moisture-sensitive trees can provide insight into subsurface water availability and ecosystem sensitivity to drought across complex terrain.Understanding how topography mediates tree growth response to climate has long been an important topic in dendroclimatology and more recently in dendroecology (e.g.Fritts et al 1965, LaMarche 1974, Bunn et al 2018).Much of this work has been tailored towards applications in paleoclimatology, focusing on elevational gradients where growth-climate relationships vary between water-and energy-limited environments (Salzer et al 2014).In water-limited settings, tree-ring studies indicate that wetter microsites related to fine-scale differences in hillslope position can mediate growth sensitivity to precipitation (Bunn et al 2005, Adams et al 2014).The mediating effects of hillslope topography are greatest in semiarid climates, where changes in aspect, slope, and elevation result in consequential differences in effective precipitation for forest productivity (Hoylman et al 2018, Tai et al 2020).However, assessments of the joint effects of elevation and hillslope position on growth sensitivity to drought have been limited by a relative lack of tree-ring data collected along both elevational and hillslope gradients at a single study area.
Valley oak (Quercus lobata, Née) is a longlived, groundwater dependent tree species endemic to California's oak woodlands (Lewis andBurgy 1964, Griffin 1973).Prior to their widespread clearing for human land use, valley oak trees grew most commonly in dense riparian forests but also in lower densities in upland woodland communities across complex terrain (Pavlik 1991).Arguably a keystone species because of the habitat and food it provides for wildlife, it has experienced ecosystem-scale decline due primarily to its preferred habitat being cleared and converted to use for agricultural production, along with other human and socio-environmental stressors (Brown and Davis 1991, Pavlik 1991, Browne et al 2019, Rohde et al 2021).Recent valley oak mortality has been observed and linked to extreme drought in California, although at lower frequencies than other species that share similar climatic distributions (McLaughlin et al 2020).Individual tree and grove resilience of valley oak has been linked to its tendency to occupy areas with persistent access to groundwater (McLaughlin et al 2020).Areas with reliable groundwater may therefore be climate change refugia for valley oaks in a drier future.Understanding how access to groundwater buffers valley oak from drought is becoming increasingly important as this region experiences decadal-scale drought conditions (Williams et al 2022).
Meticulous dendrochronology of valley oak has potential to provide a detailed understanding of the species' growth patterns, growth plasticity, and response to climate across complex terrain.Remnant patches of valley oak woodland usually occur at elevations less than 600 m, but in some places, like the Tehachapi Mountains, valley oak survive across a wide range of elevations (up to 1700 m; Twisselman 1967) and hillslope positions with variable water table depths (Pavlik 1991).Therefore, it could be an ideal species for investigating topographic mediation of tree growth and climate sensitivity, and for testing climate refugia predictions related to groundwater.To our knowledge, previous tree-ring research with valley oak has been quite limited.Four datasets exist on the International Tree-Ring Data Bank (ITRDB 2024), but published literature is restricted to an isotope study of one tree (Feng and Epstein 1995) and a recruitment study based on saplings (Mclaughlin and Zavaleta 2013).
The overarching goal of our research was to answer the question: How do valley oak growth patterns and growth sensitivity to meteorological drought vary along elevation and hillslope gradients?We addressed this question in three stages of research.First, we present a new network of well replicated and precisely dated valley oak tree-ring data, produced from 95 trees at six sites varying in elevation and hillslope position, along a transect spanning 1300 m (4265-feet) in elevation in the Tehachapi Mountains of California (figure 1).Second, we analyze population-level tree-growth signals in these data to provide basic information on the species' capacity for growth plasticity across complex terrain.Finally, we use these and gridded climate data, to evaluate topographic influences on growth sensitivity to meteorological droughts (i.e.precipitation deficiency) and to test predictions of climate change refugia for this species related to groundwater availability.We predicted that trees at riparian sites would be less sensitive to precipitation variability due to greater access to groundwater and that this riparian buffering effect would be enhanced at lower elevations in the watershed.

Study area
The Tehachapi Mountains divide the southern San Joaquin Valley from the upper Antelope Valley and the Mojave Desert.Valley oaks there grow at elevations from ∼400 to 1800 m, sparsely spaced on steep ridges and hillslopes down into denser groves along riparian corridor habitats (figure 1(A)).The Tehachapis are at the southern and dry edge of the species distribution, with mean annual climatic water deficit (CWD; potential minus actual evapotranspiration) estimates exceeding 1300 mm.Valley oak is dependent on groundwater (Jepson 1910, Lewis and Burgy 1964, Griffin 1973), and seeps and springs are not uncommon throughout the watershed, which has perennial surface flow and is near the seismically active Garlock Fault complex (Wood 1997).1(A); table 1).The highest elevation sites were sampled from the Cottonwood Creek Watershed, which is adjacent to Tejon Creek but is more accessible from the road.Both creeks have headwaters in the Tehachapi mountains, and a dozen years of field observations indicate that these streams had perennial surface flow at all the riparian sites along the elevational gradient until the winter of 2021-2022, when Tejon Creek lacked surface water at the middle site.At the low elevation sites, the valley geomorphology includes an alluvial fan that overlies the Tejon Canyon fault.Where Tejon Creek transects the White Wolf fault, there is a geologic dike that acts as a natural weir, restricting down valley subsurface flow and creating a localized shallow groundwater aquifer that supports many valley oak trees upstream of a constricted gorge.One core sample was extracted from approximately 15 trees at each site (table 1), and trees were sampled in open canopy stands to minimize potential influences on growth related to competition.Treering data were processed with meticulous methods of crossdating in dendrochronology (Glock 1937, Douglass 1941), and numerical ring width indices were derived using the detrending and autoregressive modeling approaches common in dendrochronology (Cook and Kairiukstis 1990;text S1).

Environmental data
To characterize relative water availability, with no groundwater observations available, we used three topographic metrics (figures 1(B)-(F)).To evaluate surface water balance, we utilized topographicallyinformed estimates of climatic water deficit (CWD; potential minus actual evapotranspiration) from the Basin Characterization Model, a California-wide product with 270 m resolution and monthly time steps from 1895 to 2020 (Flint et al 2021).To characterize catchment-scale soil moisture conditions, we estimated topographic wetness index (TWI) using a 10 m digital elevation model and the ESRI ArcMap spatial analyst toolbox.TWI, calculated as the natural log of the ratio of upslope catchment area to local slope, has been used as a proxy for soil moisture and groundwater (Beven and Kirkby 1979, Western et al 1999, Hoylman et al 2018).Toward a more effective descriptor of draining potential localized at the hillslope scale, we also used the height above nearest drainage (HAND; Nobre et al 2011, Liu et al 2020) to describe differences in hydrologic conditions between upland and riparian sites.HAND is calculated as the vertical distance of the land surface above the drainage channel to which it flows and has been used as a proxy for water table depth (Schietti et al 2014).
We acquired monthly precipitation and maximum temperature data from the 1/24 • nClimGrid product (Vose et al 2014; data available at www.ncei.noaa.gov/data/nclimgrid-monthly/access/) because long and temporally continuous climate station records were not available near the study area.We preferred nClimGrid to other gridded climate products because it has built-in bias corrections to adjust for instrumentation change (Quayle et al 1991).We used the single gridpoint closest to our sites from the gridded precipitation field to assess growth response to climate.We selected 1937-2016 as the common period of growth-climate comparison because >90% of all sampled trees extended back to at least 1937, prior to which tree-ring data sample depth declined.Eight tree-ring series that did not extend to the year 1937 were excluded from analysis.The number of tree-ring series used to calculate each site-level chronology is shown in table 1.

Statistical analysis
We compared growth patterns among all six site chronologies using Pearson's correlation coefficients.Correlations were then plotted as a function of distance between sites to assess the strength of the common signal across space.Tree-ring chronology statistics including mean ring width (mm), mean sensitivity, mean interseries correlation (RBAR), standard deviation, the order of the autoregressive model required to remove autocorrelation (AR model order), and the percent variance explained by the AR model were calculated using ARSTAN (Cook 1985).
The relationship between tree growth and monthly precipitation and maximum temperature data was assessed with simple and partial correlations using SEASCORR (Meko et al 2011).As described below, this analysis identified total precipitation from previous October to current April as the strongest influence on current year tree growth across our sites.Based on this seasonal climatic signal, we used an analysis of covariance (ANCOVA) to compare the slopes of ring-width index (RWI) vs. precipitation between upland and riparian sites (categorical variable) at each elevation.We standardized the RWI and precipitation data to a mean of zero and a standard deviation of one for this regression analysis.We use the slope coefficient of tree growth statistically regressed onto precipitation as our definition of (meteorological) drought sensitivity.Recognizing the complexity of drought beyond precipitation deficiency (Redmond 2002), we repeated growth-climate correlation analyses using monthly CWD data to further understand drought sensitivity.As the correlation patterns were nearly identical, but with a negative sign (not shown), we based our subsequent analyses and interpretation solely on precipitation-growth correlations.
To reveal the degree of tree growth coupling with regional precipitation in riparian vs. upland positions and how this relationship varies with elevation, we calculated spatial correlations between RWI and the precipitation field.The large number of correlations calculated with this approach increases the risk of type I errors (Wilks 2016).Consequently, we applied the False Discovery Rate procedure (Benjamini and Hochberg 1995) with a q-value of 0.05 to correct for test multiplicity and reduce the risk of identifying spurious correlations.Pre-whitened residual tree-ring chronologies were used for these growthclimate analyses because the amount of autocorrelation in the standard chronologies was highly variable among sites and often greatly exceeded the amount of autocorrelation in the precipitation time series.
Given the strong evidence that valley oak relies on groundwater, we wanted to assess whether the high levels of autocorrelation in tree growth at some sites reflected low-frequency variability in precipitation that could propagate through the hydrologic system to groundwater levels (e.g.Changnon 1987, Fan 2015, Van Loon 2015).To investigate whether high levels of autocorrelation in tree growth could be explained by a lagged multi-year response to precipitation, we used an inverse modeling approach by inducing lag-1 autocorrelation in the precipitation time series.The following iterative equation was used to add firstorder autocorrelation to the precipitation data: where 'RED.precipitation'refers to the reddened precipitation data in year t, λ is the amount of first-order autocorrelation added, and 'precipitation' refers to the unaltered precipitation series.We then compared the resulting 'reddened' precipitation series with the standard chronologies using correlation analyses.Finally, we used a data simulation approach to evaluate the significance of correlations between tree growth and reddened precipitation.This statistical procedure is outlined by van der Sleen et al (2018) and additional details are available in the supplementary material (text S2).We replicated this procedure for the low, mid, and high sites but limited our analyses to the upland chronologies because these sites were more sensitive to meteorological drought, as described below.Limiting our analysis to the upland chronologies also allowed us to use 1895-2016 as the common period because these sites had a stronger common signal and higher sample depth back in time than those from the riparian sites.

Water balance estimation
Landscape patterns in water balance indices are summarized in table 1 and illustrated in figure 1. CWD was found to increase with decreasing elevation, as expected (figure 1(B)).Larger TWI values occur along the stream network and near the low elevation sites, which have a relatively low slope angle (figure 1(C)).
Mean HAND values were ∼2-6 m at the riparian sites and ∼55-69 m at the upland sites, suggesting that shallow water tables occurred in riparian settings and deeper water tables occurred upslope (figures 1(D)-(F)).

Tree-ring growth patterns
Common dendrochronological statistics varied among sites and with hillslope position (figures 2 and 3).The mean sensitivity, mean interseries correlation (RBAR), and standard deviation statistics were consistently higher for upland trees than for their riparian pairings.Mean ring width was higher for riparian trees at mid and low elevation sites but marginally lower for riparian trees at the high elevation sites.These statistics did not show a clear trend with respect to elevation.Autoregressive models revealed large differences in autocorrelation structure among the site  chronologies (figure 3).Second order autoregressive models were optimal for fitting the autocorrelation structure in four of six chronologies, and a first-order and null model were selected for the low riparian and high upland chronologies, respectively.The variance explained by autoregressive models ranged widely from 0% to 63% and was generally inversely related to elevation (figure 4).A similar pattern was found for autoregressive models fit to the individual tree series (figure S1).
Correlations among site chronologies for the common period 1937-2016 revealed distinct growth variability along the elevational gradient (figure 5).Nearby chronologies showed strong positive correlations, particularly between upland and riparian pairings at the high (r = 0.79) and mid (r = 0.71) elevations.Correlations declined with distance between sites, and were weak between the lowest and highest elevation chronologies.

Climate-growth relationships
Hillslope position mediated tree growth sensitivity to October-April precipitation, which was the strongest climatic influence on growth across the sites (figures S3-S8).Across elevations, upland sites had consistently greater drought sensitivities than riparian sites, denoted by the steeper slope of RWI vs. precipitation (table 2; figure 6).However, the relative influence of hillslope position on drought sensitivity was related to elevation.An ANCOVA indicated that the differences in the slope of RWI vs. precipitation between upland and riparian sites were smallest at high elevation (df = 1, F = 0.38, p = 0.54) and became progressively larger downslope at the mid elevation (df = 1, F = 3.16, p = 0.08) and low elevation sites (df = 1, F = 9.36, p < 0.01).The low riparian site was the least drought sensitive of all sites.A similar pattern of climate correlation results was found for individual tree-ring series aggregated by site (figure S2).
Spatial correlation fields between ring width and gridded wet season precipitation revealed a large swath of positive, meaningful correlation coefficients centered over the study area.Across elevations, precipitation correlations were stronger for upland sites and weaker for riparian sites.Growth-precipitation correlations were most similar between upland and riparian sites at the high site, while the magnitude and spatial extent of correlations diverged increasingly between upland and riparian sites at middle and low elevations.No correlations between the low riparian chronology and regional precipitation passed the 95% significance test after adjusting for multiple comparisons.
Inducing lag-1 autocorrelation in the precipitation time series significantly strengthened the relationship with upland tree growth at the mid and low elevation sites but not for the high elevation site (figure 7).The strength of the growth-precipitation relationship peaked when adding λ = 0.3 to observed precipitation for the high site, λ = 0.8 for the mid site,  and λ = 0.9 for the low site (figure 7(C)).Comparing the relationship between growth and unaltered precipitation vs. reddened precipitation, the correlations increased only marginally at the high site but significantly at the middle and low sites (figure 7).The Monte Carlo significance tests indicated that strengthened relationships between tree growth and precipitation at the mid and low upland sites were likely not an artifact of adding autocorrelation (table S1).

Discussion
Our first-of-its-kind tree-ring dataset reveals how valley oak growth patterns and drought sensitivity have historically been mediated by hillslope position and elevation.Collectively, these results suggest that subsurface water availability has mediated how valley oak experienced meteorological drought over the last eight decades on average.Our findings provide novel and needed observational evidence that topography and its local controls (including geology), at hillslope to catchment scales, exert a strong influence on forest ecosystem sensitivity to climate in semi-arid regions.

Topographic effects on tree growth
We were surprised to find that valley oak is dramatically plastic in terms of its interannual growth patterns, as quantified using autoregressive modeling.At essence, with this analysis, we attempted to understand: how systematically dependent is tree-growth in a given year (lag-0) on tree growth that occurred in previous years (lags −1 to −10)?At the low elevation upland site, which must experience dramatic variability in water table depth, a lag-2 autoregressive model explains over 60% of the variability in tree growth.In contrast, at the high elevation upland site, we detected zero autocorrelation and a growth pattern that is most closely matched by unfiltered precipitation variability.Differences in growth persistence across environmental gradients have been observed in other tree-ring studies (Fritts 1974, LaMarche 1974, Bunn et al 2018) and are often attributed to biological processes such as needle retention in conifers (LaMarche 1974), different climatic factors limiting growth (Kipfmueller and Salzer 2010), or autocorrelation related to the climate system (Meko 1981).Yet, these explanations appear unlikely here because we observed large differences in persistence structure for one tree species across sites that were less than 20 km apart and shared similar climatic variability, stand structure, and a common climatic factor limiting growth (i.e.wet season precipitation).These results highlight valley oak as a tree species that appears to have great plasticity for low-order autocorrelation and capacity to persist across a wide range of environmental conditions.Other general patterns of valley oak growth varied among sites in association with elevation and hillslope position.Valley oaks in riparian positions exhibited lower interannual-scale growth variability and lower levels of among-tree growth coherence than those growing up adjacent hillslopes, while mean ring width was higher at riparian sites for two out of three elevation site pairings.These results align with the classic principle of site selection in dendroclimatology: trees growing on steep, well-drained slopes are expected to have higher growth synchrony, higher year-to-year growth variability, and narrower growth on average due to greater moisture limitation (i.e., 'sensitive' series ;Fritts et al 1965).Conversely, trees growing near streams, springs, and other water bodies that have consistent access to the water table are expected to have the opposite growth characteristics (i.e.'complacent' series; figure 2).
Pairwise correlations among the site-level chronologies revealed a wide range in the amount of shared variance in growth, including a lack of correlation among the low riparian and higher-elevation chronologies (figure 5).Declining growth synchrony at more distant sites follows the expected pattern, although we note that the rapid decline in correlation with distance observed here is uncommon for treering studies in non-treeline environments.

Changing drought sensitivity with hillslope position and elevation
Wet season precipitation had the strongest influence on tree growth in all six populations sampled, however the magnitude of response varied by hillslope position.In contrast to upland trees, riparian trees were relatively decoupled from local-to-regional precipitation (figure 6).Because prevailing climate, stand structure, and soil characteristics were similar between sites, we inferred lower drought sensitivity at riparian sites to reflect greater access to subsurface water.Previous studies identified that areas of high groundwater availability supported valley oak recruitment (Mclaughlin and Zavaleta 2012) and adult survival (Brown andDavis 1991, McLaughlin et al 2020).Our results align with these findings and provide another line of evidence that access to shallow water tables, i.e. low vertical distance to the stream channel, may ameliorate drought stress and provide relative hydrologic refugia for valley oak (McLaughlin et al 2017).
Low elevation riparian trees were the least sensitive to precipitation.Local geomorphology provides important context for this buffering effect at our low elevation sites.Here, trees were sampled above a geologic dike that likely forms a natural weir and restricts down-valley subsurface flows (figure 1(A)), resulting in locally elevated water tables.This interpretation is consistent with field observations of perennial flow in the stream channels adjacent to the low riparian site during recent severe drought.Low elevation sites occur with high TWI values (figure 1(C)), indicating the potential for greater subsurface water availability in this area.This finding supports the hypothesis that vegetation in convergent hillslope positions with large upslope catchment areas will have greater capacity to buffer regional climate change due to the accumulation of hydrologic subsidies in downslope positions (Fan 2015, Hoylman et al 2018, Tai et al 2020).
Differences in vertical distance to the stream channel altered valley oak drought sensitivity most strongly at lower-elevation sites.Upland trees were more sensitive to precipitation than riparian trees across elevations, and this difference was progressively larger at lower-elevation sites.This pattern may be explained by the location of the site pairings along a topoclimatic gradient.Drought sensitivity differed the most at the low sites, which had the highest CWD, suggesting that local aridity might influence the relative importance of hillslope position in mediating tree drought sensitivity.This result aligns with previous studies which showed that in settings with greater water limitation, topography became a stronger influence on forest carbon loading (Swetnam et al 2017), hydrometeorological processes (Hoylman et al 2019a), ecosystem productivity (Tai et al 2020), and ecosystem climate sensitivity (Hoylman et al 2019b).

Growth response to precipitation on annual to decadal scales
Our modeling results suggest that precipitation had a strong influence on tree growth at upland sites at annual to decadal time scales, depending on elevation.High elevation trees exhibited little autocorrelation and were primarily sensitive to interannual precipitation variability.In contrast, trees at mid and low elevation upland sites exhibited substantial autocorrelation and growth dynamics that were most closely approximated by quasi-decadal smoothing of the precipitation record.Trees at these lower elevation sites had a modest correlation with interannual precipitation but correlations strengthened by adding autocorrelation to the precipitation time series (figure 7).Higher correlations are a natural outcome of adding autocorrelation due to reduced degrees of freedom (Hu et al 2017).However, the significance test we used indicated that the strengthened relationship between tree growth and precipitation was likely not an artifact of adding autocorrelation (table S1; van der Sleen et al 2018).The strong growth response to reddened precipitation provides additional evidence for a groundwater signal in valley oak radial growth, as groundwater often represents a delayed and smoothed response to climatic inputs (Van Loon 2015).
Importantly, the potential groundwater signal identified here was site-specific, in that the trees with the strongest decadal memory of precipitation were located in upland hillslope positions on a low elevation alluvial fan (figure 1).At the upland end of this unconfined aquifer, we believe that decadal scale accumulations in upslope drainage are likely driving long-term fluctuations in the water table depth and therefore, moisture available to valley oak.Because valley oaks are known to use groundwater directly (Lewis and Burgy 1964, Griffin 1973, McLaughlin et al 2020), it is both plausible and reasonable to expect that growth at these lower elevation sites is driven by groundwater that has a delayed response to precipitation.This response corresponds to our current understanding of groundwater acting as a lowpass filter of precipitation variability as it propagates through the hydrological system (Changnon 1987, Fan 2015, Van Loon 2015).Under this assumption, increasing tree growth persistence (autocorrelation) downslope could indicate a short response time of groundwater to precipitation at high elevations and long response time at low elevations as precipitation inputs high in the watershed infiltrate the soil and flow downslope.This interpretation is supported by studies showing that groundwater can respond slowly to precipitation (years to centuries) in lowland environments, particularly in drier areas (Fan 2015, Hellwig et al 2020).
Individual winters with extreme precipitation are infrequent but fundamental to recharging subsurface water supplies, which can be relevant on decadal timescales in some aquifer complexes.In California, wet winters are often linked through atmospheric teleconnections to El Niño episodes, as has been the case in the current 2023-2024 winter.Indeed, across our study sites, strong positive anomalies and subsequent positive trends in tree growth were evident in major El Niño years, like 1998Niño years, like , 1983Niño years, like , and 1973.Even if erratically distributed in time, future winters with extreme wetness will likely be critical for sustaining refugial habitats in riparian hillslope positions.Our results collectively suggest that there is intriguing potential for paleohydrology with valley oak, which routinely lives for 200-400 yr.Careful sampling strategies might facilitate quantitative reconstruction of subsurface moisture conditions across space and back through time.
We hypothesize that valley oaks growing on the geographic margins of large groundwater stores (e.g.alluvial valleys fed by upslope drainage) at the lowest elevations and hot treeline could be most sensitive to variation in water table depth.Those sites may be the most vulnerable to future climate change.In contrast, riparian sites across elevations exhibit tree growth patterns that suggest strong reliability in subsurface water availability, which would provide the strongest refugial conditions from future climate change, as theorized by McLaughlin et al (2017).Conditions in higher elevation uplands are less predictable.Across the region of our study area, high elevation upland valley oak health status varied a great deal.We observed some populations that seemed to be thriving and others that exhibited clear signs of mortality and decline, presumably a function of subsurface water supply at the localized scale.

Conclusion
This study showed that across gradients of elevation, hillslope topography and watershed position strongly mediated valley oak growth and drought sensitivity.We find that riparian microsites buffered valley oak from historical precipitation variability, especially for an area at low elevation with large upslope catchment area.This result supports predicted occurrences of groundwater refugia for valley oak that could aid in conservation efforts in a drier future (McLaughlin et al 2017).Additionally, our findings demonstrate the value of tree-ring data as a tool to validate predicted occurrences of climate change refugia, a critical step to managing refugia for climate adaptation (Barrows et al 2020).Without direct observations of groundwater conditions at our sites, we were left to infer hydrologic processes using tree rings and topographic proxies for subsurface water.Future research should pair tree-ring data with in situ groundwater observations to improve processbased understanding of the hydrological processes likely to influence valley oak growth.Valley oak, which has received relatively little study with dendrochronology, may have potential for reconstructing past hydrogeologic variability and for providing unique information about past hydrological droughts and groundwater dynamics at local to regional scales.

Figure 1 .
Figure 1.Maps of study area topography, climate, and hillslope position characteristics.(A) Hillshade map of the tree-ring sampling sites across three elevations (black boxes), with riparian (blue triangles) and upland (orange triangles) hillslope position pairings for each site.(B) Climatic water deficit (CWD; annual total averaged for 1981-2010).(C) Topographic wetness index (TWI).(D)-(F) Height above the nearest drainage (HAND) for the upland and riparian site pairing at each elevation, with individual tree locations (black crosses) and aerial images from the USGS NAIP Plus dataset.

Figure 2 .
Figure 2. Photograph of valley oak growth rings from the upland and riparian site pairing at low elevation.Tree rings at the upland site are characteristic of climatically sensitive growth patterns, with high growth variability and strong crossdating.Tree rings at the upland site also show high temporal autocorrelation, as evidenced by prolonged periods of growth suppression and release that are visible by the naked eye.In contrast, riparian trees show complacent growth patterns.

Figure 3 .
Figure 3. Selected chronology statistics for the high, mid, and low elevation sites.Mean ring width was calculated from raw ring widths averaged by tree and site.Mean sensitivity (relative year-to-year variation in ring width) and RBAR (mean interseries correlation) were calculated from standard indices averaged by tree and site.Standard deviation, AR model order (order of the autoregressive model required to remove autocorrelation), and AR model R 2 (variance explained by the AR model) were calculated from the standard chronologies.All statistics were computed for the common period 1937-2016.Error bars are the standard error of mean.

Figure 4 .
Figure 4. Z-score time series of wet season precipitation (top, black line) and standard tree-ring chronologies (detrended not pre-whitened; black lines) for 1937-2016.Individual tree indices are plotted with blue lines.

Figure
Figure Tree growth comparisons among the six sites.(A) Correlation matrix among the six valley oak standard tree-ring chronologies (detrended not pre-whitened).Color bar indicates Pearson correlation coefficients.(B) Correlations between standard tree-ring chronologies plotted as a function of distance, for pairs of chronologies (illustrated with an offset).Analysis period: 1937-2016.

Figure 6 .
Figure 6.(A) Linear relationship between pre-whitened residual ring-width index chronologies (RWI) and October to April precipitation for the closest nClimGrid grid point to each site.Shaded areas represent the 95% confidence intervals of the linear fit.(B) Spatial correlations of RWI and October to April precipitation.Correlation coefficients represent tree growth coupling with regional precipitation and are mapped using approximately 5 km pixels.Black crosses indicate site location.Black dots indicate correlations that did not pass the significance test at the 95% level after adjusting for multiple comparisons.Analysis period: 1937-2016.

Figure 7 .
Figure7.Pearson correlations between upland standard chronologies (detrended not pre-whitened) and total October-April precipitation from 1895-2016.(A) Time series and correlation of tree growth (black) with unaltered precipitation (blue).(B) Time series and correlation of tree growth (black) with reddened precipitation (red).(C) Correlations of tree growth with precipitation as a function of the amount of autocorrelation added (λ) to the precipitation time series.The amount of autocorrelation added to maximize the correlation between tree growth and precipitation (using a λ of 0-1) is annotated (red).Reddening the precipitation data produces a better match with the observed decadal fluctuations in tree growth at mid and low elevation sites.

Table 1 .
Site characteristics.Elevation m: mean elevation of sampled trees in meters; HAND: mean height above the nearest drainage of sampled trees in meters; TWI: mean topographic wetness index of sampled trees; CWD: climatic water deficit (annual total averaged for 1981-2010) of sampled trees in millimeters.See the methods section for descriptions of HAND, TWI, and CWD.