Increasing precipitation variability and climate-growth responses of five tree species in North Carolina, USA

We examined the effects of increasing summertime precipitation variability from 1950 to 2022 on the radial-growth responses of five tree species native to central North Carolina, USA. Tree-ring data were collected from chestnut oak, post oak, longleaf pine, shortleaf pine, and Virginia pine and processed following standard dendrochronology procedures. Adjusted latewood chronologies for each species were created and correlated with either monthly or multi-monthly combinations of summertime precipitation for above average (AA, > 1 σ above mean), below average (BA, <−1 σ below mean), and near average (NA, <−1, and 1 > σ) precipitation years. June–September precipitation variability and other summertime monthly combinations significantly increased during the study period, with a 10.2% increase in AA/BA years during the 21st century. Climate-growth correlations ranged from 0.40–0.51 using all years within the study period. However, using AA and BA years exclusively, climate-growth responses ranged from 0.44–0.71, with post oak and longleaf pine experiencing significantly higher correlations. No significant changes in climate-growth responses occurred for chestnut oak, shortleaf pine, and Virginia pine. These findings suggest the effects of increased precipitation variability on climate-growth responses are species-dependent and affected by the precipitation classification (i.e. AA or BA years). These responses help explain temporal variations in the strength of climate-growth responses, particularly for some species, and offer additional considerations for dendroclimatological research.


Introduction
A tenet in dendrochronology is the Uniformitarian Principle, which states that the climatic and biological processes that influence radial growth remain temporally stable (Speer 2010, Wilmking et al 2017).However, it is common to find that the strength of climate-growth relationships varies at multiple time scales (Carrer 2011, Schwab et al 2018) and correlations between climate and radial growth can oscillate, sometimes altering statistical significance.Among the most discussed alterations of climate-growth relationships is the divergence problem where correlations between radial growth and temperature decreased beginning in the mid-20th century in the Northern Hemisphere latitudes (Briffa et al 1998, D'Arrigo et al 2008, Zhang et al 2019).This change in climate-growth relations can significantly impair dendroclimatic reconstructions (Cerrato et al 2020, Peltier and Ogle 2020, Büntgen et al 2022).Other long-term changes in climate-growth relationships have been linked to atmospheric carbon dioxide fertilization (Knapp andSoulé 2008, Wang et al 2019), nitrogen fertilization (Cienciala et al 2018), insect infestations (Knapp et al 2013, Sangüesa-Barreda et al 2015) and forest management practices such as thinning, prescribed fire, and fire suppression (Anning and McCarthy 2013, Battipaglia et al 2014, Čada et al 2020).
Climate-growth relationships are principally predicated on the extent of stress inherent to the local climate and either modified or exacerbated by topoedaphic site conditions.In general, extreme events (typically ± two standard deviations from average), including drought (Rakthai et al 2020), temperature extremes (Briffa et al 2004), and intense rainfall (Mitchell et al 2020) are coincident with higher climate-growth correlations.Conversely, in environments where interannual climatic variability is comparatively modest and/or topoedaphic conditions protect from soil-moisture extremes, climate-growth responses are less sensitive.Thus, chronologies developed from trees sampled at sites providing more stressful conditions farther from the ecological optima of the species often express greater interannual ring-width variability and have stronger climate-growth responses (i.e.non-complacent) than trees sampled at sites providing less stressful conditions and less interannual variability (i.e.complacent) (Speer 2010).
An artifact of climate change in many locations is an increase in extreme events that modify precipitation totals, including drought (Meng et al 2022), atmospheric rivers (AR) (Payne et al 2020, Mukherjee et al 2021), and tropical cyclones (Maxwell et al 2021).For example, during 2000-21, southwestern North America experienced the most extreme drought since 800 C.E. (Williams et al 2022).In California, USA, Gershunov et al (2019) found that an evaluation of the 'five most realistic [GCM] models' suggests future increases in atmospheric river precipitation and less non-AR precipitation and will continue to contribute to the state's volatile water resources.Heavy precipitation events from tropical cyclones affecting the southeastern USA have increased since the mid-20th century (Touma et al 2019) with some storms such as Harvey (Wanders et al 2017) and Florence (Reed et al 2020) producing rainfall totals exceeding 800 mm.
The frequency of extreme events has a large spatial extent-including the southeastern USA-and has increased in the last several decades (Seneviratne et al 2021).Yet little focus has been placed on how these events may alter the climate-growth sensitivity of trees.Specifically, how would an increase in interannual precipitation variability, which likely creates less complacent growing conditions, affect climate-growth relationships.In the southeastern United States, interannual variability of summer droughts and extreme rainfall events has increased since the mid-20th century and intensified in the last several decades (Wang et al 2010).In central North Carolina, USA, there has been an increase in droughts since 1970 (Soulé 2022) whereas Patterson found that summers have become significantly warmer and drier since 1960 (Patterson 2014).The timing of these changes is important as much of the interannual variability of radial growth of southeastern tree species (e.g.longleaf pine, shortleaf pine) is associated with latewood growth, which is principally produced mid to late summer and typically expresses the highest correlation of any seasonal climatic variable.Here, we address this question by evaluating the climate-growth sensitivities of five tree species native to the southeastern USA under increasing summertime precipitation variability.Specifically, we (1) assess changes in rainfall variability during 1950-2022, (2) compare climate-growth relationships between above-average/below-average or near-average years, and (3) discuss the implications for future dendroclimatic studies.

Field and laboratory methods
We collected tree-ring data from five species: chestnut oak (Quercus montana L.) (n = 38 samples from 21 trees), post oak (Q.stellata Wang.) (n = 20, 10), longleaf pine (Pinus palustris Mill.) (n = 69, 41), shortleaf pine (P.echinata Mill.) (n = 32, 20), and Virginia pine (P.virginiana Mill.) (n = 20, 13), during 2018-2021 in the Uwharrie Mountains of central North Carolina, USA (figure 1).Chestnut oak, longleaf pine, and shortleaf pine were sampled at the Fraley Grove Tract, God Mine Branch, and Skyway Grove Tract while post oak trees were sampled from Nichols Tract (figure 1).Except for post oak samples, which were collected on a sandy upland, all the chronologies were developed from trees growing on southerly slopes to maximize climate sensitivity (Fekedulugn et al 2003).Trees were sampled following standard dendrochronological methodology (Stokes and Smiley 1996), obtaining two cores per tree from opposite sides of mature trees at approximately 1.3 m height.Trees with unusual growth forms (e.g.snapped bole, lightning or fire scars, and visible heart rot) were excluded from collection.Tree location, crown height, and coring height diameter were recorded for each tree.Core samples were dried, glued into wooden mounts, and sanded progressing from 150 µm-600 µm.Sanded cores were scanned at 1200 dpi, measured for earlywood, latewood, and totalwood using WinDENDRO (Guay 2012), and crossdated using COFECHA (Holmes 1983).Two chronologies-earlywood and latewood-were developed for each species via negative exponential detrending using the program ARSTAN (Cook and Holmes 1986) and retaining the standardized (STD) outputs.Additionally, we created adjusted latewood chronologies for each species to account for the inertia effect of earlywood on latewood following the methodology of Meko and Baisan (2001) and used the adjusted latewood output for each species (n = 5) for the analyses.

Climate data
Monthly precipitation and temperature data from North Carolina, Climate Division 5 (central North Carolina), were obtained from National Oceanic and Atmospheric Administration (NOAA) (2022) from 1931 through 2022.We excluded data from 1895 to 1930 because these values were derived from statewide averages (Guttman and Quayle 1996) and included data through 2022 for a longer period to document ongoing changes in climate variability.For central North Carolina, the strongest climate-growth relationships have been documented for adjusted latewood (Soulé et al 2021), which captures radial growth from mid-June through September (Catherwood et al 2022).Conversely, relationships between radial growth and summer temperature were either weak or nonsignificant for all species and thus temperature was not included as a variable for analysis.

Data analysis
We evaluated summer precipitation variability based on the ratio of variance for a 20 year period ending in the year of the variance over the variance for the entire period .For example, the variability value for 1950 would represent the precipitation variance of 1931-1950 (n = 20) divided by the precipitation variance of 1931-2022 (n = 92).Because the months that produced the strongest climate-growth responses varied among the five species, we created separate summer precipitation variability values based on monthly or multi-monthly combinations specific to each species.For each climate combination, we tested for temporal trends during 1950-2022 in precipitation variability using Pearson correlation as the variability data were normally distributed.Additionally, we tested for regime shifts of variance following the methodology of Rodionov (2004) and using the r-package 'r-shift' (Room et al 2023) with a standard probability of 0.95 and a length of 20 years.
Adjusted latewood and monthly or multi-monthly precipitation sums were converted to z-scores (Z = x−µ σ ) to facilitate comparison between species.We defined summer precipitation as near-average (z-score < 1.0 and >−1.0, hereafter NA years), above average (z-score > 1.0, hereafter AA years) and below average (z-score <−1.0, hereafter BA years) and then correlated (Pearson r) the years occurring in those categories with latewood growth.The z-score thresholds of < −1.0 and >1.0 represent the top and bottom 15.9% (sum = 31.8%) of all observations, while the values within 1.0 to −1.0 represent 68.2% of observations.We tested for significant differences in correlation values between groups (e.g.full chronology vs. combined AA and BA years) using a Fisher transformation test (Fisher 1921).Finally, to evaluate whether AA or BA years produced a stronger climate-growth response and if these differences varied across species, we compared the goodness of fit of AA and BA years along a trendline.We derived absolute values for standardized residuals and calculated the mean standard deviation value of AA years and BA years.

Results and discussion
Overall summer precipitation (June-September) variability increased (p < 0.05) during the study period (1950-2022) (figure 2(A), supplemental figure 2(A)).Of the four either month or monthly combination periods, variance significantly (p < 0.05) increased (figures 2(B)-(D), supplemental figures 2(B)-(D)) except for July-September (figure 2(E)).However, the variance was also significantly (p < 0.05) above average for JulySeptember during the 21st century (figure 2(E), supplemental figure 2(E)).Our regime shift analysis identified significant shifts in precipitation for each monthly or multi-monthly combination (supplemental figure 2).These regime shifts may help explain temporal variations in dendroclimatic sensitivity for each 20 year period.Our final chronologies consisted of chestnut oak (n = 38 samples from 21 trees), post oak (n = 20,10) longleaf pine (n = 69, 41), shortleaf pine (n = 32, 20), and Virginia pine (n = 20, 13).Using each chronology with all years included, climate-growth correlations ranged from 0.40-0.51(p < 0.01) based on multi-monthly periods except for post oak, which best responded to June precipitation (table 1, supplemental table 1 and figure 1).Using only AA and BA years, climate-growth correlation ranged from 0.44-0.71and significantly increased for longleaf pine and post oak (table 1, figure 3).Tree response to NA years was of lower magnitude (0.29-0.48, p = 0.05) (table 1, figure 3).Further, we found substantial variability among species in how closely AA vs. BA years fit along the slope line.In general, BA years more closely fit the trend line (standardized residuals range: 0.44-0.77,x = 0.56) than AA years (range: 0.83-1.11,x = 0.97).For the two species (post oak and longleaf pine) with the largest increases in climate-growth correlations, BA years were closer to the trend line than for the other species.In short, considerable climate-growth variability exists among the five species.This variability is likely in part a function of site conditions as the three pine species and chestnut oak were sampled from a co-occurring stand on a south-facing slope with a vertical range of approximately 60 m.The populations are partially stratified by elevation, with chestnut oak most common along ridgelines, longleaf pine and shortleaf pine most common along mid-slope, and Virginia pine common along lower-slope sites.Additionally, the post oak chronology was developed from trees growing in sandy soils on flat terrain (<5% slope).
These results suggest that an increase in AA and/or BA years either increases or does not affect climate-growth relationships but is species-dependent and may be modulated by site conditions.Species-specific root structural characteristics also may be important.Comparatively, longleaf pine and post oak both have large and deep taproots (Boyer 1990, Stransky 1990, McClain et al 2010), shortleaf pine has no taproot under stony, rocky site conditions (Gibson et al 1986, Lawson 1990) and Virginia pine and chestnut oak have shallow taproots (Carter andSnow 1990, McQuilkin 1990).We posit that the interaction of site-related soil-moisture retention coupled with a species' root structure affected the climate-growth sensitivities of the five species (e.g.Montpellier et al 2020), with species with deep taproots being most affected by increasing summertime precipitation variability.
From these findings, we make three conclusions.First, the effects of increased precipitation variability on climate-growth responses are species-dependent.Post oak and longleaf pine were significantly affected by increasing summertime precipitation variability, whereas chestnut oak, shortleaf pine, and Virginia pine exhibited no significant changes.Second, the precipitation classification (AA/BA years) that is principally driving the increase in variability is important to climate-growth responses.A higher frequency of BA years should promote stronger climate-growth responses in comparison to a higher frequency of AA years as our data suggests.However, coupled with this, the type of rainfall event is important as several studies (Knapp et al 2016, Mitchell et al 2019, Mitchell and Knapp 2022) found that 'soaker' events caused by summertime slow-moving precipitation events (e.g.stalled frontal system or tropical cyclone, stationary front) almost entirely controlled latewood growth of longleaf pine.Further, Mitchell et al (2020) found that summertime precipitation reconstructions were positively associated with the frequency of intense rainfall events (IREs) and that the occurrence of narrow latewood years was typically associated with fewer IREs.These effects may also relate to post oak, but not the other species.Last, these findings contribute to our understanding of why some species' climate-growth responses vary temporally, often at decadal to multi-decadal scales.Ring-width variability is partially a function of the amount and type of precipitation event, but the frequency of these event types can vary at decadal or longer phases affecting sensitivity.These results offer additional considerations for future dendroclimatological research when addressing the potential effects of a more variable climate.

Figure 1 .
Figure 1.The Uwharrie National Forest in North Carolina with sites labeled.

Figure 2 .
Figure 2. Summer precipitation (mm., AA years (green), BA years (red), NA years (blue) variability 1950-2022 (solid orange) based on the ratio of variance for a 20-year period ending in the year of the variance over the variance for the entire period (1931-2022).Higher ratios indicate more precipitation variability.All trendlines (dashed orange) are significant (p < 0.05) except for panel E (July-September) .

Figure 3 .
Figure 3. Correlation between adjusted latewood growth (z-score) and summer precipitation (z-score) for NA years (z-scores <1 and >−1) and AA/BA years (z-scores >1 or <−1) with accompanying trendlines.Significant (p < 0.05, one-tailed t-test) differences between correlations are noted by asterisks.Correlations for the full chronologies are shown in table 1. Mean residuals (MR) are shown for AA and BA years.

Table 1 .
Correlation values between adjusted latewood (z-score) and summer precipitation (z-score) 1950-2018 except for post oak.Superscript 'a' indicates a significant difference (p < 0.05, 1-tail Fisher transformation) between AA/BA years and NA years.
During 1950During  -2000, there , therewere 11 (21.6%)BA years for June-September, and 7 (13.7%)AA years.However, post-2000, BA frequency increased to 27.3% and AA years increased to 18.2%.Combined, AA and BA year frequency increased from 35.3% of all observations 1950-2000, to 45.5% of all observations 2001-2022; a net increase of 10.2%.These findings are consistent with the results of other studies in the southeastern United States (e.g.Wang et al 2010, Li et al 2011, Diem 2012) including North Carolina (e.g.Burt et al 2017) that have documented increasing rainfall variability and/or drought frequency although considerable spatial variability exists (e.g.Moraglia et al 2022).