Growth Response of Red Oaks to Climatic Conditions in the Lower Mississippi Alluvial Valley: Implications for Bottomland Hardwood Restoration with a Changing Climate

Abstract

Bottomland hardwood forests (BHFs) offer a wide range of ecosystem services that are of high environmental and socioeconomic value. Yet, nearly 70% of BHFs in the southern United States have been lost during the past 100 years primarily due to land use change including agricultural expansion, calling for restoration efforts. We estimated the statistical relationship of the annual radial growth rate of three red oak species with climatic conditions and tree age using the tree ring data collected from a BHF plantation in the Arkansas Delta region. These species were Cherry bark oak (Quercus pagoda), Shumard oak (Quercus shumardii), and Nuttall oak (Quercus texana). The destructive sampling method was employed to obtain tree growth data and the cross-dating method was used for tree age determination. A log-linear regression model was estimated to uncover the statistical relationship between annual tree ring growth rate and climatic conditions. We identified the most critical time windows of climate variables that affect the growth of these trees. We found that the average temperature in October of the previous year and the minimum temperature between December of the previous year and January of the current year were positively associated with the radial growth rate in the current year although the maximum temperature from January to August and total precipitation from April to July of the current year were negatively correlated with the growth rate. Compared to Cherry bark and Shumard oaks, Nuttall oak was less sensitive to a rise in the minimum temperature between December and January. The projected climate change is likely to create slightly more favorable overall climatic conditions for these oak species in the region. Our findings suggest that these three red oak species are well suited for the study region for restoring BHFs, especially with a changing climate.

1. Introduction

The bottomland hardwood forests (BHFs) in the floodplains of the southeastern United States are critical for the provision of a wide range of ecosystem services such as carbon sequestration, soil and water conservation, timber, wildlife habitat, biodiversity, and recreational resources, among others [1,2,3,4]. These forests once covered over 12 million ha in the region; however, land-use change including agricultural land expansions has resulted in a 70% loss of their area and severe degradation of their quality in the past 100 years [4,5,6,7,8]. This unique type of ecosystems is threatened by human and natural forces, and changes in the BHFs in turn can have environmental and socioeconomic consequences [4,6,8]. This calls for the restoration of BHFs in the region. Yet, guidelines for successfully restoring these forests are lacking. For instance, what tree species will be most suitable for the BHF restoration? How does the growth of different tree species respond to climatic conditions especially climate extremes given the projected climate change? Answers to these questions can enhance the successful establishment and maintenance of BHFs in the region.

Among the most important tree species in BHFs in the Lower Mississippi Alluvial Valley are red oaks [2,9,10], which provide abundant food for wildlife and lumber and veneer for the production of high-quality furniture, flooring, and other products. For instance, acorns of red oaks are a critical food resource for mallards and wood ducks as they contain high levels of saturated fatty acids and other micronutrients [11]. Red oaks are also among the most valuable timber species grown in the region [2,9,10,12,13]. Thus, bottomland red oaks are desired by both wildlife and forest managers for enhancing wildlife habitat and producing high-quality wood.

However, few studies have attempted to quantify the growth/yield of red oaks in BHFs although tremendous efforts have been made in assessing the growth/yield, as well as their response to climate variability, of highly commercialized softwood timber species like loblolly pine in the region [14,15]. Studies that quantify the linkage between tree growth in BHFs and climatic conditions are even rarer. Analyzing annual tree ring growth has been widely employed to investigate the relationship between tree growth and climate variations, which can then be used to disclose and project the impact of climate change on tree growth [16,17,18,19,20]. Global climate change is expected to increase the global aggregate forest productivity largely due to higher CO₂ concentration in the atmosphere although the impact will vary across regions, tree species, and forest types [21]. In the United States, forest productivity is projected to decrease in most regions including the southern region due to more frequent and intensive drought [19].

Several studies have probed the effects of climatic conditions on the radial growth of oak species [22,23,24,25,26,27,28]. Most of these studies used monthly average temperature and total precipitation in the previous and current growing seasons to estimate potential correlations between climate variations and annual ring width increment. The growth response to climate climatic conditions was found to vary with tree species and study sites. However, to our knowledge, none of the previous studies has targeted red oaks species of BHFs in the Lower Mississippi Alluvial Valley. There is a lack of understanding about the suitability of red oaks for restoring BHFs in the region in general and the growth response of red oaks in BHFs to changes in precipitation and average and extreme temperatures in particular.

This study aims to assess the impact of climatic conditions including average and extreme temperatures and precipitation in different periods of the year on the annual radial growth rate of three red oak species planted in the Arkansas Delta region. These species are Cherry bark oak (Quercus pagoda), Shumard oak (Quercus shumardii), and Nuttall oak (Quercus texana). We collected and used tree ring data that document annual tree radial growth over 15 years alongside historical weather data to disclose the statistical linkage between the annual radial growth rate of the three oak species and climatic conditions. We incorporated climate factors into tree growth modeling, expanding traditional approaches for modeling tree and forest growth and yield. Additionally, we employed a unique approach for selecting the most appropriate/critical window (time period), rather than a predetermined time period (i.e., daily, monthly, seasonal, annual), for each climate variable based on its influence on tree growth. Our findings derived from the field-based data enrich the literature on the influence of climatic conditions on the growth rate or productivity of individual trees and forest stands while providing guidelines for ensuring the successful restoration of BLFs in the region and beyond with a changing climate.

2. Materials and Methods

2.1. Study Site

The study site is situated at the University of Arkansas’ Division of Agriculture Pine Tree Research Station (AgPTRS) in the Arkansas Delta region (35°07′48″ N, 90°55′45″ W) (Figure 1). The AgPTRS is the largest research station in the state with a total area of 4795 ha (11,850 acres), of which approximately 1416 ha (3500 acres) are tillable and the remaining 3379 ha (8350 acres) are BHFs, pine plantations, and layout areas. This region has rich and fertile soil, with the Calloway, Calhoun, Henry, and Loring soil series being predominant. It is one of the world’s most productive agricultural regions. In addition, the temperate climate makes the region suitable for the growth of many crop and tree species, with a mean January temperature of 2.0 °C, an average July temperature of 27.6 °C, and an average annual precipitation of 1173 mm.

A red oak plantation of 12.1 ha (30 acres) was established on an agricultural (soybean) field that was enrolled in the Conservation Reserve Program (CP3A). The dominant soil types on the site are silt loam, Calloway silt loam, and Zachary silt loam. Three red oak species, Cherry bark oak (Quercus pagoda), Shumard oak (Quercus shumardii), and Nuttall oak (Quercus texana) were planted in February 2004 after site preparation using the subsoiling (ripping) process to a depth of 0.91 m (36 inches) in November 2003. The intervening 3 months between ripping and planting allowed the trench to close. The seedlings were hand-planted in 12 plots with four replications for each species with a 3.05 m × 3.05 m (10 feet × 10 feet) spacing between tree rows. Due to space limitations, plot sizes were reduced from 0.81 ha (2 acres) to 0.65 ha (1.6) acres. The 12 plots with a spacing of 30.48 m × 30.48 m (100 feet × 100 feet) were established by randomly assigning one of the three red oak species. A one-tenth-ha (one-fifth-acre) square measurement subplot was established in the center of each experiment plot. The square measurement subplots were accompanied by an 18.29 m (60 feet) buffer zone on all sides of the subplots to minimize the edge effects.

2.2. Data

Sixty (20 trees per species) oak trees were randomly selected from each of the four replications (5 per replication) within the measurement subplots established in the plantation and harvested for tree ring analysis in 2020. Diameter at breast height (DBH) and total height were measured for the sampled trees. DBH was measured in two crossing directions and average DBH was computed for each sample tree. Those trees were felled at ground level by leaving a stump height of approximately 0.3 m (1 foot). The felled stems were then divided into sections of 1.83 m (6 feet), and disks of 2.5–5.1 cm (1–2 inches) in thickness were extracted at the lower and upper ends of each section. These collected disks were marked with species, tree number, and disk number (representing the stem position where it was extracted) and then brought to the laboratory for further analysis. In the laboratory, the disks were sanded prior to measurement. The cross-dating method [29] was used for the determination of tree age (count of rings), and ring width was measured for each growing season with two different directions and then average width was calculated for each.

Daily meteorological data were retrieved from the Applied Climate Information System (ACIS) operated by the National Oceanic and Atmospheric Administration Regional Climate Centers (NOAA RCCs) [30]. The ACIS facilitates the use of complex meteorological data from multiple data collectors and generators under one management system. For the observed historical weather data, we used the gridded weather dataset derived from the Parameter-elevation Regressions on Independent Slopes Model (PRISM) [31]. We retrieved the meteorological data from 2005 to 2018 using the coordinates of the center of the study site. We also downloaded the projected future climate data from the ACIS dataset generated by applying the Localized Constructed Analogs (LOCA) downscaling method to 32 global climate models (GCMs) [30]. We obtained the climate projections for two greenhouse gas (GHG) emission scenarios, Representative Concentration Pathways (RCPs) 4.5 and 8.5, at our study site for the period of 2020–2099. The RCPs are a set of scenarios that depict different potential trajectories of GHG emissions and atmospheric concentrations [32,33]. RCP 4.5 is a moderate emission scenario with global GHG emissions peaking around 2040 whereas RCP 8.5 represents the highest emission scenario with continuously rising GHG emissions throughout the 21st century. Then, bias correction was performed to remove systemic model errors in GCMs. Various methodologies for bias correction have been developed including quantile mapping (QM) and scaled distribution mapping (SDM) [34,35,36,37,38]. These methods calibrate the distributions of modeled climate variables using their corresponding observations. We used the SDM method with a gamma distribution because it can maintain the changes and trends of the raw climate variables projected from the models [35]. The projected future temperatures and precipitation in the study area are shown in Figure 2 and Figure 3, respectively.

2.3. Selecting the Most Appropriate Windows for Climate Variables

We attempted to select climate variables that represent the most critical limiting factors such as temperatures and water availability for tree growth for possible inclusion in our regression model. Tree ring growth has been found to depend on the regional pattern of temperature and precipitation [16,39,40,41]. As such, climate drivers considered for inclusion in our model were average temperature (T_avg), maximum temperature (T_max), minimum temperature (T_min), and total precipitation (P). We calculated the monthly average, maximum, and minimum from the daily average, maximum, and minimum temperatures, respectively and the monthly total precipitation from the daily precipitation. The monthly temperatures and precipitation were then aggregated over different windows (time periods) for each climate variable.

Different windows of a climate variable can have different impacts on tree growth. Hence, for each climate variable considered in this study we further determined its most appropriate/critical window that had the most statistically significant impact on tree growth. The window was an annually repeating period over which annual tree growth was significantly affected by a climate driver. To reduce the unnecessary large number of inessential windows, we selected the maximum time span of 15 months with several months before and after the growing season. The selection of the maximum time span was based on the aggregate effect of climate drivers on tree growth [16,41,42]. The range and location of the most appropriate/critical window were then identified within the maximum time span.

We used the climwin package [43] in R to determine the most appropriate/critical windows of the climate variables. In this process, we attempted to identify the window in which the climate variable can best explain the variations in the tree ring growth using a nonlinear model that assumes that annual tree ring growth rate varies with tree age and climatic conditions. We prepared all the permutations of the starting and ending months for each climate variable within the maximum time span. Using the climwin package, we identified and estimated the best-fit nonlinear model based on the lowest value of AIC (Akaike Information Criterion corrected for small sample size). We also applied 5-fold cross-validation to calculating AICs to avoid the over-fitting of the model. For each climate variable, we chose a single best window when there was a distinct choice; otherwise, we selected multiple windows that were almost equally appropriate/critical.

2.4. Estimating the Statistical Relationship between Annual Tree Ring Growth Rate and Climatic Conditions

We regressed the annual radial growth rate of tree rings on climate variables, tree age, and tree species. The regression model takes the following form:

$$ln\left(G{R}_{it}\right)={\alpha }_0+{\alpha }_1{D}_i+{\displaystyle \sum }_{j=1}^J{\beta }_{j}Cli{m}_{jt}+{\displaystyle \sum }_{j=1}^J{\gamma }_j{D}_i*Cli{m}_{jt}+\delta t+{\epsilon }_{it}$$

(1)

where i denotes individual trees, t represents the number of years since the tree was planted; j stands for each of the climate variables included in the model (j = 1, 2, …, J); ln denotes natural logarithm; D is a dummy variable representing the three red oak species; Clim represents the climate variables; * denotes the interaction between the two variables; α, β, γ, and δ are the regression coefficients, and ε is the error term. The annual growth rate of tree rings was calculated using the ratio of the annual radial change in the tree ring to the accumulated tree ring growth (i.e., the radius of the tree ring at the beginning of the year). The climate variables considered include the average, minimum, and maximum temperatures and total precipitation in their respective most appropriate/critical time windows.

The regression model was specified using the bidirectional stepwise variable selection approach. This approach starts with the base model consisting of one constant (foreword selection) or all potential explanatory variables (backward elimination), then attempts to add or delete one variable in each step/iteration to improve the statistical quality of the model. This process continues until adding or deleting one more variable can no longer improve the model. Metrics used for gauging the quality of the model include p-values from extra-sum-of-squares F tests, Bayesian Information Criterion (BIC), and Akaike Information Criterion (AIC) [44]. The AIC values were further adjusted/corrected for the sample size [45]. The corrected AIC (AICc) was then used along with other indicators for detecting multilinearity such as multicollinearity tolerance and the variance inflation factor (VIF). If multilinearity was detected, then the causing variable was dropped from the model, and the variable selection process continued without the eliminated variable. The statistical analysis was conducted using packages in R, primarily the olsrr package [46].

2.5. Estimating the Impact of Climate Change on the Tree Ring Growth

We estimated the impacts of future climate change on the annual radial growth rate of the three red oak species based on the regression results (Equation (1),

Table 1. The estimated statistical relationship between the annual radial growth rate of three red oak species and climate and other variables. (Year—the number of years since the trees were planted; T_{min(pDec, cJan)}—the minimum temperature between December of the previous year and January of the current year; T_{avg(pOct, pOct)}—the average temperature in October of the previous year; P_{(cApr, cJul)}—the total precipitation between April and July of the current year; T_{max(cJan, cAug)}—the maximum temperature between January and August of the current year; T_{min(pDec, cJan)}:D_NUT—the interaction term between the minimum temperature from December of the previous year to January of the current year and the dummy variable indicating Nuttall Oak).

Independent Variable						Collinearity Statistics
	β	Standard Error	Standard Beta	t-Value	p-Value	Tolerance	VIF
Constant	51.463	1.918		26.84	<0.001
Year	−0.026	0.001	−0.761	−25.68	<0.001	0.57	1.75
Tmin(pDec, cJan)	0.005	0.001	0.172	5.54	<0.001	0.52	1.92
Tavg(pOct, pOct)	0.007	0.001	0.168	5.71	<0.001	0.58	1.72
P(cApr, cJul)	−0.011	0.003	−0.151	−3.99	<0.001	0.35	2.84
Tmax(cJan, cAug)	−0.007	0.003	−0.108	−2.57	0.010	0.29	3.49
Tmin(pDec, cJan):DNUT	−0.001	<0.001	−0.155	−6.87	<0.001	0.99	1.01

Note: F(6,763) = 204.3, p < 0.001, R² = 0.616, n = 770.

) and the projected future climate change (Figure 2 and Figure 3). From Equation (1), we have

$${\widehat{GR}}_{kt}=e^{{\widehat{\alpha }}_0+{\widehat{\alpha }}_1{D}_i+{\displaystyle \sum }_{j=1}^J{\widehat{\beta }}_{j}Cli{m}_{jt}+{\displaystyle \sum }_{j=1}^J{\widehat{\gamma }}_j{D}_i*Cli{m}_{jt}+\widehat{\delta }t}$$

(2)

where ˆ denotes “estimated” and e is the base of the natural logarithm. Then, at a given future time (t) with a projected climate change (denoted by clim_jt1 for all j’s) compared to the reference case of no climate change (clim_jt0), the ratio of change in ${\widehat{GR}}_{kt}$ due to the climate change can be expressed as

$$\frac{{\widehat{GR}}_{kt1}-{\widehat{GR}}_{kt0}}{{\widehat{GR}}_{kt0}}=e^{{\displaystyle \sum }_{j=1}^J{\widehat{\beta }}_j\left(Cli{m}_{jt1}-Cli{m}_{jt0}\right)+{\displaystyle \sum }_{j=1}^J{\widehat{\gamma }}_j{D}_i*\left(Cli{m}_{jt1}-Cli{m}_{jt0}\right))}-1$$

(3)

Thus, Equation (3) can be used to estimate the ratio of change in the annual radial growth rate in response to changes in future climatic conditions.

3. Results

3.1. Most Appropriate Time Windows for Climate Variables

From all the possible time windows within the maximum time span, we found the most appropriate/critical window (starting and ending months) for each of the climate variables (Figure 4). It was the month of October of the previous year (T_{avg(pOct, pOct)}) for average temperature, from January to August of the current year (T_{max(cJan, cAug)}) for maximum temperature, from December of the previous year to January of the current year (T_{min(pDec, cJan)}) for minimum temperature, and from April to July of the current year (P_{(cApr, cJul)}) for precipitation. These selected windows had distinctive ΔAIC_c values (change/difference in AIC_c) when compared to all other windows. In other words, there were no other windows that were nearly as good as these windows for the climate variables considered.

3.2. Relationships between Annual Radial Growth Rate and Climatic Conditions

The bidirectional stepwise variable selection process led to the inclusion of six independent variables in the final regression model. They were the number of years since the trees were planted, the average temperature in October of the previous year (T_{avg(pOct, pOct)}), the maximum temperature between January and August of the current year (T_{max(cJan, cAug)}), the minimum temperature between December of the previous year and January of the current year (T_{min(pDec, cJan)}), the total precipitation between April and July of the current year (P_{(cApr, cJul)}), the interaction term between the minimum temperature from December of the previous year to January of the current year and the dummy variable indicating Nuttall Oak (T_{min(pDec, cJan)}:D_NUT). The model was overall a statistically good fit to the data with R² = 0.616, F(6,763) = 204.3, and p < 0.001. All the independent variables were significant (p < 0.05), and there was no evidence of multicollinearity (all VIFs < 5). The estimated regression model is shown in Table 1.

The number of years since the trees were planted, the maximum temperature between January and August of the current year (T_{max(cJan, cAug)}), the total precipitation between April and July of the current year (P_{(cApr, cJul)}), and the interaction term between the minimum temperature from December of the previous year to January of the current year and the dummy variable indicating Nuttall Oak (T_{min(pDec, cJan)}:D_NUT) had negative impacts on the annual radial growth rate. However, the average temperature in October of the previous year (T_{avg(pOct, pOct)}) and the minimum temperature between December of the previous year and January of the current year (T_{min(pDec, cJan)}) had positive impacts on the tree ring growth rate.

The estimated regression coefficients (Table 1) represent the ratio of change in the growth rate associated with a one-unit increase in each climate variable. For example, a one-unit (1 °C) increase in the average temperature in October of the previous year (T_{avg(pOct, pOct)}) would lead to a 0.7% increase in the growth rate for all species. A one-unit (1 °C) increase in the minimum temperature between December of the previous year and January of the current year (T_{min(pDec, cJan)}) would increase the growth rate by 0.5% for Cherry bark and Shumard oaks and by 0.4% (0.005–0.001) for Nuttall oak, respectively. For all three species, an increase in the maximum temperature between January and August of the current year (T_{max(cJan, cAug)}) by 1 °C would reduce the growth rate by 0.7%, and a 1 mm increase the precipitation between April and July of the current year (P_{(cApr, cJul)}) would decrease the growth rate by 1.1%.

3.3. Impacts of Climate Change on the Radial Growth Rate of Red Oaks

Compared to the historical benchmark (2005–2018), the annual radial growth rate of the three red oak species was estimated to rise slightly and gradually with the projected climate change under RCPs 4.5 and 8.5 (

Table 2. Estimated average percentage changes in the annual radial growth rate of red oaks under climate change relative to the historical average climatic conditions (2006–2018).

Species	RCP 4.5		RCP 8.5
	2030–2048	2080–2099	2030–2048	2080–2099
Cherry bark & Shumard Oaks	0.27	1.76	1.04	4.66
Nuttall Oak	0.28	1.41	0.94	3.63

). The increase in the growth rate would become more noticeable from 2030–2043 to 2080–2099 as the average, maximum, and minimum temperatures go up further. The growth rate would increase more from RCP 4.5 to RCP 8.5. Additionally, Cherry bark and Shumard oaks would have a larger increase in the growth rate than Nuttall oak except during 2030–2048 under RCP 4.5.

4. Discussion

Our results show that the average temperature in October of the previous year (T_{avg(pOct, pOct)}) and the minimum temperature from December of the previous year to January of the current year (T_{min(pDec, cJan)}) had statistically significant, positive impacts on the annual radial growth rate of all the three red oak species (Table 1). The influence of the average temperature in the previous October on the tree ring growth could be attributed to the carbohydrate reserve of trees prior to the growing season. A warmer October and a rise in the minimum temperature between December and January could lead to the increased accumulation of carbohydrates during the fall and winter seasons [42], which will stimulate radial growth in the following spring [47]. As much as 30% of the annual radial growth of red oaks can be achieved before their bud bursts [48]. Thus, carbohydrate accumulation in the fall and winter seasons can play a vital role in boosting tree ring growth in the coming growing season.

The maximum temperature between January and August, however, was found to be negatively related to the annual radial growth rate of the three red oak species (Table 1). The study site has a very mild climate with ample precipitation except for some summer months. Monthly maximum temperatures on the site can reach above 30 ℃ between June and August and 9 ℃ even in January (Figure 2), and August is the driest month (Figure 3). An increase in the maximum temperature would result in higher plant evapotranspiration and soil moisture evaporation, reducing the radial growth of trees [23,41,49]. Although the average, minimum, and maximum temperatures had different directions and magnitudes of effects on the radial growth of the red oaks, their aggregate impact was positive (Table 1). In other words, an increase in the average, minimum, and maximum temperatures by the same magnitude would increase the annual radial growth rate of the three red oak species.

Likewise, the total precipitation from April to July had a negative impact on the tree ring growth (Table 1). The plantation of this study is located next to the Mississippi River with annual precipitation ranging from 1500 mm to 1750 mm (Figure 3), among the regions with the highest precipitation in the U.S. Excessive rainfalls can raise the water table, reducing oxygen availability to the root system and potentially causing hypoxia, oxygen deficiency of plant tissues [28,50,51,52]. Additionally, increased precipitation is usually associated with more cloudy days, reducing solar radiation available for tree growth. All these may help explain the negative relationship between the annual radial growth rate and the total precipitation from April to July.

The radial growth of the three red oak species responds differently to the minimum temperature between December of the previous year and January of the current year (Table 1). The radial growth of Nuttall oak tends to be less responsive to an increase in the minimum temperature than the other two species (Cherry bark and Shumard oaks). This also indicates that Nuttall oak appears to be more cold-tolerant than the other two species as it can perform better in terms of radial growth than Cherry bark and Shumard oaks in lower minimum temperatures in December and January.

Finally, our results show that the annual radial growth rate diminishes as trees grow older. With a semi-log model used, this indicates a nonlinear trend of diminishing radial growth over time for all three red oak species (Table 1). This result is consistent with the findings of previous studies [53,54,55].

Our findings have useful implications for restoring and managing BHFs in the Arkansas Delta region, especially with a changing climate. All three red oak species seem well suited for the region even with the projected climate change. A warmer climate would in general promote the growth of these red oak species, especially Cherry bark and Shumard oaks (Table 2). However, according to the projected future climate in the region, the average, maximum, and minimum temperatures would increase by different magnitudes (Figure 2). The maximum temperature would rise more than the average temperature, which would increase more than the minimum temperature. Specifically, the maximum temperature between January and August of the current year (T_{max(cJan, cAug)}) is projected to increase by 2.1 °C during 2030–2049, which is higher than a 1.5 ℃ increase in the average temperature in October of the previous year (T_{avg(pOct, pOct)}) and a 0.9 ℃ rise in the minimum temperature between December of the previous year and January of the current year (T_{min(pDec, cJan)}) (Figure 2, [30]). With the increase in the minimum temperature under climate change, the advantage of Cherry bark and Shumard oaks over Nuttall oak in terms of diameter growth will become more obvious. Nevertheless, because Nuttall oak can endure lower minimum temperatures in the winter, it seems more suitable for the northern part of the region than Cherry bark and Shumard oaks.

Precipitation in the region is projected to decrease in most months of the year (Figure 3). For instance, precipitation between April and July (P_{(cApr, cJul)}) is expected to decrease by 11.9 mm during 2030–2049. The reduced precipitation in the spring and early summer could promote the diameter growth of the three oak species (Table 1). However, the reduced precipitation in August (the driest month) and October (though outside of the most appropriate/critical precipitation window) coupled with increased temperatures may impose water stress on the trees. This needs to be further monitored and investigated.

5. Conclusions

We estimated the statistical relationship of the annual radial growth rate of three red oak species with climatic conditions and tree age using the tree ring data collected from a BHF plantation in the Arkansas Delta region in the Lower Mississippi Alluvial Valley. We identified the time window of each climate variable that is most critical to the tree growth. We found that the average temperature in October of the previous year and the minimum temperature in December of the previous year and January of the current year were positively associated with the diameter growth of Cherry bark, Shumard, and Nuttall oaks, although the maximum temperature from January to August and the total precipitation from April to July of the current year were negatively correlated with the tree growth. Compared to Cherry bark and Shumard oaks, Nuttall oak was less sensitive to a rise in the minimum temperature between December and January. The projected climate change is likely to create slightly more favorable overall climatic conditions for these oak species in the region. Our findings suggest that these three red oak species are well suited for restoring BHFs in the study region, especially with a changing climate.

This study is based on the data collected from a plantation with 15 growing seasons. Future studies can attempt to obtain and incorporate tree growth data with a longer time series and from both plantations and natural BHFs. This study can also be expanded by exploring the dynamics of total biomass including belowground biomass and other ecosystem services (e.g., carbon, water regulation, wildlife, and biodiversity) of BHFs in response to climate variability.