Stem Defect Rates and Ice Storm Damage for Families of Pinus taeda from Coastal and Piedmont Provenances Planted on a North Carolina Piedmont Site

Abstract

Twenty Pinus taeda L. families from both the Coastal Plain and Piedmont provenances in the southeastern United States were planted on an upper Piedmont site that experienced a severe ice storm at age 3 years. Storm damage and defect rates through age 11 years were compared with the seed transfer distance and the seed parents’ breeding values to develop prediction models for storm damage and rates of forking, stem break, and sawtimber potential. Warmer-source families had higher probability of limb or stem breaks and foliage injury from the storm. Taller trees were more likely to experience breaks and foliage injury, even after accounting for seed transfer distance. Trees with forks or fusiform rust (Cronartium quercuum f. sp. fusiforme) infection had a higher probability of breaks. Trees with limb breaks or foliage injury did not have reduced sawtimber potential, but broken stems reduced sawtimber potential. The storm did not cause immediate mortality, but trees with major limb breaks, stem breaks, or foliage injury were less likely to be alive at age 8 years. At age 11 years, families with the best combination of breeding values for forking, straightness, and rust resistance had a predicted 60% of stems having sawtimber potential, whereas families with the worst combination had 30%.

Study Implications: Planting warmer-source Pinus taeda (loblolly pine) families farther north and inland may lead to greater growth but poses a risk of damage from cold temperatures and ice storms. Trees grown for solid-wood products must be relatively defect-free and require a longer rotation, whereas bioenergy and pulpwood can use smaller, defective trees. This analysis presents predictions of defect rates through age 11 years based on the seed source and breeding values using data from a planting in the upper Piedmont of North Carolina. Land managers can use these models to weigh the benefits and risks when choosing families for reforestation.

Pinus taeda L. (loblolly pine) is the most economically important tree species in the southern United States, accounting for 80% of the 1.15 billion seedlings planted in the region in 2021–2022 (Newell and Enebak 2022). Its economic importance can be attributed to its fast growth, broad adaptability, and responsiveness to silvicultural treatments. Tree improvement programs have made significant genetic gains in stem volume, straightness, and fusiform rust resistance (caused by the fungus Cronartium quercuum f. sp. fusiforme) over nonimproved wild material (McKeand et al. 2021).

The earliest efforts of tree improvement programs in the southern United States were to understand provenance variation in growth and adaptability. Provenance testing of P. taeda began in 1927 (Lambeth et al. 2005; Wells and Wakeley 1966). The results indicated there can be a significant advantage to using a nonlocal provenance to achieve gain by capturing the natural genetic variation present among provenances (Lambeth et al. 1984). Provenance studies in the southeastern United States demonstrated that the mean annual minimum winter temperature (MWT) of the seed source is the most important factor influencing growth, and sources from warmer origins generally grew faster than seed originating from colder locations (Schmidtling 1997, 2001). For example, Atlantic Coastal Plain sources have demonstrated faster growth than more northern and inland sources from the Piedmont (Kegley et al. 2004). Piedmont sources, however, tend to display improved stem form and increased cold tolerance compared with their more southern or coastal counterparts (Farjat et al. 2017; Lantz and Kraus 1987; Schmidtling 2001; Shalizi et al. 2022; Wells and Wakeley 1966; Zapata-Valenzuela et al. 2015).

The successful movement of Coastal sources inland may yield large returns, but the movement includes the risk of cold damage (Lambeth et al. 2005). The difference between the seed source MWT and the planting area MWT can be described as a seed transfer distance. Summarizing several provenance tests, Schmidtling (2001) reported no increased risk of cold damage in the southeastern United States for transfer distances up to 2.8°C (5°F) and a high risk of cold damage for transfer distances greater than 5.6°C (10°F). Cold damage may be caused by several factors, including late-season growing stress, midfall and early winter temperatures, and rapidly falling winter temperatures accompanied by storms with high wind and ice (Lambeth et al. 2005). The degree of winter damage may be minimal in the form of foliage injury or broken limbs from snow or ice or more severe, resulting in broken stems or even tree mortality from extreme cold. Trees that are bent or leaning greater than 40° are unlikely to recover (Pickens 2015). Ice storm severity is viewed in terms of the amount of ice accumulation, duration of accumulation, and resulting damage (Irland 1998). Ice accumulation is recognized as one of the most damaging climatic factors to forests in temperate regions (Bragg et al. 2003). Although severe ice storms are infrequent in much of the southern United States (Anonymous 2024), when they do occur, they may cause significant yield loss (Forgrave 2001; Halverson and Guldin 1995; White 1944). Even when total yield is not affected, broken limbs and stems from ice or snow damage can reduce tree quality from a higher-value solid-wood product like sawtimber to a short-length low-grade product.

Although P. taeda demonstrates better winter hardiness than many other southern pines (Schmidtling 2001), certain provenances are particularly susceptible to ice damage. Jones and Wells (1969) demonstrated that warmer provenances had a higher ice damage frequency than colder, more inland sources. Although ice damage was greatest among trees from warmer provenances, these trees still produced greater volume than trees from inland provenances at age 15 years. This and other studies found very low mortality in both provenances, although warmer sources had poorer stem form due to damage (Jones and Wells 1969). When evaluated through age 19 years, Shalizi et al. (2022) reported higher growth but reduced sawtimber production for the Coastal provenance than the Piedmont provenance when planted in the upper Piedmont, yet the Coastal provenance outperformed the Piedmont provenance for sawtimber production in the lower Piedmont.

Many tree factors, such as stem size, forking, and leaf area, may contribute to the type and severity of damage in forest trees (Aubrey et al. 2007; Bragg et al. 2003). Forking may contribute to irregularly shaped crowns and weaker upper stems, increasing their susceptibility to ice damage (Amateis and Burkhart 1996; Jones and Wells 1969). Fusiform rust disease may also affect susceptibility to ice damage through weakening wood quality in and around galls, as well as increased overall stress on the tree. Breakage on the main stem at the rust gall is common and can amount to substantial financial losses in a plantation setting (Powers et al. 1974). Decades of progeny testing have determined that within a provenance, families of P. taeda vary considerably in their breeding values for defect traits like forking rate, rust resistance, and stem straightness. When choosing among seedling families for reforestation, landowners can use the Performance Rating System (PRS) published by the Cooperative Tree Improvement Program at NC State University that summarizes the breeding values for the majority of families planted in the southeastern United States (McKeand 2019; Shalizi et al. 2023). Choosing families with lower forking tendencies and higher rust resistance might be viable options to reduce defect risk in ice-prone regions.

The potential benefits and risks of a particular genetic seedling stock must be evaluated in the context of the intended products. For instance, a landowner growing P. taeda for sawtimber may be cautious of moving warmer-source material too far north/inland, given the risk of cold or ice damage that may reduce the quality of the timber harvested. Further, trees grown for products that require longer rotations, such as large sawlogs, will be more likely to encounter an extreme weather event. A landowner growing wood solely for bioenergy or for pulpwood should be less concerned with stem quality and may be willing to accept the risk of cold damage to achieve faster growth or more total biomass yield. Land managers require estimates of the types of defects and their frequency when determining which seedling genetics to plant.

This study is part of a larger experiment designed to evaluate the bioenergy production potential for P. taeda in the upper Piedmont of North Carolina, USA, using different genetic options (provenances and families within provenance) and silvicultural thinning regimes that include sawtimber production alongside bioenergy (Maynor et al. 2021). The objectives of the study were to evaluate the differences among P. taeda provenances and families for storm damage and defect rates immediately following a major ice storm at stand age of 3 years as well as the lasting impact through age 11 years. Individual tree characteristics measured immediately before the storm were compared with the damage incurred. Defect rates were characterized multiple times through age 11 years and modeled based on the effects of seed source movement. Further, prediction models for stem defects were made using the genetic scores (breeding values) estimated from regional progeny tests and published in the PRS for land managers to use when choosing families for reforestation. Finally, the impacts of the storm damage assessments were compared with defect assessments at older ages to characterize damage as transient or having a lasting impact on product potential.

Materials and Methods

Plant Material and Experimental Design

A detailed explanation of the experiment is given in Maynor et al. (2021) and briefly summarized here. Twenty open-pollinated families of P. taeda were planted at the NC Department of Agriculture and Consumer Services Umstead Experimental Farm in Butner, Granville County, North Carolina (figure 1) to assess genetic variation in biomass at young ages and yields of solid wood products at traditional pine plantation harvest ages. Biomass yield is an important trait for bioenergy production, so the study used ten of the fastest growing families from each of the Piedmont and Coastal breeding populations of the NC State University Cooperative Tree Improvement Program. The MWT of the seed parents ranged from -13.9°C to -8.3°C, and the average by provenance was -8.8°C for the Coastal-source families and -11.3°C for the Piedmont-source families.

The difference between the family seed source MWT and the planting area MWT was used as a seed transfer distance. Figure 1 includes the source locations for the parents (ancestors of second-generation parents) used in this study, with colors indicating their seed transfer distance. The MWT of the county where the study site is located is -14.0°C, and seven of the twenty families had transfer distances within 2.8°C, and one family had a transfer distance greater than 5.6°C. It was noted that the family with the smallest seed transfer distance was sourced from one of the farther geographical distances from the planting site (western Tennessee) (figure 1).

The experiment was planted in the spring of 2012 with a randomized split-split plot design with five experimental blocks, each with harvest treatments as main plots; provenance (Coastal and Piedmont) assigned at random to split plots, and families (ten open-pollinated families per provenance) randomly assigned to thirty-six-tree rectangular subplots within each provenance plot. Although the provenance was a treatment in the experimental design, this analysis focuses on seed transfer distance as a metric for cold-hardiness because there was much variation for MWT within a provenance and overlap among provenances. For measurements taken prior to the thinning and harvesting treatments (implemented at age 9 years), there were effectively ten blocks in the study. Trees were planted at a higher-than-typical density of 2,563 trees per hectare with a spacing of 1.83 m × 2.13 m. There were 7,200 experimental seedlings planted (360 seedlings per family) and surrounded by two rows of buffer trees around each provenance subplot and four rows around each block for a total of 12,408 trees.

Ice Storm Event and Measurements

During February 2015 (3 years after planting), two severe ice/snow storm events accompanied by extremely low temperatures (between 0°C and -15°C) affected the study site over multiple days. Precipitation in the form of snow, sleet, and freezing rain (ranging from 1.3 to 2.5 cm) and prolonged periods of cold resulted in damage in the form of broken limbs, main stem breakage, and foliage injury. Wind was not extreme. One month prior to the winter storms, trees were assessed for status (alive, dead, or damaged), total height, presence of fusiform rust disease, and the presence of forks or ramicorn branches. Immediately after the storm, trees were assessed for damage in the form of limb/stem breaking, foliage injury, and stem lean. Very few trees suffered excessive stem lean (4.6% of trees had main stem leaning greater than 30°), so we did not further analyze this trait. For limb/stem breaking, a four-point scale was used (photos in Supplemental Figure S1) where

• Limb/stem score 1: no damage with main stem and branches intact,

• Limb/stem score 2: minor damage with main stem intact and loss of one or two branches,

• Limb/stem score 3: more intensive damage with multiple branches broken out, but tree and main stem are intact,

• Limb/stem score 4: main stem broken or major crown damage where tree is not expected to recover or grow into a merchantable product.

The extent of foliage injury (colloquially referred to as needle “burn”; photos in Supplemental Figure S1) was also assessed with a four-point scale where

• Foliage injury score 1: green foliage with no visible damage,

• Foliage injury score 2: minor presence of brownish/reddish foliage, typically towards the top of the crown,

• Foliage injury score 3: extensive presence of brown/red foliage on much of the crown,

• Foliage injury score 4: all foliage is brown.

The counts and percentage of trees in each class are shown in Supplemental Table S1. The majority of trees had a score of 1 for limb/stem breaking (71%) and for foliage injury (86%), so the scores greater than 1 were lumped for both measurements to make binary traits for the presence or absence of limb/stem breaking and foliage injury due to low counts in the individual categories.

Trees were assessed again at ages 5, 6, and 8 years for status and measurements of defect, including the presence/absence of forking, ramicorn branches, fusiform rust disease, and stem breakage. At ages 6, 8, and 11 years, trees were assessed with a sawtimber potential score following (Cumbie et al. 2012) where

• Sawtimber potential score 1: No quality defects in the stem in the bottom 1.5 logs (7 m) of the tree,

• Sawtimber potential score 2: Minor defects in the first 1.5 logs but still likely to be a sawtimber tree; minor defects may include small or few ramicorn branches, high fork (above first log), or minor sweep,

• Sawtimber potential score 3: Defects present such as low forks, large or multiple ramicorn branches, or large sweep that preclude the tree from growing into a solid-wood product within a typical rotation length. The tree may still be used for pulpwood or biomass,

• Sawtimber potential score 4: Cull or nonmerchantable because major stem defects such as stem rust, multiple forks, or extremely poor growth and stem form that prevent the tree from being used for pulpwood or biomass.

For analysis, the score was converted to a binary trait by combining scores 1 and 2 into one class (tree has sawtimber potential) and combining scores 3 and 4 in another (tree does not have sawtimber potential). Plots that were thinned were excluded from the age 11 data analysis. The counts and percentages of trees by defect class and age are shown in Supplemental Table S2.

To see how well damage and defect could be predicted, we included the breeding values published in the Cooperative Tree Improvement Program’s 2023 Piedmont PRS as covariates. This database includes breeding values estimated from the Cooperative’s second, third, and fourth cycle progeny tests across the Piedmont region. The forking, straightness, and rust breeding values are all on the probability scale and centered on 0.50. Although labeled as “forking,” the breeding values for forking also include the probability of having a ramicorn branch, as the two are lumped together because the incidence is usually lower than ideal for estimating breeding values, and the two have a positive genetic correlation (Xiong et al. 2010).

Statistical Analysis

Modeling Storm Damage Occurrence Based on Tree Characteristics

Using the measurements made at age 3 years immediately prior to the ice storm, models were developed for probability of limb or stem breaks and probability of foliage injury from the storm. These traits are binary (present or absent) with binomial distributions, so a generalized linear model was used to model the effects of block (a categorical variable), the seed transfer distance and tree height prior to the storm (covariates), and the presence or absence of a fork, ramicorn branch, or fusiform rust gall prior to the storm (incidence variables). The model is given by

where, π is the probability of the response (limb/stem breaks or foliage injury), log[π/(1-π)] is the logit (log transformed odds), $μ$ is the overall mean, $Bi$ is the fixed effect of block (⁠$i=1,2,…,5$⁠), $D$ is the seed transfer distance (in units of °C), $H$ is the tree height prior to the storm (in meters), and $K$⁠, $R$⁠, and $G$ indicate the presence of a fork, ramicorn, and rust gall prior to the storm, respectively, with values of one if the defect is present and zero if absent. The $βp$ are coefficients that estimate the effect of the covariates and indicator variables on the logit scale. The $εijkl$ is the error term associated with individual tree $l$ in family plot $k$ in the whole plot $j$ in block $i$⁠.

Modeling Defect and Sawtimber Potential Through Age 11

The defect assessments for forks, stem breaks, and sawtimber potential were made at multiple ages, so a mixed-effects model was used to consider the lack of independence among the repeated measurements of the same tree. In addition to the seed transfer distance, we wished to evaluate how breeding values affected the probability of defect. The model used was

where $Tijkl$ is the random effect of tree $l$ with $T∼N(0,σT2)$⁠, $Am$ is the fixed effect of age (categorical variable), $KBV$⁠, $GBV$⁠, and $SBV$ are the covariates for forking, rust, and straightness breeding values from the Piedmont 2023 PRS, and all other terms as previously defined. Age was treated as a categorical variable because we did not expect defect rates to change linearly with time for a few reasons. The most obvious reason is that defects tend to be associated with particularly strong weather events, which do not occur at regular intervals. Also, we observed instances of trees evaluated as forked at younger ages assessed as not forked at older ages. These trees oftentimes were assessed as having a ramicorn branch at the older assessment, and one stem presumably recovered apical dominance. Further, sawtimber potential score assessments on young trees are not straightforward, because it can be difficult to determine which defects will be recovered as the tree grows.

Determining the Effect of Storm Damage on Product Potential

Another objective was to determine whether the effects of storm damage had a transient or lasting impact on defect assessed at older ages. This was done by modeling the sawtimber potential score at age 11 years with terms indicating the limb/stem score and foliage injury score that were measured immediately after the ice storm at age 3 years. The model is given by

where $π$ is the probability of sawtimber potential at age 11 years, $C$ is the limb/stem break score after the storm at age 3 years, $I$ is the foliage injury score after the storm at age 3 years, and other terms as previously defined. Both storm damage scores were included in the model as the four-point scales used during assessment (rather than collapsing to binary traits as done in the previous analyses). We also modeled the probability of survival at age 8 years using the same model. Age 8 year survival was used rather than age 11 because we observed competition-induced mortality at age 11 due to crown closure and the onset of the stem-exclusion stage in stand dynamics.

For all of the models presented, effects were tested for significance using the analysis of deviance type III tests for fixed effects (Hastie and Pregibon 1992). Parameters estimates and their 95% confidence intervals were converted to the odds ratio scale for interpretation. The least square means were predicted for combinations of the significant fixed effects and transformed to the probability scale for graphing. These probabilities can also be interpreted as rates or the proportion of trees. Model parameters were estimated using the glm() function in R (R Core Team 2022), and for mixed-effects models, the glmer() function in the lme4 package was used (Bates et al. 2014). The car::Anova() function was used to carry out the type III analysis of deviance for the significance of fixed effects (Fox and Weisberg 2019). Least square means were estimated with the emmeans() function in the emmeans package (Lenth 2023).

Results

Storm Damage at Age 3 Years

The plot-level percent survival at stand age 3 years (before the storm) averaged 96.7%, and all plots had at least 83.3% of trees surviving, with a quarter having 100% survival (Table 1). No trees were noted as having immediate mortality following the storm. The seed transfer distance and tree height had a significant effect on both the probability of limb or stem breaks after the storm as well as foliage injury (Table 2). The presence of a fork, ramicorn branch, or rust gall also had a significant effect on the probability of limb or stem breaks but not foliage injury. For a 1°C increase in seed transfer distance, the odds of limb or stem breaks was estimated to increase by 7% and the odds of foliage injury by 11% (holding height and defect incidence constant) (Table 3). A 1 m increase in height almost tripled the odds of limb or stem breaks and increased the odds of foliage injury by 26%. For a given seed transfer distance and tree height, the presence of a fork or a ramicorn more than doubled the odds of limb or stem breaks.

The predicted probability of limb or stem breaks increased rapidly for trees taller than the average height (2.72 m) (figure 2). The differences among seed sources were most evident for taller trees and trees with defect. For trees with average height and no defect, the farthest seed transfer distance tested (6°C) had a predicted probability of breaks of 0.24, whereas the local source (seed transfer distance of 0°C) had a predicted probability of 0.17. For an average tree height with a moderate seed transfer distance (3°C), trees with a fork or ramicorn branch and a fusiform rust gall had a predicted probability of limb or stem breaks of 0.45, compared with a probability of 0.21 for a tree without these defects.

The probability of foliage injury also increased rapidly for trees with heights taller than the average (figure 3). For shorter trees, such as 2.0 m, the predicted probability of foliage injury for local-source material was 0.08 compared with 0.15 for the warmest source tested. For taller trees, such as 3.5 m, the local seed source had a probability of 0.11 compared with 0.20. It is notable that even the most cold-hardy trees in the study suffered limb/stem breaks and foliage injury from the cold storm, although significantly less than those with farther seed transfer distance.

Defect Rates Through Time

The probability of forking was significantly affected by stand age, seed transfer distance, and the breeding values for forking and stem straightness (Table 4). Forking probability increased through time to a mean of 0.22 at age 8 years (figure 4a). Differences among the seed sources increased through time, and at age 8 years, the predicted probability of forking was 0.27 for the warmest-source material compared with 0.17 for the local source. After age 5 years, the family forking rates did not always increase through time, as many trees assessed as being forked at a younger age were assessed as not forked at older ages, oftentimes being assessed as having a ramicorn branch. In fewer instances, forked trees died. The effect of forking breeding values on forking probability was greater at older ages (figure 4b). At age 8 years, a fork breeding value of 0.75 had a predicted probability of 0.28, whereas a fork breeding value of 0.35 had a predicted probability of 0.17. Note that the PRS breeding values are based on progeny tests across the Piedmont region and are centered on 0.50, whereas the probabilities predicted here are for this particular study site. The odds of forking increased by 9% for a one-degree increase in seed transfer distance when holding the forking breeding value constant. The odds of forking increased by 10% for a ten-point increase in forking breeding value (while holding the seed transfer distance constant) (Table 5).

The probability of stem break was significantly affected by stand age, seed transfer distance, and rust breeding value (Table 4). At age 8 years, the predicted probability of stem break was 0.16 for the farthest seed transfer distance compared with 0.09 for the local source (figure 5a). The most rust susceptible (higher rust BV) family at the site had a predicted stem break probability of 0.18, and the most resistant material had a predicted probability of 0.10. The odds of stem break increased by 13% for a 1°C increase in seed transfer distance, and the odds increased by 10% for a ten-point increase in rust breeding value (while holding other model factors constant) (Table 5).

The probability of a tree having sawtimber potential was significantly affected by stand age, seed transfer distance, and all three breeding values considered (fork, rust, and straightness BV) (Table 4). The sawtimber potential rates did not decrease monotonically with age, presumably due to the difficulty in assessing the impact of defect at young ages on future product potential (figure 6). Regardless, the ranking of family means was very consistent through time. The local source had a sawtimber potential rate of 0.55 at age 11 years whereas the most distant source had a rate of 0.39.

Holding other model factors constant, a ten-point increase in fork BV reduced the odds of sawtimber potential by 14% (Table 5). A ten-point increase in straightness BV increased the odds of sawtimber potential by 12%. A ten-point increase in rust BV reduced the odds of sawtimber potential by 4%. The predicted sawtimber potential rates based on varying levels of breeding values for fork, rust, and straightness are given in figure 7. The predicted sawtimber potential probability for families with the best combination of breeding values (low forking BV, high straightness BV, and low rust BV) was 0.60 and as low as 0.30 for families with the worst combination of breeding values. The effect of rust BV was much less than that of fork BV and straightness BV, presumably due to the low hazard of this upper Piedmont test site and mortality of severely infected trees. The parameter estimates for the sawtimber probability prediction model are given in Supplementary Table S3 as are the parameter estimates for the fork and stem break probability models. These can be used to predict probability of defect at this site based on breeding values published in 2023 Piedmont PRS.

Lasting Impact of Storm Damage

For a given seed transfer distance, the limb/stem break scores of 2 and 3 (main stem is intact but there is limb damage) did not have a significant impact on sawtimber potential at age 11 years (P = .3101 and P = .9419, respectively), whereas a score of 4 (main stem broken) did (P < .0001). Trees assessed as having the main stem broken had 97% lower odds of sawtimber potential (odds ratio of 0.07 with 95% confidence limits of 0.04 to 0.14). The foliage injury score from the storm had no significant effect on sawtimber potential at age 11 years (P = .7518).

Survival at age 8 years was also affected by the limb/stem break assessments made immediately after the storm at age 3 years when accounting for seed transfer distance. Limb/stem break scores of 3 (major limb damage, stem intact) and 4 (main stem broken) both had significantly lower survival rates than the score 1 (no limb or stem damage) (P = .0154 and P < .0001, respectively). A limb/stem score of 3 had a 66% lower odds of survival than the damage-free class (odds ratio of 0.34 with 95% confidence limits of 0.14 to 0.81), and a limb/stem score of 4 had 83% lower odds of survival compared with the damage-free class (odds ratio of 0.17 with 95% confidence limits of 0.13 to 0.21). Further, foliage injury from the cold storm was indicative of mortality at age 8 (overall test P = .0002), even after accounting for seed transfer distance. Although the model term for foliage injury was significant, the survival effect estimates for the individual scores had poor precision (high standard errors) and were not significantly different. This implies that foliage injury was related to mortality, but we do not have enough information to determine the extent of foliage injury that can be tolerated.

Discussion

Similar to reports by Jones and Wells (1969), the cold storm at age 3 years did not appear to cause direct mortality, but there were apparent differences in damage among provenances and vigor was reduced on trees incurring damage. The stem breaks at age 3 years had a lasting impact on sawtimber product potential, and trees with a major limb break or main stem break from the storm were less likely to be surviving at age 8 years. Foliage injury from the storm did not affect sawtimber product potential, but the presence of foliage injury reduced the odds of survival at age 8 years, even after accounting for seed transfer distance. Although not formally assessed, repeated foliage injury was noted at older ages, and the injury measured at age 3 years was apparently an indicator of a lack of cold hardiness. In a 20-year-old stand, Belanger et al. (1996) found that damaged trees that were vigorous prior to a cold storm often became suppressed, whereas adjacent trees with less damage filled in the gaps left by broken limbs/stems and foliage dieback. Even at age 8 years, when intertree competition was low, we observed lower survival on storm-damaged trees, presumably due to lingering effects of the storm or repeated cold damage on those trees lacking cold hardiness.

Taller trees were more likely to experience limb or stem breaks and foliage injury, even when accounting for seed source, presumably due to the increased exposure above the canopy compared with shorter trees, or possibly the larger crown surface area subjected them to greater ice load (Aubrey et al. 2007; Belanger et al. 1996). Similar to results reported by Jones and Wells (1969), differences in storm damage among seed sources were significant even after accounting for height, suggesting that other factors such as crown shape, branch angle, leaf density, or wood properties influence resistance to breaking. Stem taper affects susceptibility to breaking from snow/ice loads, with trees having more taper able to withstand heavier snow/ice loading (Petty and Worrell 1981). However, previous investigations into provenance variation on stem taper has not found differences between the Coastal and Piedmont sources used in this study (Buford and Burkhart 1987; Schmidtling and Clark III 1989).

It should be noted that even the most cold-hardy seed sources in this study, which had minimal seed transfer distance, incurred considerable damage from the storm, although their defect rates were lower and sawtimber potential rates much higher through age 11 years (mean of 55% compared with 39% for the warmest source tested). Although specific to this site, we present predictions of defect rates for forks and broken stems (through stand age 8 years) and sawtimber product potential (through age 11 years) based on seed transfer distance and breeding values that should be useful for risk evaluation.

Trees forked prior to the storm had nearly triple the odds of limb/stem breaks. The irregularly shaped crowns and weaker upper stems associated with forked trees likely contribute to the higher probability of damage compared with single-stemmed trees (Amateis and Burkhart 1996). Throughout the evaluation period, warmer-source families had more incidence of forking, with the warmest-source having a forked-stem rate of 27% compared with 17% for the local source. Within provenance, family variation in forking was present and could be accounted for using the breeding values published in the 2023 Piedmont PRS, which are estimated from progeny tests throughout the Piedmont region. The family with the most forking in progeny tests (BV = 0.75) had a predicted forking rate of 28%, whereas the family with the least forking in previous testing (BV = 0.35) had a rate of 17%. The forking breeding value directly influenced the observed sawtimber potential, with sawtimber potential rates fifteen percentage points lower for families in this study, with the poorest forking BV compared with those with the best forking BV. Strict emphasis on forking BV is not typically recommended for family selection because of the relatively low genetic control of the trait (Xiong et al. 2010). However, there is clearly a benefit of avoiding families with poor forking BV on sites prone to loss of leaders (e.g., due to storm damage) when sawtimber is the intended product.

Trees with fusiform rust had higher odds of experiencing a stem break throughout the evaluation period. Van Lear and Saucier (1973) reported that stem breaks often occur at stem galls caused by fusiform rust. This finding differs from Belanger et al. (1996), who found little association between ice damage and occurrence of rust galls in 20-year-old stands, noting that most breaks were high in the canopy, and rust was low on the stem. The younger age of the trees in this study likely explains the difference in these findings. In this study, we did not formally assess the location of stem breaks, but field crews commented that many of the breaks at ages 3 and 8 years occurred at or near a fusiform rust gall, suggesting that infected wood was weaker and more prone to break. The rust breeding value influenced the probability of stem breaks through age 8 years, with the most susceptible material having a 40% higher odds of experiencing a stem break. The influence of rust BV on sawtimber potential score was not as important as forking or straightness, with about five-point reduction in sawtimber production between the most resistant and most susceptible sources. The fusiform rust rates observed at this site were considerably higher than expectation from historic progeny testing data (Walker and McKeand 2018), perhaps due to the intensive site preparation (which included fertilization) and the abundance of natural mature mixed pine-oak stands adjacent to the study site.

Jones and Wells (1969) reported that although more ice damage was observed on the warmer-source material after a storm at age 10 years, the losses were offset by the increased growth of the warmer seed sources 5 years later at age 15 years. The present study provides important assessments of the response to storm damage and resulting defect rates based on seed transfer distance, but more research is required to evaluate growth differences among the sources in conjunction with product potential. Further, thinning provides an opportunity to realize value from damaged stems and release growing space to nondamaged final crop trees to grow into higher-value solid wood products. Simulations of scenarios with varying product objectives and investment periods (e.g., rotation lengths) should be evaluated.

Conclusions

This study provides further evidence that ice storms can cause significant and lasting damage to Pinus taeda plantations, but the impact can be mitigated by choice of seed transfer distance and progeny test breeding values. Using the prediction models presented here, land managers can weigh the benefits and risks when choosing among the wide variety of families that are available for reforestation as they consider their product objectives and timeframe. These models expand on the guidelines given by Schmidtling (2001) by predicting the percentage of trees with sawtimber quality as a function of the progeny-tested breeding values for defect traits (forking, rust, straightness) in addition to seed transfer distance. Although ice storms are not predictable events, landowners can choose seedling genetics that reduce the risk of damage.

Supplementary Material

Supplementary material is available at Forest Science online.

Acknowledgments

The assistance of many graduate students and staff from NC State University is acknowledged.

Funding

This work was funded by the North Carolina Department of Agriculture and Consumer Services Bioenergy Research Initiative Grant Program, Contract Number 17-072-4008 and by members of the North Carolina State University Cooperative Tree Improvement Program. Additional support came from the following USDA National Institute of Food and Agriculture programs: McIntire-Stennis Project NCZ04214; the Integrated Biomass Supply Systems (IBSS), Subcontract from University of Tennessee, award 2011-68005-3041; and the Pine Integrated Network: Education, Mitigation, and Adaptation Project (PINEMAP), a Coordinated Agricultural Project (CAP) # 2011-68002-30185. Financial support from the Department of Forestry and Environmental Resources and the College of Natural Resources at North Carolina State University is also acknowledged.

Conflict of Interest

None declared.

Literature Cited

Author notes