Abstract
Background Along electric distribution corridors in urban-exurban landscapes, forest edges are susceptible to damage associated with storm events. Disturbances and management interventions designed to preempt their effects (e.g., tree trimming) alter characteristics of tree structure and morphology (e.g., branch and crown structure), which may be associated with tree failure and likelihood of associated infrastructure damage. This study assessed the relationship between 3-dimensional tree crown structure and tree biomechanics and characterized the effect of utility tree trimming on tree sway dynamics using terrestrial laser scanning (TLS).
Methods In this study we extracted and analyzed measures of crown structure (i.e., crown asymmetry, crown area, total volume, crown diameter to height ratio, and crown evenness) for individual trees during leaf-off conditions before and after implementation of tree trimming and linked these measures to tree biomechanics data, to evaluate how commonly implemented trimming practices affect both tree sway frequency and displacement—important indicators of tree stability.
Results Results illustrated the effects of common tree trimming practices on tree crown structure, but there were not consistent changes to tree movement characteristics directly following tree trimming across our 24 study trees. However, we found that the associated changes in crown structure through tree trimming affected tree displacement in moderate wind conditions. Additionally, we found there were no significant differences between frequencies across treatment types.
Conclusions This pilot project lays the foundation for understanding the intricate relationship between 3-dimensional crown structure and tree biomechanics following roadside tree trimming.
Introduction
Climate change is significantly impacting forests across the globe, altering disturbance regimes and their impact on ecosystems (Dale et al. 2001). Windstorms are a key disturbance agent in forests throughout the world and are increasing in severity and frequency. These changes are likely to result in increased damage to trees and forests (James et al. 2006; James et al. 2018; Jackson et al. 2021) and substantial effects on ecosystem structure and functioning. When wind exceeds the critical resistance capacity of branches, tree stems, and root systems, they can break and uproot (Gardiner and Quine 2000; Spatz and Bruechert 2000; James et al. 2006; Mitchell 2013). Windthrow can affect and alter the landscape, stand and microsite scale, and leave legacies lasting for centuries or millennia. Wind disturbances can affect stand condition, patch size, and distribution of stand types across the landscape (Mitchell 2013; Gardiner et al. 2016).
Wind disturbances can pose a high risk to human society, in part because tree failures can cause damage to built infrastructure leading to losses in environmental, economic, and social services (Gardiner et al. 2016). Electric distribution corridors in urban-exurban landscapes are highly susceptible to storm damage, especially as storm severity and intensity increase under climate change (Parent et al. 2019). Annually in New England, trees can be responsible for approximately 25% of electric service interruptions through branch loss, snapped stems, and uprooting (Eversource 2019). In storms with high winds and precipitation, tree failures may increase to cause > 90% of power outages (Eversource 2019). As branches and limbs encounter power lines, outages and equipment damage create public safety hazards and can lead to extended power outages.
Tree trimming and removal are the primary mitigation techniques employed by utilities and other stakeholders in limiting the effect of storms on electric distribution corridors. Utility vegetation management programs aim to reduce tree-related power outages through the removal and pruning of trees along distribution system corridors (Goodfellow et al. 1987; Hammerling 2012). Within the United States, there are approximately 10,190,746 linear km of distribution lines (Warwick et al. 2016). Pruning and removal procedures are implemented along these distribution lines (primarily located along residential roads and highways) in order to mitigate power outages and associated utility infrastructure damage (Kuntz et al. 2002; Yahner and Hutnik 2005; Eversource 2019). In 2001, US business sectors lost approximately 104 to 164 billion dollars due to power outages alone (Lineweber and McNulty 2001). Vegetation management can be extremely costly for utilities both on a national and local scale. For example, vegetation management nationally costs between 2 to 10 billion dollars annually in the United States (Guggenmoos 2003) while the small state of Connecticut costs 50 to 60 million dollars per year in the years 2012 to 2014 (Hammerling 2012), but roadside vegetation management greatly reduces liability costs associated with tree risk problems (Hasan et al. 2016). Prior studies have analyzed the impacts of vegetation management on the resilience of electric distribution grids, identifying and predicting where “trouble spots” in the grid system may occur based on vegetation management data (Cerrai et al. 2019; Taylor et al. 2022; Walker and Dahle 2023; Wedagedara et al. 2023). Other studies have found significant effects of trimming strategies and increased management intensity (e.g., Enhanced Tree Trimming; ETT) on reducing minor outages (i.e., blown fuses, tripped recloser, etc.), but not for major damage and outages (i.e., broken poles or downed wires)(Parent et al. 2019).
The structure of tree crowns is an important driver of their response to wind forcing, with complex 3-dimensional architecture benefitting trees by minimizing sway motion and stress due to branches dynamically interacting with the trunk’s mass (James et al. 2006). In particular, the crown volume ratio— the ratio of whole tree volume to crown volume—is an essential predictor of wind damage risk (Jackson et al. 2019a). Branching geometry, including both angle and length, has a strong effect on the forces branches must resist as they change the center of mass, particularly on branches with leaves (Loehle 2016). Altering crown structure can have significant effects on the dynamic biomechanical properties of trees, such as the natural frequency and damping ratio, which are used to accurately assess wind-induced failure (Kane and James 2011). The mechanical stability of a tree is its ability to withstand dynamic forces such as wind throughout the course of its life (James et al. 2006). As trees interact with the wind and transfer energy, they display characteristic and recognizable swaying patterns that can be unique to the individual, known as a tree’s natural frequency (James 2003; Jackson et al. 2021). Prior work on the relationships between tree architecture and natural sway frequencies has found that slender trees have a single natural frequency that dominates their motion—acting as a simple pendulum which increases their vulnerability to wind damage (Jackson et al. 2019b). Also, greater slenderness has been associated with susceptibility to failure during storms (Ancelin et al. 2004; Bunce et al. 2019; Snepsts et al. 2020). Roadside forests may benefit from management promoting lesser slenderness and height to crown ratio, and thus, an increased fundamental vibrational frequency (FVF) (Bunce et al. 2019).
Vegetation management along roadsides has been shown to reduce outages (Parent et al. 2019), but the specific effects of tree trimming on near-term tree stability are not well understood. While vegetation management is a practical necessity in urban-exurban systems, tree trimming alters tree crown structure and mass (Jaeger et al. 2022). Variable trimming specifications and rotations are implemented by utilities, each consisting of distinct criteria calling for different clearance zones between trees and infrastructure (Eversource 2019). These different treatment types are likely to have varying impacts on tree crown structure and biomechanics as well as the sway response of trees which can indicate stability (Kane and James 2011). For example, tree trimming can affect crown top-heaviness and, subsequently, the mass distribution throughout the tree thus changing the biomechanical response. Tree displacement after trimming is a key indicator of tree stability during wind events. Past research on roadside forests has explored silvicultural techniques to enhance forest resilience and reduce utility damage (Ward et al. 2017; Bunce et al. 2019). However, understanding individual tree displacement is crucial for mitigating utility infrastructure damage and minimizing power outages. Altering crown and forest structure through tree trimming could potentially cause increased movement and instability, leading to an increased risk of failure during storm events. This is underscored by a previous study which examined displacement and the tilt/sway behaviors, aiming to enhance our understanding of the likelihood of tree failure under wind loads (Yang et al. 2021). While an immediate increase in displacement post-trimming may suggest reduced stability, the long-term effects of trimming on stability may vary throughout the trimming cycle. Therefore, continuous monitoring of displacement is essential to evaluate tree stability and assess the risk to utility infrastructure, especially with the growing severity and intensity of wind disturbances (Dale et al. 2001). Understanding relationships between crown structure and biomechanics is crucial to effectively manage roadside forests to mitigate power outages, equipment damage, and foster stable resilient trees.
To better understand the potential effects that vegetation management practices have on tree structure and linkages with stability and storm resistance, our broad objective was to characterize relationships between 3-dimensional tree crown structure and tree biomechanics following implementation of utility tree trimming. Our specific research questions were: (1) How do common roadside tree trimming practices affect 3-dimensional tree crown structure? (2) How do trimming practices affect tree biomechanics and to what extent can changes be attributed to alteration of crown structure? We hypothesized that tree trimming would increase tree movement in wind in the near term due to reduction in surface area to dissipate wind energy. In addition, we hypothesized that crown asymmetry would be a key predictor of tree displacement and biomechanical change.
Materials and Methods
Site Description
We selected 24 roadside trees across 3 sites located around the University of Connecticut in Storrs, Connecticut, USA (41.8084° N, 72.2495° W)(Figure 1) including oaks (Quercus alba and Quercus velutina), hickories (Carya ovata and Carya tomentosa), and maples (Acer rubrum and Acer saccharum) (Appendix). The 3 sites (PHSE, IPB, TNPK), included 7, 8, and 9 trees respectively (Appendix). Roadside edge trees were chosen based on their proximity to the roadside (1 to 5 m) and examined to confirm their health, ensuring they were devoid of apparent defects like decay or disease. This examination involved visually inspecting tree stems for any evident holes or damage, as well as checking the canopy for dead branches. Tree diameters ranged between 32.9 and 76.4 cm with a total of 9 Carya spp., 9 Acer spp., and 6 Quercus spp. of which 9 were control trees where no trimming practices were implemented. In the study location, the mean annual temperature is 8.77 °C and mean annual precipitation is 117.5 cm/year, while through the growing season (May to October), precipitation is 54.9 cm and the average temperature is 18.05 °C (NOAA 2024).
Study site location at the University of Connecticut, Storrs, Connecticut, USA.
Tree Trimming Specifications
Two common roadside tree trimming treatments, Enhanced Tree Trimming (ETT) and Scheduled Maintenance Trimming (SMT)(Figure 2) were implemented in this study. ETT, also referred to as ground-to-sky, is characterized by trimming 2.43 m (8 ft) horizontally from electrical equipment from the ground up (Eversource 2019). SMT is characterized by trimming 4.57 m (15 ft) above utility equipment, 3.04 m (19 ft) below, and 2.43 m (8 ft) horizontally (Eversource 2019). Tree trimming in this study was implemented by a local professional tree care practitioner (Tennent Tree Service Inc., Windham, Connecticut) on 15 trees across the 3 sites in early March 2023 with no trimming occurring on the 9 control trees. At each site a walk through was conducted with the tree company manager prior to trimming to discuss the implementation of ETT and SMT on the study trees. To our knowledge, and based on evaluation of the trees, trimming had not been implemented on these trees prior to this study.
Tree trimming specifications include Scheduled Maintenance Trimming (SMT) and Enhanced Tree Trimming (ETT). SMT is characterized by trimming branches 15 feet (4.5 meters) above, 8 feet (1.2 meters) to the side, and 10 feet (3 meters) below utility equipment while ETT is characterized by trimming branches to 8 feet (1.2 meters) on both sides of utility equipment.
Biomechanics: Instrumentation, Data Collection, and Processing
To collect biomechanics data for each tree individually, data loggers were constructed using Arduino nano boards, Adafruit microSD card breakout boards+, and an Adafruit DS3231 Precision RTC Breakout connected to an Adafruit ADXL 345 triple-axis accelerometer. Accelerometers captured tree displacement in angles of tilt (degrees) across both x - and y-axes where rotation in the x-axis is tilt forward and backward while rotation in the y-axis is tilt side to side. Accelerometers were positioned between 7 to 10 m up the tree stem from ground level or just below the start of the crown (i.e., major branching or junction) in weather-proof boxes. Sensor heights were recorded and measured using a TruPulse® 200 Laser RangeFinder. Accelerometers were positioned facing north across all trees. Data were collected at a frequency of 10 Hz (i.e., 10 measurements per second) for leaf-off pre- and post-trimming conditions. Seven days of pre- and post-trimming data were selected for analysis in 2022 April and early May and 2023 March respectively. Accelerometer data were filtered through a bandpass Butterworth filter with a passband of 0.01 to 2 Hz, attenuating low and high frequency noise using MATLAB (R2O22a)(MathWorks Inc. 2022). Ten-minute z-axis 
samples during wind speeds greater than 2 m/s (4.47 mph) and between 0.2 to 0.5 Hz were selected for analysis to extract the fundamental frequencies of each tree before and after trimming. This range ensured some additional accelerometer noise below 0.2 Hz was removed to avoid affecting the estimation of fundamental frequencies. Fundamental frequencies were extracted using the fft.rff function within Python’s (version 3.10)(Python Software Foundation 2021) module “numpy” (Appendix). For each day, modes were recorded for each tree and then averaged across the 7-day study period for both pre- and post-trimming conditions to estimate a fundamental frequency for each tree. The change in frequency (Δfreq) was calculated by the difference between pre- and post-trimming frequencies for each tree. Unfiltered raw accelerometer data in angles of tilt was converted into meters of displacement to obtain displacement metrics for each tree using the sensor height, live crown height, and tree height. Meteorological data were collected from a nearby meteorological tower equipped (< 1.6 km) on campus with a propeller vane (R.M. Young Wind Monitor, model 05103, Campbell Science, Rockford, Illinois, USA) anemometer approximately 30 m above the ground which measured and recorded wind speed and direction at 1 Hz (Campbell Scientific datalogger, model CR1000, Campbell Scientific, Logan, Utah, USA). Wind metrics including average speed, maximum speed, and direction of average speed were obtained from these measurements.
Terrestrial Laser Scanning (TLS) Data Collection and Processing
Individual tree terrestrial lidar scans were acquired using a FARO Focus3D X130 (FARO Technologies, Inc; Lake Mary, FL, USA) scanner for leaf-off pre- and post-trimming conditions in 2022 and 2023 March. Scans were collected at 1.3 m above the ground at 4 scan positions located approximately 30° NE, 120° SE, 210° SW, and 300° NW around each tree (Appendix). Distances between scans did not exceed 14 m while distance from the tree trunk did not exceed 12 m to ensure high-resolution around focal trees. Scanning positions were adjusted between 6 to 12 m and between +/− 30° to obtain optimal viewing angles decreasing occlusion effects. Scanning resolution of the FARO Focus3D was set at one-fourth resolution with 4× quality. Registration targets were set between all 4 scan points consisting of 3 spherical foam targets placed approximately 1 m apart from other targets between each scan point. Target sets were used to register and merge the 4 scans into a local coordinate system. To ensure adequate point cloud collection and quality, scans were collected during near windless conditions during leaf-off conditions in 2022 and 2023 spring. Scan co-registration was completed in FARO SCENE version 6.2.5.7 (FARO Technologies, Inc. 2021). Point clouds were exported into CloudCompare v2.11.3 (Girardeau-Montaut 2020), and noise above the canopy and below-ground level were removed manually. Cleaned tree point clouds were exported into Computree v.5.0.221b for Dikstra-based tree segmentation using the SimpleForest plugin (version 5.3.2) and an optimized workflow (Hackenberg et al. 2015). Segmentation results were then loaded into CloudCompare to isolate the focal tree as well as ensure no overlap between tree segmentations occurred. Any erroneous segmentation result was carefully examined and removed or stitched to the correct tree using the merge cloud tool within CloudCompare. Individual trees were extracted and cleaned using the Statistical Outlier Removal (SOR) Filter Tool with base parameters of 6 points for mean distance estimation and 1.00 standard deviation multiplier threshold (nSigma) in CloudCompare after segmentation to prepare as inputs for constructing quantitative structural models (QSMs).
Quantitative Structural Models (QSMs) and Structural Metrics
To obtain individual crown structural metrics quantitative structure models (QSMs) were built using the GitHub repository InverseTampere/TreeQSM (Raumonen et al. 2013) in Matlab (version 2022a). Crown architectural metrics were extracted for each QSM using the R package ITSMe (Terryn 2022), including ‘Crown Height,’ ‘Crown Evenness,’ and ‘Crown Diameter to Height Ratio,’ and the TreeQSM_Architecture repo in Matlab, ‘Crown Area,’ ‘Crown Volume Asymmetry,’ and ‘Total Volume’ (Appendix). Structural metrics are defined in detail by Terryn et al. (2020) and were adapted from Åkerblom et al. (2017) (Appendix). Metrics were extracted for both pre- and post-trimming conditions for the 15 trimmed trees and only for the “pre” sampling period for the 9 control trees. Control trees were examined in the field and compared against the LiDAR scans to ensure no significant branch loss or damage occurred between the period of pre- and post-accelerometer data collection. Tree heights and DBH were derived from the ITSMe R package (Terryn 2022).
Statistical Analysis
All statistical analysis was conducted in R (version 4.2). To analyze the effect of tree trimming on fundamental frequencies, analysis of variance (ANOVA) was performed on the change in frequencies between (Δfreq) pre- and post-trimming condition across the 3 trimming type groups (Control, SMT, and ETT) using the ‘lme4’ and ‘lmerTest’ packages. In the model, Δfreq was set as the response variable while trimming types were set as fixed effects with species and plot as random effects.
To analyze the effect of tree trimming on tree displacements, data for both pre- and post-trimming were first compiled to create a histogram of number of observations (10-minute periods) by maximum wind gust. See Appendix for maximum crown displacements. From this distribution of data, 2 response variables, ‘Δ_median_maxD’ (the change in maximum displacement around the median wind speed in the dataset) and ‘Δ_90_maxD’ (the change in maximum displacement around the 90th percentile of wind speed in the dataset), were derived by selecting all displacement data within +/− 1 m/s around both values. For Δ_median_maxD, all data between 6.33 to 8.33 m/s were analyzed while displacement values for wind speeds between 11.33 to 13.33 m/s were analyzed for Δ_90_maxD. As above, ANOVA was used to evaluate differences between trimming types with Δ_median_maxD and Δ_90_maxD set as response variables and trimming type as fixed effects with species and plot as random effects. An additional model evaluated the effect of genera on displacement where Δ_median_maxD and Δ_90_maxD were set as response variables with genera and trimming types as fixed effects and plot as a random effect.
Multicollinearity among crown structural metrics was assessed before inclusion as potential predictors of biomechanical response. This was achieved through pairwise correlation matrices developed using Pearson’s linear correlations (Appendix) using the ‘corrplot,’ ‘Hmisc,’ and ‘tidyr’ packages. Additionally, the final set of predictors were analyzed for multicollinearity using the Variation Inflation Factor (VIF) function. Based on these assessments one structural metric, ‘Canopy Height,’ was removed as a predictor from the final regression analysis as it was found to be highly correlated with the ‘Crown Diameter to Crown Height Ratio’ metric. After determining the final set of predictors to be included to assess changes in these metrics between pre and post, Welch’s 2-sample t -tests were performed in order to evaluate changes to crown structure among ETT and SMT for each of the 5 structural predictors.
Finally, to evaluate the effects of structural predictors on changes in frequency and displacement associated with trimming, mixed effects models were built using the ‘lmer’ package. The response variables for the 3 models were Δfreq, Δ_median_maxD, and Δ_90_maxD, with species included as a random effect. Model comparisons were performed for all structural predictor combinations and the models with the lowest Akaike Information Criterion (AIC) were evaluated as the most highly supported models.
Results
For the change in frequency across trimming treatment types, there was no statistically significant difference among treatments (n = 24, P = 0.472), and individual comparisons between treatments and Controls were non-significant for both ETT (P = 0.645) and SMT (P = 0.424)(Figure 3). There was also no significant difference across the treatment groups for displacement at high wind levels (Delta_90_maxD; n = 24, P = 0.788) nor for individual comparisons with controls for either ETT (P = 0.497) or SMT (P = 0.742)(Figure 3). However, there was a significant difference among treatments for displacement at moderate wind levels (Delta_median_maxD; n = 24, P = 0.032) and individual comparisons displayed a significant difference of treatment groups from controls for both ETT (P = 0.042) and SMT (P = 0.015) (Figure 3). Subsequent pairwise comparisons revealed a statistically significant difference between controls and SMT (P = 0.045), however no statistically significant differences between control and SMT (P = 0.106), nor between ETT and SMT (P = 0.8515). Further analysis for the effect of genera (oaks [Quercus spp.], maples [Acer spp.], and hickories [Carya spp.]) for displacement at high wind levels (Delta_90_maxD) while including treatment type showed no statistically significant differences for oaks (P = 0.927), maples (P = 0.898), or hickories (P = 0.117) and ETT (P = 0.476) or SMT (P = 0.687). However, there was a significant difference for species and treatment type for displacement at moderate wind levels (Delta_ median_maxD) while including treatment type for hickories (P = 0.001), ETT (P = 0.039), and SMT (P = 0.0135) but not for oaks (P = 0.112) or maples (P = 0.376).
Δ (change) across displacement and frequency for trimming types, Control, Enhanced Tree Trimming (ETT), and Scheduled Maintenance Trimming (SMT) with a dashed line representing baseline at zero. Top: Change in maximum displacement in the 90th percentile in meters between 11.33 to 13.33 meters per second (m/s) wind speed conditions across trimming types. No statistically significant difference across treatment groups (n = 24, P = 0.788) or for individual comparisons with controls for either ETT (P = 0.497) or SMT (P = 0.742). Middle: Change in maximum displacement around the median in meters between 6.33 to 8.33 m/s wind speed conditions across trimming types. Statistically significant difference across treatment groups (n = 24, P = 0.032) and for individual comparisons with controls for ETT (P = 0.042) and SMT (P = 0.015). Bottom: The change in frequency between pre and post conditions across trimming types. No statistically significant difference among treatments (n = 25, P = 0.472) and individual comparisons between treatment and controls (ETT; P = 0.645) and (SMT; P = 0.424).
Analysis comparing changes in crown structural metrics between the treatment groups found that only 1 of 5 structural parameters ‘Crown Diameter to Crown Height Ratio’ (n = 15, P = 0.05067) changed differently between ETT and SMT treatments (Figure 4). The ratio between crown diameter and crown height varied more following ETT, which shifted the starting point of the crown to a point further up the stem and decreased the crown diameter (Figure 4). The remaining 4 structural predictors did not change differently across the 2 treatment groups; ‘Crown Area’ (n = 15, P = 0.2575), ‘Crown Volume Asymmetry’ (n = 15, P = 0.7193), Crown Evenness’ (n = 15, P = 0.5609), and ‘Total Volume’ (n = 15, P = 0.8193) (Figure 4).
Welch’s 2-sample t-test between Enhanced Tree Trimming (ETT) and Scheduled Maintenance Trimming (SMT) for each Δ (change) in structural predictor with dashed line representing baseline at zero; (n = 15). Top: Change in Crown Diameter to Crown Height Ratio (P = 0.050); Second: Change in Crown Area (P = 0.257); Middle: Change in Crown Asymmetry (P = 0.719); Fourth: Change in Crown Evenness (P = 0.560); Bottom: Change in Total Volume (P = 0.819).
Mixed effects modeling supported a single model, with ‘Crown Diameter to Height Ratio’ as a predictor, showing statistical significance (P = 0.05) in predicting frequency change (Table 1). For predicting change in displacement during average wind conditions (Delta_median_maxD), 3 models were highly supported based on information criteria statistics (all within +/− 2 units of the lowest AIC score), although all had low explanatory power (Table 1). The first model included ‘Crown Volume Asymmetry’ as the predictor for the change in displacement during average wind conditions while the second model included ‘Crown Diameter to Height’ and the third model included ‘Crown Evenness’ as important predictors (Table 1). Similarly for predicting change in displacement during high wind conditions (Delta_90_maxD), 4 models were highly supported based on information criteria statistics (Table 1). The first model included both ‘Crown Volume Asymmetry’ and ‘Crown Evenness’ as important predictors for the change in displacements during high wind conditions while the second model included ‘Crown Diameter to Height Ratio’. The third model included only ‘Crown Volume Asymmetry,’ and the fourth model only included ‘Crown Evenness’ (Table 1).
Coefficients for linear models for displacement and frequency for study trees (n = 24). Average wind speed corresponds to wind data between 6.33 to 8.33 m/s; high wind speed corresponds to wind data between 11.33 to 13.33 m/s. AIC = Akaike information criterion.
Discussion
Trimming on Effects on Crown Structure and Variable Crown Form
In this study, we compared the effects of 2 commonly implemented trimming types used in the northeastern United States (ETT and SMT), on tree biomechanics and crown structure. We found that ETT generally resulted in more substantial changes to crown form and tree symmetry. This finding is congruent given SMT tends to be a less intensive treatment, largely focused on removing smaller branches between trimming cycles white ETT generally removes all branches from the ground up on one side of the tree. The effect of treatment types on crown structure varied across individuals and species likely due to variable crown form. Initial tree forms differed across genera as a result of variable crown form and positioning of branches throughout individual tree crowns. Tree crown architectures are intricately variable, influenced by their developmental sequence, biomechanical constraint of shapes, and the growing environment (Echereme et al. 2015). The growing environment of these trees likely drove their initial crown form and shape as, being edge trees, they were exposed to more sunlight and relatively free of neighboring crowns on one side of the tree (facing the roadside). Individual trees also differed in the amount of branches located in the trimming area, with the result that some study trees had more branches removed than others within the same trimming type. Furthermore, prior research has shown that intra- and inter-specific competition can affect crown diameter, and height relating to differences in shade tolerance among individuals (Del Río et al. 2019). As these study trees have been growing along roadsides, there is likely high inter-specific competition for resources, especially light, affecting their overall form and number of branches. Additionally, initial tree form also differed among individuals of the same species suggesting intra-specific variability. Hickories (Carya spp.) in this study exhibited simpler crown architecture, with narrower crowns starting higher up the stem. Maple (Acer spp.) and oak (Quercus spp.) crowns tended to be larger overall with more branches although maple crowns began much lower on stems. Moreover, larger oaks had more dead branches than maples and hickories, which were subsequently removed by trimming and oak branches in general were larger in diameter. Overall, initial individual tree form and crown characteristics (such as crown shape and branching arrangement resulting from competition, biomechanical constraints, and growing environment) can affect the outcome of trimming on canopy structure.
Effects of Tree Trimming on Sway Frequency and Displacement
We found a lack of consistent changes in tree movement characteristics directly following tree trimming across 24 roadside trees represented by 3 genera and 5 species with variable initial tree and crown form. As wind loads impact a tree’s crown, branches respond through complex sway motion moving strategically to prevent the development of dangerous large sway motions (James 2003). Branch removal causes less energy to be dissipated through a tree’s branching architecture and is rather directed towards the stem, potentially causing an increase in movement. Branches within the crown act to dampen wind effects, however, when damping is reduced through crown removal the effects of movement on the tree bole may become more pronounced (Moore and Maguire 2005). Based on this, we expected there to be a change in sway frequency of trees following trimming. However, our results indicated no significant difference in sway frequency for trimmed trees relative to pre-trimming conditions or controls (Figure 3). Prior research has indicated > 80% of the crown mass may need to be removed to produce any significant difference in frequency (Moore and Maguire 2005). Change in crown area for our trees following trimming ranged between 22% to 55% for ETT and 14% to 35% for SMT with some outliers (Figure 4). Therefore, standard trimming practices alone may not be enough to alter the frequency of tree movement due to the relatively low proportion of crown mass removed.
We expected a stronger effect of the more intensive ETT treatment relative to SMT, but did not find a strong consistent change in frequency between trimming types (Figure 3). One factor potentially limiting differences between trimming types was that treatments as actually applied in the field were not particularly disparate. There was often overlap, as trees classified as ETT in practice aligned more closely with SMT specifications which was observed directly in the field following trimming. Branch removal for some trees classified as ETT closely resembled that of the SMT treatment. Conversely, certain trees classified as SMT showed significantly more branch removal, with one SMT outlier experiencing over a 60% change in crown area (Figure 4). Lack of differentiation may be attributed to differences in initial tree crown form, with some species naturally having crowns starting higher in the tree with less lower branches to be removed during trimming (e.g., hickories). Lack of change in frequency may also be attributed to the absence of leaves. Prior research has shown that the presence of leaves increases the size of the sail area and thus changes the fundamental frequency (Kane and James 2011; Bunce et al. 2019; Jackson et al. 2019b). This may suggest that while the results ofthis study do not show consistent changes in frequencies, a study incorporating more trees in addition to the presence of leaves could potentially see more pronounced differences across treatment types.
We hypothesized that crown architectural traits such as crown asymmetry would be key predictors of changes in tree biomechanics following trimming. However, although crown architecture was altered by trimming, crown asymmetry and other architectural traits did not have a discernible effect on frequency. This may be attributed to prior tree acclimation to wind, as the edge trees in this study have likely adjusted to repeated mechanical stimuli which can change their xylem structure effectively causing them to become more wind-firm (Badel et al. 2015). To this point, trees in this study generally were larger in diameter and less slender than surrounding interior forest trees based on visual observations in the field. Prior studies have found that tall slender trees tend to have a simpler architecture resulting in a higher fundamental frequency placing them at a higher likelihood of wind damage (Jackson et al. 2019b). Interior trees growing taller to compete for limited resources are thus more slender and sheltered by surrounding trees, thus not acclimating to strong winds like individuals growing along a forest edge (Sellier and Fourcaud 2009). The edge trees in this study may, therefore, have previously been able to allocate more resources for secondary growth and stability relative to neighboring interior trees, potentially limiting their responsiveness to trimming.
We found that immediately following trimming, trees that were trimmed increased in displacement relative to control trees during moderate but not high wind conditions particularly for hickories. Hickories may have been affected in particular due to their slender architecture over oaks and maples. The greater change in displacement in the moderate wind conditions over high wind conditions may also be related to the limited range of wind conditions (< 15 m/s), however, this may indicate that trimming effects could be overwhelmed in greater wind speed conditions. High wind conditions (11.33 to 13.33 m/s) were high for the period of this study but were not storm-level wind conditions. For example, recent notable storms that caused widespread power outages and significant damage in the northeast United States included Tropical Storm Isaias in 2020 (gusts of 17.8 to 22.3 m/s)(LeComte 2021), Tropical Storm Irene in 2011(sustained winds 17.8 to 22.3 m/s and gusts up to 29.9 m/s)(Hart 2011), and Hurricane Sandy in 2012 (gusts up to 40.2 m/s)(National Weather Service 2012). The effects of such wind conditions were not considered in this study (no events of this magnitude occurred). Our results also suggest that changes in crown structure associated with tree trimming can have some effect on tree crown displacement in the wind. Crown evenness and asymmetry were found to be significant, but not necessarily strong, predictors of displacement change for both moderate and high wind speed conditions (Table 1). As noted above, treatment overlap, low relative proportion of crown area removed, and prior wind acclimation may also have reduced the effects of trimming on displacement.
This pilot project represents an initial assessment of the potential relationships between tree crown structural changes and tree biomechanics following trimming. Both accelerometer-based assessment of tree movement characteristics and TLS-based analysis of tree crown characteristics are time consuming and difficult to implement at scale. Therefore, other approaches could be used to supplement these techniques. Accelerometer-based data loggers are relatively cheap and accessible, but it can be difficult to derive clean meaningful data on fundamental frequencies (relative to inclinometers or strain gauges) due to inherent accelerometer noise. While TLS can provide highly-detailed information on crown structure, it is often constrained to small spatial areas due to setup and scan acquisition time (Calders et al. 2020). To incorporate entire roadside forest stands with more trees into analysis of trimming effects on crown and canopy structure, unmanned aerial vehicle-(UAV-) LiDAR could be tested for extracting similarly detailed crown structure. Scaling up could be valuable if strong and consistent relationships between crown architecture and biomechanical change are illustrated in future studies, limiting the need to monitor sway directly at the individual tree or species level. For this study, we focused on the immediate response of tree biomechanics to trimming during leaf-off conditions. Further assessment of tree sway dynamics in leaf-on conditions may reveal more significant effects of trimming and provide additional insights into changes to tree biomechanics.
Conclusions
This study comprised a novel analysis characterizing relationships between 3-dimensional crown structure and tree biomechanics on trees subjected to variable tree trimming practices. We focused on the immediate response of tree biomechanics after trimming during leaf-off conditions across 3 roadside forest sites. We found there were not consistent changes to tree movement characteristics directly following the application of tree trimming. However, we found that the associated changes in crown structure as a result of tree trimming affected tree displacement in moderate wind conditions. This study suggests that the effects of trimming on crown architecture may not have negative effects on tree stability in the near term, but our results are limited to leaf-off conditions where tree movement is generally less pronounced and the linkage between crown architecture and movement is likely to be less significant. Future work should focus on leaf-on conditions and changes in tree movement and crown architecture over the course of a utility trimming cycle, which will provide more complete information on the effects of tree trimming on roadside tree stability. Despite the lack of strong patterns observed in this study, understanding changes to crown structure following tree trimming will be important in promoting resilient trees during wind disturbances along with decreasing damage to utility infrastructure while reducing power outages through targeted vegetation management intervention.
Conflicts of Interest
The authors reported no conflicts of interest.
Acknowledgements
The authors thank Isaac Betts, an undergraduate research technician, for assisting with field work and data collection. This project was supported by funding from Eversource Energy in Connecticut through the Eversource Energy Center at UConn.
Appendix
Description of the 24 study trees. DBH = diameter at breast height.
Selected and analyzed structural metrics obtained from QSMs (quantitative structural models) from the ITSMe R package (Terryn et al. 2023). Descriptions from Åkerblom et al. (2017), Terryn et al. (2020), and using the TreeQSM_Architecture matlab functions (Jackson et al. 2019b).
Spectral diagram showing the estimated dominant fundamental frequency for Tree #4902, which is a 45.2 cm Carya tomentosa (Mockernut Hickory) tree, using a fast fourier transform in R.
Terrestrial Laser Scanning field scan position layout design for each focal tree.
All study trees maximum crown displacement (meters) and maximum speed (m/s) per 10-minute period for all available data pre- and post-trimming across the 3 treatment types control (top), Enhanced Tree Trimming (ETT)(middle), and Scheduled Maintenance Trimming (SMT)(bottom). Displacement curves are colored by conditions, pre- (blue) and post- (orange).
Correlation matrix for structural predictors against the 3 response variables Delta_Freq (change in frequency between pre- and post-conditions across trimming types), Delta_median_maxD (maximum displacement around the median – change in median displacement in meters between 6.33 to 8.33 m/s wind speed conditions), and Delta_90_maxD (maximum displacement in the 90th percentile – change in maximum displacement in meters between 11.33 to 13.33 m/s wind speed conditions) with P-values displayed for non-significant correlations.
- © 2024 International Society of Arboriculture
Literature Cited
Jump to section
Related Articles
Cited By...
- No citing articles found.