The Anatomy and Biomechanical Properties of Bifurcations in Hazel

(Corylus avellana L.)

A thesis submitted to the University of Manchester for the degree of DOCTOR OF PHILOSOPHY in the Faculty of Life Sciences

2015

Duncan Slater

This page intentionally left blank

Table of Contents Preliminary Sections Page No. Abstract 15

Declaration 16

List of abbreviations 17

Acknowledgements 18

Preface 19

Dedication 20

Chapter 1: Introduction

1.1 Introduction 22

1.2 Literature review 22

1.2.1 Definition of a tree bifurcation 1.2.2 Definitions of mechanical properties related to this study 1.2.3 Mechanical failure of bifurcations in trees 1.2.4 Bifurcations with included bark 1.2.5 Previous research into the mechanical performance of bifurcations in trees 1.2.6 The mechanical properties of greenwood (xylem) in relation to bifurcations 1.2.7 Previous research into the anatomy of junctions in trees 1.2.8 Trade-offs in xylem 1.2.9 Literature review summary 1.3 Research aims and objectives 42

1.3.1 Selected species and junction type for investigation 1.3.2 Thesis structure

1.4 References 49

Chapter 2: Determining the mechanical properties of bifurcations in hazel (Corylus avellana L.) by testing their component parts

2.1 Chapter Abstract 58

2.2 Introduction 59

2.3 Materials and Methods 65

2.3.1 Sample collection and organisation 2.3.2 Rupture tests 2.3.3 Calculation of bifurcation breaking stress 2.3.4 Three point bending tests 2.3.5 Sample size 2.3.6 Sampling for basic density testing 2.3.7 Statistical analysis

2.4 Results 74

2.4.1 Rupture tests of hazel bifurcations 2.4.2 Three point bending tests

2.5 Discussion 79

2.6 References 83

Chapter 3: The anatomy and grain pattern in bifurcations of hazel (Corylus avellana L.) and other tree species

3.1 Chapter Abstract 86

3.2 Introduction 87

3.3 Materials and Methods 90

3.3.1 Superficial examination 3.3.2 Internal anatomical investigations

3.4 Results 97

3.4.1 Superficial examination 3.4.2 Internal anatomy

3.5 Discussion 107

3.6 References 113

Chapter 4: Interlocking wood grain patterns provide improved wood strength properties in bifurcations of hazel (Corylus avellana L.)

4.1 Chapter Abstract 116

4.2 Introduction 117

4.3 Materials and Methods 122

4.3.1 Sample collection 4.3.2 Compression tests 4.3.3 Tensile tests 4.3.4 Basic density testing 4.3.5 Statistical analysis

4.4 Results 128

4.4.1 Compression tests 4.4.2 Tensile tests 4.4.3 Basic density

4.5 Discussion 131

4.6 References 135

Chapter 5: The level of occlusion of included bark affects the strength of bifurcations in hazel (Corylus avellana L.)

5.1 Chapter Abstract 138

5.2 Introduction 139

5.3 Materials and Methods 143

5.3.1 Sampling 5.3.2 Rupture tests 5.3.3 Three point bending tests 5.3.4 Measurements of included bark 5.3.5 Statistical analysis

5.4 Results 149

5.4.1 Effects of the extent and location of included bark

5.5 Discussion 156

5.6 References 162

Chapter 6: An assessment of the remodelling of bifurcations in hazel (Corylus avellana L.) in response to bracing, drilling and splitting 6.1 Chapter Abstract 166

6.2 Introduction 167

6.3 Materials and Methods 171

6.3.1 Selection of hazel bifurcations 6.3.2 Modification to the hazel bifurcations 6.3.3 Observations 6.3.4 Rupture testing 6.3.5 Basic density testing

6.3.6 Statistical analysis

6.4 Results 180

6.4.1 Specimen losses and mean specimen dimensions 6.4.2 Observations of bifurcations prior to testing 6.4.3 Rupture testing 6.4.4 Basic density at hazel bifurcations 6.5 Discussion 191

6.5.1 Discussion of results by bifurcation type 6.5.2 Basic density 6.5.3 Limitations of the study 6.5.4 Conclusions

6.6 References 197 Chapter 7: An assessment of the movement behaviour of bifurcations in hazel (Corylus avellana L.) under dynamic wind loading using accelerometers 7.1 Chapter Abstract 200

7.2 Introduction 201

7.3 Materials and Methods 205 7.3.1 Selection of hazel bifurcations 7.3.2 Accelerometry 7.3.3 Wind speed assessment 7.3.4 Observations 7.3.5 Rupture tests 7.3.6 Statistical analysis

7.4 Results 213

7.4.1 Summary of primary data 7.4.2 Observations 7.4.3 Regression analysis 7.4.4 Differences in movement related to bifurcation type 7.4.5 Rupture tests 7.4.6 The influence of bifurcation morphology and position

7.5 Discussion 225

7.6 References 229

Chapter 8: Discussion 8.1 Synthesis of research findings 234

8.2 Estimating the factor of safety for bifurcations in trees 238

8.3 Critique of methodologies used 243

8.4 Recommendations for further research work 246 8.4.1 Causes of differing bifurcation anatomy 8.4.2 Extending the study to other plant species 8.4.3 The relative risk of failure for bifurcations in trees 8.4.4 Opportunities for biomimicry of bifurcation anatomy 8.4.5 Advanced techniques for assessing bifurcations 8.4.6 Creating innovative remedial treatments for flawed bifurcations in trees

8.5 Implications of the findings of this study for arboricultural 253 practices 8.5.1 Pruning 8.5.2 Bifurcations and tree hazard management 8.5.3 Bark inclusions 8.5.4 Bracing 8.5.5 Split bifurcations

8.6 References 258

Table of Figures

Chapter 1: Fig. 1.1 A bifurcation in hazel (Corylus avellana L.) 23 Fig. 1.2 Illustration of a yield point and breaking point of a hazel 26 bifurcation Fig. 1.3 Location of maximum bending stresses in branches and 27 bifurcations under bending Fig. 1.4 Failure of a bifurcation under wind-loading 31 Fig. 1.5 3D diagram of xylem cells formed in an angiosperm 34 Fig. 1.6 Shigo’s model of branch attachment 36

Chapter 2: Fig. 2.1 Types of tensile failure in tree bifurcations 60 Fig. 2.2 Observation of grain orientation on the fracture surface 61 of a hazel bifurcation Fig. 2.3 Location of the pivot point in calculating the maximum 62 breaking stresses of bifurcations Fig. 2.4 Illustration of the three components that contribute to 64 the mechanical strength of a bifurcation in a tree Fig. 2.5 Material removed in the two component tests 67 Fig. 2.6 Diagram illustrating morphological measurements taken 68 for each bifurcation Fig. 2.7 Rupture testing of hazel bifurcations and additional 69 measurements Fig. 2.8 Diagram of the three-point testing rig used 72 Fig. 2.9 Locations for extracting 5 mm wood cores from 25 73 hazel bifurcations for basic density testing and slide production Fig. 2.10 Boxplot of maximum breaking stress of all sample sets 75 in this study Fig. 2.11 Comparison of the anatomy of ‘junction wood’ and 78 normal stem wood

Chapter 3: Fig. 3.1 Implausible wood grain arrangements at the bifurcations 88 of woody plants Fig. 3.2 Sample preparation for CT scanning 93 Fig. 3.3 Determination of the inter-vessel tortuosity of the 96 vessels segmented out from the scanned hazel volumes Fig. 3.4 Visual observations of the wood grain orientation at de- 98 barked junctions and surfaces of fractures of split bifurcations Fig. 3.5 Visualisation of CT Scan output 104 Fig. 3.6 Contrasting form of rays in tangential view 106 Fig. 3.7 Simplified pattern of interlocking wood grain 109 Fig. 3.8 Schematic diagram of the arrangement of cell types at a 110 bifurcation Fig. 3.9 Whirled grain at a bifurcation in common ash (Fraxinus 111 excelsior L.)

Chapter 4: Fig. 4.1 Two improbable wood grain arrangements at 118 bifurcations in trees Fig. 4.2 Comparison of vessel shapes in stem wood and junction 119 wood of hazel Fig. 4.3 Interlocking wood grain patterns at bifurcations in trees 120 Fig. 4.4 Diagram of compression test methodology 124 Fig. 4.5 Diagram of tensile test methodology 125 Fig. 4.6 Orientation of wood strength testing at a bifurcation 126 Fig. 4.7 Location for excision of tensile testing samples for bark- 127 included bifurcations Fig. 4.8 Remedial xylem growth at a bark-included bifurcation 133

Chapter 5: Fig. 5.1 Type I, Type II and branch failure modes of tree 140 bifurcations under tension across the bifurcation.

Fig. 5.2 Potential development pathways for a bark-included 141 bifurcation Fig. 5.3 Measurements of the fracture surfaces of bark-included 146 bifurcations carried out in ImageJ Fig. 5.4 Categorisation of bifurcations with included bark into 147 two types Fig. 5.5 Failure modes in relation to the diameter ratio of the 150 bifurcation Fig. 5.6 Typical force/displacement graphs for specimen types 151 Fig. 5.7 Boxplot of mean yield stress of branches compared with 152 the mean breaking stresses of the normally-formed bifurcations and bifurcations with included bark Fig. 5.8 Boxplot of mean breaking stress of bifurcation types 154 Fig. 5.9 Suggested contrast in bending behaviour between a low 158 diameter ratio bifurcation and a high diameter ratio bifurcation Fig. 5.10 Weaker and stronger forms of bifurcations with 160 included bark

Chapter 6: Fig. 6.1 Wood grain patterns at the apices of bifurcations 169 Fig. 6.2 Artificially-modified bifurcations left to grow in-situ for 174 two to four years Fig. 6.3 Three types of bark-included bifurcation found in hazel 178 trees Fig. 6.4 Features of the development of the artificially-modified 183 bifurcations Fig. 6.5 Boxplot of mean breaking stresses for the main 184 bifurcation types and mean yield stress for branches tested

Fig. 6.6 Boxplot of mean breaking stresses of the three types of 185 bark-included bifurcation tested Fig. 6.7 Boxplot of mean breaking stresses of the three types of 187 pre-drilled bifurcation tested Fig. 6.8 Comparison between the breaking stresses for normally- 189 formed and braced bifurcations Fig. 6.9 Boxplot of mean basic density of samples excised from 191 the apices and sides of normally-formed and braced bifurcations

Chapter 7: Fig. 7.1 Diagram of the three tri-axial accelerometers affixed to a 206 hazel bifurcation using cable ties Fig. 7.2 Method of chronological synchronisation of 208 accelerometer readings

Fig. 7.3 Scatterplot of Va against a1 for seven normally-formed 216 bifurcations over seven sessions

Fig. 7.4 Scatterplot of Va against a2 for seven normally-formed 217 bifurcations over seven sessions Fig. 7.5 Differences in synchronisation in branch movement by 219 bifurcation type (all measuring sessions) Fig. 7.6 Differences in synchronisation in branch movement by 220 bifurcation type (three windiest measuring sessions) Fig. 7.7 Acceleration profiles in the y-z plane for three hazel 221 bifurcations under minor and major wind loading

Fig. 7.8 Scatterplot of a1 against h for all seven sessions 223

Chapter 8: Fig. 8.1 Attempt to create a natural brace traversing a bifurcation 251 Fig. 8.2 A naturally-formed brace above a bark-included 252 bifurcation Fig. 8.3 A mature split bifurcation that is a significant hazard 257

List of Tables

Chapter 2: Table 2.1 Criteria for the selection of the drill size used on 66 modified bifurcations

Chapter 3: Table 3.1 Species and specimen numbers for bifurcations that 91 were subject to visual observations as part of this study Table 3.2 Comparisons in the anatomy of xylem from the stems 101 and junction apices of hazel bifurcations Table 3.3 The abundance and average height of rays in the two 106 tissue types scanned

Chapter 4: Table 4.1 Compressive strength of excised cubes with standard 129 errors Table 4.2 Tensile strength of excised dumbbells with standard 130 errors

Chapter 5: Table 5.1 Analysis of failure modes observed 149

Chapter 6: Table 6.1 Determination of drill size for modifying drilled hazel 173 bifurcations Table 6.2 Summary of bifurcation specimens tested 175 Table 6.3 Basic density of samples taken from different 190 bifurcation types tested, by location

Chapter 7: Table 7.1 Summary data for each measuring session 214-5

Chapter 8: Table 8.1 Compatibility of the anatomical model with previous 235-237 findings Table 8.2 Mean bending stresses for branches and bifurcations of 240 hazel tested in this study

Word Count: 61733

Abstract Academic Centre: The University of Manchester Candidate: Duncan Slater Supervisor: Prof. A. Roland Ennos Degree Programme: PhD in Plant Sciences Title of Thesis: The Anatomy and Biomechanical Properties of Bifurcations in Hazel (Corylus avellana L.) Date of Submission: 30th September 2015

The anatomy of bifurcations in trees requires further scientific investigation as the current anatomical model for them is logically flawed. The provision of a better model will assist in scientific studies of woody plants, the risk assessment of junctions in mature trees and provide bio-inspiration for Y-shaped joints in composite materials.

In this study, the xylem formed in the central axis of a hazel (Corylus avellana L.) bifurcation is shown to provide a disproportionately greater amount of its tensile strength. CT scanning identified that this centrally-placed xylem was 28.1% denser, with 63% less vessels formed in this tissue, such vessels being 50.5% of the diameter and 32.5% of the length of those formed in adjacent stem tissues. The wood grain pattern at the bifurcation apices were 22 times more tortuous, forming interlocking patterns that acted to resist tensile forces by requiring the extraction or breaking of wood fibres along their length (the axial tensile strength of wood). Subsequent tests confirmed that this conferred more than 100% additional tensile strength to these specialised xylem tissues. These findings provided the basis of a novel anatomical model for bifurcations in woody plants.

Further to this, the effects of several factors upon junction strength and biomechanical behaviour were assessed in bifurcations of hazel, identifying the weakening effect of bark inclusions and three types of artificial modification as well as differences in wind- induced movement between bifurcation types.

This study concludes that further investigations of bifurcations in a wider range of woody plants and observations of the developmental stages of the interlocking wood grain patterns found at bifurcations would usefully add to existing knowledge.

Declaration

The candidate declares that no portion of the work referred to in this thesis has been submitted in support of an application for another degree or qualification of this or any other university or other institute of learning.

Copyright Statement i. The author of this thesis (including any appendices and/or schedules to this thesis) owns certain copyright or related rights in it (the “Copyright”) and he has given The University of Manchester certain rights to use such Copyright, including for administrative purposes. ii. Copies of this thesis, either in full or in extracts and whether in hard or electronic copy, may be made only in accordance with the Copyright, Designs and Patents Act 1988 (as amended) and regulations issued under it or, where appropriate, in accordance with licensing agreements which the University has from time to time. This page must form part of any such copies made. iii. The ownership of certain Copyright, patents, designs, trademarks and other intellectual property (the “Intellectual Property”) and any reproductions of copyright works in the thesis, for example graphs and tables (“Reproductions”), which may be described in this thesis, may not be owned by the author and may be owned by third parties. Such Intellectual Property and Reproductions cannot and must not be made available for use without the prior written permission of the owner(s) of the relevant Intellectual Property and/or Reproductions. iv. Further information on the conditions under which disclosure, publication and commercialisation of this thesis, the Copyright and any Intellectual Property and/or Reproductions described in it may take place is available in the University IP Policy (see http://documents.manchester.ac.uk/DocuInfo.aspx?DocID=487), in any relevant Thesis restriction declarations deposited in the University Library, The University Library’s regulations (see http://www.manchester.ac.uk/library/about/regulations) and in The University’s policy on Presentation of Theses.

List of Abbreviations

ANOVA – analysis of variance statistical test BSI – British Standards Institution CT - computerised tomography DF – degrees of freedom in statistical tests FEA – finite element analysis GLM ANOVA – general linear model analysis of variance statistical test ISA – International Society of Arboriculture ITFD – International tree failure database MFA – Microfibril angle (of cellulose strands in the walls of xylem cells) NTP – Natural target pruning SE – Standard error of the mean UCLan – University of Central Lancashire, England UTM – universal testing machine UTS – ultimate tensile strength 2D – two dimensional 3D – three dimensional

Acknowledgements

It is not practicable to provide here a list of all the contributors to this research, as there are so many names dutifully recorded in the twelve project books I have used over the six years of this part-time PhD programme. Every person on my collated list has made a significant contribution to the outcomes of this project, should that be through the sharing of their ideas, thoughts, resources or connections to other researchers. My apologies that I cannot name you all here, but I thank you all for your help in my pursuit of knowledge on bifurcations in trees.

As my supervisor, Professor Roland Ennos has been pivotal to the success of this research by imparting his excellent experience in researching and reporting the biomechanics of animals and plants and always casting an astute eye over the project’s progress and outcomes. I have learnt a great deal under his kind supervision which I can take forward when undertaking further research.

Support from staff of the University of Manchester has been essential; I would especially like to acknowledge the contribution of Dr. Robert Bradley of the Manchester X-ray Imaging Facility in creating novel coding for analysis of the CT scans of wood samples.

Many thanks must go to the undergraduate students at UCLan who undertook side tasks related to this research, namely: Joe Barnes, Gareth Buckley, Matthew Dumelow, Owen Haines-Myers, Claire Harbinson, Peter Lowes, Laurence Smith, Sam Turner and Ian Williams.

Funding of this six year part-time PhD programme was provided by Myerscough College, and I would like to express my gratitude to the College for its continued support over such a long period of study.

Preface

A brief statement of the author’s current degrees and research experience is included here, for the benefit of external examiners.

The author of this thesis currently holds the following degrees:

Degree Type Subject Awarding Body Year of Award

BSc (Ord) Forestry University of 1996 Aberdeen

BA (Hons) Humanities with Open University 2001 Philosophy

MSc Resource Management Middlesex 2003 (Arboriculture) University

The candidate has worked for the University of Central Lancashire as a lecturer in arboriculture at Myerscough College, Lancashire since 2006, including the supervision of undergraduate research projects. This supervisory role has led to the production of three co-authored research publications concurrently with the candidate’s research reported within this thesis, and a list of these publications is provided here:

Slater D and Harbinson C J (2010) Towards a new model for branch attachment; Arboricultural Journal 33, 95-105.

Turner S, Slater D and Ennos A R (2012) Failure of bifurcations in clonal varieties of Platanus x acerifolia (Aiton) Willd; Arboricultural Journal 34, 179-189.

Andrew C and Slater D (2014) Why some UK homeowners reduce the size of their front garden trees and the consequences for urban forest benefits as assessed by i-Tree ECO; Arboricultural Journal 36, 197-215.

Dedication

Alison Barbara Slater (1947-2010) For selflessly sharing her hobby of growing trees with her youngest son.

Chapter 1

Introduction

1.1 Introduction

This thesis describes a six year part-time PhD study into the anatomy and biomechanical properties of bifurcations in the stems of hazel trees. A key outcome of this research is the reporting of a new anatomical model for bifurcations in hazel trees, based upon analysis of wood grain patterns and tracheary element orientations to be found at such junctions. Further to this, a series of experiments are presented that assessed the mechanical performance of bifurcations in hazel subject to natural and artificial defects and under dynamic wind loading.

A key purpose of this research work is to facilitate better scientific modelling of tree anatomy and biomechanical performance of components of a tree’s crown. A further purpose for this research work had a more practical outcome: informing arboriculturists and tree surveyors how to assess the factor of safety of a junction in a tree, so that they can successfully manage risks arising from trees growing near to people or property.

This thesis is presented in alternative format and contains six result chapters that report the main research outcomes from this study.

1.2 Literature Review

1.2.1 Definition of a tree bifurcation

A bifurcation in the stem of a tree is a junction where the stem divides in two, allowing bi-directional growth (Fig. 1.1). Bifurcations are often informally categorised into two types depending on the ratio of the diameters of the two branches which are conjoined; where a smaller diameter branch conjoins with a much larger diameter tree stem or trunk this is often referred to as a ‘branch-to-stem attachment’ or ‘branch attachment’(Harris et al., 2004); where the two conjoined branches are of similar

22 diameter, the bifurcation is commonly referred to as a ‘fork’ (OED online, 2015) or the two branches as ‘co-dominant stems’ (Shigo, 1986, 1991; Smiley, 2003; Kane, 2014). This is not a strict botanical division between bifurcation types, as there is a continuous series of diameter ratios that can occur between conjoined branches – it is akin to the common division in woody plants between ‘shrubs’ and ‘trees’ in that, however you choose to define them, some species and specimens exhibit intermediate characteristics.

Figure 1.1: A bifurcation in hazel (Corylus avellana L.)

A branch junction with a high diameter ratio (c. 80%), which is frequently formed in the stem structure of hazel; author’s own image (2015). The apex of the bifurcation is labelled, as the specialised xylem formed at this location will be the focus of much of this study. The union of these two branches is formed by wood that lies between the bifurcation of the pith within the branch and the bifurcation apex, and this zone is visible externally by virtue of the appearance of a ridge of disturbed bark, referred to as the ‘branch bark ridge’ (Shigo, 1991). For this thesis, the axes directions, as shown in the bottom right hand side of this image (x = horizontally in-line with the bifurcation; y = horizontally perpendicular with the bifurcation; z = vertically ascending the main stem), will be used to explain the plane of view of all relevant subsequent figures.

Bifurcations are a feature of all woody plants; high diameter ratio bifurcations are very common features in the crowns of trees that grow in a decurrent pattern (e.g. Acer, Aesculus, Quercus, Ulmus, Zelkova), and appear less frequently where trees grow in an excurrent pattern (e.g. Abies, Picea) (Kozlowski et al., 1991). Raimbault (1991) identified 25 architectural models for ornamental shrubs and trees, the majority of which contain these high diameter ratio bifurcations.

1.2.2 Definitions of mechanical properties related to this study

There are a number of attributes of materials and material testing that initially need to be defined in order to discuss the mechanical performance of structural components of a tree.

The strength of a material is often reported in terms of both its yield strength and its ultimate tensile strength (UTS) (Howatson et al., 1991). Further to this, it is also important to note that the yield strength of a material under compressive or shear stresses may be very different from its tensile yield stress, and that maximum yield stresses for a material may vary in the x, y and z axes.

Yield stress, σy, measured in pascals, is the stress level at which a material starts to undergo plastic deformation (Gere and Timoshenko, 1996). Applying a stress level below the yield stress to a solid material would typically give rise to elastic behaviour, with the material exhibiting a linear relationship between its displacement and the force applied (Hooke’s Law), and when the stress was relieved then the material would return to its original state as it was when at rest. Once the yield point is surpassed, however, permanent deformation of the material will occur. Consequently, the yield strength of a material is an important measure, as it defines the level of stress at which damage to a material or structure begins (Ennos, 2011).

UTS is also measured in pascals and identifies the stress level at which a material breaks or wholly fails under tension (Gere and Timoshenko, 1996). The UTS of a material is typically higher than its yield stress, but in very brittle materials that exhibit no plastic deformation they can coincide (Beer et al., 2014).

Figure 1.2 identifies the yield point and breaking point of a hazel bifurcation to illustrate these definitions.

Figure 1.2: Illustration of yield point and breaking point of a hazel bifurcation

Force/displacement diagram of a normally-formed bifurcation of hazel undergoing a rupture test (method described in section 2.3.2 of this thesis), identifying the yield point and breaking point (usually where the peak force is applied), for which the bending stresses at these points can be estimated.

Further to these introductory definitions, there are two main complications in investigating the mechanical properties of component parts of trees; first, that the xylem (wood) formed by trees is a material that is highly anisotropic, having very different mechanical properties axially, radially and tangentially under tension and compression (Panshin and De Zeeuw, 1980); second, most failures of temperate trees are caused by bending stresses (Niklas, 1992), so it is often more appropriate to test the component parts of trees by bending them, rather than crushing, pulling, shearing or tearing them.

When a tree branch or bifurcation undergoes bending, maximum bending stresses occur in the outermost wood fibres (Ennos and van Casteren, 2010), with maximum tensile stresses occurring at the convex outer edge of the bent member and maximum

26 compressive stresses occurring at the concave inner edge (Fig. 1.3). Unlike many artificial materials, the greenwood of temperate angiosperms mostly has a significantly lower yield stress in compression than under tension (Dresch and Dinwoodie, 1996), resulting in the initial yielding of a bending branch on its concave side, at the point of maximum compressive stress (van Casteren et al., 2012). It is also important to note that a bending test will often give a different result than a tensile test of the same material or structure because of the presence of surface defects (resulting in a lower maximum bending stress when compared with the UTS) or internal defects (resulting in a lower UTS when compared with the maximum bending stress) (Hodgkinson, 2000).

Figure 1.3: Location of maximum bending stresses in branches and bifurcations under bending

A: Location of maximum bending stresses in a branch undergoing bending, with the maximum tensile stress at the outer (convex) edge of the branch’s curvature, and the maximum compressive stress at the inner (concave) edge. Initial yielding of the branch, when the yield stress (σb) is reached, is most likely to occur on the inner edge under compressive forces due to the anisotropic properties of the wood formed in the branches of most temperate tree species.

B: Location of maximum tensile and compressive stresses in a bifurcation when the two branches are being bent away from each other in the xz plane. F represents a force acting on one branch of the bifurcation to pull it away from the other branch, and FR represents the reaction force as the other branch is being restrained, a method used in the rupture tests reported in this thesis to assess the bending strength of bifurcations. Initial yielding of the bifurcation may occur either where the compressive stresses are highest at the outer edge of the Y-shape (Type I failure) or where the tensile stresses are highest on the inner edge of the Y-shape (Type II failure). This difference in failure mode is related to the diameter ratio of the two branches that are conjoined at the bifurcation (as shown in Chapter 2 of this thesis), with Type II failures becoming more common when the two branches are of near-equal diameter, as shown in this figure.

In this study the maximum bending stresses of bifurcations (σa) and the yield stresses of branches (σb), calculated in megapascals (MPa), are used to report differences in the bending strength (in other literature sometimes referred to as ‘flexural strength’) of branches and different types of bifurcation. An assessment is also made as to whether there are significant differences in the mean breaking stresses of different types of bifurcation.

1.2.3 Mechanical failure of bifurcations in trees

Arboriculture is a relatively new discipline, arising from the horticultural and forestry industries and focussed specifically on tree management practices (Harris et al., 2004). Many of the tree management practices carried out by arborists are based upon the

28 reported experiences of arboriculturists in what one would term ‘grey literature’ and technical information has often been disseminated by trade organisations. As a consequence, scientific studies within arboriculture are currently limited in scope and number. Dahle et al. (2014) have recommended that the arboricultural industry coordinates research into tree biomechanics, as, for example, there is often an insufficient link between identification of a defect in a tree and scientific evidence of the increased likelihood of the tree failing due to this defect.

A number of authors published in arboricultural texts have identified junctions in trees as potential failure points in stormy weather that tree surveyors should be aware of and assess carefully (Shigo, 1991; Matheny and Clark, 1994; Lonsdale, 1999; Hartman et al., 2000; Harris et al., 2004) and some authors have gone so far as to advise that bifurcations are potentially hazardous if their diameter ratio is high (Kane et al., 2008; Ryder and Moore, 2013; Gilman, 2015).

Both Gilman (2003) and Kane et al. (2008) found that the diameter ratio (or ‘aspect ratio’) between the two branches of a bifurcation had a substantial effect on the bifurcation’s bending strength, with bifurcations with a ratio of branch diameters near to 100% being approximately half as strong as those with a ratio of 50%. This is strong evidence that attachments of two co-dominant branches at a bifurcation have a lower load-bearing capacity than branch-to-stem attachments but does not directly show that their factor of safety is reduced, as the biomechanical behaviour of such junctions under wind loading has not been taken into account, which would be a logical continuation of this experimental work.

There is no substantial body of scientific work that supports this latter assertion and this seems at odds with the fact that nearly all crowns of decurrent trees contain multiple high diameter ratio bifurcations in order that a single trunk can give rise to a sufficiently wide canopy to provide the trees’ requirement for photosynthates. If the factor of safety for such bifurcations was much lower than for other components of the tree’s crown (i.e. branches, twigs, trunk) then one would anticipate that bifurcation failure would be more

29 frequent in storms than any other failure type, but there is a lack of collated industry- based evidence to support this proposition (Dahle et al., 2014; ITFD, 2015). In a relatively small sample, Kane (2014) found no significant difference in breaking stress between normally-formed co-dominant stems and branches/boughs in the crowns of red oaks (Quercus rubra L.) when carrying out destructive pulling tests on mature trees. It may well be the case that this is a misperception of the relative safety of bifurcations that comes from the high frequency of failures of bifurcations with included bark and potentially some misinterpretation of the few previous scientific studies that have focussed on the strength of junctions in trees. These two factors are considered below.

1.2.4 Bifurcations with included bark

A common malformation of a bifurcation in a tree is the inclusion of bark within the union of the two branches (Fig. 1.4).

Figure 1.4: Failure of a bifurcation under wind-loading

A split bifurcation in common ash (Fraxinus excelsior L.) exhibiting included bark at the bifurcation apex, which failed during a winter storm in Lancashire, England; author’s own image (2010).

These bark-included bifurcations most frequently occur in the mid-crown of trees, where two ascending near-vertical branches conjoin at a tight angle (Lonsdale, 2000). If the secondary thickening of both branches is quicker than the development of wood at the bifurcation apex, bark then becomes trapped at the apex of the bifurcation, weakening the join. The two arising vertical limbs of such a bifurcation will also result in a greater likelihood of bark being included as the bifurcation apex will not be constantly loaded with substantial leverage from either limb due to the disposition of each limb’s gravitational loading.

There is ample industry-based evidence that these malformed bifurcations are frequent points of failure in trees under dynamic wind-loading (Shigo, 1986; James, 1990; Matheny and Clark, 1994; Lonsdale, 1999, 2000; Hartman et al., 2000; Smiley, 2003; Harris et al., 2004; Helliwell, 2004; Gilman, 2011, 2015; Dahle et al., 2014; Kane, 2014) – however, there is currently no satisfactory means to assess such bifurcations for gradations in strength, so as to identify those that are more likely to fail and which constitute a substantial defect in a tree. As this is a significant issue in the management of trees for their associated risks and hazards, Chapter 5 of this thesis directly addresses this question and provides a suggested means to assess the strength of such a bark- included bifurcation.

1.2.5 Previous research into the mechanical performance of bifurcations in trees

The mechanical performance of bifurcations in trees has mostly been limited to static testing of specimens to the point of failure by carrying out rupture tests (also referred to as ‘tensile tests’) to obtain their maximum bending stress (Lilly and Sydnor, 1995; Gilman, 2003; Smiley, 2003; Pfisterer, 2003; Kane et al., 2008). Whilst this approach is informative in that it provides a measure of the bending strength of the bifurcation, it does not replicate the dynamic wind loading that would most frequently act to break a bifurcation in a temperate tree. As a consequence, care is needed in the interpretation of

32 such studies, if one were to apply their findings to a standing tree whose risk of wind- induced failure had to be assessed.

Both Smiley (2003) and Pfisterer (2003) examined the mechanical performance of bifurcations with included bark. Smiley (2003) experimented with red maple (Acer rubrum L.) bifurcations, and found that the mean breaking stress of bark-included bifurcations was 20% lower than that of normally-formed bifurcations. Pfisterer (2003) experimented with common hazel (Corylus avellana L.) and concluded that bark- included bifurcations and bifurcations with different angles showed no significant differences in breaking stress but that normally-formed bifurcations exhibited a sudden break upon failure whereas the bark-included bifurcations did not. Pfisterer (2003) also carried out three point bending tests on branches of hazel, and found that the bifurcations broke at stresses approximately 20% lower than the stresses needed to break these branches.

It is important to note that these previous mechanical studies only assessed the anatomy of their bifurcations superficially, most citing another published branch attachment model (Shigo, 1985) to introduce or discuss their findings. This anatomical model is critiqued in section 1.2.7 of this chapter. This is a substantial omission in this previous work, as it is important to relate the strength of a biological component to its anatomical features.

1.2.6 The mechanical properties of greenwood (xylem) in relation to bifurcations

The xylem formed by woody plants in their stems and branches is a complex biological material with anisotropic mechanical properties (Cannell and Morgan, 1987; Carlquist, 2001; Fratzl, 2007). In stems and branches, the axial strength of the greenwood is often several times higher than its radial or tangential strength (Panshin and De Zeeuw, 1980). This difference in strength is the result of the principal wood grain arrangement (Fig. 1.5), with most cells aligned axially (along the length of the branch or stem). The radial

33 strength of greenwood is typically ~20% greater than the tangential strength (although this does vary considerably by species), which has been shown to be primarily due to the radial arrangement of medullary rays (Burgert et al., 1999; Reiterer et al., 2002).

Figure 1.5: 3D diagram of xylem cells formed in an angiosperm

(Image adapted from Grosser, 1977). This diagram illustrates the main cell types and features to be found in the xylem formed in woody fruiting higher plants (angiosperms). The anisotropic nature of the xylem can readily be seen in this 3D representation, with most cells arranged axially and the ray cells arranged radially, resulting in different yield strengths for this xylem in the axial, radial and tangential directions. The following features are labelled by letters in this diagram: LP: scalariform plate; P: pits that allow sap transfer between vessels; PA: parenchyma cells; R: ray.

Xylem production in woody plants occurs on the inner edge of the vascular cambium (Philipson et al., 1971) which is the innermost layer of the inner bark. The xylem formed by the vascular cambium has been shown to adapt to the strains experienced by the vascular cambium (Jaffe, 1973), and those adaptations are multi-factorial – the

34 qualities of the xylem produced may be altered, as well as the quantity of xylem being produced (Brayton and Archer, 1979). Such alterations can involve increased wood density (van Gelder et al., 2006), changes in the frequency of cell types within the new xylem (Christensen-Dalsgaard et al., 2007), alterations in grain orientation and cell form (Burgert et al., 1999), as well as sub-cellular changes, especially of the microfibril angle of cellulose strands in the cell walls of tracheids and fibres (Boyd and Foster, 1974; Brayton and Archer, 1979; Lichtenegger et al., 1999; Burgert, 2006; Donaldson, 2008).

The anisotropic nature of xylem as a material, its substantial change in mechanical properties in different planar orientations and its ability to remodel when subjected to mechanical strains means that changes in wood grain direction which occur at a bifurcation are likely to prove to be a very important factor in explaining the bifurcation’s bending strength. One means by which trees can avoid producing a weak material in a key location in their structure is by wood grain re-orientation, such that the xylem elements are orientated parallel to the forces they have to withstand so the xylem can provide greater strength at that location (Kramer and Borkowski, 2004; Jungnikl et al., 2009). However, the wood grain patterns to be found at high diameter ratio bifurcations (‘forks’) are not comprehensively reported, with authoritative publications such as the ‘Atlas of Woody Plant Stems’ (Schweingruber et al., 2006) providing insufficient illustration and explanation of the wood grain arrangement at such branch junctions. Shigo’s model of branch attachment (Shigo, 1985) does attempt to illustrate the wood grain orientation at low diameter ratio bifurcations and this model is critiqued in the next subsection of this introduction.

1.2.7 Previous research into the anatomy of junctions in trees

Published in the Canadian Journal of Botany in 1985, Shigo’s paper entitled ‘How tree branches are attached to trunks’ has been cited frequently and its illustrations used repeatedly in arboricultural literature, including the majority of studies on bifurcations cited in this introduction. In particular, the expanded diagram of overlapping branch

35 and stem tissues described in the paper is very familiar to trained arborists and often used to illustrate branch attachment in other publications (Fig. 1.6).

Figure 1.6: Shigo’s model of branch attachment

A pictorial representation of Alex Shigo’s model of branch attachment, showing three seasons of growth and three ‘collars’ from the trunk overlapping incremental growth at the base of a lateral branch (Redrawn from: Shigo, 1985).

Shigo (1985) describes a model of branch attachment reliant upon the formation of a series of overlapping collars produced in succession by the base of the branch and then the trunk in each growing season. The base of the branch is also said to produce a ‘tail’ of xylem tissue that is then, in most instances, completely overlapped by the trunk collar (Fig. 1.6). This model of attachment was determined by chainsaw dissection, the de-

36 barking of branch junctions, branch pulling, branch junction splitting, testing of water transport patterns using dyes and analysis of decay patterns after pruning.

In a direct challenge to this model of branch attachment, Neely (1991) carried out conduction tests with dye at different times of the year and discounted the presence of the overlapping ‘tails’ of tissue proposed in Shigo’s model (1985). Kramer and Borkowski (2004) have also questioned the existence of ‘tails’ at the base of branches, and consider Shigo’s model to overstate the case for grain realignment at junctions of trees. In the main, however, there has been little debate on Shigo’s model of branch attachment, and it has only recently been challenged as the accepted model for teaching branch-to-stem attachment in arboricultural education (Slater and Harbinson, 2010).

In attempting to prove this anatomical model, Shigo (1985) relied upon an experiment in which thirty branch junctions in oak (Quercus spp.) were debarked in June and then this process was repeated with a further set of thirty branch junctions in August of that year. It was upon this experiment that the conclusions were based about the alternate seasonal overlapping of branch and trunk tissues. However, this experiment does not give conclusive proof that the trunk collar tissues envelope the branch collar tissues during one growing season, followed by the overlaying of another branch collar at the start of the next season. As this experiment was destructive to the samples being examined and was only carried out over one growing season, it only illustrates that different branch junctions at different times of the year showed different levels of occlusion of the smaller branch into the larger adjoining stem.

The vascular cambium is continuous from trunk to branch, forming a complete sheath around tree junctions (Philipson et al., 1971). No scientific research since 1985 has shown that the vascular cambium is capable of producing the ‘overlapping collars’ illustrated in Shigo’s branch attachment diagram (Fig. 1.6). Research prior to and cited within Shigo’s paper (Larson and Richards, 1981; Larson and Fisher, 1983) does not describe such a layout of xylem tissues.

This model purports to explain both the sap conductance routes found at junctions and the mechanical strength of these junctions. Shigo (1985) states that branches are mechanically attached to trunks through the overlapping of branch collars with trunk collars, season upon season and even dismisses the ‘tails’ of tissue below the branch as being too insubstantial to provide significant mechanical support. This supposition raises two obvious problems with this anatomical model: First, where trees grow in tropical climates, cambial growth may be continuous or erratically periodic and this casts doubt on the assumption that there are seasonal differences in the timing of growth within the vascular cambium situated at the base of the branch and in the adjacent stem; Second, given that the structure of a high diameter ratio bifurcation intergrades with the structure of a low diameter ratio bifurcation (i.e. they are topologically equivalent), this model offers no explanation as to how high diameter ratio bifurcations in trees are held together. For branch-to-stem attachments, Shigo (1985) asserts that the sole significant mechanical attachment is the seasonal overlaying of trunk collars upon branch collars, and that these are absent in co-dominant stems (Shigo, 1986). For high diameter ratio bifurcations, it follows that there will be no such overlaying of tissues as both arising branches will be growing at a roughly equal rate and at the same time.

It could be argued that there are two forms of limb attachment in trees – one for high diameter ratio bifurcations and another for low diameter ratio bifurcations – but this proposition would need to overcome two difficulties. The first difficulty would be defining these two attachment types, when such attachments come in all possible diameter ratios and show no substantial morphological differences apart from varying levels of occlusion of the smaller diameter branch into the larger one. It is highly implausible that there would be an absolute cut-off point between one type of attachment anatomy and another based solely on the diameter ratio of the two branches that were conjoined. The diameter ratio between branches is not static and can be significantly affected by growth rate, the branches’ position in the tree crown and by pruning (Gilman, 2015). The second difficulty would be providing a complementary anatomical model for high diameter ratio bifurcations that does not alter Shigo’s model for low diameter ratio bifurcations by implication: for example, by creating an anatomical model for high diameter ratio bifurcations that is reliant on the xylem tissues

38 formed under the branch bark ridge for the mechanical strength of such bifurcations but not for low diameter ratio bifurcations, which also have a branch bark ridge.

Research by plant anatomists has reported the common occurrence of tortuous grain patterns and whirled wood grain at the junctions of woody plants (Lev-Yadun and Aloni, 1990; Kramer 1999; Lev-Yadun, 2000). Lev-Yadun and Aloni (1990) found circular tracheary elements were a common feature of branch junctions in fifteen woody plant species. This phenomenon of circular vessels and tracheary elements has also been found to be frequent at the nodes of herbaceous plants (André, 2000).

It is noticeable that there has not been a synthesis of these observations, in that Shigo (1985) gives no account of the presence of whirled grain at junctions in his model, and that Lev-Yadun and Aloni (1990) ascribe the role of whirled grain at bifurcations only in terms of its hydraulic function and neglect that this xylem may have changed mechanical properties and perform a structural function too. Their only suggestion in their conclusion is that whirled grain may result in branches falling off more readily where this wood grain pattern occurs (Lev-Yadun and Aloni, 1990), which has not been proven and seems improbable.

With whirled grain patterns being observed so frequently at the apex of bifurcations in a wide range of woody plants, an anatomical model for bifurcations that provides an explanation of its mechanical strength needs to take this frequently-found anomalous wood grain pattern into account.

1.2.8 Trade-offs in xylem

In the biological sciences, it is well-reported that xylem tissues in trees (the wood formed in the tree’s branches, stems and roots) perform multiple functions (Tyree and Zimmermann, 2003; Nobel, 2009; Schweingruber et al., 2011). Sapwood provides mechanical support to the tree’s structure, the storage of solid carbohydrates such as

39 starch and it also functions as a conduit for sap to ascend to the tree’s crown to facilitate transpiration of its foliage (Mencuccini et al., 1997; Ennos, 2001; Dahle and Grabosky, 2009; Hacke, 2015).

In terms of the structure that the xylem produces to perform these functions, it is always a compromise, as these functions require conflicting attributes in the sapwood (Chiu and Ewers, 1992; Carlquist, 2001; Hacke and Sperry, 2001; McCulloch and Sperry, 2005; Sellin et al., 2008; Chave et al., 2009). For instance, the conduction of sap is far more efficient through larger tracheary elements with thinner cell walls, but the more that the wood contains such elements, the less it can supply the requisite mechanical support. Gartner (1995) identifies a number of trade-offs exhibited in the xylem of woody plants, including the trade-off between mechanical strength and efficiency in sap conduction found in many woody plant species, and she illustrates this further with extreme examples of woody stem distortion to favour factors other than mechanical strength (Gartner, 2001). Harris et al. (2004) state that the structure of a branch attachment must also be a compromise between sap transport and mechanical support.

In biomechanical studies, this state of trade-off between functions of the xylem is sometimes not acknowledged. The most frequent manifestation of this phenomenon is where the term ‘mechanical optimisation’ is used in biomechanical studies of trees to describe how the xylem tissues or aerial structures produced by trees perform (Mattheck and Kubler, 1995; Mattheck and Bethge, 1998; Lichtenegger et al., 1999). It is considered more appropriate by the author of this thesis to discuss mechanical performance in terms of its efficiency, effectiveness and its impact upon other ‘goals’ of the organism or other functions of the tissues (Farnsworth and Niklas, 1995). Crawley (1994) stated that trees represent efficient designs related to their evolutionary history and their development within certain environmental constraints: he dismissed the proposition that tree design is ‘optimal’.

This is an important distinction to make within this thesis, as bifurcations in trees must act as conduits for sap as well as mechanical components in a tree’s crown. It is known

40 that junctions in trees are zones of restricted sap conductance within trees (Larson and Isebrands, 1978; Zimmermann, 1978; Tyree and Alexander, 1993), which suggests that such junctions may be biased to trading off efficiency in sap conductance to ensure they have sufficient mechanical strength. Their ‘design’ is most likely to be an effective and efficient trade-off of these two functions, having undergone many millions of years of natural selection. The means by which this co-efficiency is achieved is the most important lesson that can be learnt from investigating the anatomy and mechanical performance of bifurcations. Again, this approach to the research requires a synthesis between previous reported findings by specialists in mechanics and those findings reporting upon the hydraulic architecture of trees.

1.2.9 Literature Review Summary

From this literature review it is found that there is no current anatomical model for high diameter ratio bifurcations in woody plants and that the widely-accepted anatomical model for low diameter ratio bifurcations in trees is not logically consistent and most probably errant. Overall, there is a need to draw together biomechanical and biological studies in order to synthesise a new anatomical model for bifurcations that is consistent with the findings of both these disciplines. Of greatest importance is to outline the wood grain orientation at bifurcations, as this will be key to a satisfactory explanation of their mechanical properties. Once this new model is tested, it can be used to relate anatomical features of bifurcations meaningfully with their biomechanical properties, leading to a better comprehension of the subject for biologists and arborists.

1.3 Research Aims and Objectives

The key aim of this study (Aim 1) was to produce a model for the anatomy of junctions in trees that was logically consistent with the findings reported in published scientific literature on this topic and which provided a plausible explanation of the known mechanical performance of these components of tree crowns.

To achieve Aim 1 required a critique and synthesis of previous published findings and a methodical approach to assessing a number of attributes of such junctions by visual observation and mechanical testing. The subsequent anatomical model would need to explain how the three factors of wood grain orientation, tissue arrangement and mechanical properties of the xylem formed at bifurcations provide a sufficient load- bearing capacity to the junction. To do this required that the anatomical model was verified through scientific experimentation, which gave rise to the following inter- linked objectives:

Objective 1: To determine by experiment which component of a bifurcation contributed most to its bending strength

Objective 2: To determine by observation of wood grain orientation and tissue arrangement if there were differences in the anatomy of the xylem tissues forming the relevant component identified by fulfilling objective one when compared with xylem tissues in stems and branches

Objective 3: To determine by experiment if any anatomical differences identified by fulfilling objective 2 result in different mechanical properties for the xylem tissues forming the relevant component within junctions when compared with xylem tissues in stems and branches

Once this anatomical model gave a sufficient plausible explanation for the mechanical behaviour of normally-formed bifurcations, it was determined that factors that might affect the relative strength of such junctions would be assessed (Aim 2), as this could

42 provide a range of practical applications of the anatomical modelling for use by tree owners, arborists and others. To achieve Aim 2, the following objectives were set for this study:

Objective 4: To determine by experiment how the diameter ratio of a junction affected the bending strength by relating this to the junction’s anatomy

Objective 5: To determine by experiment the extent by which the presence of included bark weakens a junction, and whether gradations in the strength of bark- included junctions can be predicted

Objective 6: To determine by experiment what the effect of bracing and other artificial interventions have upon junction strength

Objective 7: To determine by experiment whether junction strength relates to the relative level of mechanical perturbation experienced by assessing bifurcations for differences in their movement under dynamic wind loading

Overall the two aims and associated seven objectives, if met, would provide a detailed assessment of junctions in trees that would elucidate how their anatomy supplied their biomechanical properties, and would assess the strengthening or weakening effects of some common variations in junction morphology. This would not only inform scientific work on the modelling of tree form and physiology, but also usefully inform the arboricultural and forestry industries that have to manage some junctions in trees as potential hazards.

1.3.1 Selected Species and Junction Type for Investigation

Early experiments in this study involved examination and testing of a range of woody species, but to achieve a coherent and consistent narrative for this thesis it was decided to concentrate upon analysing sample bifurcations of one woody plant species: The species selected was common hazel (Corylus avellana). There were a considerable number of factors in favour of this decision. First, hazel is prone to producing many bifurcations in its crown structure which have a high diameter ratio and are very frequently not marred by the presence of other knots or secondary lateral branches, so that obtaining high numbers of similar Y-shaped replicates from a selected wooded area was achievable. Second, hazel is locally common in the north west of England and is frequently coppiced, resulting in the regular and sustainable local supply of bifurcations which are essentially ‘waste’ from the coppice-worker’s viewpoint. Third, hazel has a mid-range wood density and in stature lies between what a layperson may commonly call a ‘shrub’ and a ‘tree’ depending on its growing environment, so that findings in this species may well be applicable to a wider range of other deciduous broadleaved shrubs and trees. Finally, the wood of hazel is not toxic nor would samples of bifurcations be difficult to handle or give rise to additional safety precautions due to the presence of thorns, resins or other biohazards.

Basic density was calculated for wood samples extracted from stems and bifurcations. Basic density is defined as the weight of the dried wood of a sample divided by the volume of the sample when it was greenwood (Osazuwa-Peters and Zanne, 2011). This was the preferred density measure because it more closely informs the failure of bifurcations when in a ‘greenwood’ state and conducting sap.

Bifurcations were selected as the type of junction to test as these represented the most common form of a junction to be found in a tree. In addition, published research has concentrated on reporting upon this type of junction, as referenced in the literature review supplied in this chapter, and this would then allow findings of this study to be set in the light of this previous work. Junctions with a greater number of arising branches

(e.g. trifurcations, quadfurcations, etc…) are far less common than bifurcations in the crowns of trees and their assessment by mechanical testing would be necessarily more complex and open to criticism whichever methodology was used.

A critique of these decisions is provided in the discussion chapter of this thesis (Chapter 8).

1.3.2 Thesis Structure

This thesis is presented in the alternative format under the rules and regulations of the University of Manchester. The six results chapters within this thesis are presented in a style suitable for the journals in which they have been placed or where the work has been submitted. At the time of writing this thesis, four of these papers have been published and one is in press to be published by mid-2016. At the start of each results chapter the basic details of the associated paper are provided where relevant, including the journal to which it was submitted and its current status.

These result chapters have been edited and reformatted to achieve consistency in the presentation of figures and references. Editing of the original papers has also involved the removal of repetitive details within the materials and methods sections of chapters 5, 6 and 7. However, there is necessarily some repetition of citations and closely- interrelated themes in the introductions and discussions of these results chapters, which is a common feature when presenting a PhD thesis in the alternative format.

Presented below is an outline of the results chapters in this thesis that identifies the research aims and objectives which each chapter addresses. Details are also given of the associated journal paper or journal submission and the contributions of all authors to the work.

Chapter 2: Determining the mechanical properties of bifurcations in hazel (Corylus avellana L.) by testing their component parts

Authors: Slater D and Ennos A R

Journal: Trees: Structure and Function

Status: Published 2013

Fulfilled aims and/or objectives: Aim 1: Objective 1

Contributions of all authors: The idea for this initial experiment came about through discussion between the authors, to find a suitable method to assess the contribution to bending strength of different components of a bifurcation. The first author selected and cut samples of hazel, carried out the experiment and analysed the data. The first author also wrote the initial draft manuscript in its entirety, which Prof. Ennos then commented upon. The first author then revised the manuscript, submitted it to the chosen journal and revised it further in response to reviewers’ comments.

Chapter 3: The anatomy and grain pattern in bifurcations of hazel (Corylus avellana L.) and other tree species

Authors: Slater D, Bradley R S, Withers P J and Ennos A R

Journal: Trees: Structure and Function

Status: Published 2014

Fulfilled aims and/or objectives: Aim 1: Objective 2

Contributions of all authors: The first author initiated this investigation, making observations of de-barked bifurcations and supplying samples for CT scanning. Dr. Bradley assisted in the carrying out of the CT scanning process, and wrote novel coding in MATLAB so that the arising data could be used to compare xylem tissue attributes. The first author carried out the necessary data handling processes to apply the MATLAB code and wrote an initial draft paper, to which Dr. Bradley added a section to the method which provided the detail of the novel coding used. The resulting draft manuscript was then reviewed by Prof. Withers and Prof. Ennos prior its submission to

46 the selected journal. The first author then responded to the reviewer’s comments and revised the paper accordingly.

Chapter 4: Interlocking wood grain patterns provide improved wood strength properties in bifurcations of hazel (Corylus avellana L.)

Authors: Slater D and Ennos A R

Journal: Arboricultural Journal

Status: Published 2015

Fulfilled aims and/or objectives: Aim 1: Objective 3

Contributions of all authors: This experimental work was carried out at the recommendation of Prof. Ennos. The first author sourced the hazel material, processed them into the appropriately-shaped samples for testing, and carried out all the necessary experimentation. The first author drafted the entirety of the paper, upon which Prof. Ennos commented, and then it was submitted to an appropriate journal.

Chapter 5: The level of occlusion of included bark affects the strength of bifurcations in hazel (Corylus avellana L.) Authors: Slater D and Ennos A R

Journal: Journal of Arboriculture and Urban Forestry

Status: Published 2015

Fulfilled aims and/or objectives: Aim 2: Objectives 4 and 5

Contributions of all authors: This experiment was initiated by the first author, having had a considerable amount of experience of the failure of bark-included junctions in trees in his previous career. Sample collection was facilitated by Mike Carswell, a coppice worker based in Manchester. On collection and analysis of data, a draft manuscript was written by the first author, upon which Prof. Ennos commented, and this was then submitted to a journal thought suitable for this arboricultural topic. The first author then revised the paper on the basis of the reviewers’ comments.

Chapter 6: An assessment of the remodelling of bifurcations in hazel (Corylus avellana L.) in response to bracing, drilling and splitting

Authors: Slater D and Ennos A R

Journal: Journal of Arboriculture and Urban Forestry

Status: Submitted 09.06.2015

Fulfilled aims and/or objectives: Aim 2: Objective 6

Contributions of all authors: The first author initiated this experiment, as a logical progression from the experiment reported in Chapter 2. All practical work and data analysis was carried out by the first author, and a draft manuscript was presented to Prof. Ennos for review prior to submission to the selected journal.

Chapter 7: An assessment of the movement behaviour of bifurcations in hazel (Corylus avellana L.) under dynamic wind loading using tri-axial accelerometers

Authors: Slater D and Ennos A R

Journal: Agricultural and Forest Meteorology

Status: Submitted 16.07.2015

Fulfilled aims and/or objectives: Aim 2: Objective 7

Contributions of all authors: An initial pilot study, at the suggestion of and under the supervision of the first author was carried out by Laurence Smith, an undergraduate student of Myerscough College in 2013-14. However, the data arising from this pilot study was deemed insufficient so the first author also initiated this further experiment, as a logical progression of the findings reported in Chapters 5 and 6. Contributions of the authors were exactly as outlined for Chapter 6 (above).

Several figures used in the associated papers and within this thesis were created by David Elwell and Mike Heys, both former employees of Myerscough College, at the direction of the author and their contributions to the quality of the original illustrations within this thesis is duly acknowledged here.

1.4 References

André J P (2000) Heterogeneous, branched, zigzag and circular vessels: unexpected but frequent forms of tracheary element files: description, localization, form; In: Savidge R., Barnett J. and Napier, R. (ed.s) Cell and molecular biology of wood formation; Oxford, Bios Scientific Publishers, 387-395.

Beer F P, Johnston Jr. E R, DeWolf J T and Mazurek D F (2014) Mechanics of materials; 7th Edition; Boston, McGraw-Hill Education.

Boyd J D and Foster R C (1974) Tracheid anatomy changes as responses to changing structural requirements of trees; Wood Science and Technology 8, 91-105.

Brayton F W and Archer R R (1979) Tree design: some biological solutions to mechanical problems; BioScience 29, 293-298.

Burgert I, Bernasconi A and Eckstein D (1999) Evidence for the strength function of rays in living trees; Holz als Roh – und Werkstoff 57, 397-399.

Burgert I (2006) Exploring the micromechanical design of plant cell walls; American Journal of Botany 93, 1391-1401.

Cannell M G R and Morgan J (1987) Young’s modulus of sections of living branches and tree trunks; Tree Physiology 3, 355-364.

Carlquist S (2001) Comparative Wood Anatomy; 2nd Edition; Springer Series in Wood Science; Berlin Heidelberg New York, Springer-Verlag.

Chave J, Coomes D, Jansen S, Lewis S L, Swenson N G and Zanne A E (2009) Towards a worldwide wood economics spectrum; Ecology Letters 12, 351-366.

Chiu S-T and Ewers F W (1992) Xylem structure and water transport in a twiner, a scrambler and a shrub of Lonicera (Caprifoliaceae); Trees: Structure and Function 6, 216-224.

Christensen-Dalsgaard K K, Fournier M, Ennos A R and Barfod A S (2007) Changes in vessel anatomy in response to mechanical loading in six species of tropical trees; New Phytologist 176, 610-622.

Crawley M J (1994) Life History and Environment; In: Crawley M J (ed.) Plant Ecology; 2nd Edition; Oxford, Blackwell Science.

Dahle G and Grabosky J (2009) Review of literature on the function and allometric relationships of tree stems and branches; Journal of Arboriculture and Urban Forestry 35, 311-320.

Dahle G, Grabosky J, Kane B, Miesbauer J, Peterson W, Telewski F W, Koeser A and Watson G W (2014) Tree Biomechanics: a white paper from the 2012 international meeting and research summit at the Morton Arboretum (Lisle, Illinois, US); Journal of Arboriculture and Urban Forestry 40, 309-318.

Donaldson L (2008) Microfibril angle: measurement, variation and relationships – a review; IAWA Journal 29, 345-386.

Dresch H E and Dinwoodie J M (1996) Timber; Structure, Properties, Conversion and Use; London, MacMillan Press Ltd.

Ennos A R (2001) Trees; London, Natural History Museum.

Ennos A R and van Casteren A (2010) Transverse stresses and modes of failure in tree branches and other beams; Proceedings of the Royal Society B 277, 1253-1258.

Ennos A R (2011) Solid Biomechanics; New Jersey, Princeton University Press.

Farnsworth K D and Niklas K J (1995) Theories of optimization, form and function in branching architecture in plants; Functional Ecology 9, 355-363.

Fratzl P (2007) Biomimetic materials research: what can we really learn from nature's structural materials?; Journal of the Royal Society Interface 4, 637-642.

Gartner B L (1995) Patterns of xylem variation within a tree and their hydraulic and mechanical consequences; In Gartner B L (ed.) Plant Stems; Physiological and Functional Morphology; New York, Academic Press.

Gartner B L (2001) Multitasking and tradeoffs in stems, and the costly dominion of domatia; New Phytologist 151, 311-313.

Gere J M and Timoshenko S P (1996) The Mechanics of Materials; 4th Edition; Boston, PWS Publishing Co.

Gilman E F (2003) Branch to stem diameter affects strength of attachment; Journal of Arboriculture 29, 291-294.

Gilman E F (2011) An illustrated guide to pruning; 3rd edition; Independence, Kentucky US, Cengage Learning.

Gilman E F (2015) Pruning severity and crown position influence aspect ratio change; Journal of Arboriculture and Urban Forestry 41, 69-74.

Grosser D (1977) Die Hölzer Mitteleuropas; New York, Springer-Verlag.

Hacke U G (ed.) (2015) Functional and Ecological Xylem Anatomy; Cham, Switzerland , Springer International Publishing.

Hacke U G and Sperry J S (2001) Functional and ecological xylem anatomy; Perspectives in Plant Ecology, Evolution and Systematics 4, 97-115.

Harris R W, Clark J R and Matheny N P (2004) Arboriculture: Integrated Management of Landscape Trees, Shrubs and Vines; New Jersey US, Prentice Hall.

Hartman J R, Pirone T P and Sall M A (2000) Pirone’s Tree Maintenance; 7th edition; New York, Oxford University Press.

Helliwell R (2004) A discussion of the failure of weak forks; Arboricultural Journal 27, 245-249.

Hodgkinson J M (2000) Mechanical testing of advanced fibre composites; Cambridge, Woodhead Publishing Ltd.

Howatson A M, Lund P G and Todd J D (1991) Engineering tables and data; 2nd Edition; Norwell, USA, Kluwer Academic Publishers.

ITFD (2015) Failure Type Report; California, International Tree Failure Database; accessed 15/06/2015 via http://svinetfc8.fs.fed.us/natfdb/reports.aspx

Jaffe M J (1973) Thigmomorphogenesis: The response of plant growth and development to mechanical stimulation; Planta 114, 143-157.

James N D G (1990) The Arboriculturalist’s Companion: A guide to the care of trees; London, Wiley Blackwell.

Jungnikl K, Goebbels J, Burgert I and Fratzl P (2009) The role of material properties for the mechanical adaptation at branch junctions; Trees: Structure and Function 23, 605-610.

Kane B, Farrell R, Zedaker S M, Loferski J R and Smith D W (2008) Failure mode and prediction of the strength of branch attachments; Journal of Arboriculture and Urban Forestry 34, 308-316.

Kane B (2014) Determining parameters related to the likelihood of failure of red oak (Quercus rubra L.) from winching tests; Trees: Structure and Function 28, 1667- 1677.

Kozlowski T T, Kramer P J and Pallardy S G (1991) The Physiological Ecology of Woody Plants; New York, Academic Press.

Kramer E M (1999) Observation of topological defects in the xylem of Populus deltoides and implications for the vascular cambium; Journal of Theoretical Biology 200, 223-230.

Kramer E M and Borkowski M H (2004) Wood grain patterns at branch junctions: modelling and implications; Trees: Structure and Function 18, 493-500.

Larson P R and Isebrands J G (1978) Functional significance of the nodal constricted zone in Populus deltoides; Canadian Journal of Botany 56, 801-804.

Larson P R and Richards J H (1981) Lateral branch vascularisation: its circularity and its relation to anisophylly; Canadian Journal of Botany 59, 2577-2591.

Larson P R and Fisher D G (1983) Xylary union between elongating lateral branches and the main stem in Populus deltoides; Canadian Journal of Botany 61, 1040- 1051.

Lev-Yadun S and Aloni R (1990) Vascular differentation in branch junctions of trees: circular patterns and functional significance; Trees: Structure and Function 4, 49-54.

Lev-Yadun S (2000) Whirled grain in wood and topological defects; Journal of Theoretical Biology 205, 511-514.

Lichtenegger H, Reiterer A, Stanzl-Tschegg S E and Fratzl P (1999) Variation of cellulose microfibril angles in softwood and hardwoods – a possible strategy of mechanical optimization; Journal of Structural Biology 128, 257-269.

Lilly S and Sydnor T D (1995) Comparison of branch failure during static loading of silver and Norway maples; Journal of Arboriculture 21, 302-305.

Lonsdale D (1999) The Principles of Tree Hazard Assessment and Management; London, TSO.

Lonsdale D (2000) Hazards from Trees: A General Guide; Edinburgh, Forestry Commission.

Matheny N P and Clark J R (1994) A Photographic Guide to the Evaluation of Hazard Trees in Urban Areas; Urbana US, ISA.

Mattheck C and Bethge K (1998) The structural optimization of trees; Naturwissenschaften 85, 1-10.

Mattheck C and Kubler H (1995) Wood – the Internal Optimization of Trees; Berlin, Springer-Verlag.

McCulloch K A and Sperry J S (2005) Patterns in hydraulic architecture and their implications for transport efficiency; Tree Physiology 25, 257-267.

Mencuccini M, Grace J and Fioravanti M (1997) Biomechanical and hydraulic determinants of tree structure in Scots pine: anatomical characteristics; Tree Physiology 17, 105-113.

Neely D (1991) Water transport at stem-branch junctures in woody angiosperms; Journal of Arboriculture 17, 285-290.

Niklas K J (1992) Plant Biomechanics: an Engineering Approach to Plant Form and Function; Chicago US, University of Chicago Press.

Nobel P S (2009) Physicochemical and Environmental Plant Physiology; 4th edition; Oxford, Academic Press.

OED Online (2015) "fork, n."; Oxford University Press; webpage last accessed June 15, 2015 at: http://www.oed.com/view/Entry/73379?rskey=xHPRfF&result=1

Osazuwa-Peters O and Zanne A E (2011) Wood density protocol; webpage last accessed March 29, 2016 at: http://www.publish.csiro.au/prometheuswiki/tiki- pagehistory.php?page=Wood density protocol&preview=11

Panshin A J and De Zeeuw C (1980) Textbook of Wood Technology; New York, McGraw-Hill.

Pfisterer J A (2003) Towards a better understanding of tree failure: Investigations into bending stresses of branch junctions and reiterates of European Filbert (Corylus avellana L.) as a model organism; In: B. Hartmut, K. H. Strauch, and Backhaus G.F. eds. Second International Symposium on Plant Health in Urban Horticulture (Berlin) 394, 272–279.

Philipson W, Ward J M and Butterfield B G (1971) The vascular cambium: its development and activity; London, Chapman & Hall Ltd.

Raimbault P (1991) Quelques observations sur les systèmes racinaires des arbres de parcs et d’alignements: diversité architecturale et convergence dans le développement; In : L’arbre, biologies et développement; Edelin C (ed.), Naturalia Monspeliensia 1991: 85-96.

Reiterer A, Burgert I, Sinn G and Tschegg S (2002) The radial reinforcement of the wood structure and its implication on mechanical and fracture mechanical properties—A comparison between two tree species; Journal of Materials Science 37, 935-940.

Ryder C M and Moore G M (2013) The arboricultural and economic benefits of formative pruning street trees; Journal of Arboriculture and Urban Forestry 39, 17-24.

Schweingruber F H, Börner A and Schulze E-D (2006) Atlas of Woody Plant Stems: evolution, structure and environmental modifications; Berlin Heidelberg, Springer-Verlag .

Schweingruber F H, Börner A and Schulze E-D (2011) Atlas of Stem Anatomy in Herbs, Shrubs and Trees; Berlin Heidelberg, Springer-Verlag.

Sellin A, Rohejärv A and Rahi M (2008) Distribution of vessel size, vessel density and xylem conducting efficiency within a crown of silver birch (Betula pendula); Trees: Structure and Function 22, 205-216.

Shigo A L (1985) How tree branches are attached to trunks; Canadian Journal of Botany 63, 1391-1401.

Shigo A L (1986) A New Tree Biology Dictionary; Durham, New Hampshire US, Shigo & trees, associates.

Shigo A L (1991) Modern Arboriculture; Durham, New Hampshire US, Shigo & trees, associates.

Slater D and Harbinson C (2010) Towards a new model of branch attachment; Arboricultural Journal 33 (2), 95-105.

Smiley E T (2003) Does included bark reduce the strength of co-dominant stems?; Journal of Arboriculture 29, 104-106.

Tyree M T and Alexander J D (1993) Hydraulic conductivity of branch junctions in three temperate tree species; Trees: Structure and Function 7, 156-159.

Tyree M T and Zimmermann M H (2003) Xylem structure and the ascent of sap; 2nd Edition; Berlin New York, Springer-Verlag. van Casteren A, Sellers W, Thorpe S, Coward S, Crompton R and Ennos A R (2012) Why don’t branches snap? The mechanics of bending failure in three temperature angiosperm trees; Trees: Structure and Function 26, 789-797.

Van Gelder H A, Poorter L and Sterck F J (2006) Wood mechanics, allometry, and life- history variation in a tropical rain forest tree community; New Phytologist 171, 367-378.

Zimmermann M H (1978) Hydraulic architecture of some diffuse porous trees; Canadian Journal of Botany 59, 2286-2295.

This page intentionally left blank

Chapter 2

Determining the mechanical properties of bifurcations in hazel (Corylus avellana L.) by testing their component parts

Authors Slater D and Ennos A R

Status of Associated Paper Published in 2013

Journal and Edition Trees: Structure and Function 27 (6), 1515-1524

2.1 Chapter Abstract

The ability to accurately predict the load-bearing capacity of tree bifurcations would improve tree surveying and tree surgery techniques and assist with the biomechanical modelling of a tree’s structure.

In this study, the bending strength of bifurcations of hazel (Corylus avellana L.) was investigated by assessing the mechanical contributions from three component parts of each bifurcation. Intact bifurcations and ones in which either central or peripheral xylem lying under the branch bark ridge at the apex of the bifurcations had been removed were subjected to rupture tests. The bending strength of these bifurcations was compared with that of the arising branches by carrying out a three-point bending test on the smaller arising branches of the intact specimens.

All bifurcations failed in tension, splitting between the arising branches. By removing the centrally-placed xylem, constituting approximately a fifth of the width of the fracture surface, the bifurcations’ bending strength was reduced by around 32%, while removing the outer four fifths reduced the bifurcations’ bending strength by 49%. Intact bifurcations had around 74% of the yield strength of the smaller arising branch.

It is concluded that the tensile strength of the centrally-placed xylem at the apex of a tree bifurcation is a critical strengthening component. This helps to explain the weakness of bifurcations with included bark, which lack this component. This study concludes that tree bifurcations should not by default be considered flaws in a tree’s structure.

2.2 Introduction

Branch junctions in the crown of a tree in which there are two arising branches are formally referred to as bifurcations. Previous studies have identified high diameter ratio bifurcations as a structural weakness within trees (Kane et al., 2008); in particular, those bifurcations where bark has been included within the join between the two branches are considered dangerously weak (Smiley, 2003). The failure of a bifurcation in a tree will cause damage to the tree’s structure, could lessen its longevity and aesthetic value, and may cause injury to people as well as damage to adjacent property (Lonsdale, 1999).

Studies by Pfisterer (2003) and Kane (2007) identified that the angle of a bifurcation in a tree was not a significant factor that determined the load-bearing capacity of that bifurcation. A detailed study by Kane et al. (2008) identified that the mode of failure, the presence of included bark and the morphology of the attachment were all significant factors in bifurcation strength, but that the best predictor for bifurcation strength was the ratio of the diameters of the two arising branches of the bifurcation, with high diameter ratio bifurcations being significantly weaker. Mattheck and Vorberg (1991) have outlined the morphology of well-formed tree bifurcations, where stress notches at the bifurcation apex are minimised by differential growth of the cambium within the inner contour of the bifurcation and thus failure rates are reduced. However, there is currently no satisfactory explanation as to how the wood grain arrangement at tree bifurcations assists in providing adequate strength to the join. The most widely accepted anatomical model for branch attachment (Shigo, 1985) does not provide a satisfactory explanation of how wood grain orientation provides mechanical support in the case of a bifurcation formed in the crown of a tree (Slater and Harbinson, 2010).

Previous observations suggest that tree bifurcations fail in tension down the centre of the bifurcation (Fig. 2.1), though the process is somewhat different depending on the diameter ratio of the two arising branches of the bifurcation (Kane et al., 2008). The authors of this current study categorise the failure of tree bifurcations into two types. The Type I mode of failure (Fig. 2.1a) is more commonly exhibited where there is a

59 large difference in diameter between the two arising branches of the bifurcation (e.g. the diameter of one branch is only half of the other; a diameter ratio of 50%). Xylem at the outer edge of the smaller branch starts to buckle under compressive forces before a crack is initiated at the bifurcation’s apex. Where the two arising branches are of approximately equal diameter (a diameter ratio of 80% or more), it is more common that a Type II failure occurs, where the first visual sign of failure is a crack starting at the apex of the bifurcation. (Fig. 2.1b).

Figure 2.1: Types of tensile failure in tree bifurcations

A: In type I mode of failure, the first stage of failure exhibits yielding of wood under compression at the outer edge of the smaller branch of the bifurcation, prior to the second stage of failure, which is the splitting of the wood at the apex of the bifurcation;

B: Type II mode of failure in tree bifurcations exhibits only one stage of failure, immediately starting with the splitting of the tissues under tension at the bifurcation apex.

When the fracture surfaces of broken bifurcations are examined (Fig. 2.2) it can be seen that there is a narrow central zone of xylem at the centre of the bifurcation which is of higher density and has a more tortuous wood grain pattern than the surrounding xylem

(Lev-Yadun and Aloni, 1990; Pfisterer, 2003). This central zone exhibits a wood grain orientation whereby some wood fibres are axially aligned with the tensile forces that may act to split the bifurcation apart, and this xylem tissue should therefore provide a larger proportion of the strength of the bifurcation, by area, than the peripheral xylem of a bifurcation. If this were the case, it would help to explain why bifurcations with bark inclusions are substantially weaker, as the included bark would prevent this centrally- placed and denser xylem providing a connection between the two conjoined branches of a bifurcation.

Figure 2.2: Observation of grain orientation on the fracture surface (yz plane) of a hazel bifurcation.

The sides of the fracture surface exhibit a standard radial grain orientation; zone1 has some fibres broken along their length and jutting out perpendicular to the fracture surface, representing an axial grain orientation; zone 2 represents a transition zone between these two grain orientations.

As the bifurcation is pulled apart, the two arising branches will rotate about a point at the centre-line below the bifurcation apex (point Xb, Fig. 2.3b), pulling apart the wood at the apex and compressing the wood on the outer edges of the bifurcation. This can easily be seen in the case of Type I failures (Fig. 2.1a), where buckling of the sapwood at the outer edge of the bifurcation can be seen significantly below its apex. The outer edge of the bifurcation where the smaller branch joins the parent stem is far more likely to suffer compressive failure during rupture tests than the wood on the opposite side, where the larger branch joins the parent stem, as higher compressive stresses will be experienced at this edge when the bifurcation is pulled apart.

Figure 2.3: Location of the pivot point in calculating the maximum breaking stresses of bifurcations (xz plane).

For each illustration, A is the larger-diameter branch and B is the smaller-diameter branch. Fpeak is the peak force applied to the fork before failure, with FR being the equal and opposite reaction force. L is the length between the point of application of the force Fpeak and the top of the join between the smaller branch and the bifurcation.

B1 is the diameter of the smaller branch at this joining point. L1 is the suggested additional length required to take into account the lever point within the junction and

Xb is the diameter of the bifurcation across which the moment (Fpeak (L+L1)) acts.

A: The standard measurements taken for estimating the breaking stresses at Y-shaped components (Gere and Timoshenko, 1996), based upon a pivot point situated along the middle of line B1;

B: Author’s suggestion of an alternative location for the pivot point used for the calculation of the maximum bending stress of a bifurcation, based on observations of compressive failure of tissues on the outer edge of smaller branches of bifurcations.

This alternative pivot point is at the midpoint of line Xb, an additional distance L1 into the bifurcation.

These observations suggest that the strength of bifurcations under tension can be considered to be due to component parts just as the root systems of plants contribute to several different components of their anchorage (Ennos, 2000).

In the case of bifurcations, three primary components can be considered; the centrally- placed xylem within the bifurcation, the peripheral xylem perpendicular to the plane of the bifurcation (yz plane, Fig. 1.1), and the peripheral xylem in the plane of the bifurcation adjacent to the attachment of the smaller arising branch (Fig. 2.4).

Figure 2.4: Illustration of the three components that contribute to the mechanical strength of a bifurcation in a tree.

Note that the branch in the foreground is treated as the smaller diameter branch in this perspective drawing, with the grey stem being the larger diameter branch.

In this diagram, three components of a bifurcation in a tree are identified and highlighted in red. Component 1 is the tortuous xylem at the centre of the join that resists tensile failure; component 2 is the remaining xylem under the branch bark ridge that resists tensile failure, and component 3 is the outer edge of the smaller member, which resists compressive failure.

The relative contribution of each of the three principle components can be estimated by comparing the bending strength of intact bifurcations with that of ones in which individual components have been removed. A similar approach was previously successfully used for the more complex system of root anchorage (Coutts, 1983; Ennos, 1991; Crook and Ennos 1993). This modelling assumes that when one of the components is destroyed, the bifurcation behaves in a similar way when it is tested to breaking point, so that the other two components resist the same forces.

The relative bending strength of the bifurcation compared with that of the arising branches can also be estimated by removing branches from the bifurcations after they

64 are tested and subjecting them to three point bending tests. This should give a better understanding of the mechanical properties of tree bifurcations, and determine whether bifurcations are a significant weakness to a tree’s structure.

2.3 Materials and Methods

2.3.1 Sample collection and organisation

On 3 November 2010, 106 bifurcations of hazel (Corylus avellana L.), each approximately 450 mm in length, were cut from trees in a hazel coppice in Prestwich Country Park, Manchester. No more than three bifurcations were harvested from any one tree. Each bifurcation had two arising branches of similar diameter and all the bifurcations were between three to eight years old. Each specimen was cut so that the length of the parent stem retained was at least twice the length of the visible branch bark ridge and each arising branch was at least 220 mm in length. The mean diameter of the parent stem (the average of PS1 and PS2, Fig. 2.6) for the specimens collected was 31.15 mm (± 0.53 SE). Hazel was chosen due to its propensity for producing bifurcations and because the biomechanical behaviour of bifurcations in this species has also been investigated by Pfisterer (2003).

Cut specimens were immediately placed in separate plastic bags to prevent drying and stored in a cold store at 2 ºC until subjected to rupture testing, within ten days of their collection.

In order to determine the relative importance of the three components of the bifurcation structure, the bifurcations were placed in three groups: a control group which were pulled apart intact; a second group with the central 20% of the xylem width removed by drilling out material at the apex of the bifurcation (Fig. 2.5a); and a third group with the outer 80% of the xylem width removed by sawing them perpendicular to the plane of the two arising branches, along the line of the branch bark ridge, to leave 20% of the width of connecting material in the centre of the bifurcation (Fig. 2.5b). To ensure a

65 range of similar bifurcation sizes were used for each treatment, the collected specimens were graded into three sizes based on the diameter of their parent stem (24-29.9 mm; 30-35.1 mm; and 35.2-49.6 mm) and divided equally between the three treatment groups.

To remove the central 20% of the width of the parent stem at the bifurcation apex, specimens were clamped in a vice, and the centre of the apex of the bifurcation was drilled to a minimum depth of 40 mm. The drill size used on each of these modified specimens was determined by selecting a drill approximately 20% of the diameter of the parent stem perpendicular to the plane of the bifurcation (PS2), as shown in table 2.1:

Table 2.1: Criteria for the selection of the drill size used on modified bifurcations The drill size used to drill out the central xylem on the second set of samples was based on the diameter of the parent stem as measured just below the branch bark ridge.

Diameter of parent stem perpendicular to Drill size used to modify the specimen the plane of the bifurcation (mm) (mm)

< 22.5 4

22.5 – 27.49 5

27.5 – 32.49 6

32.5 – 37.49 7

37.5 – 42.49 8

A third group of specimens were modified to remove the outer 80% of the width, by cutting into both edges of the bifurcation perpendicular to the plane of the bifurcation to the depth of 40% of the diameter of the parent stem with a hacksaw. This procedure left only c. 20% of the diameter of the parent stem intact down the centre of the bifurcation attachment.

Figure 2.5: Material removed in the two component tests. The parameter d is the under-bark diameter of the parent stem perpendicular to the plane of the bifurcation at the point where the pith of the parent stem divides into the two piths of the two branches. Areas highlighted by diagonal lines represent the cuts or drill holes made to the modified bifurcations in the yz plane. A: The bifurcation apex has been drilled to remove a width of material down the centre of the bifurcation, using a drill bit approximately 20% of the width d, of the attachment. B: The bifurcation apex has been sawn with a hacksaw from both sides, to leave a residual attachment at the centre of the bifurcation equivalent to 20% of the width d.

2.3.2 Rupture tests

Each bifurcation was then prepared for rupture testing by drilling two 6 mm diameter holes approximately 200 mm from the top of the branch bark ridge of the bifurcation, one hole at the top of each member arising from the bifurcation. The holes were set perpendicular to the plane of the bifurcation. Excess material above the drilled holes was cut away.

The over-bark diameters of the parent stem and the basal diameters of the two arising branches were also measured using digital callipers, both in the plane and perpendicular to the plane of the bifurcation (Fig. 2.6).

Figure 2.6: Diagram illustrating morphological measurements taken for each bifurcation in xz and yz planes.

The larger member was labelled as ‘A’, the smaller branch as ‘B’, for ease of reference.

Lengths B1 and A1 were the shortest diameters across the two branches just above the point of attachment. Measurements of the parent stem diameter were also taken (as illustrated) for analysis of bifurcation morphology.

The distances between the two drill holes and the top of the branch bark ridge of the bifurcation were measured using a metal rule to give the distances a, b and c (Fig. 2.7). Each specimen was then attached between the crosshead and base of an Instron® 4301 universal testing machine, using two purpose-built steel ‘U’ shape brackets. 5 mm bolts held the two arising branches of the bifurcation within the centre of each ‘U’ bracket, using the pre-drilled holes (Fig. 2.7). The crosshead of the Instron® testing machine was then moved upwards at a rate of 35 mm min-1, pulling the bifurcation apart, while the force required was measured at a rate of 10 measurements per second with a 1 kN load cell and an interfaced computer plotted a graph of force against displacement. This rate of testing was used previously by van Casteren et al. (2012) to determine the

68 mechanical performance of greenwood branches of a range of species and similar rates of testing are recommended by the British Standards Institute (BSI, 2004).

Figure 2.7: Rupture testing of hazel bifurcations and additional measurements

The sample is shown attached to the base and cross-head of an Instron® universal testing machine using bespoke U-shaped brackets. The smaller branch B was always attached to the crosshead of the machine; this allowed close observation of the behaviour of the outer edge of the smaller branch, which assisted in categorising the mode of failure of the specimens tested. Distances a, b and c were measured using a metal rule.

The peak force (Fpeak) and displacement at peak force (ext) were both recorded. The mode of failure of each bifurcation was also carefully observed, and those observations recorded. No local deformations were observed around the drill holes on any specimens tested.

After the rupture test, for the modified specimens, the width of the drill hole or width of the material remaining after sawing were measured using digital callipers. Any anomalies on the fracture surfaces were also recorded, such as the presence of included bark or dense knots from small lateral branches.

2.3.3 Calculation of bifurcation breaking stress

The moment, Mpeak, required to break each bifurcation was calculated using equation 2.1.

M peak  Fpeak b Sin  (Eq. 2.1) where b is the length of the line from the drill hole in the smaller branch B of the specimen to the mid-point of the section at B1 and Ө is the angle at which the force is applied relative to the bearing of length b (Fig. 2.3b).

Ө in turn was calculated in degrees using equation 2.2.

(a  ext )2  b2  c2   Cos 1 (Eq. 2.2) 2(a  ext )b where (a + ext) is the distance between the two drilled holes in the members of the bifurcation at the point when peak force was recorded, b is the distance between the drill hole in the smaller branch and the apex of the branch bark ridge and c is the distance between the drill hole in the larger branch and the apex of the branch bark ridge (Fig. 2.7).

To normalise the bending strength of the bifurcations of different sizes, the maximum bending moment was divided by the product of the section modulus of the elliptical cross section of the thinner branch of the bifurcation at its point of attachment to the other branch. The maximum bending stress for each bifurcation, σa, was calculated using the equation 2.3.

32 푀푝푒푎푘 휎푎 = 2 (Eq. 2.3) 휋 퐵1 퐵2 where B1 and B2 are the diameters of the thinner branch at its base (Fig. 2.6) (Gere and Timoshenko, 1996).

2.3.4 Three point bending tests

For the control group of intact bifurcations, a three point bending test was then carried out on the smaller arising branch to compare its strength with that of the bifurcation, and so determine the extent to which the bifurcation was a weakness in the structure of the tree. The thinner branch was placed upon steel supports, 292 mm apart, and subjected to a three-point bending test (Fig. 2.8). A semi-circular plastic probe of 30 mm diameter, attached to a 1 kN load cell in the crosshead of the Instron, was lowered until it touched the branch halfway between the supports. The crosshead was then driven downwards at a rate of 35 mm min-1, bending the branch until it failed, while an interfacing computer recorded a graph of force vs displacement.

This test was used to calculate the yield stress, σb, of the branch using equation 2.4.

8 푃푚푎푥 퐿푠푝푎푛 휎푏 = 2 (Eq. 2.4) 휋 퐷푚푖푑 푊푚푖푑

where Pmax was the maximum load and Lspan was the distance between the supports,

Dmid and Wmid were the diameters of the branch in-line and perpendicular to the load

71 respectively, measured where the plastic probe was in contact with the branch during the test (Gere and Timoshenko, 1996).

The completion of the rupture tests and three-point bending tests allowed a comparison between the maximum bending stresses of the bifurcations tested with the yield stresses of the smaller branches that arose from the bifurcations.

Figure 2.8: Diagram of the three-point testing rig used

This method was used to find the yield strength of the smaller-diameter branches of the hazel bifurcations being tested. The rubber covers to the three-point testing rig assisted in stopping the branches from rolling during testing, as did correctly orientating the branches for the test; however, a small number of specimens had to be rejected for persistently rolling when tested, despite these adjustments.

2.3.5 Sample size

The size of the sample deemed suitable for analysis of their mechanical properties was reduced from the 106 bifurcations originally collected to 94 bifurcations as, after splitting, 12 were found to have significant areas of included bark within their structure.

These losses to the sample size resulted in the following distribution of valid data within the groups: 35 specimens successfully tested in the control group, 31 of which underwent successful three point bending tests of their thinner branches; 30 specimens in the drilled group of modified bifurcations; and 29 specimens in the sawn group of modified bifurcations.

2.3.6 Sampling for basic density testing

To determine alterations in basic density and wood grain orientation, 5 mm diameter wood cores were extracted from 25 other hazel bifurcations collected from the same site in 2011. A SuuntoTM increment borer was used to extract wood cores from the apex of each bifurcation and from the stem just below the bifurcation (Fig. 2.9). The arising wood cores had their volumes estimated by the water displacement method then were oven- dried at 60 ºC for 96 hours so that their wood basic density could be calculated. Subsequently, wood anatomy slides were made from a selection of these cores using a GSL 1 sledge microtome, in order to determine wood grain orientations at these two locations.

Figure 2.9: Locations for extracting 5 mm wood cores from 25 hazel bifurcations for basic density testing and slide production (xz plane view)

The core from the stem was extracted just below the visible end of the branch bark ridge. The core from the bifurcation was taken at the apex of the branch bark ridge, in the plane of the bifurcation.

2.3.7 Statistical analysis

Statistical analysis was undertaken using the software packages Minitab 15 and Excel 2007, using a paired t-test, a one-way ANOVA and a general linear model ANOVA. Normality of the residuals was verified using the Anderson-Darling test.

2.4 Results

2.4.1 Rupture tests of hazel bifurcations

The bifurcations failed in only two ways. Four specimens of the control group suffered a Type I mode of failure (Fig. 2.1a). All the other test specimens exhibited a Type II mode of failure (Fig. 2.1b). No failures occurred in any of the branches in the specimens during the rupture testing of the bifurcations.

A one-way ANOVA of the maximum bending stress to break apart the bifurcations for the three different size classes set for this experiment showed that there were no significant differences (F2 = 0.46; p = 0.637). This shows that the factor of the size of each specimen had been satisfactorily normalised by the calculations used to determine stress at the outer edge of the bifurcations pulled apart.

After the rupture tests had been completed, it was possible to accurately verify the extent of the modifications carried out to the two sample groups where the bifurcations

74 had been drilled or sawn. Drilled samples had 21.45% (± 0.65 SE) of the length d (as shown in Fig. 2.5) removed by the drill bit. Sawn samples had 77.37% (± 0.95 SE) of the length d cut away, leaving 22.63% intact at the centre of the fracture surface.

A General Linear Model ANOVA showed significant differences in σa for the intact, drilled and sawn bifurcations and σb for the branches (Fig. 2.10) (F3 = 200.8; p < 0.001); a post hoc Tukey test with 95% simultaneous confidence intervals identified that all sample group means were significantly different from each other (p < 0.001) and the residuals were found to be normally distributed using an Anderson-Darling test (AD125 = 0.265; p = 0.690). The maximum bending stresses experienced by the drilled and sawn bifurcations were on average only 68.33% and 50.87% that of intact bifurcations, respectively.

Figure 2.10: Boxplot of maximum breaking stress of all sample sets in this study

Boxplot showing differences in bending stress that caused the breaking of bifurcation or the yielding of branches (σa and σb), calculated using equation 2.3 for all bifurcation samples and equation 2.4 for all branch samples. Branches (n = 31); Drilled

75 bifurcations (n = 30); Intact bifurcations (n = 35); Sawn bifurcations (n = 29). Boxplots show the mean bending stress, the upper and lower quartiles and the full range of bending stresses for each group. The different letters above the boxes denote significant differences between groups.

These data can then be used to provide estimates for the contribution of all three components of the bifurcation to its overall strength. Tensile resistance of centrally- placed xylem (component 1) contributed 31.67% and tensile resistance of outer xylem (component 2) contributed 49.13%. By subtracting these two components, this gives a figure for the compressive resistance of the outer edge of the bifurcation (component 3) of 19.2%.

2.4.2 Three point bending tests

Branches were stronger than intact bifurcations (Fig. 2.10), the bifurcations having on average a maximum bending stress of only 73.57% (± 2.27 SE) that of the yield stress of the thinner branches, a difference which a paired t test showed was significant ( t30 = 9.63; p < 0.0001 ). However, there was considerable variation, with one bifurcation having 97.1% of the bending strength of the thinner branch, whereas the weakest bifurcation, by this comparison, had only 47.3% of the strength of its thinner branch.

The Type I mode of failure, only observed in four of the tested specimens, was associated with bifurcations that were on average 88.2% (± 2.5 SE) of the strength of their thinner branch under bending.

2.4.3 Basic density tests and grain orientation observations

In the study of hazel bifurcations, basic density was 18.9% higher (± 0.5% SE) in the 5 mm diameter wood cores extracted from the centre of bifurcations (649 kg/m3 ± 6.1 SE) than from the adjacent stem (546 kg/m3 ± 9.5 SE) and this was found to be a statistically significant difference using a paired t-test (t23 = 11.99; p < 0.0001). All the cores taken from the bifurcations were denser than their paired samples from the stem, with the range being between 6% and 35% denser.

Examination of the wood slides made from these extracted cores showed that grain orientations at the bifurcation apex were complex and twisting, with the grain taking very tortuous routes (Fig. 2.11), when compared with the standard tangential view exhibited from slides made from the cores extracted from the stem.

Figure 2.11: Comparison of the anatomy of ‘junction wood’ and normal stem wood

Sections approximately 20 μm thick, stained with 1% Toluidine blue O

A: Tortuous grain pattern exhibited from wood extracted from the apex of the branch bark ridge of a hazel bifurcation;

B: Tangential view of wood extracted from a hazel stem just below the branch bark ridge.

2.5 Discussion

The components testing worked well; removing the outer or inner area of the bifurcation had no observable effect on the kinematics of failure, in that the modified bifurcations flexed about the same axis of rotation as the control specimens, and there appeared to be no change in how cracks were initiated in the specimens. This helps to justify the pragmatic approach of this experiment, that the three components examined could be treated as additive contributors to the overall strength of each bifurcation. Our modifications to the specimens (drilling or sawing) did reduce the strength of the bifurcations, giving clear indications about the relative importance of the three components. We did not destroy the third component directly, the compressive resistance of the outer edge of the smaller branch just below the bifurcation apex, because destroying this component would change the way in which the bifurcation deformed when stressed.

Firstly, it is clear that the centre of the bifurcation contributes substantially to the bifurcation’s load-bearing capacity. The central fifth of the diameter of the bifurcations’ fracture surface contributed almost a third of the strength to the specimens tested, whereas the outer edges that made up nearly four-fifths of the fracture surface contributed just under half.

We consider that the higher contribution of the central zone relative to its area is due to the altered wood grain orientation in this central zone (Fig.s 2.2 and 2.11a); this arrangement prevents the bifurcation splitting apart better than if the grain was straight and aligned perpendicular to the plane of the bifurcation at its apex, which would only supplying the radial or tangential strength of the xylem. The heightened basic density in this central zone will also be a contributory factor to this component’s relative strength. The changes in grain orientation and basic density are good examples of trade-off in xylem (Tyree et al., 1994; Gartner, 1995; Chave et al., 2009), whereby some efficiency in conductance is foregone for the sake of some gain in mechanical strength within the

79 xylem that forms at the apex of these bifurcations. Further to this study, a more detailed analysis of the wood grain pattern and tissue composition in this location is warranted.

The importance of the central xylem is one tangible reason why, when a tree bifurcation has bark included in its structure, it is found to be weaker; in such bifurcations no connections will be made at the centre of the bifurcation due to the obstruction caused by the bark inclusion. Further anatomical studies of the xylem and its grain orientation in this critical area of tree bifurcations are needed to work out exactly how it supplies the apex of the bifurcation with its high relative strength.

The component testing approach could also be applied to trees with included bark and to larger tree bifurcations, to further investigate the factors that determine bifurcation strength, particularly relative to the strength of their arising branches.

Secondly, observations of the Type I mode of failure in this and other experiments carried out by the authors strongly suggests that the peak moment for the failure of tree bifurcations has previously been underestimated by other researchers. We have observed the pivot point for forces acting to split the bifurcation is further into the attachment; the evidence for this comes from the location of the compressive yielding seen in the Type I mode of failure. A better estimate of bifurcation strength will be obtained if a pivot point further into the point of attachment is selected to calculate the peak moment (adding length L1 to length b as shown in Fig. 2.3b, when calculating the peak moment (Equation 2.1)). This observation implies that the calculated bifurcation bending strength is underestimated here by approximately 5-10%.

The Type I mode of failure occurred where the maximum bending strength of these bifurcations was closer to the yield strength of the thinner branch, compared to the sample mean. A lower branch diameter ratio (e.g. 75% or below) is associated with this mode of failure, and this may confer strength to the bifurcation through the higher incorporation of the wood of the thinner branch into the tissues of the thicker branch,

80 forming a ‘knot’. This agrees with the findings of Kane et al. (2008) on the influence of diameter ratio on the strength of tree bifurcations.

Thirdly, if all the material at the branch bark ridge were cut away, to separate the two parts, then only limited bending of each branch is needed for the crack to propagate further down the stem. This helps to explain why, once a tree bifurcation suffers an initial crack at its apex in a storm, the structure is very greatly weakened and should be considered dangerous.

Finally, our results of the bending tests on arising branches suggest that tree bifurcations can be almost as strong as their arising branches, despite being considered as structural flaws by several authors (Matheny and Clark, 1994, Gilman, 2003, Smiley, 2003, Kane et al., 2008). Our branch arm lengths were approximately 200 mm in length; using a much shorter or longer length of branch would result in different kinematics of failure, and if the centre of gravity of the full length of arising branches were used as the point where the load was applied, one would anticipate more type I failures and branch failures occurring in such tests. Furthermore, the occasions when two arising branches of a tree bifurcation undergo substantial bending in opposite directions from each other during windy weather may be infrequent, given the recognised mechanical behaviour of branches on trees to dampen potentially damaging oscillations (Spatz et al., 2007). This suggests that well-formed bifurcations are not more likely to break than any other part of a tree.

However, there was considerable variability in the strength of the bifurcations, when compared to the bending strength of the arising thinner branch. Some of this variation might be due to environmental effects. The bifurcation will develop its strength primarily in response to the bending stresses it experiences; if it rarely experiences significant tensile stresses at its apex, the bifurcation may not develop strong xylem connections in this region (Jaffe, 1973). That does not mean that the lack of reinforcement is a structural flaw; rather it represents a best fit to the environment in which it is growing (Farnsworth and Niklas, 1995). Further studies could contrast the

81 relative bending strength of more mature tree bifurcations from both wind-exposed and sheltered locations with their arising branches to assess the extent of their environmental adaptation to bending stresses.

Smiley (2003) found that included bark in red maple (Acer rubrum L.) bifurcations lowered their bending strength by approximately 20%. Given that a small drill hole at the centre of these hazel bifurcations reduced their bending strength by 31.7%, it is probable that the presence of naturally-occurring flaws in bifurcations results in trees remodelling them to address this structural weakness as they grow and experience more wind-loading (Mattheck and Vorberg, 1991). To test this idea, a further experiment could be carried out, modifying living tree bifurcations by drilling out the central section of each one and comparing their subsequent growth with that of intact bifurcations. This would help determine how bifurcations adapt to induced flaws and where they put on additional wood growth to aid that adaptation.

2.6 References

BSI (2004) British Standard EN 789:2004 Timber structures. Test methods. Determination of mechanical properties of wood based panels; British Standards Institute, London.

Chave J, Coomes D, Jansen S, Lewis S L, Swenson N G and Zanne A E (2009) Towards a worldwide wood economics spectrum; Ecology Letters 12, 351-366.

Coutts M P (1983) Root architecture and tree stability; Plant and Soil 71, 171-188.

Crook M J and Ennos A R (1993) The mechanics of root lodging in winter wheat, Triticum aestivum L.; Journal of Experimental Botany 44, 1219-1224.

Ennos A R (1991) The mechanics of anchorage in wheat, Triticum aestivum L. Journal of Experimental Botany 42, 1607-1613.

Ennos A R (2000) The mechanics of root anchorage; Advances in Botanical Research 33, 133-157.

Farnsworth K D and Niklas K J (1995) Theories of optimization, form and function in branching architecture in plants; Functional Ecology 9, 355-363.

Gartner B L (1995) Patterns of xylem variation within a tree and their hydraulic and mechanical consequences. In: Gartner, B L (ed.) Plant stems; physiological and functional morphology; New York, Academic Press.

Gere J M and Timoshenko S P (1996) The Mechanics of Materials; 4th Edition; Boston, PWS Publishing Co.

Gilman E F (2003) Branch to stem diameter affects strength of attachment. Journal of Arboriculture 29, 291-294.

Jaffe M J (1973) Thigmomorphogenesis: The response of plant growth and development to mechanical stimulation; Planta 114, 143-157.

Kane B (2007) Branch strength of Bradford pear (Pyrus calleryana var. 'Bradford'); Arboriculture and Urban Forestry 33, 283-291.

Kane B, Farrell R, Zedaker S M, Loferski J R & Smith D W (2008) Failure mode and prediction of the strength of branch attachments; Arboriculture & Urban Forestry 34, 308-316.

Lev-Yadun S & Aloni R (1990) Vascular differentiation in branch junctions of trees: circular patterns and functional significance Trees: Structure and Function 4, 49-54.

Lonsdale D (1999) Principles of Tree Hazard Assessment and Management; London, DETR.

Matheny N P & Clark J R (1994) A Photographic Guide to the Evaluation of Hazard Trees in Urban Areas; Urbana US, International Society of Arboriculture.

Mattheck C & Vorberg U (1991) The biomechanics of tree bifurcation design. Botanica Acta 104, 399-404.

Shigo A L (1985) How tree branches are attached to trunks. Canadian Journal of Botany 63, 1391-1401.

Slater D & Harbinson C (2010) Towards a new model of branch attachment; Arboricultural Journal 33, 95-105.

Smiley E T (2003) Does included bark reduce the strength of co-dominant stems? Journal of Arboriculture 29, 104-106.

Spatz H-C, Brüchert F & Pfisterer J A (2007) Multiple resonance dampening or how do trees escape dangerously large oscillations? American Journal of Botany 94, 1603-1611.

Tyree M T, Davis S D & Cochard H (1994) Biophysical perspectives of xylem evolution: Is there a tradeoff of hydraulic efficiency for vulnerability to dysfunction? IAWA Journal 15, 335-360.

van Casteren A, Sellers W, Thorpe S, Coward S, Crompton R and Ennos A R (2012) Why don’t branches snap? The mechanics of bending failure in three temperature angiosperm trees; Trees: Structure and Function 26, 789-797.

Chapter 3

The anatomy and grain pattern in bifurcations of hazel (Corylus avellana L.) and other tree species

Authors Slater D, Bradley R S, Withers P J and Ennos A R

Status of Associated Paper Published in 2014

Journal and Edition Trees: Structure and Function 28 (5), 1437-1448

3.1 Chapter Abstract

Wood grain arrangements at the bifurcations and other junctions within a tree must be arranged to mechanically join together the two or more branches, yet not adversely restrict sap-flow. The grain orientation at junctions therefore represents a trade-off in xylem performance between the functions of efficient sap conductance and the provision of adequate load-bearing capacity.

Initial observations of wood grain orientation were made on the surfaces of several dozen debarked and fractured bifurcations of a wide range of tree species, both by eye and using a scanning electron microscope. Subsequently, small volumes of wood were sampled from two locations within the junctions of hazel, at the junction apex and on the outer section of the junction. Wood was imaged in 3D using high resolution X-ray tomography, and the scanned volumes were analysed for their wood grain patterns.

It was found that the wood at the apices of hazel bifurcations contained only 37% of the number of vessels contained in wood within the adjacent stem. The vessel elements formed at the junctions were only 32.5% the length of those in the stem, had a mean diameter only 50.5% of the stem vessels and consequently only 26.3% of their lumen volume. The passage of the vessels through the junction wood deviated from a straight line (Euclidean) distance by 14 times more than the stem wood vessels did. The interweaving of vessels in the junction wood was over 22 times greater than in the stem wood. A survey of rays showed them to be 58% more abundant in junction wood but only 62% of the height of rays in the stem wood.

These results suggest that where two tree branches of similar diameter join to form a bifurcation, an interlocking wood grain pattern is formed at its apex, which provides it with a higher tensile strength. Breaking of a normally-formed hazel bifurcation requires wood fibres to be stretched axially and broken across, which requires greater stress than if the wood at the apex was arranged radially or tangentially.

3.2 Introduction

The anatomy of xylem in the stems of woody angiosperms has been the subject of much study (Jane, 1962; Carlquist, 2001); however, the anatomy of the xylem tissues that are formed at a junction in a tree has not received as much scientific attention – no illustration of wood anatomy of a junction at the tissue or cellular level is provided, for instance, in authoritative publications such as Schweingruber et al. (2006). As a consequence, knowledge about the anatomy of junctions in trees is sparse; how do they function as both mechanical components and conduits for sap?

If the wood grain arrangement were such that the grain transits across the apex of the junction, thus providing a great deal of strength to it (Fig. 3.1a), this would be breaking the biological rule of ‘from source to sink’ (Kramer and Kozlowski, 1979), in that some grain would be orientated so as to carry sap from one set of foliage to another, so from ‘sink to sink’. This wood grain pattern would supply the necessary mechanical strength to hold the junction together because many wood fibres or tracheids would have to be stretched axially and broken across or extracted along their length to split the junction apart; however, this cannot be the wood grain pattern of a bifurcation in a tree, as it foregoes entirely the conductance role of the xylem at the apex of the junction. Although in angiosperms there is substantial division in roles between the vessels that efficiently carry sap and the wood fibres that supply mechanical support, for the most part these different cells are still orientated in the same direction in any given location within the xylem.

If the wood grain arrangement was optimised for hydraulic conductivity (Fig. 3.1b), then the load-bearing capacity of the junction would primarily come from the tangential strength of the wood at the join. Wood, as an anisotropic material, has a relatively low tangential strength compared with axial strength; the tangential tensile strength of the wood of beech trees (Fagus sylvatica L.), for example, is typically around 9 MPa, whereas its axial tensile strength is typically 50 MPa (Dresch and Dinwoodie, 1996); the

87 arrangement shown in figure 3.1b would result in a very weak junction that would readily snap apart under moderate loads.

Figure 3.1: Implausible wood grain arrangements at the bifurcations of woody plants (xz plane views)

A: The “engineer’s solution” – the wood grain at the apex is continuous between the two arising branches;

B: The “hydrologist’s solution” – the wood grain is continuous from the parent stem into both branches without any mechanical reinforcement of the bifurcation.

The most cited model for the anatomy of wood grain at junctions in trees is that of overlapping collars which grow sequentially during an annual cycle of growth as described by Shigo in his 1985 paper entitled ‘How tree branches are attached to trunks’ (Fig. 1.6, Chapter 1). In this model, new xylem layers are initially produced along the length of the smaller branch and then subsequently in the same growing season an overlapping layer is produced by the larger branch or stem to which the smaller branch is joined; Shigo states that it is this seasonal overlapping of the two wood layers that forms

88 the swollen join at the point of attachment and also provides the requisite mechanical strength to the junction.

Shigo’s model requires different timings for the initiation and cessation of cambial growth in different parts of the junction, something which it is easy to envisage in seasonal climates, but less so where tree growth is continuous, such as in the tropics (Turner, 2001). Similarly, as Slater and Harbinson (2010) pointed out, it is hard to see how this overlapping of tissues can occur when the two branches are approximately equal in diameter. A study examining the wood grain pattern in bifurcations of temperate trees could therefore shed light on the general applicability of Shigo’s model.

In their study of the mechanics of the bifurcations of hazel trees, Slater and Ennos (2013) found that centrally-placed xylem at the apex of the bifurcations was a critical component in providing each junction with its load-bearing capacity; in removing 20% of the width of the xylem at the apex of hazel junction with a drill, they showed that this component was providing approximately 32% of the mechanical support to the junction and the remaining 80% to either side of this central zone provided only 49% of the mechanical support. Analysis of the pattern of wood grain at the apex of the branch bark ridge may also therefore provide an explanation as to why this central apical zone in the xylem of hazel and other woody species is particularly important in conferring strength to a bifurcation.

Although the axially-arranged fibrous tissues provide the majority of the support needed for a tree’s stem or branches, an important subsidiary role is played by the ray tissues (Burgert et al., 1999; Ennos and van Casteren, 2010); they act to reinforce the shoot radially, preventing the branches from splitting when they are bent, except along their centre-point (van Casteren et al., 2012), which results in greenstick fracture. The form, abundance and orientation of ray tissues at the junctions of trees has not yet been characterised in published academic work, however, and it is likely that their morphology will have some implications for the strength of such junctions.

3.3 Materials and Methods

3.3.1 Superficial Examination

Junctions from 20 woody plant species were examined by carefully peeling back their bark and snapping selected small bifurcations by hand. Visual observations were made of the wood grain pattern around the junction and upon the two fracture surfaces of each snapped specimen. No junction where a branch was epicormic in origin was examined. The surfaces of the fractures of 34 bifurcations of a range of species were then scanned with a FEI XL30 ESEM-FEG electron microscope, in order to be able to see individual wood cells and their orientation. A list of the species examined and the types of examination used are given in table 3.1.

Table 3.1: Species and specimen numbers for bifurcations that were subject to visual observations as part of this study (excluding CT scanning)

Details of specimens Species Plant family De-barking Splitting of ESEM examined and visual junction and imaging observation visual of of wood grain examination fracture patterns of the surface surfaces around of fractures junction by by eye and/or eye and/or under x 10 under x 10 magnification magnification

Acer Sapindaceae 20 8 - pseudoplatanus Betula pendula Betulaceae 20 5 2 Chamaecyparis Cupressaceae 10 - - lawsoniana Corylus Betulaceae 50 50 6 avellana Crataegus Rosaceae 20 10 2 monogyna Eucalyptus Myrtaceae 12 - - gunnii Fagus Fagaceae 10 10 4 sylvatica Fraxinus Oleaceae 60 20 - excelsior Larix decidua Pinaceae 15 10 5 Pinus sylvestris Pinaceae 10 10 - Platanus x Platanaceae 40 40 - acerifolia Prunus avium Rosaceae 15 12 - Quercus robur Fagaceae 32 16 5 Salix viminalis Salicaceae 50 20 - Sorbus Rosaceae 16 - 2 aucuparia Sorbus Rosaceae 6 - 2 intermedia Taxus baccata Taxaceae 15 15 2 Tilia cordata Malvaceae 18 6 - Ulmus glabra Ulmaceae 30 30 2 Viburnum Adoxaceae 10 - - opulus

3.3.2 Internal Anatomical Investigations

Further to these initial visual observations, common hazel (Corylus avellana L.) was chosen as a case study species because it produces bifurcations most readily. Sample material can also be gathered easily from a sustainable resource due to the regular coppicing cycles utilised by forestry contractors to manage stands of hazel trees. It was also the species used by Slater and Ennos (2013) to examine the mechanical performance of tree bifurcations and the relative importance of their different mechanical components.

Six bifurcations of hazel with parent stem diameters between 30.5 mm to 48.8 mm were cut from six separate hazel stools growing in Prestwich, Manchester, England. For these specimens, the diameter ratios between the two branches ranged from 53% to 98%, in order to examine high diameter ratio bifurcations rather than those that might be characterised as branch-to-stem attachments. Two cubes of wood with sides approximately 1.5 mm x 1.5 mm x 1.5 mm were excised from each bifurcation, one cube cut from the side of the stem approximately 25 mm below the bifurcation and the other cut from the apex of the bifurcation (Figure 3.2a). These locations were chosen to contrast the wood grain pattern in the tissues at the edge of the bifurcation where one might anticipate seasonally overlapping layers of xylem (Shigo, 1985) with the patterns at the bifurcations’ apices. These samples were then placed in thin plastic tubes, with six cubes mounted vertically one above the other in each tube, in order to facilitate faster CT scanning (Figure 3.2b).

CT Scanning

The Xradia microXCT-400 X-ray tomography system (Xradia, Pleasanton, California, USA) at the Manchester X-ray Imaging Facility was used to produce one region of interest scan within each of the twelve wood cubes. The voltage was set at 70 kV, the current at 10 W. Scanning time was approximately 12 hours per block, with 2,001 images being taken as the specimen was rotated over 182°, with an exposure time of 17 seconds per radiograph. In-line propagation-based X-ray phase contrast was used to improve contrast of the internal boundaries within the specimens (Bradley, McNeil and

Withers, 2010). This contrast is generated by the interference of X-rays that have passed either side of a boundary, resulting in the formation of interference fringes at the detector. Generally, one bright fringe and one dark fringe are formed either side of the boundary, with the intensity of the detected fringes being controlled through the positions of the X-ray source and detector relative to the sample. It was found that a sufficient amount of phase contrast was obtained by setting the X-ray source to 70 mm from the sample and the detector to 30 mm from the sample, with the x 20 objective lens selected. The arising reconstructed volumes were 947 x 947 x 947 voxels in dimension, with each voxel being approximately 0.97 µm in width, height and depth.

Figure 3.2: Sample preparation for CT scanning

A: Location of the excision (xz plane) of cubes of wood from hazel bifurcations for CT scanning; B: Bespoke plastic tube used to hold six excised cubes for CT scanning.

Anatomical Analysis

Post-scan analysis was carried out by segmentation of randomly selected vessels within each scanned volume. We consider that the paths that vessels (each consisting of a series of interconnecting vessel members) took through the cropped volumes should act as a good proxy for wood grain orientation in their immediate vicinity. Individual vessel members within the specimens excised from stems were segmented with the aid of the 3D region-growing tool in Avizo 7.1 (VSG, Burlington, MA, USA). However, an

93 alternative strategy was required for the specimens cut from the bifurcations, as the vessel members were highly interconnected and tortuous in nature.

The selected vessels were tracked though the volume slice-by-slice using sequential 2D region-growing from randomly selected points (seed points) within the vessel volumes

. The seed point for region growing on each slice was generated automatically for each vessel by calculating the centre of mass of the segmented vessel on the previous slice. Seed points on the first slice were placed automatically by applying a threshold to select all regions having a grey-level value below that of the surrounding vessel walls. The local thicknesses of these regions were calculated and seed points placed at local maxima above a given threshold. The local thickness is a means of measuring the local diameter of each vessel and is calculated for a given pixel by determining the largest circle that fits within the region and also contains that pixel (Hildebrand and Rüesgsegger, 1996). The threshold was set such that smaller cell types (e.g. fibres and ray parenchyma) were ignored.

Small side connections to the segmented vessels on each slice were removed immediately after region-growing by a novel object-separation technique. The standard watershed separation approach was found not to work well due to the occurrence of small projections out of the vessel walls into adjacent vessel lumen (Meyer, 1994). Instead, the novel technique involved calculating the local thickness for the segmented region and identifying boundaries areas (i.e. connections) where the local thickness reduced to below a given threshold. Boundary areas which form connections between two or more larger regions were then removed. By these means, the larger regions were separated from each other and the region which contained the seed point was identified as the selected vessel lumen. Manual adjustment and surface smoothing was then applied in Avizo to remove larger side connections and produce the final segmentations. In particular, any voxels that represented elements of the scalariform plates formed at vessel member joins in the hazel’s xylem were removed manually to facilitate analysis.

To reduce the data processing time, volumes were cropped to be approximately 0.485 mm in height (z-axis) and 2D region growing was terminated when 8,000 pixels had been

94 selected for each seed point. Observations were then made of these segmented vessels using Avizo 7.1 visual analysis software. Overall, 36 vessels were segmented within the stem volumes and 42 vessels segmented within the volumes excised from the bifurcations. The volumes of the vessels were measured using the material statistics option within Avizo 7.1.

Parametric data on the vessels was collected by the use of a novel algorithm encoded within MATLAB® software. This coding measured the vessels’ dimensions by producing a chain of spheres spanning the vessel. The chain was produced from a given starting point by determining sequentially the largest diameter sphere contained within the vessel volume which also adjoins the previous sphere in the chain. The chain of spheres thereby forms a sampling of the vessel along its centre line on a scale related to the local vessel diameter. Therefore, the path defined by the centre of the spheres is less sensitive to small scale surface features (and hence possibly noise in the tomography data) than that produced by the standard medial axis transform (Katz and Pizer, 2003). Consequently, measurements of vessel length and tortuosity are much less sensitive to these small scale features; unrealistically high values of these parameters were calculated when using the standard medial axis approach. Reiterations of this process from randomly chosen start points along the vessel centre line (medial axis) provided an average length for each vessel and an average diameter. The length of the vessel was calculated by generating a spline through the centres of the spheres and the points at which the sphere surfaces touched. The tortuosity (τ) of the vessel was calculated from the path length (L) as follows:

L   (Eq. 3.1) C where C is the shortest (Euclidean) distance between the start and end of the path taken.

A means of measurement was also developed to characterise the relative paths of the vessels which we term the ‘inter-vessel tortuosity’ (Fig. 3.3): this parameter is calculated for a pair of vessels by determining the difference in positions between the two vessels’ paths (generated by spline interpolation of the chain of spheres) on planes perpendicular

95 to an axis of the volumetric data along which one of the vessels (i.e. the reference vessel) is primarily aligned. The inter-vessel tortuosity is then calculated by dividing the length of this ‘difference path’ with the corresponding path length along the primary axis. In this way, the ‘difference path’ length is compared to the projected length of this path along the primary axis (Fig. 3.3). Using the projected length along the primary axis ensures that the measure is applicable to vessels which double back on themselves (Fig. 3.6f) and hence have centre-line coordinates which are multi-valued with respect to one or more axes.

This inter-vessel tortuosity therefore captures the extent to which the vessels varied their proximity to each other as they transited each volume analysed, to quantify any ‘interweaving’ that may be occurring (Zimmerman and Brown, 1971). An average inter- vessel tortuosity measure was achieved by taking values of inter-vessel tortuosity for all possible pairings within each scanned volume, and then averaging the six resultant values for each sample type. The higher level of vessel termination and short or negligible length of overlap between vessels in the scanned volumes from bifurcations meant that only vessels that travelled at least 20 μm alongside each other down a neutral axis were utilised, and only the minimum tortuosity values were used, to avoid over-estimation of this parameter.

Figure 3.3: Determination of the inter-vessel tortuosity of the vessels segmented out from the scanned hazel volumes

A: Two vessels selected to compare their inter-vessel tortuosity;

B: The line created by subtracting the positions of one vessel from the other along planes (e.g. x-y planes as shown) perpendicular to the primary axis (e.g. the axis z as shown). The tortuosity of this constructed line in relation to its projected length along the primary (z) axis provides our measure of inter-vessel tortuosity.

To determine the abundance of rays in the stem and junction tissues, a tangential view was taken through the centre of all twelve scanned volumes and a count of the number of rays that bisected the 919 μm width of the scan was carried out. Average ray heights were determined by using ImageJ imaging software to measure five randomly selected multiseriate rays from each of these tangential views. Visual observations were also made on the morphology of the rays present in both tissues.

3.4 Results

3.4.1 Superficial Examination

Observations by eye, under low magnification and using the electron microscope found that the wood grain appeared more tortuous at the apex of all the junctions examined than in other adjacent areas (Fig. 3.4b). This more tortuous zone lies between the point where the one pith of the parent stem divides into the two piths of the arising branches and the apex of the junction but varies in width and intensity considerably between samples. In agreement with previous observations of xylem in this zone (Shigo, 1985), the wood grain is orientated at close to 90 degrees from the axial orientation of the parent stem. From observations of the surfaces of fractures, wood grain orientation in this zone was such that some fibres or tracheids had been stretched along their axes and were either extracted from one side of the bifurcation or were broken across their length in order to break these bifurcations apart (Fig.s 3.4a, 3.4c, 3.4d and 3.4e). Exceptions were found where included bark lay within some bifurcations tested; these junctions with included bark all had a small internal angle between the two branches. In these cases, analysis of the

97 fracture surfaces showed that no significant area of wood fibres or tracheids were broken across their axis in order to split these bark-included bifurcations apart (Fig. 3.4f).

Figure 3.4: Visual observations of the wood grain orientation at de-barked junctions and surfaces of fractures of split bifurcations

A: A yz plane view of a bifurcation from a common oak (Quercus robur L.) pulled apart to reveal a centrally-placed ‘spur’ of xylem that clearly bridged between the two members (white arrow);

B: A de-barked bifurcation of common ash (Fraxinus excelsior L.) with a zone of tortuous grain evident at its apex (white arrow);

C: ESEM image of the fracture surface (yz plane) of a broken bifurcation of common larch (Larix decidua Mill.) exhibiting a spine of tracheids that have been broken across their length to break apart the bifurcation (white arrow);

D: ESEM image of the central region of a fracture surface (yz plane) in common beech (Fagus sylvatica L.) showing fibres and vessels broken across their length to separate the bifurcation (white arrow);

E: ESEM image of the central region of a fracture surface (yz plane) of a bifurcation of common hazel (Corylus avellana L.) that was pulled apart, showing a vessel and fibres that were broken across their length when the bifurcation was split (white arrow);

F: Digital scan image of the fracture surface (yz plane) of a hazel bifurcation exhibiting included bark (white arrow), set on a grid of 1 cm square card.

3.4.2 Internal Anatomy

In the CT scanned volumes excised from six hazel bifurcations, it was frequently the case that the vessels terminated in the samples taken (only 17 of the 42 vessels analysed transited these volumes, two of which deviated to exit along a different plane). In contrast, the vessels were relatively straight and continuous in the samples taken from the sides of the stems just below the bifurcation, where all 36 vessels fully transited the scanned volumes of xylem and exited along the same plane as they entered. The orientation of vessels from the stem varied in angle by less than 1 degree from the axis of the parent stem: in contrast, the vessels analysed from the apex of the bifurcations were found to be orientated between 73.7 and 88.5 degrees from the parent stem’s axis. One vessel selected within a sample excised from a bifurcation exhibited a circular pattern with an infinite tortuosity; what has been described by researchers as ‘whirled grain’ (Lev-Yadun and Aloni, 1990). The parameters for the tortuosity of this particular circular vessel were excluded from the dataset, as this would make statistical analysis of this data invalid.

Through the visualisation of the CT scanned volumes, the segmentation of randomly- selected vessels in each volume and the use of our algorithm written in MATLAB®, we were able to quantify the following parameters and contrast them between the samples excised from the side of the stem and those cut from the apex of each hazel bifurcation: the average number of vessels found in the scanned volumes; the average percentage of the scanned volumes that was interpreted as cell wall; the number of selected vessels whereby the series of vessel members terminated within the scanned volume; the average length of vessels in the six scanned volumes, counting only those vessels that did not terminate and did not alter their axial plane whilst transiting the scanned volume; the average number of joins between vessel members along each selected vessel that did not terminate in the scanned volume; the average diameter of the lumen of the vessels; the average volume of the lumen of the selected vessels; the average tortuosity of the selected vessels; and the average inter-vessel tortuosity between vessels in the same scanned volume (Table 3.2).

100

Table 3.2: Comparisons in the anatomy of xylem from the stems and junction apices of hazel bifurcations The cell wall percentage and the number, average length, average diameter and morphological characteristics of vessels in the stem and junction wood of hazel, determined by observation of the scanned volumes using Avizo 7.1 and the bespoke MATLAB code that generated the chains of spheres through the segmented vessels. Stem Junction Statistical test and significance Anatomical parameter Mean percentage of 54.01% 69.19% Paired T Test scanned volume t = 10.052, df = 77, interpreted as cell wall P < 0.001 Number of vessels 36 42 N/A analysed in the six scanned volumes Mean number of vessels 80.5 ± 6.21 SE 29.8 ± 4.51 SE Paired T Test in the six scanned t = 9.401, df = 77, volumes P < 0.001 Number of vessels that None 25 Χ2 test for terminated within the association Χ2 = 32.505, df = 77, scanned volumes P < 0.001 Mean length of vessels 478.5 μm ± 3.33 SE 564.9 μm ± 12.68 SE Two sample T Test that fully transited the (n = 15) t = 5.9155, df = 50, volume P < 0.0001 Mean number of vessel 1.50 ± 0.08 SE 5.44 ± 0.71 SE Two sample T Test element joins visible (n = 15) t = 7.6053, df = 50, along the vessels that P < 0.0001 fully transited the volume Mean diameter of the 19.68 μm ± 0.44 SE 9.93 μm ± 0.54 SE Two sample T Test lumens of vessels t = 13.354, df = 77, P < 0.0001 Mean volume of the 9.36 x 10-13 m3 ± 0.34 2.46 x 10-13 m3 ± Two sample T Test lumens of vessels SE 0.37 SE t = 13.418, df = 77, P < 0.0001 Mean tortuosity of 1.013 ± 0.002 SE 1.190 ± 0.029 SE Two sample T Test vessels (n = 41: one vessel t = 5.2827, df = 76, was circular) P < 0.0001 Mean inter-vessel 1.022 ± 0.005 SE 1.486 ± 0.112 SE Paired T Test tortuosity, averaged (n = 6) (n = 6) t = 3.2912, df = 11, over the six scanned P < 0.05 volumes

Cell wall volume was 28.1% higher in the wood excised from the junctions than in stem wood. Vessels were scarcer in the wood from the junctions, having only 37% of the quantity of vessels found in the scanned stem wood samples. The termination of vessels

101 was very frequent in the junction wood: 59.5% of the vessels analysed in the six examined volumes of junction wood terminated within the volume. In contrast, there were no terminations for the 36 vessels analysed within the six volumes of stem wood.

As noted above, the scanned volumes were truncated to be only 0.485 mm in height for ease of analysis. The vessels that crossed stem wood volumes were axially orientated with very little deviation: none of them exhibited a bearing of more than 0.85 of a degree from the axial orientation of the stem. As a consequence, their mean length as measured by the chain of spheres was very close to the scanned volume’s height, at 478.5 μm. In contrast, the orientation of the vessels in the wood from the hazel bifurcations was between 73.7 and 88.5 degrees from the axis of the stem, necessitating the rotation of the scanned volumes by 90 degrees in order to carry out visual comparisons between these apical tissues and those of the hazel stem. The vessel lengths were 16% greater than the scanned volume height, showing that the mean vessel route in junction tissue was much more tortuous than in the stem tissue.

The visualisation of the segmented vessels allowed for a count of vessel joins. In the hazel stems, the vessels typically exhibited 1.5 joins along their length, the mean vessel element length being 319 μm. In the junction wood, the mean number of joins seen was 363% higher, resulting in mean vessel element lengths of only 103.8 μm, only around a third the length of the vessel elements in the stem wood.

The mean diameters of vessels analysed in the junction wood was only 50.5% of the mean diameter of the vessels in the stem wood. The combination of shorter vessel element lengths and smaller diameter vessels resulted in the lumen volumes in the junction wood being only 26.3% of those found in the stem wood.

Measures of the tortuosity of the vessels in both tissues showed statistically significant differences. Excluding the one circular vessel, the vessels in the junction wood deviated from a straight line by 14.62 times more than the vessels in the stem. A measure of the

102 inter-vessel tortuosity showed that the vessels in the junction wood varied much more in their orientation towards and away from each other, when contrasted with the vessels in the stem wood, the mean deviation of vessels from each other’s paths in the junction wood being 22.09 times higher than in the stem wood.

Example visualisations of the CT scanning output, the segmented vessels examined and the analysis using chains of spheres are provided in Figure 3.5. It can be seen from Figures 3.5a and 3.5b that the transverse sections of stem wood and junction wood contrast because of the substantial difference in orientation of the wood grain, the tissues from the junction having been excised near the junction apex, within the saddle-shape of the join between the two branches of the bifurcation. Figures 3.5c and 3.5d visually contrast typical segmented stem wood and junction wood vessels, where the different orientation at nearly 90 degrees of the vessels in the junction wood can be seen, as well as their smaller diameter and higher tortuosity. Figures 3.5e and 3.5f contrast the shape of the vessels in stem and junction, with the shape of the vessel analysed from the apex of the bifurcation (Fig. 3.5f) having a noticeable zig-zag pattern, which contributes substantially to its greater tortuosity.

103

Figure 3.5: Visualisation of CT scan output:

104

A & B: Transverse cross sectional view of the region of interest X-ray tomography scan of the stem (A) and the centrally-placed xylem at the bifurcation (B) of hazel (xy plane);

C & D: 3D visualisation of segmented vessels in Avizo 7.1 for those in the stem (C) and at the apex of the bifurcation (D);

E & F: Illustration of the chain of spheres generated by our MATLAB® algorithm used to measure the parameters of the vessels segmented in the stem (E) and at the apex of the bifurcation (F), the adjacent scale is in microns.

Note that the vessel illustrated in Fig. 3.5f has been rotated 90º, to facilitate a visual comparison with Fig. 3.5e.

The data from the sampling of ray abundance and height are given in table 3.3. Rays were 58% more abundant in the scanned junction tissues but the multiseriate rays sampled there were only 62% as tall as those measured in the stem tissues. The rays that had formed within the tissue of these hazel bifurcations were noticeably distorted, often forming more elliptical or c-shaped profiles in tangential view when compared with the lens-shaped rays found in normal stem wood (Fig. 3.6). In addition, the rays in the junction tissue followed a much more tortuous route through the material, although the ray cells were too small (due to the resolution of our scans) for us to quantify their degree of tortuosity using the technique we had developed to analyse the vessels.

105

Table 3.3: The abundance and average height of rays in the two tissue types scanned Analysis was based on sampling the middle of each volume in tangential view. Ray abundance was assessed by determining the number of rays that were bisected when traversing the tangential view perpendicular to the wood grain axis. Heights of rays were measured in tangential view using ImageJ and data are based on measuring five randomly-selected multiseriate rays from each scanned volume. Stem Junction Statistical test RAYS and significance Anatomical parameter Abundance of rays 13.42 ± 0.73 SE 21.22 ± 1.34 SE Two sample T Test per millimetre in t = 5.12, df = 7, P = 0.001 tangential view Average height of 302.26 μm ± 12.16 SE 187.26 μm ± 15.81 SE Two sample T Test multiseriate rays in t = 5.77, df = 54, P < 0.001 tangential view

Figure 3.6: Contrasting form of rays in tangential view

A: typical lens-shaped rays formed in stem wood of Corylus avellana;

B: distorted rays formed in wood at the apex of a bifurcation in the same species.

106

3.5 Discussion

From our analysis of the wood grain patterns examined, it is clear that the centrally-placed xylem tissues within the branch bark ridge situated at the apex of the junctions examined are quite different from grain patterns seen in normal woody stems. For the species examined, it has long been known that it has a higher density; Shigo referred to this zone as ‘compacted xylem’ but did not allocate it any significant mechanical role in his branch attachment model (Shigo, 1985), while Slater and Ennos (2013) found that the wood basic density was 18.9% higher than wood outside the junction in samples approximately 25 mm3 in volume. In this study we found that in our scanned hazel samples of approximately 0.5 mm3, the cell wall content of samples excised from junctions was around 28% higher than typical stem wood.

However, the main difference between the vessels in the stem wood and at the apices of the hazel bifurcations was in the pattern of the grain itself. The severe change in vessel orientation at the apex of the bifurcations was due to this xylem being formed in the saddle-shape found at these apices, where the wood grain must divert to join to either one or other branch of the bifurcation. Though this study agrees with the findings of Zimmerman and Brown (1971) that ‘interweaving’ is a common feature of xylem in a woody stem, this occurred to a far greater extent at the apex of the bifurcation, and there were other large differences: vessels were less frequent here, their paths disrupted, the wood grain was tortuous (sometimes whirled), and the individual cells were approximately a third shorter, and approximately a half smaller in diameter than in the typical xylem of a stem or branch local to the junction.

The parameter of inter-vessel tortuosity allowed us to measure the extent of this interweaving with accuracy. In our comparison between xylem from the parent stem and from the apex of the bifurcation, the degree of deviation from a straight path was at least 22 times higher in the xylem excised from the bifurcations of hazel, analysing the segmented vessels that we selected in this study. This value probably substantially under- estimates the actual level of grain tortuosity at junctions for several reasons; the most twisted and narrow wood grain patterns exhibited no vessel content and therefore could not be followed at the resolution attained with our micro-tomography scans. The circular

107 grain pattern also had to be excluded from our tortuosity measurements to achieve a mean value for tortuosity of our samples, since a circular path has a tortuosity of infinity.

The vessels and rays situated at the apex of the bifurcations also appeared crushed, as if they were formed under substantial axial and radial growth pressures – and that may well be the origin of their tortuous routes, as they are formed in part of the vascular cambium that would come under pressure from the incremental growth of both adjacent branches that arise from the bifurcation. However, the hypothesis that the effect of being ‘squeezed from both sides’ is the ontogenesis of this contorted xylem would need to be the subject of a further study; here, we have merely examined its anatomy, not its origins.

Due to the wood grain arrangement at the bifurcation apex, to break such a bifurcation apart requires not just the separation of fibres that are aligned along their axis, but also the snapping across of wood fibres or tracheids at and near the apex of the bifurcation. Our findings complement those of Burns et al. (2012) and provide an anatomical explanation of the fibre bridging and fibre pull-out that these researchers found when they examined junction failures in Pinus radiata D. Don. Once the bifurcation is split past the point where the two piths of the branches of the bifurcation conjoin, only the tangential strength of the wood in the stem prevents the crack propagating further. Having to stretch fibres or tracheids along their length so that they are broken across or extracted from surrounding tissues requires much greater stresses than stretching them radially or tangentially (Dresch and Dinwoodie, 1996). In this way the pattern of the wood grain we found at the apices of bifurcations explains how tree junctions can have sufficient strength to stand up to storms and keep aloft large tree branches.

The tangential strength of stem wood is somewhat weakened by the presence of rays that lie in that plane and can act to propagate cracks through the stem’s centre (Ennos and Van Casteren, 2010). However, in the wood of the bifurcations examined, the rays were far from being aligned in any one plane due to the tortuous routes they took through the xylem and their distorted shapes. This arrangement of the rays may also confer greater tangential strength to the tissues formed at the bifurcation apex and so further help to

108 prevent the bifurcation from splitting. Lens-shaped rays have a tendency to cause splits to develop by concentrating stresses at their top and bottom (Ennos, 2011), but this tendency will be lessened for the distorted and narrower rays formed in the xylem at the bifurcation apex.

We represent our new anatomical model of hazel bifurcations graphically by a highly simplified pattern of wood grain orientation (Fig. 3.7). It is the interlocking of the grain pattern at the apex that makes the breaking across of some fibres/tracheids a necessity in order to split the bifurcation apart. This solution allows for all functional vessels to be running “from source to sink” in terms of sap flow, but also provides an explanation as to the appearance of fibres orientated at right angles to and jutting out from the plane of the fracture surfaces of the broken bifurcations examined.

Figure 3.7: Simplified pattern of interlocking wood grain (xy plane) formed at the bifurcation apex

This idealised pattern produces the most important component of the join between branches at the apex of a hazel bifurcation. Note that the route of each grain line passes from the parent stem into one or other arising branch, ensuring all sap-conducting routes run from ‘source to sink’ as they should.

109

We also provide here an idealised anatomical drawing of xylem tissues as they would be arranged at a young hazel bifurcation (Fig. 3.8)

Figure 3.8: Schematic diagram of the arrangement of cell types at a bifurcation

Key tissue types are illustrated at the bifurcation formed in Corylus avellana L.; the piths (yellow), vessels (blue), fibres (white) and rays (red). Note that the rays are arranged to radiate from the pith to the outer edge of the stem, which means that at the apex of the bifurcation they ascend to the apex and they do not traverse the bifurcation.

It is also tempting to suggest that this interlocking is a feature of tree junctions in general. Examination of de-barked junctions of 19 other tree species showed the same wood grain pattern. The more mature tree junctions (between 15 to 45 years old) examined for this

110 study by simple de-barking of specimens of a range of woody angiosperms and gymnosperms, exhibited a large quantity of whirled grain at their apices (Fig. 3.9). Lev- Yadun and Aloni (1990) have previously discussed whirled grain as a feature of tree junctions that provides hydraulic segmentation to sap flow. Where whirled grain is produced at the apex of tree junctions it will also have an important mechanical role to play in providing tree junctions with their load-bearing capacity. The presence of whorls in the vicinity of junctions would allow for a larger area of the junction apex to produce an interlocking grain pattern. As a result, more wood fibres/tracheids would need to be torn apart along their length to split the junction apart, as they could be anchored around and within the whirled grain patterns. This would help to explain the more frequent presence of whirled grain at the apex of more mature tree junctions, which would need a greater load-bearing capacity.

Figure 3.9: Whirled grain at a bifurcation (xy plane) in common ash (Fraxinus excelsior L.) Wood grain pattern exposed by peeling bark away from a semi-mature bifurcation of common ash.

111

It should be noted that the interwoven and whirled grain seen in tree bifurcations are quite different from the seasonally overlapping collars of low diameter ratio bifurcations described following visual inspection by Shigo (1985). Study of a wider range of branch attachments using our CT scanning method and other microscopic techniques could help determine how the pattern of the wood grain in branch attachments changes with the ratio of their diameters. Further studies into the morphology of the ray systems formed at the junctions of a range of woody plants are also recommended.

112

3.6 References

Bradley R S, McNeil A and Withers P J (2010) An examination of phase retrieval algorithms as applied to phase contrast tomography using laboratory sources; Proc. SPIE, 7804.

Burgert I, Bernasconi A and Eckstein D (1999) Evidence for the strength function of rays in living trees; Holz als Roh und Workstoff 57, 397-399.

Burns L A, Mouritz A P, Pook D and Feih S (2012) ‘Bio-inspired design of aerospace composite joints for improved damage tolerance; Composite structures 94, 995- 1004.

Carlquist S (2001) Comparative Wood Anatomy; 2nd Edition; Springer Series in Wood Science; Berlin Heidelberg New York, Springer-Verlag.

Dresch H E and Dinwoodie J M (1996) Timber; Structure, Properties, Conversion and Use; London, MacMillan Press Ltd.

Ennos A R and van Casteren A (2010) Transverse stresses and modes of failure in tree branches and other beams; Proceedings of the Royal Society B 277, 1253-1258.

Ennos A R (2011) Solid Biomechanics; New Jersey US, Princeton University Press.

Hilderbrand T and Rűesgsegger P (1996) A new method for the model-independent assessment of thickness in three-dimensional images; Journal of Microscopy 185, 67-75.

Jane F W (1962) The Structure of Wood; London, Adam and C. Black.

Katz R A and Pizer S M (2003) Untangling the Blum medial axis transform; International Journal of Computer Vision 55, 139-153.

Kramer P J and Kozlowski T T (1979) Physiology of Woody Plants; New York San Francisco London, Academic Press.

Lev-Yadun S and Aloni R (1990) Vascular differentiation in branch junctions of trees: circular patterns and functional significance; Trees: Structure and Function 4, 49- 54.

113

Meyer F (1994) Topographic distance and watershed lines; Signal Processing 38, 113- 125.

Schweingruber F H, Börner A and Schulze E–D (2006) Atlas of Woody Plant Stems; Evolution, Structure and Environmental Modifications; Berlin Heidelberg New York, Springer-Verlag.

Shigo A (1985) How branches are attached to trunks; Canadian Journal of Botany 63 (8), 1391-1401.

Slater D and Harbinson C J (2010) Towards a new model for branch attachment; Journal of Arboriculture 33 (2), 95-105.

Slater D and Ennos A R (2013) Determining the mechanical properties of hazel forks by testing their component parts; Trees: Structure and Function 28 (5), 1437-1448.

Turner I M (2001) The ecology of trees in the tropical rain forest; Cambridge England, CUP. van Casteren A, Sellers W, Thorpe S, Coward S, Crompton R and Ennos A R (2012) Why don’t branches snap? The mechanics of bending failure in three temperature angiosperm trees; Trees: Structure and Function 26, 789-797.

Zimmermann M H and Brown C L (1971) Trees: Structure and Function; Berlin Heidelberg New York, Springer-Verlag.

114

Chapter 4

Interlocking wood grain patterns provide improved wood strength properties in bifurcations of hazel (Corylus avellana L.)

Authors Slater D and Ennos A R

Status of Associated Paper Published in 2015

Journal and Edition Arboricultural Journal 37, 21-32

115

4.1 Chapter Abstract

Xylem found in the stems of woody plants has contrasting strength in different planes due to its anisotropy. At the apex of the junctions of woody plant branches, sufficient wood strength is needed to prevent the arising branches splitting apart. The main aim of this research was to address an issue so far overlooked, which is the contribution of wood grain orientation and patterns at the apex of such junctions to supplying this strength.

In this study, the wood grain patterns of xylem produced at the apex of bifurcations of hazel (Corylus avellana L.) were investigated. The mechanical properties were determined using compression and tensile tests of excised wood samples with an Instron® Universal Testing Machine. Sample strength was contrasted with the strength of xylem produced at the side of the bifurcations and in the adjacent lower stems.

Wood formed at the top of bifurcations in hazel exhibited more tortuous wood grain patterns and had twice the radial and tangential tensile strength when compared with adjacent wood formed in the stem. It also had significantly higher radial and tangential compression strength. Basic density at the apex of hazel bifurcations was 13.4% greater than in the adjacent stem.

The authors conclude that tortuous wood grain patterns formed at the apex of junctions in hazel trees supply additional strength. This mechanical arrangement should inform anatomical studies of junctions and may inform the design of manufactured Y-shaped components made from fibrous composite materials.

116

4.2 Introduction

Sapwood of temperate trees performs a number of functions, including solid carbohydrate storage for the plant; however, the two primary functions of the sapwood can be considered to be the provision of mechanical support to the structure of the tree and the supply of sap to the crown of the tree (Mattheck and Kubler, 1997).

There is often a trade-off between increasing the mechanical strength of diffuse porous wood and increasing its ability to conduct sap efficiently; this is because higher wood strength typically requires more cell wall per unit volume of wood and a higher efficiency in sap conductance requires less cell wall per unit volume (Gartner, 1995). In extreme conditions, as can be found in the wood of some tropical lianas (Ewers et al., 1990), mechanical support is greatly foregone for the sake of improved sap conductance through the production of large vessels with few supporting fibres. The reverse situation can be found in the strong ribs of wood formed at the lateral edges of buttresses of tropical trees, where conducting vessels are almost absent from the wood that will experience high tensile stresses when the tree sways in the wind (Christensen-Dalsgaard et al., 2007).

The wood of temperate forest trees is a highly anisotropic material, with compressive and tensile strength values often six or more times higher along the wood grain (the axial direction) than transversely in either the radial or tangential direction (Jane, 1970). The radial strength of wood is typically slightly higher than its tangential strength, for to pull wood apart or to crush it radially one must break or rupture the rays, narrow lenses of parenchyma tissue that run radially through the wood of all temperate forest trees (Burgert and Eckstein, 2001).

When the junctions in the aerial portions of trees and shrubs are considered, the wood grain arrangement cannot be such that it entirely transits across the junction apex from one branch to another. This would result in a wood grain arrangement that conducted

117 sap between sets of foliage (Fig. 4.1a); the foliage at the end of both branches require a supply of sap, and so the wood grain pattern must connect each branch to the roots of the plant, not to other foliage-bearing branches (Shigo, 1991).

If the wood grain simply flowed up from below a bifurcation directly into one branch or the other, then the wood grain pattern at the bifurcation’s apex would be arranged so that there was only the tangential strength of the wood in that location (Fig. 4.1b). The tangential tensile strength of wood is far lower than its axial tensile strength: bifurcations configured in this way would surely fail and break too easily, as fibres lying side-by-side would be easily peeled apart from each other with the leverage on the bifurcation from both branches. Although this latter wood grain arrangement could possibly occur, it would not be successful at providing the load-bearing capacity needed in junctions of temperate trees.

Figure 4.1: Two improbable wood grain arrangements at bifurcations in trees (xz plane views).

A: Wood grain runs directly across the bifurcation apex, which would supply ample strength to the bifurcation but would break the rule of ‘from source to sink’ for sap flow;

118

B: Wood grain runs only up into each branch from the parent stem, without providing additional strength to the bifurcation; although allowing for sap flow, a bifurcation with this wood grain configuration would be structurally very weak.

Through component testing, Slater and Ennos (2013) found that the centrally-placed xylem at the apex of the bifurcations in hazel (Corylus avellana L.) supplied a disproportionately large amount of the tensile strength to the bifurcation. Through the use of Micro-CT scanning, Slater et al. (2014) have shown the wood grain pattern found at the apex of such bifurcations to be tortuous, with a much lower proportion of the number of vessels than in surrounding stem wood. Whirled grain and wood grain with a zigzag pattern is often observed in this location (Fig. 4.2).

Figure 4.2: Comparison of vessel shapes in stem wood and junction wood of hazel

Visualisation of single vessels from CT scans of samples of hazel (Slater et al., 2014). Axes scales in microns.

A: a typical vessel formed in a hazel stem;

119

B: a vessel formed at the apex of the bifurcation just 25 mm above the location of vessel A, showing far greater tortuosity and considerably more constriction in lumen size. Note this vessel has been rotated 90º to allow a comparison with A.

Images courtesy of the Manchester X-ray Imaging Facility, University of Manchester.

As an outcome of this experimentation and analysis, a new model for branch attachment has been proposed (Slater et al., 2014). Here the wood grain passing through the apex of the bifurcation can interlock in such a way that to pull the bifurcation apart requires the breaking of wood fibres along their length (supplying the axial strength of wood to this location). Otherwise, there would be only the tangential strength of the wood grain (Fig. 4.3). The interlocking pattern is such that it allows flow of sap ‘from source to sink’ near the bifurcation apex. However, in the most central zone of the junction it has been shown to be the case that the wood grain is very tortuous, the cells smaller and thicker- walled, and so this xylem tissue is substantially less effective at sap transport (Slater et al., 2014).

Figure 4.3: Interlocking wood grain patterns at bifurcations in trees

A: simplified schematic diagram of interlocking wood grain formed at a bifurcation (xy plane view), looking down onto the bifurcation apex;

B: Diagram further identifying the location of the interlocking grain pattern at a bifurcation in a tree (xz plane view).

120

These observations are supported by other research identifying that tortuous grain patterns are reliably found at the apex of junctions of Ailanthus, Ficus, Melia, Pinus, Platanus and Quercus (Lev-Yadun and Aloni, 1990) and at the nodes of other plants (André, 2000). A simple interlocking pattern can potentially develop into a more complex pattern with age, through increment growth, to produce large and easily observed whirled grain patterns (Lev-Yadun and Aloni, 1990). Junctions in diffuse porous trees have also been found to be significant constrictions in sap flow efficiency (Zimmermann, 1978) and this may be due to the tortuous routes taken by the wood grain in this location, as well as heightened basic density.

Bark often becomes included within a bifurcation of a tree where two vertically- orientated branches conjoin at a tight angle (Gilman, 2012). This inclusion of bark into a bifurcation will initially prevent the formation of an interlocking grain pattern at the bifurcation apex. However, as the tree grows and the bifurcation receives more loading from the two arising branches, additional wood often forms at either side of the included bark, in order to support the bifurcation (Mattheck and Breloer, 1994) thus forming a cup-shaped bifurcation.

The logical next step from these findings is to test the strength of wood cut from the apices of such bifurcations in trees. Such a study could confirm that this tortuous, interlocking and potentially whirled wood grain pattern, which has been found at bifurcations of hazel and other tree species, actually supplies greater strength to them. Finding how trees produce strong junctions by altering their wood grain orientation could provide a bio-inspired pattern for the fabrication of stronger and tougher T-shaped and Y-shaped components in artificial structures made of fibrous composite materials, such as the joints in boats and aircraft (Burns et al., 2012).

In this study, we report our findings on the comparative tensile and compressive strength of wood cut from the apices of bifurcations in hazel trees, when compared with wood cut from slightly lower down in the stems and to the side of these bifurcations in

121 the same trees. We also tested wood samples cut from the ribs of mature bark-included bifurcations in hazel that had formed at either side of the included bark, to determine if the interlocking wood grain patterns found by Slater et al. (2014) were providing additional strength in these locations too.

We selected hazel as an experimental subject as it has a medium density of wood for a temperate tree, the bifurcations it forms have been subject to previous research (Pfisterer, 2003; Slater and Ennos, 2012; Slater et al., 2014) and bifurcations can be obtained sustainably through the practice of regular coppicing of hazel trees.

4.3 Methods and Materials

4.3.1 Sample collection

Between the 14th of December 2011 and the 18th of January 2012, 51 hazel branch junctions were harvested from hazel coppice situated at Prestwich Country Park, Manchester. The junctions were chosen randomly from within the coppice area and varied in age and size, ranging between eight and fifteen years in age and having parent stem diameters between 58.7 mm and 122.9 mm.

All the junctions selected were ‘bifurcations’; that is, each junction gave rise to two branches of approximately equal diameter emerging from a single parent stem. The diameter ratio of these bifurcations was determined by comparing the diameter of the larger arising branch just above the junction with that of the smaller arising branch, and for this sample the mean diameter ratio was found to be 86.5% and the range was from 78.4% to 97.8%. This selection ensured that none of the junctions consisted of a subordinate branch which had been occluded into a larger stem at its base to form a knot. Twelve of these hazel bifurcations contained bark inclusions, which formed cup- like unions with raised ‘ribs’ of wood situated either side of the included bark.

122

After being cut from each tree using a Japanese pull-saw, and retaining at least 150 mm of the length of the parent stem below the apex of the bifurcation and 150 mm of the length of both branches arising from each bifurcation, each hazel bifurcation was carefully wrapped in a plastic bag. The samples were kept in a cold store at 2-3° C to reduce sap loss, so that the wood within each specimen stayed in a fresh condition for the following mechanical tests.

4.3.2 Compression tests

For compression testing, seventeen normally-formed hazel bifurcations had nine small cubes of wood approximately 9 mm in height, width and thickness cut from them using a finely toothed hacksaw and a razor blade. Three were taken from the outer edge of the parent stem; three from the outer edge of the junction in the plane of the bifurcation, and three from the apex of the junction (Fig. 4.4a). These cubes were then kept on a zinc mesh situated above a layer of water in an airtight container to maintain the samples’ moisture content until they were tested within 24 hours. A compression test using an Instron® model 4301 Universal Testing Machine (UTM) (manufactured by Instron® Ltd., Norwood, Massachusetts) was carried out on each of these cube-shaped samples, crushing one cube from each location in the radial, tangential and axial direction respectively for each of the seventeen hazel bifurcations.

Cubes were crushed between flat steel plates at a rate of 10 mm min-1 (Fig. 4.4b), whilst an interfaced computer connected to the UTM simultaneously recorded the displacement and the load. Each test was carefully observed and notes taken on their mode of failure. A small number of the excised cubes were rejected due to cutting defects evident during testing that affected their mode of failure.

Cellular solids such as wood yield in compression well before they break due to the cell walls buckling, so we calculated the yield stress of the wood from the

123 force/displacement curve. The point of yield (Fig. 1.2) for each sample was determined using the methods described by van Casteren et al. (2012).

Figure 4.4: Diagram of compression test methodology

A: Locations (xz plane) for excision of wood cubes from 17 normally-formed bifurcations, the cubes’ edges being approximately 9 mm in length. One sample was taken for each orientation (longitudinal, radial and tangential) from every location

B: Compression test set-up on the Instron® Universal Testing Machine.

4.3.3 Tensile tests

Transverse tensile tests of wood samples were carried out on dumbbell-shaped pieces of wood cut from the parent stem just below the start of the branch bark ridge, from the outer edge of the junction in-line with the bifurcation and from the apex of the junction in both radial and tangential orientations (Fig. 4.5a). Initially, ten normally-formed hazel bifurcations were used to supply these dumbbell-shaped samples, which were cut out using a hacksaw and a small rotary sander so that each wood sample had a length of the narrow region of 11 mm. For samples taken from the parent stem or the side of the junction, the width and thickness at the centre of each dumbbell was 4.89 mm ± 0.03SE by 4.88 mm ± 0.04SE respectively. For samples taken from the apex of the junctions, where higher wood strength was anticipated, the width and thickness was made smaller

124 to match the limitations of the UTM’s load cell, at 2.44 mm ± 0.03SE and 2.58 mm ± 0.02SE respectively.

Figure 4.5: Diagram of tensile test methodology

A: Locations (xz plane) for excision of dumbbell shapes in 22 normally-formed hazel bifurcations, 25 mm in length x 10 mm breadth x 5 mm width. One sample was taken for each orientation (longitudinal, radial and tangential) from every location, apart from the apex, where longitudinal samples were not taken.

B: Tensile test set-up on the Instron® Universal Testing Machine.

The arrangement of rays in the xylem situated in the centre of the junction is such that the rays do not transit the join between the branches. Consequently, the wood strength in-line with the bifurcation is considered to be the tangential strength and the wood strength perpendicular to the bifurcation in this location is considered to be the radial strength (Fig. 4.6).

125

Figure 4.6: Orientation of wood strength testing at a bifurcation (xy plane).

Rs is the direction for testing the radial strength of the wood in the stem, Ts for the tangential strength of the stem, Rf for the radial strength of the wood at the bifurcation and Tf for the tangential strength of the wood at the bifurcation.

Image courtesy of the Manchester X-ray Imaging Facility, University of Manchester.

In January 2012 a further 24 hazel bifurcations were collected, with 12 of these bifurcations containing a relatively wide bark inclusion at their apex, the other 12 having no bark inclusion and acting as a control. Dumbbells for tensile testing were excised in a similar way, but for this subset, it was deemed necessary only to test the tangential strength of the tissues at the bifurcation apex. The purpose of this experiment was to contrast the strength of the apical xylem tissues from the bark-included bifurcations with the normally-formed bifurcations in the plane of the bifurcation, the direction in which they may split if subjected to strong wind forces. For the bark- included bifurcations, the dumbbells were excised from the apices situated either side of the bark inclusion (Fig. 4.7). Whilst the specific gauge length for these specimens was kept at 11 mm, the width and thickness at the centre of these twenty-four samples was increased slightly, at 3.05 mm ± 0.03SE by 3.15 mm ± 0.04SE respectively.

126

Figure 4.7: Location for excision of tensile testing samples for bark-included bifurcations.

For all the tensile tests, the dumbbells were clamped to the base and crosshead of the Instron® Universal Testing Machine and the crosshead moved upwards at a speed of 1mm min-1 until each sample failed, the interfacing computer recording the displacement (mm) and peak load (N) for each sample (Fig. 4.5b). The tensile strength of the wood from these samples was then calculated by dividing the peak force by the cross sectional area of the thinnest part of the dumbbell (van Casteren et al., 2012).

4.3.4 Basic density testing

After the tensile tests, both halves of each dumbbell were then kept upon a zinc mesh situated above a layer of water in an airtight container to maintain the samples’ moisture content. Their volumes were then measured within 24 hours using the water displacement method (Hughes 2005) and they were then dried in an oven at 60 °C for 48 hours and weighed. The small size of the samples used in this experiment meant this was sufficient time to dry them out completely. The basic density was determined by dividing the dry weight of the samples by their volume (Osazuwa-Peters & Zanne, 2011).

127

4.3.5 Statistical analysis

Yield strength data obtained from the tests were initially entered into Microsoft Excel 2013 for basic analysis of sample means and standard errors. Data was then transferred to Minitab v. 16 to conduct ANOVAs to identify differences in sample means. Significant differences between sample types were identified using post-hoc Tukey tests after each ANOVA, setting the confidence level at 95%. Residuals were checked for normality using the Anderson-Darling test.

4.4 Results

4.4.1 Compression Tests

Observations of the samples cut for compression testing identified no visual differences in wood grain orientation between samples cut from the parent stem and the side of the bifurcation; however, samples cut from the bifurcations’ apices displayed more tortuous wood grain patterns, which could be seen by eye.

The compressive yield strengths of the wood samples successfully tested are shown in table 4.1. In all three areas of the bifurcations from which the cubes were cut, axial wood strength was greater than transverse strength, as might be expected. However, there were significant differences between wood cut from the bifurcation apex and in the other areas; it was stronger both radially and tangentially, but weaker axially.

Significant differences in sample means between the three sample locations were identified by a one-way ANOVA and post-hoc Tukey test set at a confidence level of 5% for each direction (axial, radial and tangential) in which the samples were tested.

128

Table 4.1: Compressive strength of excised cubes with standard errors. Statistical differences were determined by a one-way ANOVA (p < 0.001) and post-hoc Tukey tests (confidence interval: 95%). Locations with significant differences in each column are denoted by different letters. Normality of data was determined using Anderson-Darling tests on the residuals of the ANOVAs.

Sample location Axial Radial Tangential Compressive Compressive Compressive Strength (MPa) Strength (MPa) Strength (MPa)

Parent stem 18.73 ± 0.91a 4.04 ± 0.23b 3.6 ± 0.19b

n=17 n=17 n=17

Side of bifurcation 18.02 ± 0.90a 4.61 ± 0.23b 3.88 ± 0.24b

n=16 n=17 n=16

Apex of normal 7.57 ± 0.28b 6.19 ± 0.2a 7.97 ± 0.45a bifurcation n=15 n=17 n=16

Normality test results AD48 = 0.189 AD50 = 0.458 AD49 = 0.668 on residuals of p = 0.895 p = 0.253 p = 0.076 ANOVA

4.4.2 Tensile tests

The tensile strength of wood from the three different locations is shown in table 4.2. One sample failed atypically and was excluded from data analysis. Radial strength of wood from parent stems and the sides of the bifurcations was somewhat higher than the tangential strength, as might be expected, due to the orientation of parenchyma rays in the radial direction. At the apices of the bifurcations, however, wood was significantly stronger both radially and tangentially. The tensile strength of the wood from the apices

129 of the bifurcations with included bark was 11.2% lower than that in normally-formed bifurcations. Significant differences between all the samples tested were identified using a one-way ANOVA and post-hoc Tukey test (F6 = 30.54; p< 0.001). An Anderson Darling normality test was used to determine that residuals from the ANOVA were normally distributed (AD82 = 0.715; p = 0.076).

Table 4.2: Tensile strength of excised dumbbells with standard errors. Statistical differences were determined by a one-way ANOVA (p < 0.001) and post-hoc Tukey test (confidence interval of 95%). Locations with significant differences are denoted by different letters.

Sample location Radial Tensile Tangential Tensile Strength (MPa) Strength (MPa)

Parent stem 10.26 ± 0.45d 8.01 ± 0.36d

n =9 n =10

Side of bifurcation 11.88 ± 0.75cd 9.03 ± 0.5d

n =10 n =9

Apex of normal 22.27 ± 1.81a 16.18 ± 0.67 b bifurcation n =10 n = 21

Apex of included bark Not tested 14.55 ± 0.59 bc bifurcation n = 12

4.4.3 Basic Density

The wood’s basic density was found to be 546.5 ± 6.7 kg m-3 SE for the dumbbell shaped wood samples cut from the parent stem, 554.1 ± 8.2 kg m-3 SE for those cut from the side of the bifurcations and 619.64 ± 3.2 kg m-3 SE for the samples cut from the apices of the bifurcations. Basic density was found to be significantly different

130 between the locations using a one-way ANOVA (F2 = 39.65; p< 0.001) and a post-hoc Tukey test showed that it was significantly higher at the bifurcations’ apices, being on average 13.4% denser than wood excised from the parent stem, with the range of density difference found to be between 5% to 23%. There was no significant difference between the basic density of samples taken from the parent stem and those cut from the side of the bifurcations.

4.5 Discussion

The compressive wood tests provide evidence of the effect of the changing grain orientation of xylem at the apex of bifurcations found by Slater et al. (2014), as the axial, radial and tangential strength of the wood excised from the bifurcations’ apices lies between values for the axial and radial strength of xylem cut from the parent stem. In addition, on the cut surfaces of the sample cubes excised from the bifurcations’ apices, a more tortuous wood grain pattern could be seen. These observations imply that tortuous interlocking grain patterns, including whirled grain, gives xylem formed at the apex of a junction specialised biomechanical qualities.

The tensile tests of the dumbbell shapes also identified that the xylem at the apex of normally formed bifurcations supplied more than twice the tensile strength to the junction than would be given if the cells were in a simple tangential arrangement in this location.

As this interlocking grain pattern is formed in the saddle-shape of a bifurcation apex, cutting out flat shapes from this apex is bound to reduce the effectiveness of any grain interlock present. It is improbable that the alignment of the cut dumbbell would match exactly with the alignment of the 3D interlocking wood grain arrangement in the bifurcation. Consequently, values of tensile strength given here for these apical tissues should be considered an estimate of the minimum to be expected, and the actual

131 strength of this material in-situ at the bifurcation apex is likely to be considerably higher.

Given the nature of the experiment, it is not possible to state that the presence of tortuous and interlocking grain patterns at the top of bifurcations alone explains all the differences in wood strength found in the samples tested. From our results, it appears likely that a heightened basic density is also a contributory factor to the higher wood strength in this zone. Other wood properties such as the proportion of different cell types in the tissue, microfibril orientation and the level of cellulose and lignin content may also be factors that are affecting wood strength in this location; these would be useful areas for future research.

The xylem at the apices of bark-included bifurcations was just over 80% stronger under tension in the tangential direction than the xylem sampled from the parent stem. However, rather than the single apex that a normally-formed bifurcation has, bark- included bifurcations have two apices, one either side of the obstructing bark. From this test, it appears that interlocking grain can form at the two apices positioned at either side of the included bark of these bifurcations, and this may provide such junctions with considerable strength where the junctions have become cup-shaped bifurcations with the two apices situated above the position of the bark inclusion (Fig. 4.8).

132

Figure 4.8: Remedial xylem growth at a bark-included bifurcation

Illustration of xylem formed around included bark within a bifurcation in hazel and the location of interlocking wood grain at the bifurcations’ apices. The left hand image shows the xz plane, the right hand image illustrates a view through the junction in the yz plane (Fig. 1.1 for axes definitions)

The xylem in the dumbbells excised from the apex of the bifurcations was significantly denser when compared to those sampled from parent stem material, and every comparison identified that the tissues from the apex were denser than tissues from the adjacent parent stem. A heightened wood density will confer greater strength to this location (Niklas, 1993; Anten and Schieving, 2010), where stresses will concentrate should the two interconnected branches of a tree sway away from each other. However, this increased strength is likely to come at the cost of much lower sap conductance efficiency through this zone due to the presence of more tortuous grain orientation, higher cell wall thicknesses, and the lessening of the number and size of vessel elements (Slater et al., 2014).

It can be concluded that the heightened basic density and tortuous nature of the wood grain found at the apices of bifurcations in hazel and in other trees makes these tissues less likely to rupture from radial or tangential loading. We consider this to be a good

133 example of functional trade-off in xylem, such as discussed by Gartner (1995), whereby a substantial loss in sap conductance efficiency is compromised in this part of the hazel tree’s structure (Zimmermann, 1978) for the sake of increased wood strength.

This finding will help with further modelling of the biomechanical and structural behaviour of trees; the authors intend to continue this research using hazel as a model organism. The wood grain patterns found at the apex of bifurcations in hazel trees should be copied to trial the production of novel composite joints using fibrous materials; these novel bio-inspired T-shaped and Y-shaped components could have much higher load-bearing capacities and greater toughness than those currently manufactured.

134

4.6 References

Anten N P R and Schieving F (2010) The role of wood mass density and mechanical constraints in the economy of tree architecture; American Naturalist 175, 250- 260.

Burgert I and Eckstein D (2001) The tensile strength of isolated wood rays of beech (Fagus sylvatica L.) and its significance for the biomechanics of living trees; Trees: Structure and Function 15 (3), 168-170.

Burns L A, Mouritz A P, Pook D and Feih S (2012) Strength improvement to composite T-joints under bending through bio-inspired design; Composites Part A 43, 1971-1980.

Christensen-Dalsgaard K K, Fournier M, Ennos A R and Barfod A S (2007) Changes in vessel anatomy in response to mechanical loading in six species of tropical trees; New Phytologist 176 (3), 610-22.

Ewers F W, Fisher J B and Chiu S T (1990) A survey of vessel dimensions in stems of tropical lianas and other growth forms; Oecologia 84, 544-552.

Gartner B L (ed.) (1995) Plant stems: physiology and functional morphology; London, Academic Press Ltd.

Gilman E F (2012) An illustrated guide to pruning; 3rd Edition; New York, Delmar Publishers.

Hughes S W (2005) Archimedes revisited: a faster, better, cheaper method of accurately measuring the volume of small objects; Physics Education 40 (5), 468-474.

Jane F W (1970) The structure of wood; 2nd Edition; London, A & C Black.

135

Lev-Yadun S and Aloni R (1990) Vascular differentiation in branch junctions of trees: circular patterns and functional significance; Trees: Structure and Function 4, 49-54.

Mattheck C and Breloer H (1994) The body language of trees: a handbook for failure analysis; London, TSO.

Mattheck C and Kubler H (1997) Wood – the internal optimization of trees; Berlin, Springer-Verlag.

Niklas K J (1993) Influence of tissue density-specific mechanical properties on the scaling of plant height; Annals of Botany 72, 173-179.

Shigo A L (1991) Modern arboriculture: A systems approach to the care of trees and their associates; Durham, New Hampshire US, Shigo and Trees, Associates.

Slater D and Ennos A R (2013) Determining the mechanical properties of hazel forks by testing their component parts; Trees: Structure and Function 27 (6), 1515-1524.

Slater D, Bradley R S, Withers P J and Ennos A R (2014) The anatomy and grain pattern in forks of hazel (Corylus avellana L.) and other tree species; Trees: Structure and. Function 28 (5), 1437-1448. van Casteren A, Sellers W L, Thorpe S K S, Coward S, Crompton R H and Ennos A R (2012) Why don’t branches snap? The mechanics of bending failure in three temperate angiosperm trees; Trees: Structure and Function 26, 789-797.

Zimmermann M H (1978) Hydraulic architecture of some diffuse porous trees; Canadian Journal of Botany 59, 2286-2295.

136

Chapter 5

The level of occlusion of included bark affects the strength of bifurcations in hazel (Corylus avellana L.)

Authors Slater D and Ennos A R

Status of Associated Paper Published in 2015

Journal and Edition Journal of Arboriculture and Urban Forestry 41, 194-207

137

5.1 Chapter Abstract

Bark-included junctions in trees are considered a defect as the bark weakens the union between the branches. To more accurately assess this weakening effect, 241 bifurcations from young specimens of hazel (Corylus avellana L.), of which 106 had bark inclusions, were harvested and subjected to rupture tests. Three-point bending of the smaller branches acted as a benchmark for the relative strength of the bifurcations.

Bifurcations with included bark failed at higher displacements and their modulus of rupture was 24% lower than normally-formed bifurcations, while stepwise regression showed that the best predictors of strength in these bark-included bifurcations were the diameter ratio and width of the bark inclusion, which explained 16.6% and 8.1% of the variability respectively. Cup-shaped bark-included bifurcations where included bark was partially occluded by xylem were found to be on average 36% stronger than those where included bark was situated at the bifurcation apex.

These findings show that there are significant gradations in the strength of bark- included bifurcations in juvenile hazel trees that relate directly to the level of occlusion of the bark into the bifurcation. It therefore may be possible to assess the extent of the defect that a bark-included bifurcation represents in a tree by assessing the relative level of occlusion of the included bark.

138

5.2 Introduction

Junctions in trees that are separated by bark being included in their union are frequently found in urban and forest trees (Lonsdale, 1999). Such junctions have a reputation of being structural flaws in tree crowns (Shigo, 1989; Lonsdale, 2000; Harris et al., 2004; Gilman, 2011), and they are commonly recorded as a defect by tree assessors and others with responsibility for the safety of people and property adjacent to trees (Matheny and Clark, 1994; Mattheck and Breloer, 1994).

Where only two branches arise from a junction in a tree, this is formally referred to as a bifurcation. It has been established that the ‘diameter ratio’ between the two branches that arise from a bifurcation in a tree has a substantial effect on its mechanical strength and failure mode (Gilman, 2003). The ‘diameter ratio’ is defined as the ratio between the basal diameters of the smaller and larger branch, measured just above the point of their attachment to each other at the bifurcation, and is often also referred to as the ‘aspect ratio’ (Gilman, 2003). Kane et al. (2008) found through rupture testing that bifurcations formed in young trees of three species (Acer rubrum L., Quercus acutissima Carruthers and Pyrus calleryana Decne.) that had a diameter ratio of 70% or higher were only half as strong as those that had a clearly subsidiary branch. Additionally, these researchers found that the fracture surfaces of bifurcations with a low diameter ratio showed that xylem tissues of the smaller branch were embedded within the larger branch; in contrast, co-dominant stems exhibited relatively flat fracture surfaces with little to no embedding of tissues.

Two distinct failure modes occur in higher diameter ratio bifurcations of hazel (Corylus avellana L.) when they are subjected to tensile loading, and these have been defined by Slater and Ennos (2013) as Type I and Type II failure modes. In the Type I failure mode, which tends to occur at intermediate diameter ratios (70% to 80%), there is compressive yielding of the xylem at the base of the smaller branch at its outer edge, before the bifurcation splits at its apex (Fig. 5.1a). In the Type II failure mode, which occurs most often when the two branches are nearer to the same diameter (diameter

139 ratios > 80%), there is no compressive yielding and the bifurcation fails by a sudden splitting of tissues at its apex (Fig. 5.1b). In much lower diameter ratio bifurcations (< 70%), yielding of the branch under compression then tearing of its tissues under tension near the bifurcation becomes a common mode of failure (Fig. 5.1c), which is termed a ‘branch failure’.

Figure 5.1: Type I, Type II and branch failure modes of tree bifurcations under tension across the bifurcation.

A: In Type I failure mode, the xylem yields initially under compressive forces on the outer edge of the bifurcation before the bifurcation splits at its apex under tension.

B: In Type II failure mode the initial failure is under tension at the bifurcation apex.

C: In branch failures, the initial failure is compressive buckling of the xylem on the underside of the branch before the top of the branch is torn apart under tension.

The strength of a normally-formed hazel bifurcation can be considered to be provided by three components: the resistance of wood at the centre of the join to tension, the resistance of wood at either side of the centre of the join to tension and the bending resistance of the wood at the side of the smaller branch as it joins the other branch. The tensile strength of a bifurcation in a tree is increased by it having a zone of interlocking wood grain in the centre of the join (Slater and Ennos, 2013; Slater et al, 2014).

140

Once bark is included into a bifurcation it is inherently weakened as the centrally-placed interlocking wood grain is absent at the apex (Slater et al, 2014). Smiley (2003) found that young tree bifurcations with bark inclusions in Acer rubrum L. were 20% weaker when pulled apart than those without bark inclusions. A bifurcation with included bark may not remain a significant defect as it matures; it may develop in ways that affect both the relative size of the bark inclusion and the shape of the bifurcation overall. A bifurcation may grow to completely occlude the bark inclusion (Fig. 5.2: embedded), so it is invisible from the outside; it may form additional xylem around and above the bark inclusion without fully occluding it (Fig. 5.2: cup-shaped bifurcation); or the bark inclusion may persist and remain at roughly the same proportion of the width of the join with every annual increment of growth (Fig. 5.2: wide-mouthed bark inclusion).

Figure 5.2: Potential development pathways for a bark-included bifurcation

Timeline showing the morphology of the xylem and bark-inclusion perpendicular to the plane of the bifurcation (yz plane, Fig 1.1), potentially leading to the formation of three distinct forms of bark inclusion, which have been given a proposed nomenclature here: embedded bark, a cup-shaped bifurcation or a wide-mouthed bark inclusion.

141

In arboricultural guidance on this commonly-occurring structural flaw, Lonsdale (2000) suggests that the length of the bark inclusion that is visible along the branch bark ridge below the apex of a bifurcation may be linked to the likelihood of its failure. Helliwell (2004) has also suggested that there may be an influence on the strength of a bifurcation with included bark from the degree of constriction of the parent stem’s diameter just below the apex of the bifurcation where the bark inclusion starts. Kane et al. (2008) found that the percentage area of the fractured attachment covered by a bark inclusion in red maple (Acer rubrum), sawtooth oak (Quercus acutissima) and callery pear (Pyrus calleryana) did not reliably predict the strength of the bifurcation, but that overall the strength of bark-included bifurcations was lower than normally-formed bifurcations.

Despite these general observations by experienced arboriculturists, there is currently no means of quantifying the heightened risk of failure of bifurcations with included bark in trees from observing their external morphology or the position and size of the bark inclusion present. In this study, therefore, we investigated the strength of bifurcations in relation to the presence or absence of bark inclusions, and, if present, the position, shape and size of bark inclusions found. We sought to find a simple rule by which the relative weakness of a bifurcation with included bark could be predicted.

We chose to model this mechanical behaviour in one species, Corylus avellana L., as similar research on this species has been carried out by Pfisterer (2003) which allows for a comparison in findings, and the wood grain orientation and mechanical contributions of different components of such bifurcations in this species have recently been uncovered (Slater and Ennos, 2013). We have favoured this species as an experimental subject as it provides a sustainable source of bifurcations and working with coppice grown material of one species limits the effects of other factors (e.g. age differences, differences in levels of exposure) that could affect bifurcation strength. Having a more comprehensive picture of the biomechanics of bifurcations in one woody species which has been well-researched in respect of its anatomy and mechanical behaviour justifies this single species choice in this study.

142

Testing the strength of young tree bifurcations may provide useful insight for tree assessors where they inspect larger-growing tree species with bark included junctions, although this approach will likely have its limitations in terms of the scale of the tree bifurcations tested.

5.3 Materials and Methods

5.3.1 Sampling

Between November 2010 and January 2012, 241 junctions of hazel were harvested from hazel coppice situated at Prestwich Country Park, Manchester. All the junctions harvested had two emergent branches, making each one a ‘bifurcation’. Collecting from only one site was necessary to limit the number of factors affecting bark inclusion formation and bifurcation strength: for example, if one collected from more exposed and more sheltered locations the strength of the individual bifurcations within the sample would vary much more widely. Collection of the samples was randomised throughout the coppice, avoiding obtaining more than two bifurcations from any one tree and not taking any bifurcations from trees growing along the edges of the coppice. This resulted in 96 samples being collected from the same tree as one other sample, and 145 samples each being the only one collected from a particular tree.

Samples were cut to retain approximately 100 mm of the parent stem and 215 mm of each branch arising from the bifurcation. Samples were wrapped separately in plastic bags and put in cold storage at 2-3 °C to reduce sap loss before testing. The hazel bifurcations had an average parent stem diameter of 33.2 mm (range 17.01 mm to 58.69 mm) and an age range of between three to eight years old

143

5.3.2 Rupture tests

Rupture tests of the hazel bifurcations were carried out as described in section 2.3.2 of Chapter 2 of this thesis.

The failure mode of each bifurcation was observed closely and recorded during this test procedure. The Type I failure mode was categorised by eye through the appearance of ripples caused by compression forces on the outer edge of the smaller branch as it joined the bifurcation, prior to the splitting of the bifurcation apex. Specimens recorded as undergoing Type II failure mode exhibited no compressive yielding in the exterior tissues prior to the bifurcation splitting at its apex. Branch failures were categorised as all those failures that occurred in the arising branch and that did not split the bifurcation apart (Fig. 5.1).

The method of measurement of the dimensions of each sample and the subsequent calculations of maximum bending moment, Mpeak, angle θ and the maximum bending stress of the bifurcation, σa, were those described in section 2.3.3 of Chapter 2 of this thesis.

5.3.3 Three point bending test

After the rupture testing, a three point bending test was carried out on the smaller of the branches arising from the bifurcation to determine the bending stress it could withstand before yielding. All the branches were carefully checked that they had not been damaged during the rupture testing prior to this three point bending, to ensure this testing gave reliable results. Limitations of the load-cell available meant that branches above the diameter of 23 mm could not be bent to their yield point, limiting the sample size for this second test to 83 branches.

144

These three point bending tests were carried out as described in section 2.3.4 of Chapter 2 of this thesis. The only variation was that the span width of the three point bending rig was adjusted from 292 mm (as reported in Chapter 2) to 295 mm. The span length available for these tests was necessarily limited to 295 mm because of the location of two side columns on the Instron® UTM. Calculation of the maximum bending stress,

σb, acting upon the branch at the point that it yielded used equation 2.4 (Chapter 2).

The completion of the rupture tests and three-point bending tests allowed a comparison to be made between the maximum bending stresses of the bifurcations tested with the yield stresses of the smaller branches that arose from these bifurcations.

5.3.4 Measurements of Included Bark

For all the bifurcations where bark inclusions were exposed during the rupture testing (n = 104), the fracture surfaces were then excised and digitally scanned using an HP Scanjet 2400® (Manufacturer: Hewlett Packard, Palo Alto, California). These samples were then categorised as either embedded bark inclusions (n = 17), cup unions (n = 57) or wide-mouthed bark included bifurcations (n = 30) (Fig. 5.2). The image analysis software ImageJ (Abramoff et al., 2004) was then used to measure the area of bark relative to that of the fracture surface (Fig. 5.3a) by the use of the polygon selection tool. The same technique was used to measure the ratio between the width of the bark inclusion at the apex of the bifurcation and the width of the parent stem at the base of the branch bark ridge, where the pith of the parent stem bifurcates (Fig. 5.3b). This second measure was chosen as we suspected that as the highest tensile stresses act at the bifurcation apex when the two branches are pulled apart, so the failure would occur more easily when a higher proportion of included bark was present in this location.

145

Figure 5.3: Measurements of the fracture surfaces (yz plane view) of bark-included bifurcations carried out in ImageJ.

A: Proportion of the area of the fracture surface (yz plane) containing included bark. The area of the bark (dark) was divided by the area of the fracture surface overall (striped lines);

B: Relative width of the bark inclusion at the apex of the bifurcation (upper white arrow) in the yz plane, when compared with the width of the parent stem (lower white arrow), at the point where the pith bifurcates.

The bifurcations with included bark that was exposed at the apex (n = 87) were also categorised as to whether they had formed a cup-like bifurcation (where two areas of xylem were found at the apex of the bifurcation, formed either side and above the bark inclusion), or whether there was included bark situated at the apex of the bifurcation (Fig. 5.4a and b). Again, this comparison was chosen to try to assess if there was a difference in the strength of these two types of bark included bifurcation because of the difference as to which material (wood or bark) was situated at the apex.

146

Figure 5.4: Categorisation of bifurcations with included bark into two types

A: fracture surface (yz plane) of a cup-shaped bark-included bifurcation with wood at its apex;

B: fracture surface (yz plane) of a wide-mouthed bark-included bifurcation with included bark at its apex.

5.3.5 Statistical analysis

A Chi-Squared test was used to determine if there was a significant difference in the frequency of failure modes between bifurcations with included bark and normally- formed bifurcations.

To analyse the relationship between different failure modes observed and the diameter ratio of the samples tested, a GLM ANOVA was carried out with one covariate (the diameter of the parent stem) and with the random factor of the tree number from which each sample was collected. A post-hoc Tukey test with 95% confidence interval was used to confirm statistical differences between groups of samples exhibiting different failure modes.

147

To analyse the relationship between the displacement of the sample prior to failing and the failure modes exhibited by the samples, a GLM ANOVA with post-hoc Tukey test was used, with the diameter of the parent stem as covariate. A subsequent one-way ANOVA was used to determine if bark-included bifurcations exhibiting a Type II failure mode had significantly shorter displacements before failure than normally- formed bifurcations.

A one-way ANOVA, alongside a post-hoc Tukey test with 95% confidence interval, was used to find differences in sample strength between normally-formed bifurcations, bifurcations with included bark and smaller arising branches.

The relationship between the maximum breaking stress, σa, and the shape of the bark inclusions in the bifurcations with included bark exposed at their apex (n = 87) was investigated using stepwise regression analysis. Samples with embedded bark (n = 17) were excluded from this analysis as they did not have a width of bark at the apex of the bifurcation. These stepwise regressions were performed to identify the best models for predicting bifurcation strength from the parameters that were measured for each sample (the diameter ratio, the parent stem diameter, the proportional area of included bark on the fracture surface and the ratio of the bark width at the bifurcation apex with the parent stem diameter) could predict bifurcation strength better.

A GLM ANOVA, alongside a post-hoc Tukey test with 95% confidence interval, were used to confirm differences between groups of categorised bark-included bifurcations and normally-formed bifurcations, again with the diameter of the parent stem as a covariate and with the number of the tree collected from as a random variable.

Residuals from these ANOVAs and regressions were tested for normality using the Anderson-Darling test to ensure the data were suitable for analysis by parametric statistical tests.

All statistical tests were carried out in Minitab® 16 statistical software.

148

5.4 Results

The range of diameter ratios found in the sample was from 53% to 100%, with the mean ratio being 81.41% ± 0.7 SE. There was no significant difference in the average branch diameter ratio between bifurcations with or without included bark; diameter ratios of the two branches were 80.8% ± 1.0 SE for the normally-formed bifurcations and 82.1% ±

1.1 SE for bifurcations with included bark. Neither did the two types of bifurcation 2 show a significant difference in the relative incidences of the three failure modes (Χ 2 = 4.224; p = 0.121) (Table 5.1); in both, Type II failure modes were the most common and branch failures the least common.

Table 5.1: Analysis of failure modes observed Instances of different failure modes experienced (n) and associated mean diameter ratios (μ) of control and bark included bifurcations subjected to tensile testing.

Specimen type Branch Type I Type II failure failure failure

Control n = 9 n = 53 n = 73

μ = 76% μ = 74% μ = 86%

Bark included n = 6 n = 29 n = 71 bifurcations μ = 66% μ = 76% μ = 86%

A subsequent GLM ANOVA showed that there were significant differences between these three modes of failure due to difference in diameter ratio (F2, 236 = 6.28; p =

0.004); the parent stem diameter was not a significant co-variant (F1, 236 = 3.82; p =

0.057) and the random factor of the tree number was not significant (F192, 236 = 0.78; p = 0.866). The higher the diameter ratio, the more common were Type II failure modes and the less common were Type I failure modes and branch failures. A post-hoc Tukey test (CI = 95%) confirmed that this difference was significant between the Type II failure mode and the other two failure modes observed (Fig. 5.5).

149

Figure 5.5: Failure modes in relation to the diameter ratio of the bifurcation

(diameter ratio expressed as the percentage of B1/A1, as defined by Fig. 2.7 of this thesis) Letters above the bars mark heterogeneity in the sample groups, as determined by a GLM ANOVA and post-hoc Tukey test with 95% confidence interval.

Mean displacements of samples prior to yielding were 135.26 mm ± 15.18 SE for branch failures, 83.04 mm ± 5.08 SE for Type I failures and 37.17 mm ± 1.55 SE for Type II failures. A GLM ANOVA identified that there was a statistical difference between these three groups in terms of the extent of their displacement prior to yielding

(F2, 236 = 89.59; p < 0.001); the parent stem diameter was not a significant co-variant (F1,

236 = 0.08; p = 0.774). A post-hoc Tukey test (CI = 95%) confirmed that this difference was significant between all three failure modes, identifying that branch failures occurred after the greatest displacement and Type II failure modes after the least displacement. The mean displacement for Type II failures of normally-formed bifurcations was 43.32 mm ± 2.29 SE, whereas the mean displacement for Type II failures of bark-included bifurcations was 30.85 mm ± 1.8. Analysis of these specimens exhibiting Type II failure mode using a one-way ANOVA and post-hoc Tukey test (CI = 95%) found that this difference was significant (F1, 142 = 18.18; p < 0.001), in that bark-included

150 bifurcations broke at 71.2% of the displacement that the normally-formed bifurcations broke at.

Figure 5.6 shows typical examples of the force/displacement graphs of the rupture tests on the hazel bifurcations that suffered the Type I and the Type II failure modes in normally-formed bifurcations, a typical branch failure and the typical failure of a bifurcation with included bark at its apex. It can be seen that a long phase of plastic yielding occurs in both branch failure and in Type I failure mode of bifurcations without included bark (Fig. 5.6), with large subsequent deflections before the maximum force is reached. In contrast, in Type II failure mode, there is a sharp drop in force due to fracture after only a very short phase of yielding, while in the bifurcation with included bark, even though it is undergoing Type II failure mode, there is apparent plastic yield at a lower force and a more gradual reduction in force after failure.

Figure 5.6: Typical force/displacement graphs for specimen types

The mean yield stresses for the branches subjected to three point bending tests (σb), and the mean breaking stresses for the normally-formed bifurcations and those with

151 included bark subjected to rupture tests (σa) are shown in Figure 5.7. Bark included bifurcations were on average 24.3% weaker than ones without included bark, which were in turn 13.6% weaker than the smaller branch. A one way ANOVA identified a significant difference in bending stresses for these three groups (F2, 320 = 112.25; p <

0.001), the residuals were found to be normally distributed (AD323 = 0.402; p = 0.358) and a post-hoc Tukey test (CI = 95%) confirmed that each group’s mean yield stress was significantly different from the other groups.

Figure 5.7: Box plot of mean yield stress of branches compared with the mean breaking stresses of the normally-formed bifurcations and bifurcations with included bark. Boxes labelled with different letters are significantly different, as determined by a one- way ANOVA and post-hoc Tukey test (CI: 95%).

152

5.4.1 Effects of the Extent and Location of Included Bark

The first regression model that identified a significant relationship used a combination of the diameter ratio (t84 = 4.42; p < 0.001) and the area of the bark inclusion (t84 = 2.38; p = 0.02). The overall model fit was R2 = 0.21 and the best fit line was given by equation 5.1:

휎푎 = 69.9 − 35.2 푟 − 24.6 푎 (Eq. 5.1) where r is the diameter ratio of the two branches of the bifurcation (as a percentage with a maximum of 100%) and a is the area of bark as a percentage of the entire fracture surface (maximum value 100%) from the point of the bifurcation of the pith to the apex. The diameter ratio predicted 15.8% of the variability in the sample, the area of the bark inclusion only a further 5.3% using this model (equation 5.1). When the factor of parent stem diameter was added to this regression model, it did not significantly improve the prediction of breaking strength (t83 = 1.04; p = 0.302).

The second regression model found to be significant using the stepwise regression approach identified a stronger relationship using a combination of the diameter ratio (t84

= 4.57; p < 0.001) and width of bark inclusion (t84 = 3.0; p = 0.004). The overall model fit was R2 = 0.247 and the best fit line was given by equation 5.2:

휎푎 = 68.5 − 35.8 푟 − 7.27 푤 (Eq. 5.2) where w is the proportional width of the bark inclusion at the apex of the bifurcation when compared with the width of the parent stem (as a percentage, no maximum limit). The diameter ratio predicted 16.6% of the variability in the sample, the width of the bark inclusion a further 8.1% using this model (Equation 5.2). When the factor of parent stem diameter was added to this second regression model, again it did not significantly improve the prediction of breaking strength (t83 = 0.67; p = 0.502).

153

The mean breaking stress (σa) of normally-formed bifurcations (n = 135) was 46.9 MPa (± 0.8 SE), the mean breaking stress for bifurcations with embedded bark (n = 17) was 44.7 (± 1.79 SE), whereas the mean breaking stress for cup-shaped bark-included bifurcations (n = 57) was 37.02 (± 1.11 SE) MPa, and for those with bark at their apex (n = 30), the mean was 27.22 (± 1.23 SE) MPa. A GLM ANOVA with the parent stem diameter as a covariate (F2, 236 = 49.4; p < 0.0001) and tree number as a random variable showed that there were significant differences between these four groups, and a post- hoc Tukey test (CI = 95%) showed that both the cup-shaped bark-included bifurcations and the wide-mouthed bark inclusions had significantly different mean breaking stresses from each other and from the normally-formed bifurcations and those with embedded bark (Fig. 5.8). Parent stem diameter was not a significant covariate that affected bifurcation strength (F2, 236 < 0.01; p = 0.989), nor was tree number a significant variable.

154

Figure 5.8: Boxplot of breaking stresses of bifurcation types.

Comparison of σa between normally-formed bifurcations, bifurcations with embedded bark, cup-shaped bifurcations and bifurcations with wide-mouthed bark inclusions at their apices. Boxes labelled with different letters are significantly different, as determined by a GLM ANOVA and post-hoc Tukey test (CI: 95%).

155

5.5 Discussion

The results from this study show that there are gradations in the strength of bark- included bifurcations in young hazel plants that relate to the scale and position of the bark inclusion and their level of occlusion within the wood formed at these bifurcations. These factors were found to be independent of the size of the specimens, where this was assessed by recording the diameter of the parent stems just below the bifurcation (which varied from approximately 17 mm to 59 mm). However, there was considerable variability in the sample that remains unexplained from the simple regression models used here, which explained only a quarter of the variation in strength found in the sample bifurcations.

Firstly, it is clear that the diameter ratio of the branches has a greater influence on the strength of hazel bifurcations in static rupture tests than does the extent of the bark inclusions. In both normally-formed and bark-included bifurcations, those consisting of two branches of similar diameter are weaker and are more likely to fail by Type II failure mode than those with a lower diameter ratio. Secondly, the presence of a bark- inclusion does weaken hazel bifurcations to a similar degree as was found by Smiley (2003) in Acer rubrum and that the extent of weakening increases with the width of the bark inclusion at the apex of the bifurcation. This second finding contradicts those of Pfisterer (2003) who reported that bark-included bifurcations failed at the same bending stresses as normally-formed ones. However, there was still a large degree of variability in this sample, so accurate predictions about the strength of a bifurcation cannot be made simply from examination of this aspect of its external morphology. The variability may be mainly due to differences in the reorientation of wood grain at the apices of the bifurcations, as this provides a key strengthening component (Slater et al., 2014).

Diameter ratio can have a significant effect on the failure mode of bifurcations in trees (Gilman, 2003; Kane et al., 2008). In the case of these hazel samples, boundaries for different failure modes can be set by their diameter ratios. For the samples tested, a diameter ratio higher than 80% most frequently resulted in Type II failure mode, a

156 lower ratio than that led to most of the Type I failure modes until the ratio of 72% was reached, where branch failures started occurring and only branch failures occurred at a ratio of 55% and below. It should be noted that the bifurcations of hazel were selected to have a relatively high diameter ratio between their two branches so as to successfully investigate bifurcation failures, so consequently the incidence of branch failures was low in the test specimens.

Type I failures of bifurcations showed a greater displacement prior to yielding than did Type II failures (Fig. 5.6): this is explained by the initial stage of Type I failure, where wood at the outer edge of the bifurcation is yielding under compression until sufficient tensile stress is concentrated at the bifurcation apex to split the xylem tissues situated there. Branch failures, using this form of rupture test, displayed a much extended displacement during testing, as there was a great deal of yielding under compression on the underside of the branch prior to any break of fibres under tension on the upper side (van Casteren and Ennos, 2010). The force/displacement graphs often showed a different behaviour where a bark inclusion was present, with a longer phase of plastic deformation as the bifurcation ‘crept apart’ rather than exhibiting a distinct breaking point – however, for those exhibiting Type II failure mode, the peak force was reached with less displacement in bark-included bifurcations than with normally-formed bifurcations. The absence of interlocking wood grain at the apex of these bark-included bifurcations is an obvious reason for this difference in mechanical behaviour (Slater et al., 2014). These results corroborate the findings of Pfisterer (2003), who also found differences in behaviour in hazel bifurcations with and without bark inclusions, but who did not differentiate between Type I and Type II failure modes.

The higher tensile strength of bifurcations with a higher diameter difference in their branches is ascribed by Gilman (2003) to the level of occlusion of the smaller branch into the other stem. However, it may be more appropriate to think about this relationship in terms of the loading caused by the different bending behaviours of the branches in the wind (Fig. 5.9).

157

Figure 5.9: Suggested contrast in bending behaviour between a low diameter ratio bifurcation and a high diameter ratio bifurcation (xz plane views)

From preliminary research work we have undertaken using accelerometers attached just above bifurcations in hazel, the frequency and extent of oscillations separating apart a smaller diameter branch and a larger diameter branch where their bases are conjoined at a bifurcation will both be greater than when two branches of equal diameter are bent in a wind of the same force. As a consequence of experiencing higher strain levels more regularly at its apex through this different bending behaviour, lower diameter ratio bifurcations are likely to develop a higher level of modification of their tissues to adequately resist those forces (Metzger, 1893; Jaffe and Forbes, 1993; Telewski, 1995). In contrast, the bifurcation with included bark is a structure where little to no strain is regularly experienced at its apex, so no substantial resources are committed by the tree to reinforcing it.

Bifurcations with bark inclusions were on average only three-quarters the strength of the normally-formed specimens, but there was a wide range of peak stress values, with some bark-included samples experiencing branch failure rather than splitting at the bifurcation itself and other bark-included bifurcations having less than 40% of the yield strength of the smaller branch.

158

A simple analysis of the strength of the bifurcations with included bark and their morphology provided two useful insights. Firstly, it can be concluded that small areas of embedded bark do not give rise to a significant difference in bifurcation strength. Secondly, cup-shaped bifurcations in hazel were significantly stronger than those that had bark at their apex. The conclusion from these findings is that the main reason why the strength of bifurcations with included bark was found to be so variable in the tested specimens was that the areas of included bark in the samples were at different stages of occlusion at the bifurcation apex: a higher level of occlusion of the bark inclusion resulted in an increase in the bifurcation’s strength. Thus the cup-shaped bifurcations tested in this study represented different stages of repair of the structural flaw that was caused by the initial inclusion of bark into those junctions.

From our testing of these hazel specimens, we can provide an interpretation of the mechanical performance of bifurcations with included bark in trees; however, it is very important to recognise the limitation of this study, in that young bifurcations of only one species that contained solely juvenile wood were tested, and the mechanical behaviour of mature bifurcations in different woody species may well vary from what we found in our samples.

Wide-angled bifurcations which are U-shaped at their apex and without bark inclusions and bifurcations with embedded bark should both be considered adequate structures as there should be interlocking wood grain present at the bifurcation apex. Where a significant width of included bark is found at the apex of the bifurcation, this indicates a significantly weaker bifurcation and a tree assessor should evaluate the proportional width of this bark in relation to the overall width of the join perpendicular to the plane of the bifurcation. They should also take into account the extent of adaptive growth at each side of the bifurcation, the extent of occlusion of the bark inclusion by the formation of a cup-shaped bifurcation and, most critically, whether the level of wind exposure of the bifurcation has been heightened by recent site changes or pruning works. The rapid formation of additional xylem that lies at either side of a bifurcation

159

(often indicated by a change in bark texture) may be an indication of instability of that bifurcation (Mattheck and Breloer, 1994).

Features to survey for in bark-included bifurcations, based on this study using hazel specimens, are identified in Figure 5.10.

Figure 5.10: Weaker and stronger forms of bifurcations with included bark.

Left hand images show xz plane, right hand images show yz plane through the centre of the bifurcation. See Fig. 1.1 for axes definitions.

A: Wide-mouthed bark-inclusion positioned at the apex of the bifurcation, with acutely pointed reaction growth forming below the inclusion.

160

B: A cup-shaped bifurcation with two rounded areas of abnormal growth at the apex of the bifurcation that act to resist bending stresses.

It would seem that a bark-included bifurcation’s notoriety as a defect in trees comes from the risk of this structure being exposed to a wind event or other loading event that causes the two arising branches to oscillate or move apart in a way that has not frequently occurred during the bifurcation’s prior development. This problem can be accentuated by arboricultural practices like crown thinning, felling of adjacent trees or the transplanting of trees into new locations, where these practices would lead to abrupt changes in the level of exposure to which the bifurcation is not sufficiently adapted (Wood, 1995).

Studies of the strength of bifurcations with included bark in trees should be taken further. As in this study we tested juvenile wood in only one species, a similar study using mature bifurcations in a range of species would assist in determining their mechanical behaviour. In addition, a better understanding of the forces affecting the modulus of rupture of these bifurcations may come from using finite element analysis to assess stress concentration levels at the apices of such bifurcations. Further study should also determine how frequently and under what particular wind conditions such damaging oscillations occur to bifurcations with included bark. It would also be informative to investigate the movement behaviour of normally-formed bifurcations during dynamic wind loading and to determine to what extent these bifurcations develop their morphology and wood properties in relation to the dynamic forces that act upon them.

161

5.6 References

Abramoff M D, Magalhaes P J and Ram S J (2004) Image Processing with ImageJ; Biophotonics International 11 (7), 36-42.

Gilman E (2003) Branch-to-stem diameter ratio affects strength of attachment; Journal of Arboriculture 29 (5), 291-294.

Gilman E (2011) An illustrated guide to pruning; 3rd edition; Independence Kentucky US, Cengage Learning.

Harris R W, Clark J R and Matheny N P (2004) Arboriculture: Integrated management of landscape trees, shrubs and vines; p. 38, 4th edition; New Jersey US, Prentice- Hall.

Helliwell R (2004) A discussion of the failure of weak forks; Arboricultural Journal, 27, 245-249.

Jaffe M J and Forbes S (1993) Thigmomorphogenesis: the effect of mechanical perturbation on plants; Plant Growth Regulation 12, 313-324.

Kane B, Farrell R, Zedaker S M, Loferski J R and Smith D W (2008) Failure mode and prediction of the strength of branch attachments; Arboriculture and Urban Forestry 34 (5), 308-316.

Lonsdale D (1999) Principles of Tree Hazard Assessment and Management; 1st edition; London, TSO.

Lonsdale D (2000) Hazards from trees: A general guide; Forestry Commission Practical Guide series; Edinburgh, Forestry Commission.

Matheny N P and Clark J R (1994) Evaluation of Hazard Trees in Urban Areas; 2nd edition; Illinois US, International Society of Arboriculture.

Mattheck C and Breloer H (1994) The body language of trees: A handbook for failure analysis; London, TSO.

Metzger K (1893) Der wind als massgebender factor für das waschstum der bäume; Mündener Forstliche Hefte 3, 35-86.

162

Shigo A L (1989) Tree pruning: A worldwide photo guide; pp. 78-83; Durham New Hampshire US, Shigo and Trees, Associates.

Slater D and Ennos A R (2013) Determining the mechanical properties of hazel forks by testing their component parts; Trees: Structure and Function 27 (6), 1515-1524.

Slater D, Bradley R S, Withers P J and Ennos A R (2014) The anatomy and grain pattern in forks of hazel (Corylus avellana L.) and other tree species; Trees: Structure and Function 28 (5), 1437-1448.

Smiley E T (2003) Does included bark reduce the strength of co-dominant stems?; Journal of Arboriculture 29 (2), 104-106.

Telewski F W (1995) Wind-induced physiological and developmental responses in trees; pp. 237-263; In: Wind and Trees, Coutts M. P. and Grace J. eds.; Cambridge England, CUP. van Casteren A and Ennos A R (2010) Transverse stresses and modes of failure in tree branches and other beams; Proceedings of the Royal Society B: Biological Sciences 277 (1685), 1253-1258.

Wood C J (1995) Understanding wind forces on trees; pp. 133-163; In: Wind and Trees, Coutts M P and Grace J eds.; Cambridge England, CUP.

163

This page intentionally left blank

164

Chapter 6

An assessment of the remodelling of bifurcations in hazel (Corylus avellana L.) in response to bracing, drilling and splitting

Authors Slater D and Ennos A R

Status of Associated Paper Submitted June 2015

Journal Journal of Arboriculture and Urban Forestry

165

6.1 Chapter Abstract

The ability of trees to remodel their woody structure after injury or strain to outer tissues greatly assists in their survival; however, this remodelling process is complex because it is influenced by many factors. The junctions in tree branches are an interesting case, as included bark occurs frequently in them and the factors that result in remodelling around such a flaw have not been sufficiently investigated.

In this study, 100 normally-formed bifurcations in semi-mature hazel trees (Corylus avellana L.) were artificially modified by being braced, drilled through the apex, or split, and left to grow in-situ. Two further groups of bifurcations, 120 normally-formed bifurcations and 70 bark-included bifurcations, were identified as controls. After two to four years, bifurcations were harvested and underwent rupture tests to determine their bending strength. Analysis of the data showed that the braced bifurcations over three growing seasons developed adverse taper in their branches and they had only 70.5% of the bending strength of the normally-formed bifurcations. Bifurcations with the central 20% of the xylem drilled out of them were capable of recovering fully from this defect over four years; split bifurcations were found to be highly vulnerable to failure during wind-loading events with 80% of them mechanically failing rather than remodelling around this defect.

This study concludes that a bifurcation may be considered compromised in its strength if the apex is compromised, but that semi-mature bifurcations in hazel do exhibit a good ability to remodel themselves after injury. Thigmomorphogenesis is considered a key factor in this remodelling process.

166

6.2 Introduction

For mechanical strengthening, plants are known to undergo the process of thigmomorphogensis, whereby plant growth adapts in response to strains experienced by the plant’s tissues (Jaffe and Forbes, 1993; Coutand, 2010). Mechanosensing and subsequent adaption of plant growth is well-reported for plant height and form, the modification of the shapes of leaves, peduncles, petioles and the selective thickening of the branches and stems of plants (Whitehead, 1963; Jaffe, 1973; Grace, 1977; Biro et al., 1980; Braam and Davis, 1990; Farnsworth and Niklas, 1995; Pruyn et al., 2000; Telewski, 2006), and it can be surmised from the reporting of this effect in many studies that the majority of plant structures are likely to have this ability to respond to strain, including the junctions of the aerial parts of woody plants.

Thigmomorphogensis is a mechanism for the remodelling of plant structures triggered by the strain experienced by meristematic cells (Philipson et al., 1971). In trees and other woody plants, thigmomorphogenesis can be a local phenomenon to parts of their structure, with secondary thickening occurring fastest where the highest mechanical strains are experienced (Mattheck and Linnard, 1998). It is important, however, to note that remodelling within woody plants may be for a range of functions and that mechanical strain is only one potential influence upon how a plant’s structure develops. In woody plants, sapwood serves a range of functions (Gartner, 1995), not solely the structural support of the plant’s stems and branches, and remodelling responses to a defect formed in the sapwood of a woody plant are potentially complex.

Junctions in the aerial parts of trees are considered to be potential failure points by arboriculturists (Shigo, 1991; Lonsdale, 1999), although scientific studies of the strength of such junctions have been restricted to static testing for practical reasons (Gilman, 2003; Kane et al., 2008; Slater and Ennos, 2013). A greater understanding of the biomechanical behaviour of such junctions and their ability to remodel around a defect would assist in tree management and the prediction of tree failures.

167

An anatomical model for junctions in trees has been outlined by Slater et al. (2014) based upon visual observation of the grain patterns found at junctions of 20 tree and shrub species. This model was supported by CT scanning of bifurcations in common hazel (Corylus avellana L.) to observe the orientation of vessels, rays and fibres at the bifurcation apex. This anatomical model emphasises the importance of the xylem lying centrally in the axil of the bifurcation and under the branch bark ridge as the main contributor to the strength of bifurcations, with the xylem tissues in this location typically being denser and exhibiting fewer vessels of a smaller diameter and shorter length when compared with adjacent xylem in the stem (Slater et al., 2014).

Slater et al. (2014) also describe how the wood grain pattern formed at the bifurcation apex results in some degree of interlocking of the grain such that wood fibres need to be stretched axially or pulled out of the tissue matrix along their length in order to break the bifurcation apart (Fig. 6.1). In mature limbs of many temperate tree species, whirled grain can be found at the apex of junctions (Lev-Yadun and Aloni, 1990) as a subsequent development of this initial interlocking pattern (Fig. 6.1b).

168

A B

Figure 6.1: Wood grain patterns at the apices of bifurcations

A: Interlocking wood grain pattern at the apex of a junction of common ash (Fraxinus excelsior L.), as exposed by de-barking (xy plane view);

169

B: Wood grain pattern at the apex of a bifurcation of common oak (Quercus robur L.) incorporating whirled grain (xz plane view);

C: Diagrammatic representation of interlocking wood grain in a normally-formed bifurcation (see in the xz plane view) in a woody plant, based upon the anatomical model of Slater et al. (2014) with inset displaying a basic interlocking pattern of wood grain incorporating whirled grain in the xy plane.

It is a common occurrence, however, that bark gets to be included in such bifurcations during their development. These bark-included bifurcations are weaker under static loading than normally-formed bifurcations (Kane et al., 2008; Slater and Ennos, 2015). In addition, if the apex of a bifurcation consists of bark, then that bark could act as a barrier to the future development of a normally-formed connection consisting of this denser tortuous sapwood.

In this study, we investigated the ability of bifurcations in hazel trees to remodel themselves around artificially-induced defects. Hazel (Corylus avellana L.) was selected as the test subject for this study because the authors have carried out a series of complementary investigations into the anatomy and biomechanical properties of bifurcations in this species and test material can be sourced sustainably through coppicing.

For this study, we investigated the loss of strength to these bifurcations caused by man- made defects, comparing them to both normally-formed and bark-included bifurcations grown in the same location. The three artificial defects studied were fixed-rod bracing of the two branches arising from bifurcations, the drilling out of the centrally-placed xylem at the apex of bifurcations and the splitting of the apex of the bifurcation by pulling the two branches apart from each other. It was reasoned that the braced bifurcations, in the absence of them experiencing significant mechanical strains, would become weaker over time, and that the drilled-out and split bifurcations may remodel

170 themselves around their artificially-induced defects. It was hoped that this would provide evidence that mechanical loading was a key factor in the development of strength in these bifurcations, as well as identifying the typical pattern of anatomical remodelling that occurred around these defect types.

6.3 Materials and Methods

6.3.1 Selection of hazel bifurcations

A wind-exposed semi-mature shelterbelt consisting of a mix of broadleaves species which contained semi-mature hazel trees was selected for this experiment. The planted area was on the southern boundary of the campus of Myerscough College, Lancashire, England – grid reference: SD497399 (Easting 349711, Northing 439982). The trees in this shelterbelt were planted as 3-year-old bare-rooted stock in 2004, making the hazel trees 13 years of age by the end of this study. All the bifurcations used for this experiment were formed less than two metres above ground level; this facilitated their modification by bracing, drilling or splitting and ensured that the age and diameters of these bifurcations were similar.

Bifurcation selection was biased towards choosing bifurcations with a high diameter ratio, as expressed by the percentage difference between the diameters of the thinner branch to the thicker branch in the plane of the bifurcation. Bifurcations were also selected so that both branches and the parent stem were ascending, all of them forming a relatively upright Y-shape, with no other significant branching to be found above or below 200 mm of the bifurcation apex. No more than three bifurcations were selected in the crown of any one hazel tree, which resulted in a random scattering of sample collecting along the 450 metre length of the shelterbelt.

171

6.3.2 Modifications to the hazel bifurcations

In December 2010 an initial experiment was devised whereby 50 hazel bifurcations had the centre of their apex drilled out and were left to develop over two to four years (Fig. 6.2b). The drill bit size was selected for each bifurcation so that 20% of the width of the apical tissues were removed (Table 6.1) based on a measurement of the parent stem perpendicular to the bifurcation taken just below the bulge formed by the branch bark ridge.

This drilling scheme matches that carried out by Slater and Ennos (2013) on bifurcations of hazel that were tested to determine the contribution of the centrally-placed xylem to the bending strength of such bifurcations. However, in this experiment, these drilled bifurcations were left in-situ, remaining as a component of the trees’ crowns, to assess whether and how the bifurcations would re-model around the induced defect of the drill hole. Each drill hole made was filled with silicon sealant to allow for easy identification of these modified bifurcations when they were mechanically tested, and each was sprayed with a standard fluorescent forestry marking paint so that they could be identified and harvested at a later date. In addition, 50 normally-formed bifurcations were also selected and spray-painted within the same wooded area, to act as a control of the bending strength of unmodified bifurcations.

172

Table 6.1: Determination of drill size for modifying drilled hazel bifurcations

Drill sizes were selected to modify 50 hazel bifurcations based upon the diameter of the parent stem measured just below the termination of the branch bark ridge and perpendicular to the bifurcation.

Diameter of parent Drill size used upon stem (mm) bifurcation perpendicular to bifurcation Up to 22.5 4 mm 22.5 – 27.49 5 mm 27.5 – 32.49 6 mm 32.5 – 37.49 7 mm 37.5 – 42.49 8 mm 42.5 – 47.49 9 mm 47.5 + 10 mm

In December 2011 the replicate number and scope of this experiment was expanded. A total of 50 further hazel bifurcations were artificially altered; 25 bifurcations had a 3 mm diameter steel rod fixed by bolts and washers fitted through the centre of both branches approximately 70 mm above the bifurcation to conjoin these branches (Fig. 6.2a); a further 25 bifurcations were carefully split by hand so that a crack (approximately half the length of the branch bark ridge) was induced at the bifurcation apex by bending the two branches above the bifurcation away from each other (Fig. 6.2c). The braced bifurcations were typically of a larger size (as measured by the parent stem diameter) than the mean of all the bifurcations at the start of the experiment, because of the need for the two branches of the bifurcation to be thick enough to accept the bracing rod and remain intact.

It was also determined at this time to add a further 70 normally-formed bifurcations to the original 50 normally-formed bifurcations, and also to identify in this shelterbelt 70 bark-included bifurcations for rupture testing. All additions were also marked with colour-coded fluorescent forest marking paint to aid their re-identification upon harvesting.

173

It was considered that a greater number of normally-formed bifurcations were required, as some would be subsequently drilled immediately prior to rupture testing to compare with those that may have remodelled themselves from their earlier drilling and subsequent regrowth. By increasing the replicates within the normally-formed group it was also hoped to reduce the variability in the mean breaking stress in that group, providing a better comparison between treatment types. The bark-included bifurcations were added as a group type to compare with the extent of any strength loss in the artificially modified bifurcations and thus give additional context to our results.

A B C

Figure 6.2: Artificially-modified bifurcations left to grow in-situ for two to four years

A: Diagram of rod-bracing created in 25 hazel bifurcations (xz plane view);

B: Diagram of drill hole created in 50 hazel bifurcations (xz plane view);

C: Diagram of split created in 25 hazel bifurcations (xz plane view).

A summary of the different types of bifurcation investigated is provided in table 6.2:

174

Table 6.2: Summary of bifurcation specimens tested

This table details the bifurcation types tested, numbers of replicates for each type, year of modification and associated growing seasons prior to rupture testing

Name of Description No. of Year of Growing bifurcation replic- artificial seasons type ates modification between modification and testing Bark- Naturally-occurring bifurcations 70 Not modified N/A included with bark found to be incorporated within the apex of the bifurcation (Fig. 6.3) Braced Normally-formed bifurcations 25 2011 3 modified by the conjoining of the two branches above the bifurcation with a 3 mm steel rod fitted through both branches, with a 7 mm washer and nut fitted at each end of the rod. These were left to grow within the tree’s crown for three years prior to testing (Fig. 6.2a) Newly- Normally-formed bifurcations 60 2015 0 drilled drilled at their apices using a drill-size as defined in table 6.1, immediately prior to rupture testing (Fig. 6.2b) Normally- Naturally-occurring bifurcations 60 Not modified N/A formed with no flaws observed in morphology Pre-drilled Normally-formed bifurcations 50 2010 2 and 4 modified by drilling at their apices using a drill-size as defined in table 6.1, and left to grow within the tree’s crown for two or four years prior to testing (Fig. 6.2b) Pre-split Normally-formed bifurcations 25 2011 3 modified by carefully splitting the apex by hand, by bending away from each other the two arising branches. These were left to grow within the tree’s crown for three years prior to testing (Fig. 6.2c)

175

6.3.3 Observations

Prior to harvesting of the bifurcations in 2013 and 2015, basic observations were recorded of the condition and morphology of the selected bifurcations, including any swellings associated with the artificially-modified bifurcations and also whether bifurcations had failed in-situ, prior to harvesting, within the shelterbelt.

6.3.4 Rupture testing

In January 2013, after two growing seasons, twenty-one of the bifurcations that were drilled in December 2010 and fifty of the normally-formed bifurcations were cut from the trees in order to carry out rupture testing. The bifurcations were cut so that there was a minimum length of 220 mm of both branches and at least twice the length of the branch bark ridge of the parent stem on each bifurcation. The bifurcations were wrapped in individual plastic bags immediately after cutting to minimise sap loss, and were stored in a cold store kept at 2 °C prior to rupture testing.

Twenty-five of the normally-formed bifurcations had the centre of their apex drilled immediately prior to rupture testing, using the drill sizes as defined in table 6.1.

Rupture testing was carried out using the same method and calculations as stated in section 2.3.2 of Chapter 2 of this thesis.

After this testing, careful observation was made of the fracture surfaces of all bifurcations, in relation to their morphology and appearance.

To assist with comparing the relative strength of the bifurcations, three-point bending tests of the smaller diameter branch of the bifurcation were carried out, testing the yield

176 strength of the middle of each branch whose structure had not been compromised by the rupture testing. This was carried out in a very similar fashion to the three point bending test reported in Section 2.3.4 of Chapter 2. However, the span for these branches was set at 215 mm for branches up to 20 mm in diameter and 275 mm for branches up to 23 mm in diameter. Due to limitations of the testing machine in terms of the span length that could be used and the maximum load (900 kN) that could be applied, branches with a mid-diameter of over 23 mm could not be tested to their yield point.

In February 2015, after four growing seasons for the original set of drilled bifurcations and three growing seasons for the braced and split bifurcations, all the remaining bifurcations were cut from the hazel trees and subjected to the same method of bagging, storage and rupture testing. A different InstronTM testing machine (Model 3344) had to be used for this second set of rupture tests, as the original UTM had suffered a breakdown in the two year period between these two tests. The parameters of the rupture tests were the same in nearly all respects; however, the rate of displacement was increased to 50 mm min-1, due to the large number of bifurcations that had to be processed.

The bifurcations with bark included within them were classified after testing in terms of the relative occlusion of the bark into the bifurcation, giving rise to three types of bark inclusion: embedded, cup-shaped and wide-mouthed (Fig. 6.3). This classification of bark-inclusions was used by Slater and Ennos (2015), who identified significant differences in breaking stress between these three morphological types of bark-included bifurcation in hazel. For each braced bifurcation tested, bolt cutters were used to cut the steel rod that conjoined their two branches in two places prior to rupture testing.

177

Figure 6.3: Three types of bark-included bifurcation found in hazel trees (yz plane view of internal structure)

Diagrams and images defining three morphological types of bark-included junctions in hazel, based on observations of the fracture surfaces of such bifurcations perpendicular to the plane of the bifurcation.

A: Embedded bark is surrounded entirely by xylem, the bark having been occluded into the junction.

B: A cup-shaped bark inclusion has sapwood formed around included bark which lies at the centre of the join – there is sapwood at the apex of the bifurcation rather than bark.

C: A wide-mouthed bark inclusion has a substantial width of included bark at the apex of the bifurcation, situated above any connecting sapwood.

6.3.5 Basic density testing

To supplement the data on the bending strength of the bifurcations, basic density tests were carried out on small samples of the xylem excised from the apices, from the side of

178 the bifurcations adjacent to their apices and from the parent stems of all the bifurcations tested in 2015. Both braced and normally-formed bifurcations could provide xylem from all three locations, whereas the drilled or split bifurcations and those with included bark could only supply xylem samples from the side of the bifurcation apex and the stem (Fig.s 6.2b, 6.2c and 6.3). Samples were cut using a pull saw and billhook blade, their fresh weight taken and their volume calculated by measuring the displacement weight when each sample was immersed in distilled water on a weighing scales. The mean volume of these samples for this basic density test was 444.4 mm3 ± 8.4 SE.

The samples were then oven dried for 96 hours at 60 °C and their dry weight recorded. Given the small size of the samples, this length of drying time was considered sufficient. Basic density was calculated by dividing the dry weight of each sample by the greenwood volume of the sample as measured by water displacement when freshly cut (Hughes, 2005).

6.3.6 Statistical analysis

All statistical tests were carried out using MiniTab® version 17.

For comparisons between bifurcation types, and for sub-sets within each bifurcation type, General Linear Model (GLM) ANOVAs were used to find differences in mean breaking stress, with the parent stem diameter (PS1) and the diameter ratio of the bifurcations as covariates where appropriate, in combination with a post-hoc Tukey test at a 5% confidence level. Residuals were assessed for the normality of their distribution using Anderson-Darling and Komogorov-Smirnov tests.

To determine if the branches of the braced bifurcations exhibited adverse taper a paired t-test comparing the diameter of the branches above the fitted steel brace and at the apex of the bifurcation was carried out.

179

To determine differences between the basic density of wood samples extracted from the apices and sides of bifurcations and the adjacent stem wood, a GLM ANOVA with sample volume as a covariate was used, in combination with a post-hoc Tukey test at a

5% confidence level. Residuals were assessed for the normality of their distribution using a Kolmogorov-Smirnov test.

6.4 Results

6.4.1 Specimen losses and mean specimen dimensions

Over the four years of this experiment, a number of the selected bifurcations (20 out of the total of 290 bifurcations) were lost prior to the mechanical testing. Fourteen of the bifurcations were removed from this study in 2012 as a length of the shelterbelt’s edge was accidently flailed when a neighbouring hedgerow was pruned; the remaining six bifurcations which were lost could not be re-found in 2015 due to the bio-degradability of the forestry marker paint used, as it was concluded that the paint had weathered away.

In addition, two types of the modified bifurcations suffered replicate losses for other reasons. Seven of the braced bifurcations grew over the three years to a size that was too large for the testing machine to break them (having started at the upper end of the parent stem diameter sizes chosen), which reduced this group’s size to 14 testable replicates. Twenty of the twenty-five split bifurcations suffered wind-induced mechanical failure over the three years they were in-situ. For this latter group, observations were subsequently made of these failures and of the morphology of the five bifurcations that remained.

The mean parent stem diameter (PS1) for the remaining 243 bifurcations was 30.35 mm

± 0.37 SE, the mean diameter of the smaller branch of the bifurcation just above its point

180 of attachment (b1) was 21.23 mm ± 0.26 SE and the mean diameter ratio for these bifurcations was 80.98% ± 0.75% SE.

6.4.2 Observations of bifurcations prior to testing

Bark-included bifurcations

Ten of the normally-formed bifurcations were found to contain embedded bark, so the data generated from these 10 bifurcations was moved to the bark-included group for analysis. To compensate for the reduction in the group size of the normally-formed bifurcations, the number of replicates allotted to the newly-drilled group was reduced to obtain a roughly equal number of replicates within these two groups. The categorisation of the remaining 58 bark-included bifurcations resulted in 36 being identified as wide- mouthed bark inclusions and 22 identified as cup-shaped bark inclusions (Fig. 6.3).

Drilled bifurcations

Observations of the pre-drilled bifurcations showed a range of remodelling responses to the initial drilling of the hole at their apices. In general, despite some initial dysfunction caused to adjacent tissues after drilling, additional sapwood was added around the induced defect (Fig 6.4a). Three of these bifurcations had fully embedded the silicon, surrounding it with new sapwood after four years of growth, and many more had started to cover over the top of the drill hole. In the majority of these bifurcations a general swelling in the location of the branch bark ridge was evident. For thirteen of these bifurcations, however, the drill-hole had initiated the development of included bark at the apex or a larger extent of associated dysfunction around the original drill-hole had resulted in a failure to occlude the drill-hole. No significant volume of decayed xylem was found in any of these bifurcations. This difference in development allowed the pre- drilled bifurcations to be classified into three sub-categories to match the bark-included ones; that the silicon in the drill-hole had become embedded, that the bifurcation was forming a cup-shape around the drill-hole, or that the drill-hole was still wide open at the bifurcation’s apex.

181

Split bifurcations

For the pre-split bifurcations, the high number of replicate losses through wind-induced failure was investigated. It was observed that for the five split bifurcations that had persisted for three years and been subjected to rupture testing, all had split further down the stem since the initial splitting was carried out in 2013, and the split had been halted either by encountering a substantial knot in the parent stem (for four of them) or a substantial bend in the parent stem (in one case only). The twenty bifurcations that had mechanically failed were all considered to have failed due to natural wind-induced movement and subsequent propagation of the original split down the parent stem, with the split at some point deviating to the edge of the stem, causing one branch to fall away from the tree.

The propagation of these splits and the failure of so many of this type of bifurcation meant that this group had to be excluded from any statistical analysis relating to the breaking stresses of the bifurcations. An image of the typical surviving pre-split bifurcation is provided in figure 6.4b.

Braced bifurcations

It was evident that the installation of the steel rod in 2011 had resulted in abnormal swelling of the branches at the point of drilling the 3 mm hole needed to fit the brace (Fig. 6.4c). All braced bifurcations exhibited some level of occlusion of the rod, nuts and washers and some had wholly occluded the nuts and washers. Measurements were taken of the diameter of the branches just above each braced bifurcation’s apex, as with all other bifurcations, but also the branch diameters were measured at the point above the bracing rod and its associated swelling, to determine if the bracing had resulted in the branches developing adverse taper.

182

A B C

Figure 6.4: Features of the development of the artificially-modified bifurcations

A: Fracture surface (yz plane view) of a pre-drilled bifurcation after two growing seasons, showing the silicon inserted into the initial drill-hole, dysfunction induced in the sapwood around the drill hole (discoloured area) and the remodelling of the sapwood to form a cup-shaped union;

B: Typical deformation of a pre-split bifurcation (xz plane view), where the crack had subsequently propagated to a knot in the parent stem and then been arrested;

C: Typical deformation of the branches of a braced bifurcation (xz plane view) around the implanted steel rod, after three years of growth, showing adverse taper in the smaller branch.

6.4.3 Rupture testing

All bifurcation types

Twelve of the tested bifurcations suffered branch failure, rather than failing at the bifurcation itself. To assess bifurcation strength, these twelve specimens that suffered branch failure were excluded from this part of the data analysis.

A statistical comparison was made of the mean breaking stresses of the main five bifurcation types and the yield stress of the smaller branches, using a GLM ANOVA

183 and post-hoc Tukey test after a natural log transformation of the data. It was found that 2 there were significant differences between groups (F5, 293 = 61.54; R = 51.23%; p < 0.001); branches yielded at the highest mean stress, and the bark-included and newly- drilled bifurcations broke at the lowest mean stress (Fig. 6.5). Residuals of the transformed data satisfied the Anderson-Darling test for normality (AD299 = 0.695; p = 0.07).

Figure 6.5: Boxplot of mean breaking stresses for the main bifurcation types and mean yield stress for branches tested

The data for the bifurcations excluded any sample which exhibited a branch failure rather than breaking at the bifurcation apex. The pre-split type is not included as its replicate number was too small to be statistically analysed (n = 2). Letters above boxes identify significant differences between groups by using a GLM ANOVA and post-hoc Tukey test at a 5% confidence limit.

184

Bark-included bifurcations

A comparison between the three sub-types of the bark-included bifurcations in relation to their mean breaking stress is provided in figure 6.6. A GLM ANOVA (F2,61 = 10.44; R2 = 38.93%; p < 0.001) with diameter ratio and the diameter of the parent stem as covariates found that there were significant differences between these groups and a post-hoc Tukey test identified that the wide-mouthed bark-inclusions broke apart at a lower tensile stress than the other two types. The diameter ratio was a significant covariate (p < 0.001) in that a higher diameter ratio resulted in a lower breaking stress, but the parent stem diameter was not a significant covariate (p = 0.381). Residuals satisfied the Anderson-Darling test for normality (AD66 = 0.359; p = 0.441).

Figure 6.6: Boxplot of mean breaking stresses of the three types of bark-included bifurcation tested (cup-shaped, embedded and wide-mouthed bark inclusions).

Letters above boxes identify significant differences between groups by using a GLM ANOVA and post-hoc Tukey test at a 5% confidence limit.

185

Drilled and pre-drilled bifurcations

The mean breaking stress of the newly-drilled bifurcations was 31.45 MPa ± 1.01 SE, whereas for the pre-drilled bifurcations that were allowed to grow for two growing seasons it was 32.85 MPa ± 1.85 SE, and for the pre-drilled bifurcations that remodelled around the drill holes for four growing seasons it was 36.64 MPa ± 2.35 SE.

It was observed that growth responses in the pre-drilled bifurcations were mixed, with some bifurcations suffering more xylem and cambial dysfunction than others, and some growing rapidly around the drill-hole with little to no dysfunction evident. As a consequence, the pre-drilled bifurcations were placed into three groups corresponding to the classification of the bark-included group in this study: three of the pre-drilled bifurcations had occluded the drill-hole and were categorised as ‘embedded’, 20 more bifurcations had partly occluded the drill-hole and were categorised as ‘cup-shaped’ and the remaining 13 bifurcations in the pre-drilled group exhibited no evidence of occlusion and had suffered dieback related to the drill-hole; these were categorised as ‘wide-mouthed’. A statistical comparison between these groups and the newly-drilled bifurcations is provided in figure 6.7.

186

Figure 6.7: Boxplot of mean breaking stresses of the three types of pre-drilled bifurcation tested, categorised as cup-shaped around the drill hole (cup), embedding the drill hole (embedded), a newly-drilled specimen or wide-mouthed around the drill hole (wide).

Letters above boxes identify significant differences between groups through using a 2 GLM ANOVA and post-hoc Tukey test at a 5% confidence limit (F 3, 88 = 5.70; R =

42.34%; p = 0.001). Residuals satisfied the Anderson-Darling test for normality (AD94= 0.698; p = 0.066). The diameter ratio was a significant covariate (p < 0.001) and the parent stem diameter was not significant (p = 0.909).

Bifurcations in the pre-drilled group that showed the most regrowth around the drill-hole (embedded or cup-shaped) had a higher strength than those where regrowth had not occurred (wide-mouthed), which had similar strength to the newly-drilled bifurcations (Fig. 6.7).

187

Braced bifurcations

For the braced bifurcations the mean diameter of the branches arising from the bifurcations was 24.59 mm ± 1.01 SE, but the mean diameter of the branches just above the bracing rod was 28.40 mm ± 1.01 SE. A paired t-test identified that a significant adverse taper had developed in the branches (T1, 13 = 4.75; p < 0.001). Data was normally distributed (AD28 = 0.643; p = 0.084). This adverse branch taper was not exhibited by any other bifurcation type.

Further to this observation, the breaking stress of these braced bifurcations was additionally calculated based on the section modulus of the smaller branch just above the steel rod brace and its associated swelling. This further assessment takes into account the size of the branch that would actually have to be borne by the bifurcation if the brace was not in place. The mean breaking stress of the braced bifurcations using the section modulus of the smaller branch at the bifurcation apex was 40.06 MPa ± 2.08 SE (Fig. 6.5), but when taking into account the section modulus of that same branch above the brace, the equivalent breaking stress reduced to only 30.47 MPa ± 1.44 SE. These two mean breaking stresses were compared with the mean breaking stresses of the normally-formed bifurcations (Fig. 6.8).

188

Figure 6.8: Comparison between the breaking stresses for normally-formed and braced bifurcations

This comparison uses the two estimates of the breaking stresses of the braced bifurcations, taking into account the section modulus of the smaller branch either above or below the bracing steel rod. Letters above boxes identify significant differences between groups by using a GLM ANOVA and post-hoc Dunnett test at a 5% confidence 2 limit (F 3, 77 = 19.32; R = 35.59%; p < 0.001). Residuals satisfied the Komogorov-

Smirnov test for normality (KS83= 0.093; p = 0.078). The diameter ratio was a significant covariate (p = 0.017), with an increasing diameter ratio resulting in a lowering of breaking stress; the parent stem diameter was not found to be a significant factor (p = 0.631).

6.4.4 Wood basic density at hazel bifurcations

The results of the basic density testing are provided in table 6.3. Statistical analysis of the data found that all the samples excised from under the branch bark ridge were significantly denser than those excised from the adjacent stem. Overall, samples from

189 the side of the bifurcation apex (n = 161) were 27.1% denser than the samples from the stem and the highest mean basic density was found at the apex of the normally-formed bifurcations.

Table 6.3: Basic density of samples taken from different bifurcation types tested, by location.

Letters (A, AB, B and C) below the mean in each entry identify differences between these means across bifurcation type and location of xylem extraction, as identified by a GLM ANOVA with the sample volume as a covariate and post-hoc Tukey test at a 5% 2 confidence limit (F13,357 = 93.94; R = 78.09%; p < 0.001). Residuals from this

ANOVA satisfied the Kolmogorov-Smirnov test for normality (KS372= 0.046; p = 0.059). Sample volume was not a significant factor in the differences found in wood density between groups (p = 0.509).

Basic density of extracted sample (kg m-3), by location Bifurcation type Apex Side Stem Normally-formed 644.9 ± 4.3 SE 632.8 ± 5.8 SE 493.0 ± 6.9 SE A A C n = 36 n = 36 n = 36 Bark Included N/A 628.6 ± 5.3 SE 490.9 ± 5.2 SE AB C n = 57 n = 57 Newly-drilled N/A 628.9 ± 5.8 SE 488.7 ± 7.6 SE AB C n = 30 n = 30 Pre-drilled N/A 595.0 ± 4.7 SE 494.8 ± 12.2 SE B C n = 19 n = 19 Pre-split N/A 614.6 ± 5.7 SE 503.2 ± 22.5 SE AB C n = 5 n = 5 Braced 619.9 ± 8.3 SE 619.9 ± 11.0 SE 480.0 ± 6.7 SE AB AB C n = 14 n = 14 n = 14

A significant difference in the wood’s basic density of normally-formed and braced 2 bifurcations was also identified using a GLM ANOVA (F3, 96 = 3.16; R = 8.99%; p = 0.028) and post-hoc Tukey test (Fig. 6.9). Residuals from this ANOVA satisfied the

Anderson-Darling test for normality (AD100= 0.512; p = 0.191). The samples from the

190 apices of the normally-formed bifurcations were 4% denser than the samples from the braced bifurcations.

Figure 6.9: Boxplot of mean basic density of wood samples excised from the apices and sides of normally-formed and braced bifurcations.

Letters above boxes identify differences between groups by using a GLM ANOVA and post-hoc Tukey test at a 5% confidence limit.

6.5 Discussion

Despite the set-back of some loss in sample size, this study has successfully identified the extent by which both the natural and experimentally-induced defects weakened these hazel bifurcations and that remodelling can potentially overcome these defects due to changes in growth probably caused by mechanical strain.

191

6.5.1 Discussion of results by bifurcation type

Bark-included bifurcations

The findings from this assessment of bark-included bifurcations support those of Slater and Ennos (2015), in that bifurcations with wide-mouthed bark inclusions were significantly weaker than those with a cup-shaped morphology and that overall the bark- included bifurcations had only 72.8% of the strength of the normally-formed bifurcations. Those bifurcations with embedded bark can be considered as ones that have remodelled successfully to occlude the bark which would otherwise have weakened them substantially.

Drilled and pre-drilled bifurcations

Interestingly, the bifurcations that were drilled at the point of testing were found to have almost the same mean breaking stress as the bark-included bifurcations. Both of these bifurcation types lack the interlocking wood grain pattern at the apex of the bifurcation, as found using CT scanning by Slater et al. (2014), the former type by having it drilled out, the latter type by failing to develop it sufficiently.

The pre-drilled bifurcations showed progressive recovery of their bending strength by remodelling around the initial drill holes created and the dysfunction in adjacent tissues (Fig. 6.4a; Fig. 6.7). The level of recovery varied substantially: among other factors, this may have been due to the different positions that these bifurcations had within the crowns of the hazel trees. Given the widely accepted principle of thigmomorphogenesis in plants and the evidence of atrophy in the braced bifurcations in this study, the bending moments experienced by these bifurcations when growing in-situ are likely to be linked to the extent of their remodelling around these drill holes; however, to verify this, such

192 bifurcations would need to be the subject of a more detailed analysis using tilt meters or accelerometers to assess them for differences in movement under dynamic wind loading.

Split bifurcations

It is clear from the observations of wind-induced failure in 80% of these modified bifurcations that their factor of safety was compromised by initially inducing the splits in their apices. The five bifurcations that remained intact had remodelled lower down the parent stem, after the propagating crack had been arrested by a major change in wood grain pattern and direction. This finding strongly suggests that the interlocking and denser wood at the apex of a hazel bifurcation is much-needed to prevent the initiation of cracks which would result in them splitting apart under the loading imposed by normal conditions (which, to some extent, will also have included some additional gravitational loading to the junctions, as well as any wind loading experienced). It should be noted that no other bifurcation type was observed to exhibit any wind-induced failures over the four year period of this experiment. This implies that the bark-included and drilled bifurcations in this study had a factor of safety high enough that they could persist under the dynamic wind-loading which resulted in 80% of the split bifurcations failing over three growing seasons.

Braced bifurcations

Despite the reduced number of replicates for this bifurcation type, the strength of the bifurcations and the basic density of the bifurcations’ apices, when compared with normally-formed bifurcations, strongly suggests that the effect of the rod bracing was that these bifurcations atrophied in terms of their mechanical development. In contrast to the drilled bifurcations, the effect of putting in place a rigid brace will have prevented the braced bifurcations from experiencing mechanical strains at their apices.

193

The atrophying effect found was significant but could be argued not to be very substantial if the diameter of the smaller branch at the bifurcation apex was used to assess the breaking stress. This implies that mechanical loading is not the sole inducer of further sapwood developing in a given location: new layers of sapwood are needed for the provision of new tracheal elements through each component part of a tree’s crown, even though some components may not experience substantial strains. However, for arboriculturists considering installing a rod brace in a tree, they should take into account the subsequent development of branches with adverse taper, the associated decline in the strength of the braced bifurcation and its increasing reliance upon the brace over time (Smiley et al., 2000). In this study, if the braced bifurcations were required to support the arising branches once the brace was removed, then these bifurcations had only 70.5% of the strength of the normally-formed group and their factor of safety would have been substantially eroded.

6.5.2 Basic density

All the xylem formed under the branch bark ridge was substantially denser than that found in the adjacent parent stem, for all bifurcation types. A heightened basic density at the bifurcation is likely to result in a higher breaking stress for this component (Slater and Ennos, 2013), although it is only one factor amongst many that will affect the breaking stress of any given bifurcation. The mean basic density of the greenwood formed at the apices of the braced bifurcations was 4% less dense than the wood at the apex of the normally-formed bifurcations, suggesting that wood quality had atrophied in response to bracing.

194

6.5.3 Limitations of the study

It is important to acknowledge that this study is based upon data collected from semi- mature bifurcations in hazel trees, which gives rise to limitations in the scope of the subsequent findings. This study is part of a series that has examined bifurcations in this particular species to provide anatomical and mechanical models which could then be compared and contrasted to the bifurcations of other woody species by further study.

6.5.4 Conclusions

The denser xylem formed at the apex of bifurcations in hazel (and in other tree species) plays a key function in preventing failure of the bifurcation (Slater and Ennos, 2013). Although the role of this modified xylem is important in supplying a higher bending strength, its absence does not necessarily result in bifurcation failure: connections formed either side of the bifurcation apex can clearly be adequate to give four years’ longevity or more to the juvenile bifurcations tested in these semi-mature hazel trees.

From the pre-drilled bifurcations in this study, it is clear that they can satisfactorily remodel around an induced injury or defect and recover their full bending strength over time. This compliments the analysis of Slater and Ennos (2015) that remodelling around included bark can also fully recover the strength of bifurcations in hazel. This process of repair was not uniform amongst the bifurcations in this study, and further research could seek to find key factors that relate to the rate of repair of such bifurcations. In contrast, if the hazel bifurcation is split at its apex, although it has the potential to remodel, it is much more likely that it will fail completely under further wind-loading due to the initial crack propagating further down the stem. If a rod brace is installed above a hazel bifurcation, then development of the bifurcation will atrophy, identifying that thigmomorphogensis plays an important role in the mechanical development of bifurcations.

195

These findings help to measure the extent and degree of the remodelling of such bifurcations with different treatments, and could assist in determining a factor of safety for this component of a tree’s crown. Further modelling needs to be extended beyond static rupture tests, to investigate the movement behaviour of bifurcations under dynamic wind loading, which is considered to be a key factor in the impetus for bifurcations to remodel after injury or occlude a naturally-occurring flaw, such as a bark-inclusion.

196

6.6 References

Biro R L, Hunt E R, Erner Y and Jaffe M J (1980) Thigmomorphogenesis: Changes in cell division and elongation in the internodes of mechanically perturbed or ethrel treated bean plants; Annals of Botany 45, 655-664.

Braam J and Davis R W (1990) Rain-, wind- and touch-induced expression of calmodulin and calmodulin-related genes in Arabidopsis; Cell 60, 357-364.

Coutand C (2010) Mechanosensing and thigmomorphogenesis, a physiological and biomechanical point of view; Plant Science 179, 168-182.

Farnsworth K D and Niklas K J (1995) Theories of optimization, form and function in branching architecture in plants; Functional Ecology 9, 355-363.

Gartner B L (1995) Patterns of xylem variation within a tree and their hydraulic and mechanical consequences; In: GARTNER G. L. (ed.) Plant stems; physiological and functional morphology; Academic Press, New York.

Gilman E F (2003) Branch to stem diameter affects strength of attachment; Journal of Arboriculture 29, 291-294.

Grace J (1977) Plant Responses to Wind; Academic Press, London.

Hughes S W (2005) Archimedes revisited: a faster, better, cheaper method of accurately measuring the volume of small objects: Physics Education 40 (5), 468-474.

Jaffe M J and Forbes S (1993) Thigmomorphogenesis: the effect of mechanical perturbation on plants; Plant Growth Regulation 12, 313-324.

Jaffe M J (1973) Thigmomorphogenesis: The response of plant growth and development to mechanical stimulation; Planta 114, 143-157.

Kane B, Farrell R, Zedaker S M, Loferski J R and Smith D W (2008) Failure mode and prediction of the strength of branch attachments; Arboriculture & Urban Forestry 34, 308-316.

197

Lev-Yadun S and Aloni R (1990) Vascular differentiation in branch junctions of trees: circular patterns and functional significance; Trees: Structure and Function 4, 49-54.

Lonsdale D (1999) The Principles of Tree Hazard Assessment and Management; TSO, London.

Mattheck C and Linnard W (1998) Design in Nature; Springer, Berlin.

Philipson W, Ward J M and Butterfield B G (1971) The vascular cambium: its development and activity; Chapman & Hall Ltd., London.

Pruyn M L, Ewers B J III and Telewski F W (2000) Thigmomorphogenesis: changes in the morphology and mechanical properties of two Populus hybrids in response to mechanical perturbation; Tree Physiology 20, 535-540.

Shigo A L (1991) Modern Arboriculture; Shigo and trees, associates, Durham New Hampshire US.

Slater D and Ennos A R (2013) Determining the mechanical properties of hazel forks by testing their component parts; Trees: Structure and Function 27 (6), 1515-1524.

Slater D, Bradley R S, Withers P J and Ennos A R (2014) The anatomy and grain pattern in forks of hazel (Corylus avellana L.) and other tree species; Trees: Structure and Function 28 (5), 1437-1448.

Slater D and Ennos A R (2015) The level of occlusion of included bark affects the strength of bifurcations in hazel (Corylus avellana L.); Journal of Arboriculture and Urban Forestry 41 (4), 194-207.

Smiley E T, Greco C M and Williams J G (2000) Brace rods for co-dominant stems: Installation location and breaking strength; Journal of Arboriculture 26, 170- 176.

Telewski F W (2006) A unified hypothesis of mechanoperception in plants; American Journal of Botany 93, 1466-76.

Whitehead F H (1963) Experimental studies of the effect of wind on plant growth and

anatomy; New Phytologist 62, 80-85.

198

Chapter 7

An assessment of the movement behaviour of bifurcations in hazel (Corylus avellana L.) under dynamic wind loading using tri- axial accelerometers

Authors Slater D and Ennos A R

Status of Associated Paper Submitted November 2015

Journal Arboricultural Journal

199

7.1 Chapter Abstract

Accelerometers are potentially a valuable tool in analysing the movement behaviour of trees and their component parts. Most previous research using similar tools have assessed the stability of tree stems and roots under wind loading, but accelerometers can be used to assess a wider range of component parts of trees, associated defects and growth forms.

In this study, the movement of sixteen bifurcations in semi-mature hazel trees (Corylus avellana L.) was assessed using tri-axial accelerometers during seven differing wind- loading events. Seven of the bifurcations were normally-formed, with the other nine bifurcations containing bark-inclusions, of which a subset of five were restricted by being in direct contact with neighbouring branches. Analysis of the acceleration data from the three windiest days showed that synchronised movement of the pairs of branches arising from the normally-formed bifurcations reduced the potential acceleration of one branch away from the other by an average of 57.72%, whereas this reduction was found to be only 40.38% for the bark-included bifurcations.

The data collected from this study identifies that normally-formed bifurcations are not necessarily structural flaws in trees, as the synchronised movement of the two branches of a bifurcation means they typically avoid damaging stresses occurring at the join under dynamic wind loading: the presence of rubbing or touching branches may, however, give rise to abnormal movement behaviour and heightened stresses at the bifurcation apex.

200

7.2 Introduction

Two major forces act as common threats to the structural integrity of trees: gravitational loading and wind loading. The former can be readily estimated by assessing the mass and leverage of component parts of a tree’s crown or directly measuring them through destructive testing; however the dynamic loading caused by the wind acting upon trees is a far more complex area of study. This complexity is multi-factorial: for example, the force applied to the tree’s structure during a wind-loading event is highly variable in terms of frequency, severity and direction (Peltola et al., 1993; Wood, 1995). In addition to this variability of the wind, there are a number of facets of the affected tree’s morphology and environment that will have a large impact on the extent of movement of the tree’s component parts, and the bending stresses caused in different loci (Vogel, 1996; Sellier and Fourcaud, 2005; Rodriguez et al., 2008; Spatz and Theckes, 2013). The effect of wind on the movement of leaves, the twigs to which these leaves are attached, the secondary branches which bear the twigs, the primary branches and the tree stem gives rise to complex interactions within the tree’s crown and this substantially affects the potential bending stresses acting upon any of these components (Niklas, 1992; Sellier and Fourcaud, 2009).

For an open-grown temperate tree, recent scientific study has sought to interpret its wind- induced movement through applying the principles of mass damping and mechanical modal compartmentalization (James et al. 2006; Rodriguez et al., 2008). By these principles the off-set movement of branches and twigs in the crowns of trees helps to lessen the oscillation of the tree’s stem, allowing such trees to avoid highly damaging harmonic movements in their stems under most wind-loading scenarios (Spatz et al. 2007). Milne (1991) found, when analysing the wind-induced movement of Sitka spruce (Picea sitchensis (Bong.) Carr.), that three key factors acted to dampen the swaying of the trees; a) the interference of branches by neighbouring branches; b) the aerodynamic drag on the foliage; and c) mechanical damping in the stem.

201

Accelerometers have been used successfully in a number of previous studies on the motion of trees in the wind (Peltola et al., 1993; Moore and Maguire, 2004). By using tri-axial accelerometers as tilt meters and with a sampling rate of 20 Hz, James et al. (2013) were able to compare the movement of the bases of Eucalyptus tree stems under dynamic wind loading with their movement during static loading. This testing showed that the use of such accelerometers has considerable potential in assessing the extent and frequency of movement of tree stems under wind loading and the detection of defective anchorage or substantial decay in standing trees.

Schindler et al. (2013) analysed the response of parts of a Norway maple tree (Acer platanoides L.) to real wind loading over a prolonged period using tilt meters. Their results suggest that energy absorbed from minor wind loading by aerial tree parts is largely dissipated by the movement of foliage and secondary branches. Additionally, they speculate that during severe wind loading events, harmonic movement of tree parts is probably of minor importance compared to high mean wind loading that can cause significant displacement of parts of the tree crown.

Published research on wind-induced tree sway has mostly concentrated on analysing the relative movement of tree stems and branches, with investigations into the validity of mass damping models. However, there have not been many applied studies on wind- induced movement that have assessed defects in the crowns of trees. James et al. (2014) have identified a need for scientific investigations into tree defects under wind loading to advance the arboricultural risk assessment process. The relative risk of the failure of junctions in trees would be a good candidate for such an investigation: in particular, bark- included junctions are well-known to be a structural defect in young trees, but the likelihood of their failure cannot currently be quantified.

In static testing, juvenile bifurcations of hazel (Corylus avellana L.) have been found to have a mean breaking stress of only 79% of the breaking stress of their smaller arising branch (Slater and Ennos, 2013; 2015), which implies that such bifurcations could be a focal point for failure in the crown of a hazel tree. However, this type of static testing

202 does not take into account the natural movement behaviour of the two branches arising from a bifurcation; i.e. the extent that they oscillate towards or away from each other under dynamic wind loading. This form of testing is also limited to assessing the strength of the bifurcation in one dimension (the axis in-line with the bifurcation) whereas movement of components of a tree’s crown in the wind is usually analysed in two dimensions, sometimes three.

It can be reasoned that the more synchronised the movement of the two branches along the axis of the bifurcation under wind loading, the less stress would be exerted across the bifurcation, and the more likely it would be for a branch failure to occur instead of a failure at the bifurcation, despite the branch’s higher bending strength. In addition, the static testing of bifurcations of hazel with included bark has shown that bark-included bifurcations are much weaker than normally-formed bifurcations, typically having only 65% of the bending strength of their smaller arising branches, and being approximately 20% weaker than normally-formed bifurcations (Slater and Ennos, 2015). This gives rise to two research questions: first, to what extent do the two branches of a normally-formed bifurcation move in sync during wind-loading events and thus reduce stresses at the bifurcation apex, and second, whether the wind-induced movement of bifurcations with included bark is different from that of normally-formed bifurcations.

For this study, tri-axial accelerometers were used to assess the degree by which the branches arising from bifurcations in trees move in synchronisation with each other and by what extent they move away from each other in the plane of the bifurcation, giving rise to tensile stresses at the apex of these bifurcations. The authors have recently carried out research on the anatomy and mechanical properties of bifurcations in hazel (Slater and Ennos, 2013; Slater et al., 2014; Slater and Ennos, 2015; Buckley et al., 2015), so this experiment was also carried out upon bifurcations of hazel to add further to this series of studies.

203

7.3 Materials and Methods

7.3.1 Selection of hazel bifurcations

A relatively wind-exposed lowland site containing semi-mature hazel trees planted amongst young trees of a mix of broadleaved deciduous species was selected for this experiment. The planted area was on the southern boundary of the campus of Myerscough College, Lancashire, England – grid reference: SD497399.

Sixteen hazel trees were chosen, all being between 3.5 and 4.5 metres tall and multi- stemmed, having been planted in 2004, ten years prior to this study. Adjacent trees in the shelterbelt were of the same age and consisted mostly of silver birch (Betula pendula Roth.), wild cherry (Prunus avium L.) and hornbeam (Carpinus betulus L.), being between approximately 3 and 8 metres in height. Sixteen bifurcations were selected for assessment of their movement behaviour, all bifurcations situated between 0.9 and 1.9 metres above ground level, and each bifurcation being a component of the crown of a different hazel tree. Eight of these bifurcations were chosen on the basis that they clearly contained included bark, whereas the other eight were considered to be normally formed, without bark included in the bifurcation. Upon splitting the bifurcations apart later in this experiment, it was found that one of the bifurcations we had assumed to be normally-formed did in fact contain bark within its union which had been hidden from view when the bifurcation was within the crown of the tree. All the data recorded for this bifurcation was subsequently allocated to the included bark group.

The mean diameter of the bases of the smaller branches of these bifurcations was 18.2 mm ± 0.6 SE, the mean diameter ratio of the two branch bases was 83.0% ± 3.3 SE and the mean diameter of the parent stems of these bifurcations was 29.9 mm ± 1.4 SE. The arising branches from these bifurcations ranged from 1.5 to 2.7 metres in length, most often bifurcating two or more times more before reaching the tree’s canopy.

204

7.3.2 Accelerometry

To assess the movement behaviour of these sixteen hazel bifurcations, three tri-axial USB accelerometers were used. The model type for the accelerometers used was the X2-2® (supplied by Gulf Coast Data Concepts, Waveland, Mississippi, USA), containing a Kionix KXRB5-2050 3-axis accelerometer sensor. These accelerometers were chosen due to their low weight (48 g) and their ability to capture dynamic movement at very low temporal resolution. For the purposes of this study, the sampling rate of all three accelerometers was set at 32 Hz, which is sufficient to pick up the dynamic movement of tree branches (James et al., 2013). The gain value was set as ‘low’ and the deadband value (the value at which recording would cease) was set at zero, so all measurable acceleration during the period of assessment was recorded and recording was continuous.

For each bifurcation, a loop of blue electrical tape was strapped around each branch and the parent stem, each loop positioned approximately 200 mm from the apex of the bifurcation. This allowed for the accelerometers to be attached in the same locations each time the movement of the bifurcation was assessed. The accelerometers were attached to these locations using plastic cable ties pulled tight to ensure that the movement recorded was that of the branch or stem that they were attached to. The accelerometers were positioned so that they were all aligned with each other, resulting in the x-axis recording acceleration in the vertical plane, the y-axis recording acceleration in the plane of the bifurcation and the z-axis recording acceleration in the plane perpendicular to the bifurcation (Fig. 7.1). The acceleration data in the y-z plane were the data that was subsequently analysed to attempt to answer the study’s research questions: accelerations in the x axes were not statistically analysed, as vertical movements would be very small in extent and would be very unlikely to give rise to substantial tensile strain at the bifurcation apex under normal wind loading.

Please note that the set of axes (x, y and z) discussed in this chapter, and as recorded by the accelerometers, necessarily differs from the terminology for the planes for a bifurcation used throughout the rest of this thesis (as defined in Fig. 1.1), due to the

205 orientation of the accelerometers and their data output. By the definition given in Figure 1.1 of this thesis, the xy plane of the bifurcation was assessed for its acceleration and movement in this chapter, but the output of the accelerometers defines this plane as the y-z plane (Fig. 7.1).

Figure 7.1: Diagram of the three tri-axial accelerometers affixed to a hazel bifurcation (xz plane view, as defined in Fig. 1.1 of this thesis) using cable ties.

Arrows and labels indicate the direction of the y and z axes readings taken by the three accelerometers, with the y-axis being in the plane of the bifurcation and the z\-axis being perpendicular to that plane.

206

After preliminary testing, measurements using the tri-axial accelerometers were taken for all sixteen selected bifurcations on seven separate dates in the autumn and winter of 2014. It was decided to test through the period when these deciduous trees would be initially in- leaf and then out-of-leaf, to determine if patterns of movement behaviour changed due to the presence of foliage (Roodbaracky et al., 1994; Baker, 1997; Schindler et al., 2013). The first measurement session was deliberately taken when there was nothing but very gentle breezes present, so that the recorded movements of the branches were very small, and this could act as a benchmark for movement under significant wind loading.

Measurements were taken sequentially, with the accelerometers attached to each bifurcation in turn, for a period of approximately 15 minutes per specimen (including time for attaching and detaching the accelerometers), and alternating between the assessment of normally-formed specimens and those with included bark.

To assist with the chronological synchronisation of the three accelerometers and to identify the starting and end point of each measuring session, each bifurcation was swiftly moved by hand three or four times back and forth in the plane of the bifurcation at the start and end of each recording period. These large artificial movements were easily found in the subsequent data from the accelerometer readings in the y-axis (ay, Fig. 7.2), ensuring good chronological synchronisation between the data collected by the three accelerometers for every session – and these artificial movements and any associated oscillations were then removed from the time series analysed.

207

Figure 7.2: Method of chronological synchronisation of accelerometer readings

Artificial oscillations in the plane of the bifurcation (x axis, Fig. 1.1) at the start of a recorded session enabled the synchronisation of recording for the three accelerometers. Note the smaller accelerations of accelerometer C, which was set on the parent stem 200 mm below the bifurcation and 400 mm below accelerometers A and B. Values of the two accelerometers attached to the two branches are off-set by +0.2 g and -0.2 g for display purposes.

This form of data collection resulted in time series mostly consisting of 20,000 measurements of acceleration rate collected from each of the three accelerometers for each bifurcation on each measuring session, equivalent to 10 minutes and 25 seconds of recording per bifurcation, for analysis of the naturally-occurring wind movement of the two branches and the parent stem. This resulted in the production of 112 time series to analyse. There were two exceptions, when circumstances resulted in two slightly shortened recording periods, of 9 minutes 12 seconds and 8 minutes 35 seconds respectively, as the change-overs for these two recording periods took slightly longer than normal. It was determined to include these slightly shortened time series in the data analysis, given the analysis mostly involved assessing average movements and these

208 shortened time series did not exhibit abnormal accelerations or movement behaviours within their data.

The mean acceleration and maximum acceleration recorded in the y-z plane from each time series was used to assess the movement of the two branches situated above the bifurcation in the horizontal plane (ay and az respectively). Analysis focussed on movements of these branches in the plane of the bifurcation, as bifurcations have a strong tendency to fail in this plane when the two branches move apart from each other (Kane, 2014). The accelerations of the parent stem in the y-z plane, as recorded by the third accelerometer, was only used as a means of visually assessing stem movement behaviour under different wind conditions and is not reported further here.

The mean acceleration for a branch during a session, without modification, would typically be zero or very near to zero, as the branches oscillated in both the positive and negative y and z axes as recorded by the accelerometer, with the cumulative accelerations cancelling out each other. To obtain a more meaningful quantitative measure of the movement of the branches, the root mean square of the acceleration for the smaller diameter branch of each bifurcation was calculated for each a session, using Equation 7.1.

1 푇 푎 = √ ∫ 푎 2 (푡)푑푡 (Eq. 7.1) 1 푇 0 푏

where a1 is the root mean square of the acceleration of the smaller branch of each bifurcation in the y-z plane, T is the length of time of the recording period and ab is the acceleration recorded for the smaller branch in the y-z plane. The root mean square gives an indication of both the mean acceleration and the standard deviation of the mean, and was used successfully by Rodriguez et al. (2012) to assess the dynamic movement of a small tree.

209

Further to this, the root mean square of the acceleration of the smaller branch away from or towards the larger branch was also determined using Equation 7.2.

1 푇 푎 = √ ∫ (푎 − 푎 )2 (푡) 푑푡 (Eq. 7.2) 2 푇 0 푏 푎

Where a2 is the root mean square of the difference in acceleration between the smaller and larger diameter branch of each bifurcation in the y-z plane and aa is the acceleration recorded for the larger branch in the y-z plane.

The ratio of these two means, a2/a1, was then used as a measure to determine the average acceleration difference in terms of how much the synchronised movement of the two branches may be affecting the apex of the bifurcation (Fig. 7.1), by using equation 7.3.

푎2 푎푑 = 1 − ( ) (Eq. 7.3) 푎1

This was considered to be an effective way to assess the degree of synchronisation of movement of the two branches of the bifurcation. However, it is important to note that although ad should theoretically be a good proxy for measuring the relative bending moment experienced at the bifurcation apex compared with the bending moment exerted upon the smaller branch, the mechanical system of the bifurcation may behave in a complex manner under some loading scenarios. This means that the corresponding reduction in bending moment due to the synchronised movement of the two branches is unlikely to be directly proportional to ad for two key reasons; first, at higher wind speeds there may be some bending occurring in the branches near the point of attachment of the accelerometer, as well as bending occurring at the bifurcation apex; second, the bending moment acting at the apex of a bifurcation will be proportional to the displacement of the branch, which is the double integral of the acceleration that was measured.

210

In addition, the maximum acceleration values for the smaller branch (a1max) and the maximum acceleration of one branch away from or towards the other branch (a2max) were also determined for each time series.

7.3.3 Wind speed assessment

A three-cup VortexTM anemometer (manufactured by Inspeed.com LLC, Massachusetts,

USA) was used to record the mean wind speed (Va) and the maximum wind speed (Vm) during each 10 minute 25 second time series. The anemometer was attached to a 4.5 metre long extending pole, setting the anemometer at an overall height of 4.7 metres above ground level: the wind speed could then be measured just above the height of the hazel trees, in close proximity to each bifurcation being assessed. This allowed some assessment of the relative exposure or shelter of each hazel tree as well as recording the wind speeds pertaining to each time series.

7.3.4 Observations

Prior to the harvesting of the bifurcations for mechanical testing, measurements were taken of the height of each bifurcation above ground level. It was considered that the height of the bifurcation within the hazel tree could have a significant influence on the overall movement of that component, as there would most likely be less movement nearer the base of each hazel stem, and larger swaying movements if the bifurcation were situated further up the parent stem.

It was also recorded as to whether any of the branches arising from the bifurcations were in physical contact with other branches within the crowns of the hazel trees or the branches of any adjacent trees when not experiencing wind-induced swaying. Again, it was considered that this would influence the overall movement behaviour of the branches and the bifurcation (Milne, 1991). These observations were checked by attempting to sway each parent stem in the plane of the bifurcation, which readily identified if the

211 arising branches of the selected bifurcations were in direct contact with other branches. Where both branches from the bifurcation were found to be in direct physical contact with other branches when at rest, the bifurcation was classified as ‘restricted’, all other bifurcations were classified as ‘unrestricted’ in terms of their initial movement.

7.3.5 Rupture tests

The sixteen bifurcations that underwent this movement assessment were then cut on the 4th of February 2015, placed in plastic bags to minimise the loss of sap, and taken to a cold store for preparation prior to rupture testing that was carried out within 48 hours. Each bifurcation was severed so that there was a minimum length of 220 mm of both branches and the parent stem on each sample, thus incorporating the locations where the accelerometers had been placed during the in-situ movement assessment.

Rupture tests were carried out on these sixteen bifurcations using the method described in section 2.3.2 of Chapter 2 of this thesis, and the breaking stress of each bifurcation (σb) was calculated using equation 2.3 (Chapter 2). In this instance there were two minor modifications to the methodology, in that an Instron® Model 3344 universal testing machine (UTM) was used and the crosshead of the testing machine was made to rise at a rate of 50 mm min-1, rather the previous rate of 35 mm min-1. Careful visual observation was made of the mode of failure of each bifurcation during these rupture tests, looking for initial yield points under compression or under tension.

7.3.6 Statistical analysis

To assess the relationships between Va and the degree of acceleration of the branches and parent stem (both averaged and maximum) regression analysis was carried out and the residuals were tested for normality using an Anderson-Darling test.

212

To assess for differences in the acceleration of the two branches in the y and z axis and the degree that they synchronised their movement, Repeated Measures ANOVAs were used, with the average wind speed for each time series (Va) set as a covariate.

For assessing whether there was a difference between the mean breaking stress of the two types of bifurcation (normally-formed and bark-included), a one-way ANOVA was used in association with a post-hoc Tukey test set at 5%, and the residuals were tested for normality using an Anderson-Darling test. A further GLM ANOVA was used to determine if the sub-set of the bark-included bifurcations that were observed as being ‘restricted’ exhibited a different breaking stress to the remainder in that group.

Regression analysis was used to assess the effect of morphological differences of the height of the bifurcation, the section modulus of the smaller branch and the diameter ratio of the two branches. These continuous predictors were used to model the movement behaviour of the branches, whilst the measuring sessions and types of bifurcation were set as categorical predictors to take into account the differences in mean and maximum wind speeds experienced during each time series and the differences in movement behaviours between groups.

All statistical tests were carried out using MiniTab® version 17.

7.4 Results

7.4.1 Summary of primary data

Table 7.1 presents the mean values for Va, Vm, a1, a2, ad, a1max and a2max for each measuring session, each of which included the assessment of the movement of all 16 bifurcations.

213

Table 7.1: Summary data for each measuring session

Summary of the approximate foliage cover of the hazel trees, the mean wind speed (Va), maximum wind speed (Vm), mean acceleration of the smaller branch (a1), mean acceleration of the two branches away or towards each other (a2) and the mean synchronisation of the two branches of the bifurcation (ad).

Session date 17.09.14 24.09.14 03.10.14 20.10.14 29.10.14 19.11.14 10.12.14 Foliage status of Fully in- Fully in- Fully in- c. 10% c. 20% c. 60% c. 100% hazels leaf leaf leaf defoliatio defoliatio defoliatio defoliatio n n n n Prevailing Easterly W-SW Westerly W-NW Southerly Westerly W-SW wind c. 90º c. 230º c. 270º c. 290º c. 190º c. 260º c. 230º direction

Va (m s-1) 0.22 1.67 2.27 1.13 0.91 2.09 4.07

Vm (m s-1) 2.28 7.78 7.39 6.83 4.75 7.06 12.78 RMS of acceleration 13.14 53.64 118.46 69.53 31.09 52.56 333.13 (mm s-2) of smaller

branch (a1) RMS of acceleration 12.16 29.62 45.7 37.98 22.36 27.95 162.2 (mm s-2) of the branches away from each other

(a2) Difference in acceleration 1.7% 39.7% 59.1% 36.6% 22.3% 38.8% 50.1%

ratio (ad)

214

Maximum acceleration 123.66 465.42 745.21 737.17 265.76 415.6 2427.83 (mm s-2) recorded in a smaller branch

(a1max) Maximum acceleration 70.12 290.47 377.46 471.6 169.26 363.53 1811.19 (mm s-2) of the branches away from each other

(a2max)

7.4.2 Observations

Observations of the hazel bifurcations prior to harvesting identified that both branches arising from five of the bifurcations were ‘restricted’, in that they were in direct physical contact with two or more other branches with diameters above 10 mm when at rest. These other branches either originated from the same hazel tree or from adjacent semi-mature trees. These points of physical contact (or ‘branch rubbing’) were relatively high up in the crowns of the hazel trees and were not observed during the initial selection of the bifurcations. It was considered that due to the size of the interacting branches this factor was likely to have affected the lateral movement of these hazel branches. It was also noted that all five of these ‘restricted’ bifurcations also belonged to the bark-included group. As a consequence, these five ‘restricted’ bifurcations were treated statistically as a sub-set of the bark-included bifurcation group, to determine if their movement behaviour was different from the other bifurcations.

215

7.4.3 Regression analysis

Initial analysis concentrated on the movement behaviour of the seven normally-formed bifurcations.

Regression analysis of the data for these seven specimens identified that there was a 2 significant positive relationship between Va for each time series and a1 (F1, 47 = 60.59; R = 56.31%; p < 0.001), in that the branches accelerated more at higher average wind speeds.

The equation from this regression was:

0.2801 푉푎 푎1 = 116.89푒 (Eq. 7.4)

The plot of this relationship is given in Figure 7.3.

Figure 7.3: Scatterplot of Va against a1 for seven normally-formed bifurcations over seven sessions

Residuals from this regression were normally distributed (AD49 = 0.478; p = 0.226).

216

2 There was also a significant positive relationship between Va and a2 (F1, 47 = 36.35; R = 43.61%; p < 0.001) for the normally-formed bifurcations.

The equation from this regression was:

0.1996 푉푎 푎2 = 84.32푒 (Eq. 7.5)

Fig. 7.4 Scatterplot of Va against a2 for seven normally-formed bifurcations over seven sessions

Residuals from this regression were normally distributed (AD49 = 0.405; p = 0.34).

The difference in regression equations 7.4 and 7.5 identifies that the accelerations experienced by the two branches were more in-sync at higher mean wind speeds.

2 In addition, there was a significant relationship between Vm and a1max (F1, 47 = 95.38; R = 66.99%; p < 0.001), in that the maximum acceleration recorded for a smaller branch of a normally-formed bifurcation in each time series increased with an increase in the maximum wind speed recorded.

The equation from this regression was:

217

1.327 푎1푚푎푥 = 6.252 푉푚 (Eq. 7.6)

Residuals from this regression were normally distributed (AD49 = 0.345; p = 0.47).

2 Similarly, there was a significant relationship between Vm and a2max (F1, 47 = 81.7; R = 63.48%; p < 0.001), in that higher maximum accelerations of branches away from each other were recorded in time series with higher maximum wind speeds.

The equation from this regression was:

1.393 푎2푚푎푥 = 3.241 푉푚 (Eq. 7.7)

Residuals from this regression were normally distributed (AD49 = 0.526; p = 0.171).

The difference in regression equations 7.6 and 7.7 identifies that, on average, the maximum acceleration experienced by the smaller branches were approximately twice that of the difference in acceleration between the two branches, across all seven measuring sessions.

7.4.4 Differences in movement related to bifurcation type

The ratio ay/az was analysed, to determine if by session or by bifurcation type there was variation in the extent of movement in the y and z axis. For normally-formed bifurcations over all seven measuring sessions the mean of ay/az was 98.6%, whereas for bark-included bifurcations it was 102.7%. A Repeated Measures ANOVA identified that there was no significant difference between sessions (p = 0.545), nor between bifurcation types (p = 0.601), but there was significant variation in movement between individual bifurcations 2 (F14, 90 = 5.59; R = 48.54%; p < 0.001), with the smaller branch of just one restricted bark-included bifurcation exhibiting significantly higher accelerations along the z axis in comparison with those along the y axis during all seven measuring sessions.

218

Several key factors varied between measuring sessions, including wind speed, wind direction and the extent to which the hazel trees were foliated. A Repeated Measures

ANOVA with Va as a covariate and a post-hoc Tukey test identified that ad varied 2 significantly between the three bifurcation types (F2, 89 = 7.81; R =53.37%; p = 0.006) and that normally-formed bifurcations were more synchronised in the movement of their branches than the other two bifurcation types (Fig. 7.5). The covariate, Va, was not found to be a significant factor in this statistical test (p = 0.725).

Figure 7.5: Differences in synchronisation in branch movement by bifurcation type (all measuring sessions)

Bar chart illustrating differences in mean ad due to bifurcation type, using measurements from all seven measuring sessions. Heterogeneity at CI: 95% is identified by different letters placed above the bars.

As there was a significant positive trend for greater synchronisation of movement with increasing wind speeds (Eq. 7.5), the three windiest measuring sessions, where average wind speeds were above 2 m s-1, were analysed using another Repeated Measures

ANOVA. Again, a significant difference in ad was found between bifurcation types (F2, 2 29 = 5.44; R =54.41%; p = 0.019), with the normally-formed bifurcations showing a

219 greater degree of synchronised movement between branches as confirmed by a post-hoc

Tukey test (Fig. 7.6). Again, the covariate of Va was not found to be a significant factor

(p = 0.93). The combined mean ad for the bark-included bifurcations over these three sessions was 0.4038 ± 0.0417 SE.

Figure 7.6: Differences in synchronisation in branch movement by bifurcation type (three windiest measuring sessions)

Bar chart illustrating differences in mean ad due to bifurcation type, using measurements from the three windiest measuring sessions. Heterogeneity at CI: 95% is identified by different letters placed above the bars.

Acceleration profiles in the y-z plane for examples of the three types of bifurcation examined, over the full 10 minute 25 second time series are shown in figure 7.7.

220

Figure 7.7: Acceleration profiles in the y-z plane (NB: xy plane if using Fig. 1.1 definition of axes) for three hazel bifurcations under minor and major wind loading.

Data for these illustrations taken from measuring sessions 2 and 7. Outer traces (light grey) represent ay and az of the smaller branch; inner traces (black) represent the difference in acceleration in the y-z plane between the two branches of the bifurcation, identifying the differences between in-sync motion of both branches and the motion of the smaller branch;

221

A: Normally-formed bifurcation under minor wind loading;

B: Normally-formed bifurcation under major wind loading. Note that synchronisation of branch movements is shown by the restricted extent of the inner trace;

C: Bark-included bifurcation under minor wind loading. This was an unusual movement profile, with extremities of acceleration both in the plane and perpendicular to the plane of this bifurcation;

D: The same bark-included bifurcation as illustrated in Fig. 7.5C under major wind loading, displaying more typical movement behaviour for that group;

E: ‘Restricted’ bark-included bifurcation under minor wind loading. Note, in contrast to A, that there is little synchronisation of branch movement (the inner trace is almost as large in extent as the outer trace);

F: ’Restricted’ bark-included bifurcation under major wind loading.

7.4.5 Rupture tests

For the seven normally-formed bifurcations the mean σa was 46.52 MPa ± 2.37 SE, whereas for the nine bifurcations with included bark mean σa was 28.74 MPa ± 2.50

SE. These means were found to be significantly different from each other (F1, 14 = 25.09; R2 = 64.19; p < 0.001), the residuals from the one-way ANOVA were normally distributed (AD15 = 0.222; p = 0.793), identifying that the bifurcations with included bark failed at only 62% of the breaking stress of the normally-formed bifurcations.

Further to the observation of five of the bifurcations being in direct contact with other substantial branches higher up in the crown of the tree when at rest, a GLM ANOVA was performed and identified that these five ‘restricted’ bifurcations were a weaker subset of 2 the bark inclusion group (F1,13 = 5.39; R = 74.65%; p = 0.037). For the four ‘non- restricted’ bark-included bifurcations, mean σa was 34.06 MPa ± 2.05 SE, whereas for the five ‘restricted’ bark-included bifurcations mean σa was 24.48 MPa ± 3.21 SE. Again, residuals of this statistical test were found to be normally distributed (AD15 = 0.391; p = 0.339).

222

7.4.6 The influence of bifurcation morphology and position

Regression analysis of all seven measurement sessions identified that the height of the bifurcation above the ground (h) in metres was a significant factor in determining a1 in 2 each measuring session (F7, 104 = 46.31; R = 75.71%; p < 0.001), with an increase in height associated with a higher mean acceleration of the smaller branch. Residuals from this regression were normally distributed (AD112 = 0.431; p = 0.302).

The equation from this regression model was:

1.089 푎1 = 퐶 ℎ (Eq. 7.8) where C is a constant which varies with the seven measuring sessions analysed in this sample (range 1.673 to 4.887).

Figure 7.8 illustrates this relationship across recording sessions.

Fig. 7.8 Scatterplot of a1 against h for all seven sessions

223

This assessment showed that wind-induced movement decreased in extent when assessing branches that were situated lower down in these hazel plants, as one might expect.

When the section modulus of the smaller arising branch in the plane of the bifurcation

(Wel) and the diameter ratio of the bifurcation were added to this regression model, these factors were not found to be significant in modelling a1 (F9, 102 = 1.75; p = 0.188; F9, 102

= 0.08; p = 0.779, respectively), nor were they significant as independent factors (F7, 104

= 3.02; p = 0.085; F7, 104 = 0.44; p = 0.507, respectively).

224

7.5 Discussion

This experiment demonstrates that it is viable to analyse the movement of all three members at a bifurcation through careful chronological synchronisation of the accelerometers used. For arboriculturists, the relative simplicity of this experimental approach compared with studies with access to much higher levels of instrumentation (e.g. Schindler et al., 2013) may make this type of in-field assessment of bifurcations financially viable. This study found that there were significant positive relationships between the wind speed (both average and maximum) affecting these hazel trees and the extent of movement of the smaller branches of each bifurcation (Eq.s 7.4 & 7.5) which helps to validate this methodology.

The degree of synchronisation of movement of the two branches of the bifurcation was found to vary in association with the average wind speed (Eq. 7.5): on a relatively calm day, no evidence for synchronisation of branch movement in these hazel bifurcations was found, but it was very evident once the average wind speed was above 0.5 m s-1 and increased in extent with higher wind speed. In the case of these bifurcations, the increase in synchronicity with higher wind speeds is explained by the fact that a higher wind speed gave rise to more movement in the parent stem to which both branches are attached: the more that the swaying of the parent stem dominated the overall movement of this Y- shaped component, the more in-sync the movement of the two branches above the bifurcation will be. One can equate this difference to shaking a Y-shape when holding it at its base, or, to mimic the effect found at lower wind speeds, shaking it when holding only one of its upper arms. That low wind speeds cause significant movement only in the minor twigs and smaller secondary branches of a decurrent tree is a well-recognised phenomenon (Schlinder et al., 2013).

In this study, we assessed the extent by which the movements of the two hazel branches of each bifurcation were synchronised and it is considered that a higher degree of synchronisation implies a reduction in the bending moment acting at the bifurcation apex during dynamic wind loading. On the windiest three days, the normally-formed bifurcations experienced on average a 57.72% reduced relative acceleration between the

225 two branches than they could have done if one branch had remained unmoving and the other was moving as recorded for that session, due to the close synchronicity of the swaying movements. There was less synchronicity exhibited for the bark-included bifurcations, on average only a 40.38% reduced relative acceleration (Fig. 7.4). This equates to the bark-included bifurcations exhibiting only 70% of the synchronicity of movement that the normally-formed bifurcations did.

Two factors are likely to have caused this difference in movement; first, the normally- formed bifurcations have a more rigid join, and some small proportion of the difference between the movements of these bifurcation types may relate to the more flexible joint formed at the bark-inclusions; second, at least a subset of the bark-included bifurcations were found to be in direct contact with other branches – it seems probable that most specimens of this bifurcation type experienced more interference from neighbouring branches, disrupting the synchronisation of movement of the branches.

The nine bark-included bifurcations in this sample exhibited only 62% of the breaking stress of the seven normally-formed bifurcations of hazel. This difference in breaking stress is similar to that found by Slater and Ennos (2015) when they tested a much larger sample of bifurcations in hazel trees. In addition, a sub-set of the bark-inclusions which were observed to be touching/rubbing against neighbouring branches when at rest, were shown to be significantly weaker than the other bark-included bifurcations.

These findings usefully add to the recent static testing of hazel bifurcations carried out by Slater and Ennos (2013; 2015) who found that normally-formed bifurcations typically have c. 21% lower yield strength than their smaller arising branch. This further testing using accelerometers showed that bifurcations in hazel will often not experience as high a bending moment through wind-loading as these branches due to their synchronisation of movement. This gives a satisfactory explanation for the apparent weakness of normally-formed bifurcations in static testing, but their robust persistence in the crown structure of hazels and probably other trees. Normally-formed bifurcations in trees have been previously considered to be structural flaws by some researchers based on the

226 evidence of static pulling tests (Kane et al., 2008), but our analysis of their dynamic movement under wind loading strongly suggests that this is an over-simplification, in the case of these hazel trees at least. To assess the factor of safety of a bifurcation, its behaviour under dynamic wind loading needs to be analysed as well as its breaking stress.

In carrying out such analysis in other trees, the factor of the height at which the branch is set is likely to affect the overall extent of movement, as it did in this study, and this should be taken into account as part of any future research. In this study, despite a positive trend for larger diameter branches to accelerate less under the same wind loading, the section modulus of the branch was not a significant factor in its relative acceleration. This latter finding is probably explained by the small sample tested and the limited branch diameter range used for this study, as one would expect larger diameter branches to accelerate less, all other factors being equal.

With regard to the sampled population reported here, it is important to recognise the limitations of the findings from this study. All the samples were of juvenile branches in semi-mature multi-stemmed hazel trees: the crown form of the trees and the relative position of each branch had a significant effect on our results. In addition, only a limited range of weather conditions were sampled, and there was variation in leaf cover during that period of assessment (Table 7.1). The rates of acceleration measured were small, none being in danger of giving rise to forces that would tear apart these bifurcations.

Despite these limitations, this experiment has answered our research questions: The branches of normally-formed bifurcations do move in synchronicity to a substantial extent under high wind-loading, most-likely reducing their risk of failure: There are differences in the wind-induced movement of normally-formed bifurcations to those with included bark, in that the two branches of the normally-formed bifurcations are more synchronised in their movements. Although the accelerations measured cannot be directly translated to provide a measure of the difference in bending moments acting at the apices of the normally-formed and bark-included bifurcations, it is still reasonable to

227 argue, given the extent of the difference found, that larger bending forces would be acting at the bark-included bifurcations.

A large subset of the bifurcations with included bark that were assessed were observed to be in direct physical contact with other branches higher up in the trees’ crowns, which we did not initially anticipate. This may not be solely coincidence: that these branches were restricted in their movement by other branches may have been a primary cause of the malformation of these bifurcations. The obviously ‘restricted’ bifurcations were the weakest bifurcation type and this suggests that they may have become somewhat reliant for support on the adjacent branches they were touching, causing their relative strength to atrophy. However, that does not discount the possibility that all the bifurcations were restricted in their movement by neighbouring branches to some degree. Development of a means of measuring the extent of movement restriction caused to branches may provide additional evidence of a link between this factor and a weakened component in a tree’s crown.

The differences found in movement behaviour between these bifurcation types implies that the mechanical failure of a bark-included bifurcation would occur much in advance of a normally-formed bifurcation if this effect continued on the same trend for higher wind speeds than were recorded in this study. This trend may not continue, however, as the interference of neighbouring branches is clearly complex, and larger branch movements will often be limited by the extent of their physical contact with other moving branches in multi-stemmed trees like these hazels. Determining the movement of these bifurcations at much higher wind speeds would be a useful area for future research, as would the analysis of the movement of bifurcations in other tree species with different crown forms. Early formative pruning of trees to remove points where branches are in physical contact with each other and are rubbing together could be undertaken to prevent the ‘restricted’ type of bifurcation from forming and subsequently failing, where defect development in trees needs to be avoided.

228

7.6 References

Baker C J (1997) Measurements of the natural frequencies of trees; Journal of Experimental Botany 48, 1125-1132.

Buckley G, Slater D and Ennos A R (2015) Angle of inclination affects the morphology and strength of bifurcations in hazel (Corylus avellana L.); Arboricultural Journal 37, 99-122.

James K R, Haritos N and Ades P K (2006) Mechanical stability of trees under dynamic loads; American Journal of Botany 93, 1522–1530.

James K R, Hallam C and Spencer C (2013) Measuring tilt of tree structural root zones under static and wind loading; Agricultural and Forestry Meteorology 168, 160- 167.

James K R, Dahle G A, Grabosky J, Kane B and Detter A (2014) Tree biomechanics literature review: dynamics; Journal of Arboriculture and Urban Forestry 40, 1- 15.

Kane B, Farrell R, Zedaker S M, Loferski J R and Smith D W (2008) Failure mode and prediction of the strength of branch attachments; Arboriculture and Urban Forestry 34, 308-316.

Kane B (2014) Determining parameters related to the likelihood of failure of red oak (Quercus rubra L.) from winching tests; Trees: Structure and Function 28, 1667- 1677.

Milne R (1991) Dynamics of swaying of Picea sitchensis; Tree Physiology 9, 383–399. Moore J R and Maguire D A (2004) Natural sway frequencies and damping ratios of trees: concepts, review and synthesis of previous studies; Trees: Structure and Function 18, 195-203. Niklas K (1992) Plant biomechanics: an engineering approach to plant form and function; Chicago USA, University of Chicago Press.

Peltola H, Kellomäki S, Hassinen A, Lemettinen M and Aho J (1993) Swaying of trees as caused by wind: analysis of field measurement; Silva Fennica 27, 113-126.

229

Rodriguez M, de Langre E and Moulia B (2008) A scaling law for the effects of architecture and allometry on tree vibration modes suggests a biological tuning to modal compartmentalization; American Journal of Botany 95, 1523-1537. Rodriguez M, Ploquin S, Moulia B and de Langre E (2012) The multimodal dynamics of a walnut tree: Experiments and models; Journal of Applied Mechanics 79, 044505.

Roodbaraky H J, Baker C J, Dawson A R and Wright C J (1994) Experimental observations of the aerodynamic characteristics of urban trees; Journal of Wind Engineering and Industrial Aerodynamics 52, 171-184. Schindler D, Schönborn J, Fugmann H and Mayer H (2013) Responses of an individual deciduous broadleaved tree to wind; Agricultural and Forest Meteorology 177, 69-82. Sellier D and Fourcaud T (2005) A mechanical analysis of the relationship between free oscillations of Pinus pinaster Ait. saplings and their aerial architecture; Journal of Experimental Botany 56, 1563–1573. Sellier D and Fourcaud T (2009) Crown structure and wood properties: Influence on tree sway and response to high winds; American Journal of Botany 96, 885-896.

Slater D and Ennos A R (2013) Determining the mechanical properties of hazel forks by testing their component parts; Trees: Structure and Function 27, 1515-1524.

Slater D, Bradley R S, Withers P J and Ennos A R (2014) The anatomy and grain pattern in forks of hazel (Corylus avellana L.) and other tree species; Trees: Structure and Function 28, 1437-1448.

Slater D and Ennos A R (2015) The level of occlusion of included bark affects the strength of bifurcations in hazel (Corylus avellana L.); Journal of Arboriculture and Urban Forestry 41, 194-207.

Spatz H-C, Brüchert F and Pfisterer J (2007) Multiple resonance damping or how do trees escape dangerously large oscillations?; American Journal of Botany 94, 1603- 1611.

Spatz H-C and Theckes B (2013) Oscillation damping in trees; Plant Science 207, 66-71.

Vogel S (1996) Blowing in the wind: storm-resisting features of the design of trees; Arboricultural Journal 22, 92-98.

230

Wood C J (1995) Understanding wind forces on trees; in Coutts M. P. and Grace J. (Eds) Wind and Trees; Cambridge, England, Cambridge University Press.

231

This page intentionally left blank

232

Chapter 8

Discussion

233

8.1 Synthesis of Research Findings

This PhD thesis contributes to a greater understanding of tree anatomy and the biomechanical qualities of trees by sequentially answering a number of discrete research questions about the branch junctions of trees, using hazel as a model organism.

In the thesis introduction (Chapter 1), it is highlighted that the currently prevalent anatomical model for branch attachment is inconsistent and must be erroneous in the case of high diameter ratio bifurcations. The subsequent components testing (Chapter 2) found that the xylem formed at the apex of the join contributed the most to the bending strength of a hazel bifurcation. This phenomenon was then investigated using visual observations, microscopy and CT scanning (Chapter 3), to identify that these apical xylem tissues were denser and formed a tortuous wood grain that formed an interlocking pattern. These observations allowed the development of an anatomical model which was subsequently verified by extracting wood samples and applying standard material tests to determine their tensile and compressive strength (Chapter 4).

The anatomical model provided within this thesis is a novel contribution to the understanding of tree form and function. Most importantly, it provides a good explanation of the findings of all previous relevant scientific studies on junctions in trees (Table 8.1). This suggests that the model is a successful synthesis of previous research, and thus it usefully provides a good explanation of the biomechanical properties of junctions in trees. Although this model supersedes the previous model for branch attachment proposed by Shigo (1985), it does not contradict Shigo’s findings that sap does not easily flow through a bifurcation from one branch to another, nor that decay progression through junctions results in the highlighting of denser and hence more persistent xylem tissues (Shigo, 1991).

234

Table 8.1: Compatibility of the anatomical model with previous findings

This table lists anatomical and functional phenomena previously reported for junctions in trees, with brief explanation of each and an assessment of the compatibility of these findings with the anatomical model reported in Chapter 3 of this thesis. References for the listed phenomena are given in the last column.

Anatomical or Summary of Compatibility References functional previous finding with anatomical phenomenon of model in Chapter bifurcations in 3 of this thesis trees Features of the That axially-arranged Interlocking wood Images provided within fracture surfaces of xylem cells can often grain would cause Kane et al. (2008) be seen projecting xylem cells to be bifurcations, after perpendicular to the stretched along their they have been fracture surface along length or extracted pulled apart the line between the from the material, often bifurcation of the pith resulting in some cells and the apex of the projecting axially from bifurcation the fracture surface in this location. Formation of the A raised ridge is very Such an externally- Shigo (1991); Harris et branch bark ridge often formed at the visible ridge indicates al. (2004) junction, marking the zone where the externally a line of interlocking wood division between the grain pattern is branches from the point developed within the where the pith xylem (i.e. in the axil bifurcates to the apex of the bifurcation of the of the bifurcation pith)

Grain capture That if a branch’s The centrally-placed Neely (1991); Kramer growth development is interlocking grain and Borkowski (2004)

235

greater than that of the pattern does not other branch to which constrict differential it is conjoined at a growth and bifurcation, a greater development of the proportion of the conjoined branches (as tracheary elements opposed to the previous from the parent stem model of sequential supply sap to this overlapping collars, faster-growing branch which implies a fixed subordinate status for one of the two branches) Hydraulic That junctions in trees The presence of denser Zimmermann (1978); constriction resist sap flow more and more tortuous Tyree and Alexander than adjacent stems and wood at the centre of a (1993); Aloni et al. branches bifurcation will result (1997); Eisner et al. in some additional (2002) resistance to sap flow at junctions Hydraulic That sap does not The interlocking wood Shigo (1985); Lev- segmentation readily flow from one grain arrangement does Yadun and Aloni branch into another at a not provide pathways (1990); Sprugel et al. junction in a tree for sap flow from one (1991) branch to another Whirled grain That whirled patterns Interlocking wood Lev-Yadun and Aloni of wood grain are grain patterns can (1990); André (2000); frequently found at the incorporate zones of Pfisterer and Spatz apex of junctions in whirled grain, as a (2006) trees when de-barked development of their tortuous patterns

Wood density That wood density is The greater proportion Gartner (2005); highest at the apex of of fibres and the Jungnikl et al. (2009) bifurcations smaller size and extent of vessels at the apex of bifurcations results

236

in denser wood and the production of tortuous cells Wood grain That xylem tissues are The interlocking wood Kramer (1999); Müller alignment to arranged at junctions in grain arrangement et al. (2006) trees in alignment to results in some xylem stresses at junctions the main stresses that tissues being axially they experience when aligned with the the two branches are stresses that could bent apart rupture the junction

With this new anatomical model for branch attachment observed (Chapter 2), illustrated (Chapter 3) and tested (Chapter 4), the study then sought to determine key factors that could affect the strength of a bifurcation and to determine the extent of these effects: bark inclusions (Chapter 5) and artificial treatments such as bracing, drilling and splitting (Chapter 6). The findings from the investigations into bark inclusions and the re-modelling of bifurcations identified that although bifurcations can be weakened by inclusions, as this disrupts the production of the interlocking wood grain pattern at the apex of a bifurcation, they are capable of responding to this defect and can fully rectify the loss of strength induced by embedding the bark within the xylem formed at the bifurcation. The remodelling process is also shown to be related to strain-induced growth, as specimens in which the central zone of interlocking grain was drilled out often recovered their strength over time, whereas the strength of braced specimens atrophied (Chapter 6).

The static pulling tests used in this theses are an artificial means of assessing a structural component of a biological organism, so it was considered important to incorporate an assessment of the mechanical performance of bifurcations in-situ. Chapter 7 provides critical insight into how there is a general trend for the two branches of a bifurcation to move in synchronicity when under significant wind-loading. This experiment, consequently, allows a further synthesis, in that although static pulling tests imply that failure at a bifurcation is more probable than along a branch, the analysis of bending and movement in the latter part of this study suggests that the structures of branches and

237 bifurcations are well-matched. This informs a discussion on the factor of safety for bifurcations, which is provided in section 8.2 of this chapter.

These complementary findings show that normally-formed bifurcations are satisfactory to serve their function in the crown of a hazel tree, which is to splay branches to facilitate light capture and give these branches sufficient mechanical support (Fisher 1986). The anatomical experiments demonstrate how this satisfactory level of strength is provided by the formation of interlocking wood grain. Normally-formed bifurcations in hazel are in general neither over-engineered nor under-engineered and they do not constrict sap conductance excessively (Zimmerman, 1978). They do this by having an effective design that requires the development of only a small central zone of specialised xylem tissues to provide additional mechanical support.

An interpretative analogy of this finding is that the zone of interlocking wood grain acts like a strong plastic clip would act to conjoin two hosepipes: this ‘clever clip’ does so by mostly avoiding constricting the hosepipes and hence not impeding the flow of fluids through them and it has been made with efficient use of the quality and quantity of material allocated to the task. Effective designs such as this have been illustrated in the anatomy of stems and branches of trees by many authors (e.g. Gartner, 1995; Crook et al., 1997; Mattheck, 1998; Baas et al., 2004; Niklas and Spatz, 2012), and this study adds one more example of the ingenuity of tree anatomy to efficiently fit form to function.

8.2 Estimating the Factor of Safety for Bifurcations in Trees

The concept of a ‘Factor of Safety’ is most frequently applied by engineers to artificial structures, where it is used to contrast the intended loading of a structure with the loading which would result in the failure of that structure. This concept has subsequently been applied by Niklas (1992) to seek to quantify the factor of safety for some component parts of plants.

238

There are a number of different definitions for the factor of safety of a structure which vary due to the type of loading that is being compared. In engineering, it is common to define the factor of safety (S) of a structure as the ratio of the yield stress of the structure

(σy) to the maximum stress predicted to act upon a structure due to its usage (σmax) as described in equation 8.1.

휎푦 푆 = (Eq. 8.1) 휎푚푎푥

Mattheck et al. (1993) estimated the factor of safety of tree stems by cutting out rectangular ‘windows’ through the stems of spruce and beech trees to produce known multiples of the existing stresses acting upon the trunks of these trees. From their experiment they reported that S ≥ 4.5 for tree stems; however, the size and number of replicates was not clarified nor the extent of the wind loading that acted upon these modified trees prior to failure. Subsequent notch stress analysis by these authors, using FEA, implied that S may be greater than 5 (Mattheck et al., 1993). Unfortunately, with no record of the wind loading that these trees experienced and the rather crude method used in the cutting of the rectangular holes in the tree stems, this estimate of S for tree stems should be considered only a very rough approximation which does not take into account significant variation in wind loading in the presence or absence of foliage.

Evans et al. (2008) assessed primary lateral branches of 40 woody plant species, and by estimating the typical gravitational loading that applied to them as idealised circular beams, they calculated the average bending stress acting along each branch. They reported that there was a very consistent mean stress acting upon the outer edges of these branches of 11 MPa, if they were treated as beams with circular cross sections under bending stresses caused solely by gravitational self-loading (Evans et al., 2008).

In this study, the hazel bifurcations and branches that underwent rupture tests and three point bending and broke or yielded at slightly different levels of stress, depending on the sample taken. Table 8.2 provides the mean bending stresses and their standard error

239 for these strength tests of bifurcations (σa) and branches (σb) of hazel sampled throughout this study.

Table 8.2: Mean bending stresses for branches and bifurcations of hazel tested in this study.

Means are provided with standard error, number of replicates and related reference to the chapter within this thesis where this data is reported.

Component Mean bending No. of replicates Reference stress (MPa) (n) 58.47 ± 1.10 SE 31 Chapter 2 Branches 54.28 ± 0.66 SE 83 Chapter 5

53.30 ± 0.70 SE 69 Chapter 6 42.87 ± 0.97 SE 35 Chapter 2 Normally-formed 46.79 ± 0.83 SE 127 Chapter 5 bifurcations 43.19 ± 0.99 SE 55 Chapter 6

46.51 ± 2.36 SE 7 Chapter 7 35.54 ± 0.94 SE 106 Chapter 5 Bark-included 31.46 ± 1.16 SE 66 Chapter 6 bifurcations 28.74 ± 2.53 SE 9 Chapter 7

If the factor of safety is defined by the ratio of the yield stress of the component and the typical gravitational loading found by Evans et al. (2008), this would result in an estimation of the factor of safety of hazel branches of S ≈ 5, the factor of safety of normally-formed bifurcations being S ≈ 4 and for bark-included bifurcations this would further reduce to S ≈ 3.

Appealing though this simplistic answer may seem, this is clearly too basic an analysis if one is interested in what the factor of safety of hazel bifurcations is under dynamic wind loading, which is more relevant in judging their mechanical performance.

240

Unfortunately, the ‘typical’ wind loading of a hazel branch or bifurcation would be very difficult to define, due to its variability over time, by location and due to the effects of micro-climate and interactions between neighbouring branches on any particular bifurcation. Although all normally-formed bifurcations formed in a hazel tree have commonalities with all others in terms of their anatomy and material properties, they are also individually unique in their wood grain pattern, their development, their gravitational loading, their diameter ratio, their relative position in the tree’s crown and consequently their different exposure to winds from different directions. In addition, their movement is also affected by neighbouring branches and other environmental features which could be very difficult to predict or model accurately.

Chapter 7 of this thesis goes some way in tackling this issue of safety assessment by sampling the wind-induced movements of sixteen hazel bifurcations, finding that movement synchronicity of the two branches at 200 mm above the bifurcation was quite high under significant wind loading, both in-leaf and out-of-leaf. Taking the mean level of branch synchronicity over the three windiest days of this analysis where average wind speeds were above 2 ms-1, this suggests that the smaller branch of a normally- formed bifurcation was on average experiencing about double the stress due to bending than was experienced at the bifurcation’s apex. However, the validity of looking at the average stresses is limited, as failure of such a bifurcation could occur from one single instance of a large asynchronous movement of the two branches. It would be very difficult, however, to predict the frequency of occurrence of such damaging movements in hazel bifurcations – they may require very specific environmental conditions to occur, which were not observed throughout this six-year study period in the hazel coppices assessed.

What can be concluded from the wind movement analysis carried out is that this phenomenon of branch synchronisation results in the successful avoidance of bifurcation failure in hazel trees under high levels of wind loading. There is no observational data that suggests that normally-formed bifurcations have a lower or higher factor of safety than their associated branches or stems in hazel when judged in- situ. Substantially more research work would be needed to give a definitive answer to

241 the factor of safety of a bifurcation, primarily because of the difficulty of defining and determining what is the ‘norm’ for wind loading to the bifurcation as a component, daily, annually, or throughout its lifetime. In the absence of evidence for failures at these normally-formed bifurcations to be a common occurrence, it seems sensible to conclude that their mechanical performance is satisfactory and their factor of safety is not significantly different to that of the arising branches.

In contrast, from the regular observation of the failure of bark-included bifurcations and the pre-split bifurcations assessed in Chapter 6 of this thesis, it is considered probable that such defectively-formed bifurcations that lack the interlocking grain at their apex have a significantly lower factor of safety than their branches. Analysis of the wind movement of nine of these bark-included bifurcations also adds credence to this assertion, as the two branches of these bifurcations were significantly less synchronised in their movements under dynamic wind loading. However, they did still exhibit some degree of synchronicity in movement, which again helps to explain their persistence in the crowns of trees which have endured major wind-loading events. The naturally- induced failure of one such bark-included bifurcation was observed in one of the hazel coppices used during this study.

The factor of safety, as a concept, was conceived for the assessment of man-made structures where the proposed loading was predictable. A much larger amount of data collection and analysis would be needed to determine a factor of safety for hazel bifurcations because of the unpredictability of wind-loading events and the subsequent movements in the components of a hazel tree’s crown and the corresponding stresses caused to these components. The extent of the problem caused by bark-included junctions in trees is significant to the arboricultural industry, so a combined approach of collating industry data and scientific modelling is recommended later in this discussion (section 8.4.3).

242

8.3 Critique of Methodologies Used

In this section, the main methodologies used in this thesis to obtain data on hazel bifurcations are critically analysed to acknowledge their associated limitations.

The use of only one species in a series of experiments is potentially a significant limitation to the applicability of the findings of this research to other woody plant species (Hedges, 2002). In the initial stages of this research over twenty woody plant species, both angiosperms and gymnosperms, were assessed for their junction anatomy and the wood’s basic density (Chapter 3); further to this, four species had a set of bifurcations tested for their mechanical properties. These preliminary tests identified that there was no substantial differences in the general anatomy or biomechanical behaviour of these bifurcations that would affect the satisfactory completion of the two aims of this research if only one species was used for further testing. Hazel was chosen as a ‘model organism’, accepting that care would need to be taken in extrapolating results in this species to any other species. This decision facilitated a much more in- depth assessment of the biomechanical characteristics of bifurcations formed by this one species, which resulted in satisfactory completion of all the research objectives. There is also one associated published study by Pfisterer (2003) that also undertook to investigate the mechanical properties of hazel bifurcations with which this study can compare its findings directly.

It was considered that the anatomy, morphology and breaking strength of bifurcations in hazel may vary substantially because of a considerable number of factors, including the degree of wind exposure in which it has grown, time of year of assessment, phenotypic variation, the angle of inclination and the internal angle of the bifurcation, the bifurcation’s diameter ratio, the presence or absence of included bark, the position that the bifurcation took in the crown of the tree and the age and diameter of the bifurcation. All sample selection in this study was arranged so that only a limited number of factors were assessed in any one given experiment, with efforts made to limit any other potentially influential factors, as listed above. For example, to negate the effect of substantially different wind exposure levels affecting bifurcation strength, the hazel bifurcations used for testing in Chapter 5 were obtained within a lowland coppice area, avoiding collecting from trees at the edge of the coppice. This careful sample selection

243 process has allowed the multi-factorial nature of hazel bifurcation strength to be elucidated through the isolation of factors, some of which are quite subtle in their effect.

Many of the results reported in this thesis are based upon static testing of bifurcations and branches using an Instron® Universal Testing Machine (UTM), or a similar test machine (Chapters 2, 4, 5, 6 and 7). These tests can be used to estimate the yield stress, breaking stress, toughness and elasticity of materials and the relative strength of test samples; their accuracy for material testing is typically high (e.g. ± 0.05% of actual force applied). One significant limitation of this methodology was that the branches and bifurcations tested could only be of a limited size, as only a 1kN load cell was available for use. As a consequence, there is the opportunity to add to this research by examining the development of bifurcation anatomy from its initiation to its mature form, and to additionally test the breaking stress of larger bifurcations with a UTM that can apply higher forces to these mature specimens. It is likely that more mature bifurcations would exhibit more whirled grain patterns and would have a lower strain to failure than those tested for the purposes of this thesis.

A further limitation of this study is that the bifurcation specimens obtained were all used with a fixed lever arm length of 200 mm during the static testing. The decision to use this location for the static loading of the two branches above the bifurcation was a choice made for pragmatic reasons of ease of sample handling and does not mirror the actual location of the loading from gravity or wind that would have occurred along the branches when they were in-situ in the hazel trees. Although this method will provide an accurate assessment of the breaking stresses of the bifurcations, the effect of substantial flexure of the branches could not be assessed effectively with this relatively short lever arm being used. It is considered that if the location of attachment to the UTM was substantially further away from the bifurcation apex (e.g. 1000 mm or more), then a far higher instance of branch failures would have occurred, providing a different perspective on the relative mechanical performance of the branches when compared with the bifurcations.

The use of a lever arm length of 200 mm for these rupture tests did allow for some flexure of the branches, however, and avoided the potential for failures caused directly by shear stresses. Shear failures can distort the results of bending tests when the lever

244 arm used is shorter than 15 times the diameter of the member being bent under three- point bending (Vincent, 2012); so, in the case of the testing of these bifurcations, the equivalent span length would be 400 mm and the diameters of their smaller branches were no greater than 25 mm, ruling out the potential influence of shear stresses in tearing apart these bifurcations.

In addition, it is important to acknowledge that static loading does not mimic the dynamic loading caused by the wind, which is the major force present in the natural world that can cause them to mechanically fail (Metzger 1893; James, 2003). In this thesis, such bending strength data has been combined with an assessment of wind- induced movement of hazel bifurcations in Chapter 7, so that the outcomes of these static tests can be set in the context of how these bifurcations move under dynamic wind loading. One could also go further and use a series of strain gauges on bifurcations whilst they move in the wind and/or are mechanically tested to failure, to assess where high strain levels and potential stress concentrations occur in such bifurcations.

CT scanning was used to obtain some results reported in Chapter 3 of this thesis which assessed wood grain patterns at the apex of hazel bifurcations. Unlike a histological approach that would produce 2D anatomical slides of xylem tissues, the use of CT scanning facilitates multiple assessments on the same small wood volume and allows for the assessment of wood anatomy in three-dimensional visualisations. CT scanning has also proven to be able to accurately measure wood density (Freyburger et al., 2009). An important limitation to this method is that unlike visual analysis of histology slides, the resolution of the CT scan cannot be enhanced beyond its original setting, as the data output is digital and pixelated. The resolution of the scans reported in Chapter 3 was sufficient to track the progress of vessels through the scanned wood volumes, but not high enough to be able to assess the thickness of the cell walls of fibres, for example. CT scanning was favoured for this experiment because of the highly variable wood grain orientations and heightened density of the xylem tissues discovered at the apex of hazel bifurcations which would have made the production of a series of histology slides of a uniform thickness very challenging and reproduction of 3D wood grain patterns in a series of 2D images very difficult to achieve.

245

Tri-axial accelerometers are used in Chapter 7 as a means of assessing the movement of branches arising from a bifurcation under dynamic wind loading. It is important to acknowledge the limitations of this assessment, in that the output data are values for the acceleration of the meters but not their displacement or speed at any particular time, so subsequent analysis is also limited to comparisons of the rates of acceleration measured.

8.4 Recommendations for Further Research Work

8.4.1 Causes of differing bifurcation anatomy

Although this study reports a novel model of branch attachment in trees, there are substantial research opportunities to further develop this model. In particular, the causal factors that lead to the formation of the interlocking wood grain patterns found in this study could usefully be investigated by the growing of trees and the monitoring of the development of these patterns. Such studies should also look to find causal factors that give rise to the anatomical difference in bifurcation development that have been observed between normally-formed bifurcations and those with included bark. In addition, the development of wood grain patterns from simple interlocking structures to those incorporating whirled grain could be traced through dissections of mature bifurcations. Work carried out at an intermediate scale could usefully add to this study’s analysis of low-scale visual observations that helped to model the interlocking wood grain patterns reported in this thesis and the very high scale CT scanning carried out as part of the anatomical study reported in Chapter 3.

From the outcomes of the experiment reported in Chapter 6 of this thesis, it seems probable that mechanical perturbation can considerably affect the development of bifurcations in woody plants. It would be logical to progress this finding by setting up a control and treatment study on young woody plants which could usefully assess the influence of mechanical perturbation on bifurcation development and morphology, as has been shown for some component parts of other plants (Jaffe, 1973; Jaffe and Forbes, 1993).

246

As this study has found that the currently most widely used model for branch attachment in arboriculture is erroneous, this highlights that there is still considerable work needed to produce satisfactory anatomical models of other features of woody plants. For example, if the branches of a bifurcation are kept together by interlocking wood grain patterns, is that also the case for bifurcations formed in the roots of woody plants? There are several classes of nodes formed in woody plants and herbs that could usefully be revisited in the light of this study to determine anatomical models that explain the nodes’ biomechanical behaviour. In addition, looking for rules of variation in this anatomy due to the diameter ratio of the two conjoined branches may also be fruitful.

8.4.2 Extending the study to other plant species

As this thesis is almost entirely focussed upon the anatomy and biomechanical properties of the bifurcations of hazel (with the exception of some observational data reported in chapter 3), there is plenty of scope to do comparative studies using similar techniques upon different plant species. When choosing tree species for such a comparison, analysis of a ring-porous angiosperm (e.g. Quercus spp.) would potentially make an interesting contrast with hazel (which has diffuse porous sapwood), as would analysis of bifurcations of gymnosperms (e.g. Taxus spp.). In addition, other plants contain fibrous junctions, such as branching monocots (e.g. Dracaena draco L.) and these have been found to contain circular patterns in the tissues they form at their junctions (André, 2000). Thus the general applicability of the findings within this study could be explored in a range of other plants, both in terms of analogous anatomies of junctions and also their biomechanical performance.

8.4.3 The relative risk of failure for bifurcations in trees

As the wind loading to bifurcations and their associated movement cannot be readily predicted to determine a factor of safety for normally-formed and bark-included bifurcations, it is recommended that this is assessed through a combination of field

247 studies of bark-included junctions and storm-damaged trees, continued scientific modelling and the collation of data on tree failures (of which the International Tree Failure Database is an example (ITFD, 2015)). The field-based studies should result in an estimate of the probability of failure of different crown components under different types of wind loading and ideally would assess trees of a variety of species growing in different environments, such as forests and urban areas. The scientific modelling should continue, as only scientific modelling can provide the important justification, reasoning and predictive power as to the ‘how and why’ junctions in trees fail. This combined approach is recommended as it is more likely to yield data that would be of practical use to tree owners and arboriculturists, rather than confining studies to mathematically modelling the complex multi-factorial mechanical performance of bifurcations and branches under wind-loading. A good example of this combined approach can be found in the Forestry Commission’s ‘ForestGALES’ computer based decision support tool, which acts to predict the risk of windthrow for forestry crops based upon both industry records of failures and scientific modelling of the windthrow process (Forestry Commission, 2016).

The probability of failure for a bifurcation in a tree is complex to calculate due to its multi-factorial nature, and to determine that probability one needs to obtain data at the point of failure for many such junctions under naturally-occurring conditions, with the added difficulty of trying to exclude or discount a number of associated contributory factors. Organised large-scale counts of instances of bifurcation failure during storm events (when assessed against bifurcations that did not fail during the same events) would help to achieve the desired end of estimating their probability of failure, upon which scientific modelling could usefully add predictive tools and refine the risk assessment for the range of factors involved. Such modelling could re-iteratively refine both the surveying and the computations of risk, resulting in a decision tool of a similar nature to ForestGALES.

The failure rate of bark-included bifurcations is considered to vary between tree species and cultivars, as previously reported by a panel of experienced arboriculturists (Lonsdale, 1999). The relative susceptibility of some woody plants to fail at these malformed junctions could be assessed by collating failure data from urban foresters (ITFD, 2015). Furthermore, from some initial testing of the bifurcations of three further

248 species carried out by the author in 2010, it is suspected that the flexibility of the species’ branches may be a significant factor in the frequency of junction failures, in that those which form stiffer, harder sapwood are more likely to suffer failures at bark- included junctions, and this could be determined through rupture tests and three-point bending of branches in a similar way to this current study. Such a study may satisfactorily explain the collated observations that failure of bark-included bifurcations is frequent in species such as Acer platanoides L., whereas bark-included junctions in maidenhair trees (Ginkgo biloba L.) rarely fail (LTOA, 2013).

8.4.4 Opportunities for biomimicry of bifurcation anatomy

A key finding of this study has been the extreme trade-off in sap conduction ability found in the xylem formed at the apex of bifurcations in hazel, in order to supplement the mechanical strength of the bifurcation. There is a great opportunity to develop bio- inspired and bio-mimetic junctions in man-made materials based upon the findings of this study (Mayer 2006; Burns et al., 2012) – although the best application would be within a component that acts as a conduit for the flow of a fluid as well as needing a strong join at a junction (e.g. junctions in 3D printed replicas of human organs).

Further research in applying the findings of this thesis to bio-mimic wood grain patterns in artificial bifurcations should be carried out in a fibrous, ductile and tougher material, such as the fibre-reinforced plastics used in the construction of aeroplanes and other vehicles.

8.4.5 Advanced techniques for assessing bifurcations

Due to budget and time limitations imposed upon this study, only a limited amount of measuring instruments and modelling software was used. This section considers the use of further instrumentation or software in advancing knowledge about bifurcations in trees.

249

In terms of the assessment of the mechanical performance of bifurcations in-situ, further work could usefully be undertaken, assessing the effects of wind loading on the component parts of bifurcations. In particular, if a series of light-weight strain gauges could be attached to a bifurcation, including one being placed at the bifurcation apex, then this data could usefully supplement associated data collected by rupture tests and accelerometers used in a similar manner as reported in this thesis (Chapter 7). Strain gauges have been successfully used in a number of other studies on tree biomechanics (e.g. Crook and Ennos, 1996; Crook et al., 1997; Bonser and Ennos, 1998; Brüchert and Gardiner, 2006). This approach could usefully associate the extent of strains experienced in xylem tissues at the bifurcation to the movement behaviour of the branches and the strain level at the bifurcation when it yields, to further determine the likelihood of failure of a bifurcation.

Finite element analysis (FEA) is a modelling technique which has been widely used to assess the mechanical performance of man-made structures (Reddy, 2005; Chaskalovic, 2008). Its use in the field of biomechanics is currently expanding (Rayfield, 2007; Dumont et al., 2009), but challenges still remain in modelling woody plant structures: first, the anisotropy of xylem as a material; second, its variation in strength due to temperature, moisture content and loading rate; and third, the local variations in MFA and wood grain orientation (Mackerle, 2005).

8.4.6 Creating innovative remedial treatments for flawed bifurcations in trees

Current arboricultural treatments for junctions in trees that are considered to be weak and potentially a hazard are to formatively prune the tree when young, to subordinate one of the two branches arising from the junction or to fix an artificial bracing system (Gilman, 2011).

When the re-modelling experiment reported in Chapter 6 of this study was commenced, an attempt was made to form natural braces across twenty hazel and twenty lime (Tilia cordata Mill.) bifurcations by selecting a small lateral branch arising from one of the

250 two branches of the bifurcation and getting it to graft to the other branch, so as to straddle the bifurcation (Fig. 8.1).

Figure 8.1: Attempt to create a natural brace traversing a bifurcation

Photograph of an example attempt to graft a lateral branch across a bifurcation in lime (Tilia cordata Mill.), as part of the re-modelling experiment reported in Chapter 6 of this thesis (Author’s own image, 2011). The white grafting tape that perished can be seen on the left-hand branch of this bifurcation, attempting to form a graft at this point between the smaller lateral branch and the main branches arising from the bifurcation.

Unfortunately, in this instance, the grafting process was unsuccessful as the grafting tape degraded very quickly when used outdoors and the physical connection between the branches was thus lost prematurely. However, it is still considered that there is substantial potential for natural bracing to resolve the issue of structurally flawed bifurcations in large-growing urban trees if the grafting process can be made more reliable. A particular advantage of this approach over the use of artificial braces is that braces formed by living tree branches should grow and adapt in response to mechanical perturbation, resulting in a long-lasting solution to the initial flaw (Fig. 8.2). In contrast, braces made from artificial materials inevitably degrade after being put in place, complicating the assessment of the likelihood of junction failure for arboriculturists and requiring intermittent replacement at considerable expense.

251

Figure 8.2: A naturally-formed brace above a bark-included bifurcation

Naturally-formed branch graft (white arrow) above a bark-included bifurcation (black arrow) formed in a mature beech tree (Fagus sylvatica L.) which will act as a brace to prevent tensile stresses acting at the bifurcation apex. Original image contributed by Mr. C. McCorkell, graduate of Myerscough College.

252

8.5 Implications of the findings of this study for arboricultural practices

This concluding section considers the implications of this thesis for arboriculturists whose responsibilities involve the management of trees. The limitations of this study identified in section 8.3 are acknowledged in relation to all recommendations made in this section.

8.5.1 Pruning

Natural target pruning (NTP) was developed on the basis of observations of the development of dysfunction and decay at junctions in trees that had been pruned in different ways (Shigo, 1991), and these observations also gave rise to the previous branch attachment model which this study’s model supersedes (Shigo, 1985). NTP is recommended to arborists as it acts to minimise the development of decay from pruning cuts made to trees and speeds up the rate of recovery of the wounds that pruning causes. Within NTP, the branch bark ridge and branch collar are two key external and observable features of junctions in trees that are used by arborists to determine the best location for the pruning cut to be made (Shigo, 1989).

The anatomical model outlined in Chapter 3 of this thesis identifies that the branch bark ridge seen on the outside of the bifurcation marks the location where tortuous interlocking wood grain patterns ‘knit’ together the two branches of the bifurcation. The adoption of this anatomical model would not change the applicability of NTP, as the branch bark ridge is still a key anatomical feature for identifying the division between the two branches and the internal wood grain arrangement, and a pruning cut that traverses the branch bark ridge will damage the tissues of the stem or branch that is to remain after the pruning operation, which is not desirable and would be contrary to NTP techniques. All that has altered by the adoption of this new anatomical model is the loss of the conception that under the bark of a bifurcation is a hidden set of seasonally-overlapping collars of xylem, which NTP does not rely upon for its effectiveness.

253

8.5.2 Bifurcations and tree hazard management

The findings of this study concur with the findings of Gilman (2003) and Kane et al. (2008) that in static pulling tests, the strength of high diameter ratio bifurcations in trees is approximately 20-30% lower than the strength of the associated branches, where the bending strength of bifurcations is assessed by applying forces only a relatively short distance away from the junction. However, this study has taken the assessment of the mechanical performance of bifurcations further by assessing their movement under natural wind-induced loading. This further experimentation has successfully explained this seemingly lower mechanical performance of bifurcations, in that they do not experience bending to the same extent as their associated branches do under wind loading. Therefore this study does not support the conclusions of these previous studies which identified normally-formed bifurcations as substantial structural flaws in trees. Static pulling tests used in this way are insufficient evidence of the mechanical performance of these bifurcations and concluding that a naturally-occurring and normal component of the majority of tree crowns is inherently weaker than all other aerial components does not fit well with evidence of the evolution of effective tree form and function over millennia (Zimmermann and Brown, 1971; Mattheck, 1998).

For arboriculturists, this thesis presents evidence that a normally-formed bifurcation in a tree is not a structural flaw, and need not be especially assessed for its likelihood of failure. A tree surveyor should look for evidence of malformations of a bifurcation or sudden changes in its environment (especially the removal of adjacent branches or trees that acts to increase the wind-exposure of the bifurcation) to justify the identification of it as a specific defect with a heightened likelihood of failure. In particular, the identification of the presence of included bark and/or cracks and splits at the apex of the bifurcation are important visual assessments to make.

254

8.5.3 Bark inclusions

A key motivation for undertaking this assessment of the mechanical performance of bifurcations was that the failure of bark-included junctions in amenity trees is a common and unwanted problem that many arboriculturists face (Helliwell, 2004).

This study confirms that bark-included bifurcations where the bark is not embedded and the bark’s presence is visible at the apices of the bifurcations can be considered structural flaws, which is current arboricultural guidance in the UK (Lonsdale, 2000) and internationally (Harris et al., 2004). The gradations in strength found in bark- included bifurcations in hazel trees described in Chapter 5 of this thesis may assist arboriculturists in their assessment of the relative safety of such junctions. In particular, tree surveyors should visually observe the relative width of the bark inclusion at the bifurcation apex, and whether bark or wood is situated at the apex.

Care should be taken not to condemn a junction in a tree as unsafe just on the basis of the presence of included bark and a visual inspection should be made of any interactions between branches in the crown of the tree situated above the bifurcation, especially the fusing of branches arising from the bifurcation (which may mitigate the weakness of the bifurcation), or the presence of rubbing branches (which may again mitigate the weakness or, conversely, exacerbate the problem by preventing a normal swaying motion and the subsequent synchronisation of movement of the branches of a bifurcation). It may be appropriate in some cases of two rubbing branches belonging to a single tree’s crown to stabilise and conjoin them artificially to encourage the development of a natural brace by anastomosis (the fusing of the parts of woody plants), rather than removing one of the two branches, which is the current industry guidance in the UK (BSI, 2010).

255

8.5.4 Bracing

As woody plants modify the structural strength of their components through experiencing strain in their tissues (Jaffe and Forbes, 1993), any artificial interventions that affect the structure of a tree’s crown need a good degree of scientific justification. Rigid bracing of branches situated above a bifurcation to prevent it from splitting could potentially cause a non-sustainable situation, as the strength of the braced bifurcation will atrophy over time, as will the materials used to form the brace due to their deterioration. This study has not assessed the use of flexible bracing to prevent the failure of bifurcations, which is becoming a more common form of tree branch bracing (Smiley and Lilly, 2014); however, there is still the potential for the effects of bifurcation atrophy and brace degradation to occur when using flexible braces, and these two factors can compound the likelihood of the failure of a bifurcation several years after the installation of a brace. As a consequence, the pruning of one or both branches arising from a structurally flawed bifurcation is probably to be preferred to the use of braces in many scenarios, unless the short-term preservation of the entirety of the tree’s crown is considered essential for aesthetic or cultural reasons.

8.5.5 Split bifurcations

Testing the component parts of hazel bifurcations identified that if the entirety of the apical tissues at the apex of a bifurcation was removed or structurally compromised, the bifurcation had only 19% of its original bending strength remaining (Chapter 2). This finding was supported further by the naturally-induced mechanical failure of twenty of the twenty-five split hazel bifurcations reported in Chapter 6 of this thesis, identifying that in the case of bifurcations with splits their factor of safety must be compromised (i.e. S ≈ 1).

It is considered that interlocking wood grain and whirled grain at the apex of a bifurcation acts to prevent the initiation of a crack in the xylem in this location and this tougher xylem will act to resist crack propagation until the crack extends along the full

256 length of the branch bark ridge. Cracks in xylem that are aligned with the wood grain direction are highly prone to propagate under quite low bending moments (Thuvander et al., 2000), which will be the case once the crack has run through this tortuous section of the wood grain at the bifurcation apex. As a consequence, a split bifurcation (Fig. 8.4) should most-often be considered a major hazard in a tree that needs immediate remediation where people or property could be endangered by a large branch falling away from a mature bifurcation. Urgent remedial actions should also be taken where the continuation of the crack would further damage the tree as a specimen, causing a loss in the tree’s amenity, longevity or aesthetic appeal.

Figure 8.3: A mature split bifurcation that is a significant hazard

A split bifurcation in a hybrid poplar (Populus x canadensis Moench) that now represents a substantial hazard on the urban site where the tree is growing.

257

8.6 References

Aloni R, Alexander J D and Tyree M T (1997) Natural and experimentally altered hydraulic architecture of branch junctions in Acer saccharum Marsh. and Quercus velutina Lam. trees; Trees: Structure and Function 11, 255-264.

André J P (2000) Heterogeneous, branched, zigzag and circular vessels: unexpected but frequent forms of tracheary element files: description, localization, form; In: Savidge R, Barnett J and Napier R (ed.s) Cell and molecular biology of wood formation; Oxford, Bios Scientific Publishers, 387-395.

Baas P, Ewers F W, Davis S D and Wheeler E A (2004) Evolution of xylem physiology; In: Helmsley A R and Poole I (ed.s) The Evolution of Plant Physiology; New York, Elsevier, 273-295.

Bonser R H C and Ennos A R (1998) Measurement of pre-strain in trees: implications for the determination of safety factors; Functional Ecology 12, 971-974.

British Standards Institution (2010) BS3998:2010 Tree work – Recommendations; London, BSI.

Brüchert F and Gardiner B (2006) The effect of wind exposure on the tree aerial architecture and biomechanics of sitka spruce (Picea sitchensis, Pinaceae); American Journal of Botany 93, 1512-1521.

Burns L A, Mouritz A P, Pook D and Feih S (2012) Bio-inspired design of aerospace composite joins for improving damage tolerance; Composite structures 94, 995- 1004.

Chaskalovic J (2008) Finite element methods for engineering sciences; New York: Springer Science and Business Media.

Crook M J and Ennos A R (1996) The anchorage mechanics of deep rooted larch, Larix europea x L. japonica; Journal of Experimental Botany 47, 1509-1517.

Crook M J, Ennos A R and Banks J R (1997) The function of buttress roots: a comparative study of the anchorage systems of buttressed (Aglaia and Nephileum ramboutan species) and nonbuttressed (Mallotus wrayi) tropical trees; Journal of Experimental Botany 48, 1703-1716.

258

Dumont E R, Grosse I R and Slater G J (2009) Requirements for comparing the performance of finite element models of biological structures; Journal of Theoretical Biology 256, 96-103.

Eisner N J, Gilman E F, Grabosky J C and Beeson Jr. R C (2002) Branch junction characteristics affect hydraulic segmentation in red maple; Journal of Arboriculture 28, 245-251.

Evans L S, Kahn-Jetter Z, Torres J, Martinez M and Tarsia P (2008) Mechanical stresses of primary branches: a survey of 40 woody tree and shrub species; Trees: Structure and Function 22, 283-289.

Fisher J (1986) Branching patterns and angles in trees; in Givnish T (ed.) On the Economy of Plant Form and Function; Cambridge University Press, 493-523.

Forestry Commission (2016) ForestGALES; webpages last accessed 29.03.2016 at: http://www.forestry.gov.uk/forestgales

Freyburger C, Longuetaud F, Mothe F, Constant T and Leban J-M (2009) Measuring wood density by means of X-ray computer tomography; Annals of Forest Science 66, 804-813.

Gartner B L ed. (1995) Plant Stems; Physiological and Functional Morphology; New York, Academic Press.

Gilman E F (2003) Branch to stem diameter affects strength of attachment; Journal of Arboriculture 29, 291-294.

Gilman E F (2011) An illustrated guide to pruning; 3rd edition; Independence, Kentucky US, Cengage Learning.

Harris R W, Clark J R and Matheny N P (2004) Arboriculture: Integrated Management of Landscape Trees, Shrubs and Vines; New Jersey US, Prentice Hall.

Hedges S B (2002) The origin and evolution of model organisms; Nature Reviews Genetics 3, 838-849.

Helliwell R (2004) A discussion of the failure of weak forks; Arboricultural Journal 27, 245-249.

259

ITFD (2015) Failure Type Report; California, International Tree Failure Database; Accessed 10/07/2015 via http://svinetfc8.fs.fed.us/natfdb/reports.aspx

Jaffe M J (1973) Thigmomorphogenesis: The response of plant growth and development to mechanical stimulation; Planta 114, 143-157.

Jaffe M J and Forbes S (1993) Thigmomorphogenesis: the effect of mechanical perturbation on plants; Plant Growth Regulation 12, 313-324.

James K R (2003) Dynamic loading of trees; Journal of Arboriculture 29, 165-171.

Jungnikl K, Goebbels J, Burgert I and Fratzl P (2009) The role of material properties for the mechanical adaptation at branch junctions; Trees: Structure and Function 23, 605-610.

Kane B, Farrell R, Zedaker S M, Loferski J R and Smith D W (2008) Failure mode and prediction of the strength of branch attachments; Arboriculture and Urban Forestry 34, 308-316.

Kramer E M (1999) Observation of topological defects in the xylem of Populus deltoides and implications for the vascular cambium; Journal of Theoretical Biology 200, 223-230.

Kramer E M and Borkowski M H (2004) Wood grain patterns at branch junctions: modelling and implications; Trees: Structure and Function 18, 493-500.

Lev-Yadun S and Aloni R (1990) Vascular differentation in branch junctions of trees: circular patterns and functional significance; Trees: Structure and Function 4, 49-54.

Lonsdale D (1999) The Principles of Tree Hazard Assessment and Management; London, TSO.

Lonsdale D (2000) Hazards from Trees: A General Guide; Edinburgh, Forestry Commission.

LTOA (2013, May 24) Symposium on junctions in trees conducted at the meeting of the London Tree Officer Association, King’s Cross, London.

Mackerle J (2005) Finite element analyses in wood research: a bibliography; Journal of Wood Science Technology 39, 579-600.

260

Mattheck C, Bethge K and Schäfer J (1993) Safety factors in trees; Journal of Theoretical Biology 165, 185-189.

Mattheck C (1998) Design in Nature – Learning from Trees; Berlin, Springer-Verlag.

Mayer G (2006) New classes of tough composite materials – lessons from natural rigid biological systems: Materials Science and Engineering 26, 1261-1268.

Metzger K (1893) Der wind als massgebender factor für das waschstum der bäume; Mündener Forstliche Hefte 3, 35-86.

Müller U, Gindl W and Jeronimidis G (2006) Biomechanics of a branch-stem junction in softwood; Trees: Structure and Function 20, 643-648.

Neely D (1991) Water transport at stem-branch junctures in woody angiosperms; Journal of Arboriculture 17, 285-290.

Niklas K J (1992) Plant Biomechanics: an Engineering Approach to Plant Form and Function; Chicago US, University of Chicago Press.

Niklas K J and Spatz H-C (2012) Plant Physics; Chicago US, University of Chicago Press.

Pfisterer J A and Spatz H-C (2006) ‘Biomechanical behaviour of branch junctions’; In: L. Salmen (ed.) 5th Plant Biomechanics Conference (Stockholm), 61-66.

Rayfield E J (2007) Finite element analysis and understanding the biomechanics and evolution of living and fossil organisms; Annual Review of Earth and Planetary Sciencies 35, 541-576.

Reddy J N (2005) An introduction to the finite element method; 3rd ed.; New York: McGraw-Hill.

Shigo A L (1985) How tree branches are attached to trunks; Canadian Journal of Botany 63, 1391-1401.

261

Shigo A L (1989) Tree pruning – a worldwide photo guide; Durham, New Hampshire US, Shigo & trees, associates.

Shigo A L (1991) Modern Arboriculture; Durham, New Hampshire US, Shigo & trees, associates.

Smiley E T and Lilly S (2014) Tree Support Systems – Cabling, Bracing, Guying and Propping; Best Management Practices series; Illinois, US, ISA.

Sprugel D G, Hinckley T M and Schaap W (1991) The theory and practice of branch autonomy; Annual Review of Ecological Systems 22, 309-334.

Thuvander F, Sjödahl M and Berglund L A (2000) Measurement of crack tip strain field in wood at the scale of growth rings; Journal of Materials Science 35, 6267- 6275.

Tyree M T and Alexander J D (1993) Hydraulic conductivity of branch junctions in three temperate tree species; Trees: Structure and Function 7, 156-159.

Vincent J (2012) Structural biomaterials; 3rd Edition; New Jersey, USA, Princeton University Press.

Zimmermann M H and Brown C L (1971) Trees: Structure and Function; Berlin Heidelberg New York, Springer-Verlag.

Zimmermann M H (1978) Hydraulic architecture of some diffuse porous trees; Canadian Journal of Botany 59, 2286-2295.

262