POPULATION LEVEL STORM RESILIENCE MODEL THREATS_SRM Julian Forbes - - PowerPoint PPT Presentation

POPULATION LEVEL STORM RESILIENCE MODEL THREATS_SRM

Julian Forbes‐Laird

BA(Hons), MICFor, MRICS, MEWI, M.Arbor.A, Dip.Arb.(RFS) www.flac.uk.com

A method produced for Network Rail Infrastructure Ltd Implemented with Astrium Geo‐Information Services

Origins

‐ 23 May 2011 ‐ 03 January 2012 On both dates Route Scotland was effectively closed during high winds due to the large number of windthrown trees External report recommended ‘enhanced tree removal’ ORR: implement recommendation or develop risk‐based model FLAC commissioned to research & design a model for identifying storm vulnerability to enable targeted tree removal

First principles

What is a dangerous tree?

Civil engineering protocol for management of faulted structures: If a load‐bearing structure has suffered primary failure (e.g. cracking has developed), it has proven unfit for purpose and should be taken out of service prior to suffering secondary failure (i.e. collapse) Being dynamic living organisms, trees’ ability to self‐repair complicates the identification of primary failure, but when this is observed, the tree can properly be described as dangerous: No predictable delay prior to secondary failure

THREATS_NR

• A bespoke adaptation of THREATS for the railway

environment

• The only railway‐specific TRA system
• Developed by JFL in 2009 for NR to use in its NLTS
• >3.4million trees inspected on foot
• Just over 24K hazard trees identified (7%)
• Tree failures onto the railway reduced by >70% in

four years

• The vast majority of ongoing failures are third party

trees & non‐defective windthrow

23m high third party larch, 1.1m stem dia., very severe basal decay (Ganoderma resinaceum), weighted towards 2 x 65mph lines, 165m from bridge abutment Failure Score Imminent/Immediate 50 x Target Score Very high value 40 x Impact Score Default 10 (STS) 10 = Total 20000 Risk Category 7 ‐ Emergency

Hazard tree found by JFL, 23.IV.09 Primary failure had already occurred!

WEST LAVINGTON, July 2010

‐ Off‐site tree failed onto the railway (Armillaria) ‐ Collision in fog with 95mph express, >300 passengers ‐ 1200m stopping distance under emergency braking ‐ Impact occurred <100m from 4m embankment ‐ RAIB investigation into accident also examined NR tree risk management system… ‐ THREATS / THREATS_NR subject to detailed scrutiny ‐ Probably the most detailed & technical review undertaken

• f a TRA system

Taiwan, April 2011 Derailment of low speed, narrow gauge train: 5 dead, 107 injured

THREATS_SRM

Prior research

Lochawe research project, December 2010: stability of trees on slopes: Failure criteria established for tree height v degrees of slope FLAC used LiDAR mapping (DTM/DSM differential) to find veg height & degree of slope Proof of concept was achieved for use of LiDAR for identification of tree failure predisposing factors

SRM rationale

Trees fall over in high winds. Right? Wrong… Some trees fall over in high winds To design the storm resilience model, I asked the question: WHY? Why only some…

FLAC investigated >40 tree failure sites to record underlying site‐related causal factors… for example: ‐ Topographic exposure ‐ Poor drainage ‐ Steep slope ‐ Width of tree belt ‐ Railway orientation Separately or in combination, the factors derive the SRM failure score

Risk Factor examples

Topographic exposure Where the tree or group of trees exceeds by >1/3 the height of

• ther trees or adjacent terrain (within 50m) to windward

(relative to the prevailing wind), including where the tree is located on the upper slopes of a cutting such that it exceeds the cutting lip by this amount Railway orientation from 90‐180˚ & from 270‐360˚ This factor seeks to express situations where the railway is flank‐

• n to the prevailing wind. Effectively, this covers railway
• rientations from NW‐SE through South

Key explanation

The objective of the SRM is to identify mechanically similar sites to those where failures are known to have occurred, and to set SR thresholds for them as the basis of network‐ wide tree management and risk control strategies THREATS_SRM does not seek to model tree risk, so much as site risk: some locations are inherently vulnerable to tree failure due to site‐related factors THREATS_SRM provides a framework for identifying & stratifying these sites

• Having identified the factors involved, the

next challenge was to map them

• NR commissioned a LiDAR, RGB and partial IR

survey of the Scotland network

• Undertaken by Astrium GIS (formerly

InfoTerra), a subsidiary of EADS

Introduction

How it works

The model relies on computer‐based automatic interrogation of the LiDAR dataset, based on the SRM parameters The LiDAR data interrogation uses three algorithms designed by FLAC: ‐ Approximation of stem position from polygon centroid ‐ Estimated tree count from height & area of canopy ‐ Method for inferring stem diameter at track impact point, based on LiDAR data for tree height

Stem dia. at track

Height / stem diameter ratios

A measured dataset of 5000 trees was interrogated to find H/D correlations that could be used to infer stem dia. from tree height (as measured by LiDAR). We found:

• 6m to <14m

H/D = 31

• 14m to <21m

H/D = 36.5

• ≥21m

H/D = 27 (modeled up to 35m)

Stem dia. at track impact point

The formula for this is:

• D = Db (1‐H/Ht)
• Where:
• Db = computed stem diameter
• Ht = total height of tree
• H = height above ground being considered (= track offset)
• D = diameter of tree at height H (= dia. of timber on track)

Worked examples

1. A tree of height (HT) 12m has a computed dia. (Db) of 372mm. For a 7m track offset (H): subtract 7m from 12m to get 5m which is 41.6%, so timber dia. at track (D) will be 372 x .416 = 155mm, falling within 150‐ 350mm size range for Impact Score 2. 19m tree with a 10m track offset has a computed dia. of 513mm. 19m minus 10m gives 9m or 47.4%, so timber dia. at track will be 513 x .474 = 244mm, also falling within 150‐350mm size range for Impact Score 3. The same tree as (2) but with a 7m track offset: 19‐7 = 12 or 63.1%, so

• dia. at track will be 513 x .631 = 323mm. This trips a mid‐point uplift

threshold and so would be scored as 350‐750mm

The survey corridor as seen by LiDAR…

Airborne data corridor – 10cm imagery

Airborne data corridor – 50cm LiDAR

Semi‐automatic tree canopy identification

Estimate the trunk position

Estimate the crown extent

Distance to rail

Trunk impact diameter

Distance to bridges

Overall tree risk (THREATS_SRM scoring)

Overall tree risk – by management unit

Outcomes

1.1million polygons identified from the LiDAR dataset Each polygon returns 88 data cells… 96.8 million pieces of data Of 1.1mi polygons, only 1500 found to be in Site Risk Categories 6 & 7, or 0.14% ORR briefed on THREATS_SRM: “Excellent” “Delighted”

Does it work?

At worst, the answer lies somewhere between maybe & probably Time – and the next storm – will tell BUT: It’s better than a) nothing and b) felling all the trees AND: Applying the model to subsequent non‐defective failures identifies the same site risk factors as causative

Other applications of THREATS & remote sensing… THREATS_EW

A risk assessment model for landslip

THREATS_WF

A risk assessment model for wildfires

Thank you for listening