In the previous post we looked at VALID's Goldilocks Likelihood of Occupancy canvas to explore categories that are not too wide, and not too narrow, that are just right.
This Likelihood of Occupancy canvas is useful to examine tree risk decision-making at the High Court in the UK’s landmark Poll v Bartholomew Judgment.
In Poll, a motorcyclist was seriously injured by a falling Ash stem. The Judge found for the Claimant because the tree was ‘High Risk.’
In their reports, the Claimant’s expert said the tree was ‘High Risk’ and the Defendant’s expert didn’t mention any risk. Yet in their Joint Statement, the experts agreed the tree was a ‘Medium Risk.’
Naturally, the expert’s opinions left the court scratching its head and it had to ask them to produce a Second Joint Statement to define what they meant by high and medium risk.
In the Second Joint Statement, the experts told the court the tree was ‘High Risk’. However, they concluded the risk was high after they'd assessed the Likelihood of Occupancy for a minor road at 50%, when in fact it was 1%. The experts overvalued the occupancy by a whopping factor of 50. This gaffe was so enormous the tree was in fact a ‘Medium Risk’ and not a ‘High Risk.’
The Judge would’ve found for the Defendant if the tree was a ‘Medium Risk.’
In the Scale of the Problem we saw the overall risk from branches and trees falling is so extremely low we need a microscope to see it.
The only sensible way to measure risks that go this low is to use a logarithmic scale. It turns out the Goldilocks logarithmic scale for tree risk that’s not too narrow, and not too wide, that's just right is log base 10. Just like the Richter Scale is for measuring earthquakes.
So far so numberwang. What does that mean for you?
Well, when it comes to Likelihood of Occupancy decision-making, the advantages of log base 10 are obvious when drawn to scale. There’s 5 colour-coded Likelihood of Occupancy categories in VALID. If we centre High, then you can only see a bit of the heel of Very High. Nearly all of it’s off the screen. You can make sense of Moderate but you can barely make out Low. And you can’t see Very Low at all.
So, the first decision a Validator makes with Likelihood of Occupancy is what 3 categories can’t it be? Which 3 make no sense? Once calibrated this is an effortless decision. Then it comes down to one of two. Usually, which one is the most obvious because they're huge canvases. If in doubt you go for the higher one.
We regularly use visuals to convey both the context of tree risk and how you can measure it. Recently, we shared a post with Professor David Ball's quote that the prospects of reducing the risk below the current level were comparable to finding a microscopic needle in a gargantuan haystack.
To help convey that finding a microscopic needle in a gargantuan haystack simile further, we've had a play around with illustrating the tree risk context using coloured spectrums.
In the graphic, the top spectrum shows the reality of tree risk to scale. We know that compared to other everyday risks we readily accept, the overall risk to us from branches or trees falling is extremely low. Our annual risk of being killed or seriously injured is less than one in a million. At this scale, we can't see the amber and red risks. We'd need a microscopic.
If we take the risk spectrum to a scale where we can just make out the risks that we're trying to find and manage, we have to overvalue the base-rate risk by a factor of more than 1000.
Taking the 'defect' out of tree risk-benefit assessment
Has been published in the spring edition of the Arboricultural Association's Arb Magazine. You're welcome to download a pdf copy by clicking the link above or the image below.
Here's the introduction to whet your appetite.
"We’ve grown up being told that when we assess tree risk we should be looking out for tree ‘defects’. The problem with this approach is what are commonly labelled as defects often aren’t defects at all."
Any publication about tree risk management lacks credibility if it neglects the overall risk. That's because the overall risk from branches and trees falling and causing death, injury, or property damage provides the 'Context' (ISO 31000 - Risk Management). It gives us a base rate.
This is the context of the overall risk in VALID's Tree Risk-Benefit Management Strategies.
"Compared to other everyday risks we readily accept, the overall risk to us from branches or trees falling is extremely low. Our annual risk of being killed or seriously injured is less than one in a million. That's so low, we're at greater risk from a 200 miles (320km) round trip drive to visit friends for a weekend than from branches or trees falling for a whole year. Given the number of trees we live with, and how many of us pass them daily, being killed or injured by a tree is a rare event; one that usually happens during severe weather."
Why is establishing context and base rate so important? A risk expert nails it.
VALID has four easy to understand traffic light coloured risk ratings, and this is where they sit in the Tolerability of Risk Framework (ToR).
The Tolerability of Risk Framework is an internationally recognised approach to making risk management decisions where the risk is imposed on the public.
The ToR triangle gets fatter and redder where more attention and resources should be allocated to managing the risk. It gets thinner and greener where less attention and resources should be allocated.
Where ToR is amber the risk is Tolerable if it’s ‘as low as reasonably practicable’ (ALARP) - where the costs of the risk reduction are much greater than the value of the risk reduction.
VALID has applied ToR to tree risk but has removed the numberwang because:
1) Tree risk has too much uncertainty to credibly measure at single figure accuracy with risks like 1/4, 1/300, 1/20 000, or 1/500 000 000.
2) Risk outputs as probabilities create friction in communication because many people struggle with numbers. Research shows that about 25-33% can't rank 1:10, 1:1000, and 1:100 risks from highest to lowest.
3) The risk assessor and duty holder are spared the complexity of numerical cost-benefit analysis in the amber ALARP zone.
Recently, we caught a podcast where a tree was declared 'safe' if it's less than 30% hollow. We think they meant 70% hollow. Either way, this isn't right for several reasons.
We've posted about this before, but as long as this kind of mistake is being broadcast we think it's worth repeating so the message gradually gets home.
The heart of the confusion is the t/R = 0.3 fallacy. t/R = 0.3 is when a residual wall thickness (t) is 30% of the stem radius (R). It's often cited as a failure threshold. It's not. The 'Why t/R Ratios Aren't Very Helpful' pdf explains why in detail.
In short, one reason is because of a geometric property called section modulus. Wind load and material properties remaining equal, if you double the diameter you increase the load bearing capacity of a tree by 8 times.
To add to the confusion, t/R 0.3 is often referred to as 70% hollow. In fact, a 0.3 t/R ratio is only 50% hollow. 70% is the radius, which is one dimension. t/R 0.3 is the area, which is two dimensions.
This graph from Paul Muir shows the relationship of central hollowing on:
A = Cross Sectional Area
Z = Section Modulus
t/R = 0.3
A = 49% loss of cross sectional area
Z = 24% reduction in load bearing capacity
To make matters worse. A tree with a t/R ratio of 0.3 can have a very high likelihood of failure, or it can have a very low likelihood of failure.
If all that wasn't enough, it's seldom that where decay is of concern we're dealing with a cross sectional area of a tree that's a circle.
"The implications of recent English legal judgments, inquest verdicts, and ash dieback disease for the defensibility of tree risk management regimes"
We've had several requests for a better quality image that's part of a discussion about this article on the UKTC (attachments on this group have to be below 180kb). Click the image to enlarge it.
You can download Jeremy's article about tree risk management here.
Since then, we've had further requests to set out the points in this big canvas with a step-by-step guide to make it easy to follow.
We're genuinely surprised the article has been peer-reviewed, let alone published in a journal. It's not research. Some obvious key points of fact don't make sense, even within the questionable logic of its own risk ecosystem. We've sketched them out in the above image so you can see the whole picture, and described them below. We're surprised they weren't picked up during the peer review.
The matrix has high risk, low risk, and medium risk outputs.
So, we've got a Tree Risk Matrix
High & High = High Risk
High & Low = Medium Risk
Low & High = Medium Risk
Low & Low = Low Risk
Somehow, we've gone from a Tree Risk Matrix world where:
High & Low = Medium Risk
And that's before we consider the really important stuff, like what does High, Medium, and Low actually mean, and how do you go about measuring them? Unless clearly defined, words like High, Medium, and Low are what Philip Tetlock calls 'vague verbiage'. They're illusions of communication, or 'bafflegab' as we call it. Further still, you can't reasonably model tree risk by adding or subtracting vague words or by mixing up traffic light colours.
Exploring the low occupancy = acceptable risk statement further.
That low occupancy has no clear definition or meaning in Jeremy's Tree Risk Management Frameworks should be particularly worrying for a duty holder. In VALID, low occupancy is clearly defined and there's no ambiguity. We don't burden the duty holder with trying to second guess what we mean by low occupancy. The reason why low occupancy = acceptable risk should be particularly worrying for a duty holder following Jeremy's advice is that in VALID we have several scenarios where low occupancy has risks that are Not Acceptable or Not Tolerable.
Infrequent or very low use is a higher level of occupancy than low
To make matters worse. In Jeremy's 1:10,000 Time Bomb he describes this footpath (below) has having infrequent or very low use. He outlines that following Health & Safety Executive guidance (Sector Information Minute), every year the path is walked by a person with a working knowledge of trees who gives them a quick visual check. Because these trees are being checked annually that means in Jeremy's tree risk management vocabulary, infrequent use or very low use is a HIGHER level of occupancy than low occupancy - remember, trees in low occupancy don't need checking.
Clearly, any duty holders following the guidance in Jeremy's Tree Risk Management Frameworks could quite reasonably classify the infrequent or very low use of this footpath as low occupancy and not check the trees.
This could be a substantial vulnerability for duty holders because in his 1:10,000 time bomb presentation, Jeremy makes a case for a claim being made against them if a small diameter deadwood branch from an Ash tree falls and causes significant head injuries to someone walking along this path. Even though he describes the risk as being at the lower end of his risk spectrum, the duty holder is expected to have removed the deadwood because it wouldn't have cost that much to do it.
These are just some of the more obvious concerns we have with Jeremy's take on tree risk management in his article.
Passive Assessment | The invisible gorilla in the room
There's a famous psychological experiment called the invisible gorilla. In it, you're asked to watch a short video of six people passing a basketball. Three of them are wearing white shirts and three of them black shirts. You're asked to count how many passes are made by the white shirts. Most people get the number of passes right. Because they're focused on this, what half the people don't see is a gorilla walk amongst the players, stop, face the camera, thump their chest, and walk off.
To half the people, this very obvious gorilla is invisible.
I recently found one of my invisible gorillas. Whilst putting a flowchart together for VALID's Tree Risk-Benefit Management Strategies, I realised my invisible gorilla was Passive Assessment.
Passive Assessment, and not Active Assessment, is a duty holder's most valuable tree risk-benefit management asset because;
This tree risk assessment review article by Peter Gray, from the Summer 2020 issue of Arboriculture Australia's 'The Bark', might be of interest to you.
1) The 'mathematics professor' and the risk model
The 'mathematics professor' isn't a mathematics professor. His name's Willy Aspinall and he's the Cabot Professor in Natural Hazards & Risk Science at the University of Bristol. He's a 'risk professor' who we worked with when developing VALID's risk model that does the hard work behind the scenes in the App. He's driven the model to breaking point and this is what he has to say about it:
“We have stress tested VALID and didn’t find any gross, critical sensitivities. In short, the mathematical basis of your approach is sufficiently robust and dependable for any practical purpose.”
2) Risk overvaluation? - Death by numberwang
"The risk of harm* for incidents involving motor vehicles (not motor cycles) appears to be high. There is little evidence of people being killed from cars running into fallen trees but this still apparently has a significant input to the calculated risk of harm."
This gets a bit detailed. In short, the risk isn't too high in VALID's risk model when it comes to vehicles.
There's a couple of points here. First, VALID's risk model doesn't try to measure a death. Death is too narrow and accurate a consequence for a risk that has this much uncertainty. Here's the long answer, and we shared a version of this with Peter after reading his thoughtful article.
The essence of the conundrum with traffic is that it's seldom that a death occurs unless a tree or a large branch hits the cab space of a moving vehicle. So, how do you go about modelling the consequences when they're usually a vehicle driving into a tree during its stopping distance, rather than tree part hitting the relatively small cab space?
The answer's quite complicated because it's a combination of risk modelling from published traffic accident data, the Abbreviated Injury Scale, Occam's Razor (have the least assumptions), running what's called sensitivity analysis, and ease of use.
Perhaps most importantly, 'a difference is only a difference if it makes a difference'. We don't have the data to confirm this because it doesn't yet exist, but we suspect VALID's risk model is over-valuing the consequences in some parts of traffic because of the safety measures that cars have to protect the occupants during accidents; the model's erring on the side of caution for consequences. However, does this matter? Does it make a difference to the actual risk output?
To test this, in the model, we can drop the consequences one or two orders of magnitude in each scenario, with sensitivity analysis, and see how much a difference that makes to the risk. Doing this, it's clear red risks aren't turning green. That means the duty holder is still going to do something to reduce the risk even if the consequences are overvalued.
Cranking up the numberwang to try and model trees and traffic more accurately is not only fraught with mathematical problems and increased uncertainty, but it's not going to make a difference to the decision-making of the assessor or duty holder. It would also add another layer of complexity to the risk model when it comes to decision-making where you combine traffic and people in high-use occupancy. So why try to do it?
An analogy we've used is that we're looking at a big risk picture, and in one part of the canvas that's dealing with traffic and consequences it's a bit blurry. We could get down on our hands and knees with a magnifying glass and spend ages trying to make that consequences part a bit sharper. But when we've finished and taken a step back to admire our work, the risk picture isn't noticeably different.
If you'd like to stay in touch and be kept up to date with an occasional short and snappy newsletter every few months or so, please subscribe.
The kind of thing we'll share with you are some of the choice
cuts from this News page. We'll also let you know when there's
new training dates.
What is VALID?
© VALID is a not-for-profit organisation