Strength Loss models

Several models for strength loss have been discussed in recent years, with attempts to quantify the amount of decay and estimate strength loss or failure potential. Matheny and Clark, in “Evaluation of Hazard Trees in Urban Areas” (International Society of Arboriculture, 1994 second edition) review several of the proposed methods, including the Wagener model and the Bartlett modification of it. Dr. Claus Mattheck has proposed a comparison of wood thickness to defect radius.

I’m curious how many arborists are using any of these methods, or if other methods are substituted. How often do you apply these guidelines?

Another aspect of this is how do you use the information to determine treatments to apply? For example, if you have a tree with strength loss indicated as close to the threshold, how do you know if pruning or other treatments will be sufficient? How much pruning? What if the tree has already been pruned, and you know that the crown is not the normal size? How do you reconcile these aspects with the thresholds?

And finally, what thresholds do you normally recommend? Do you vary from those under different circumstances?

Russ Carlson

Reply to post by Russ Carlson, on February 18, 1999 at 23:38:51:

Russ it would also be interesting to know how the Standard of Care (what another prudent and reasonable person would do) varies by region or locale for both professionals (their methodology, recommendations, etc) and property owners (duty to discover and to act). At a more basic level what are community standards for acceptable vs. unreasonable risk?

Don't just quote the books guys. What are the variable community standards of how the textbooks are understood and apply?

We've already seen on this board a range of opinions from "cut 'em all down right away" to "move the expressway to preserve that hollow tree."

Reply to post by Russ Carlson, on February 18, 1999 at 23:38:51:

There is a fundamental problem with strength-loss models that makes them all less than very useful. They only model the part of the tree that is measured. Only if there is a way to measure a part of the tree that is below the critical threshold, is the model of any use.
This is a ubiquitous sampling problem. For example, I can sample for a root disease by digging up certain roots from a tree and looking at them. If I dig up an infected one, the sample tells me that the tree is infected. If I do not dig up an infected one, I cannot know that the tree is uninfected.
If I measured a part of the tree that is in the model's hazard zone, I suppose I would cut the tree down. However, I have never been hired to apply a model to a tree. Therefore, I prefer to do it the old way.

Stochastic: 1. Random; specif: involving a random variable 2: involving chance or probability: PROBABILISTIC (Merriam Webster's Collegiate, 10th ed.)

Reply to post by Peter Torres, on February 18, 1999 at 23:38:51:

Peter,

I think the strength loss models need to be part of a greater analysis. Much of our judgment in regard to risk of failure is based on what is unseen, so the more we can add to the part of the analysis we can see or understand the better our judgment might become.

The key element is indeed probablity, the liklihood of failure. It is helpul to conceive of the analysis in terms of the old plant disease triangle (you know: host, pathogen, environment). The tree hazard triangle includes risk (or liklihood or probablity) of failure, loading and targets. Liklihood of failure is the weakest link (and really follows from the interaction of strength and loading).

If we observe trees that have not failed under similar loading conditions we establish some sort of norm or threshhold. If we observe trees that have failed and can link those failure to some departure from norm we start to develop a model. If we measure certain characteristics such as proportion of sound trunk wood or sound roots we start to develop decision criteria in the model.

So strength loss, as sketchy as it might be, is just one element of the model.

I've long felt that we indeed need a better statistical profile of the trees that exhibit certain departures from the norm (i.e. they have some decay or cracks or root damage or whatever) but do not fail. If anything our models may be quite conservative.

This brings us back to my post to Russ. What are community based standards for what acceptable and unacceptable risks of failure are? If you want to use the appraisal analogy: it's not up to the appraiser to tell society what value should be, the appraiser looks at what society considers valuable and develops an opinion for a certain object. Is it up to the arborist to tell society what risk to accept? Or does the arborist tell society what the risk or proability of failure is and let society apply it's ideas of wjat is acceptable?

Reply to post by Peter Torres, on February 18, 1999 at 23:38:51:

Peter, you say you prefer to do it "the old way". What do you consider that to be? You are obviously using some criteria to make a judgement, so I'm curious what those criteria are.

You are correct that the strength loss tests require sampling sufficient to provide a representative profile. I use some of these techniques quite a bit. I recognize that they are limited, but take those limitations into account when applying the principles. I won't say "this tree is one percent under the threshold, so it can stay." I rely heavily on my experience and knowledge, and only use these tests to collect additional information.

To be effective, the sampling must be done where the most information will be gained. If the tree has a cavity at 12 feet, I would be foolish to sample at 3 feet and say that is representative. Since I am trying to determine the condition of the weakest point, I would climb up to 12 feet and sample there. I would NOT sample at 3 feet, if I didn't think it was a weak point or at risk of failure there. I want to know where the risks are. Granted, you may miss the highest risk completely, but this is where the rest of your training in identifying tree defects comes in. Here is a good place to reference Claus Mattheck's work, and The Body Language of Trees (1994)

Reply to post by Scott Cullen, on February 20, 1999 at 22:12:08:

The biggest problem is that there is an alnmost infinite degree of variability within any one tree, and between trees in a stand. The model, as many models are, is simplistic and can only ever model a few of the many factors involved, many of which we may not even have begun to understand yet.

The best approach is still, in my humble opinion, to follow the CTF data base approach and systematically document failures and then, once enough data points are asembled, we will begin to have a good idea of what the species profile looks like. My most studied tree is the western hemlock, and after several thousand assessments I have an intuitive ides of what external features will mean in terms of internal condition, yet I am still unable to quantify my knowledge because I don't have enough hard data.

We should be making a concerted effort across NorthAmerica to establish a more global reporting mechanism for tree failure patterns. With time we could all benefit from knowing these detailed failure patterns, and what they imply for our assessments. In the meantime we all use experience and a mix of terchnical and anecdotal evidence to guide our decisions. I have great reluctance to use the strength loss models. Moreover, to my horror, the BC Workers Compensation Board has adopted the 30% figure as an acceptable threshold for hazard assessments in any species. If I find only 30% left in a hemlock it has very advanced root / butt rot and should come out immediately, yet we are directed that it is safe. The same would apply to Doug Fir, less so the red cedar. We face a danger that adoption of the strength loss model in the absence of species specific knowledge will lead us into more problems, not less.


Julian

Reply to post by Julian Dunster, on February 21, 1999 at 07:17:33:

Julian,

Intuitively, I agree with your suggestion that the most useful tool would be a large data set documenting failure patterns by species and locale AND non-failure patterns as well. If a tree with certain observeable conditions fails, how many with similar characteristics in close proximity din not fail under what could be assumed to be similar loading conditions.

It's beyond my technical and statistical abilities, but I would think that what would have been a cumbersome process a few years ago could be structured quite elegantly using this on-line medium. An infinite number of data collectors with an infinite number of data entry stations, no paper forms, no mailing, no transcription (effort or errors). I guess it would require a host staff and computer, but maybe that could be automated as well.

RE: BCWCB. I assume the workers compensation board's safety concern has to do with worker (climber?) safety. Assuming 30% remaining strength is a safe climbing work environment, it must be distinguished from what is safe in defining a safe target environment. In the latter there are two variables: risk of failure (and maybe 30% is unacceptable) and presence or value of targets.

Say your tree has 70% strength loss and by your judgment it has to go. Or maybe it has only a 50% strenght loss and you still say it has to go because of the targets. It would appear safe enough to climb to accomplish that removal (even if it's ony to hook up the creane choker). By contrast there is a 75% strength loss (or whatever other critera you would use) and you decide that because it is in a protected stand (low loading) and there are no direct targets it can stay. But the owner wants it down for aesthitic reasons. BCWCB would say it's unsafe to climb.

Is my interpretation of BCWCB's purpose correct?

We really need to be careful in mixing up rating and applications. Same goes for terms. A300 for example, I think, uses hazard with repect to worker hazard, but most consultants think of hazard to other targets.

Reply to post by Scott, on February 21, 1999 at 13:07:19:

Russ.
The WCB approach has nothing to do with climbers. It is a standard for assessing danger trees and that is my worry. It is adopted as a test to determine whether or not the tree is safe to retain. I would argue that even if the required 3 borings indicate there is at least 30 % shell thickness, such a large loss of wood in some species spells instant problems, but by setting a numerical standard we may be setting ourselves up for many problems, and trees do not lend theselves to simple quantification.

Julian

Reply to post by Scott, on February 21, 1999 at 13:07:19:

Scott, you wrote "A300 for example, I think, uses hazard with repect to worker hazard, but most consultants think of hazard to other targets"
I am a big champion of using worker safety in the overall hazard equation. For example, if a Dougfir has red ring rot, which is a heartrot, the weakest point cannot be measured to apply a model to, because it cannot be found, for that is the nature of red ring rot (alias white pocket rot).
Luckily, the loss of strength in affected wood is far less than it would be with many other rots, includy all of the brown rots. The wood retains much elasticity and plenty of compressional strength.
So in a yard that is big enough to drop a badly infected tree with a butt cut, I usually advise leaving the tree, because it will last a long time (perhaps) and also make an excellent wildlife tree.
But if the tree will need to be cut down using rigging from itself (and some shock load), I say cut it down now because it will oneday endanger a climber.
Chunking comes somewhere in-between

Reply to post by Russ Carlson, on February 20, 1999 at 22:12:08:

Russ, as Julian wrote, "trees do not lend theselves to simple quantification".
I understand the usefulness of models and of objectivity.
Julian's statement is another way of defining "stochastic". There are so many variables that, like the weather itself, it is not feasable to predict failure of a tree most of the time.
We can use knowledge of a statistical nature to bracket the likelihood of failure.
As important as that likelihood, is the risk acceptance of the equation.
What criteria do I use? Human terms. I talk about the targets alot. Do not ever park under this tree. That gives the client clues as to what I think will happen. As to how I decide, you ask a tough question. It is a cop-out to say "by experience" or related statements, because if I have the experience, then I know what it is.
So... . Let me get back to you on that.

Reply to post by Julian Dunster, on February 21, 1999 at 07:17:33:

Clue me in guys, my understanding of Matthecks work is that a shell thickness of 30% is minimal. The bartlett formula stipulates a 30% strengh loss (or 70% remaining strengh) as the minimum strengh threshold. When Matthecks minimal shell wall model is run through Bartletts formula, they agree pretty well.

Unless I have really missed something big, nobody is advocating leaving a tree standing with 30% residual strengh!

Have I got the figures right on the models or have i missed it clean and should take up something like knitting?

Wayne

Reply to post by Peter Torres, on February 21, 1999 at 17:37:59:

A very useful aspect of the analysis. Certainly the future includes not only the likey progression of risk of failure but the costs and risks of future mitigation.

I'd still suggest worker safety and target safety may have different thresholds; may not be considered in the same analysis (you get called later in the progression, the tree has to go now for safety of targets, is it safe to work in or around?).

It may also be that it's not either-or (do it now or it won't be safe to do later). The worker safety issue may shorten the progression period but not to today or it may leave the progression period the same (till target safety is an issue) but increase the mitigation cost (crane, helicopter, skyhook, whatever) and some of these may remain owner or manager decisions once you've given your best analysis.

Reply to post by Wayne Cahilly, on February 21, 1999 at 13:07:19:

Keep in mind that Mattheck's method is t/R, where t is the thickness of the remaining wood (thinnest point) over the RADIUS of the defect. This can skew things wildly when the defect is not centered, or is quite large. But as you pointed out, it is NOT the same as 30% strength loss measured by the Bartlett Modified formula or Wagener's original formula.

These are just models, as Julian and Peter point out. And they are very preliminary in scope, without a lot of significant data to back them up. But they sure beat the seat-of-the-pants approach, if you ask me. We have a little bit of empirical data, and now a few beginning steps at working with that data. A long way to go but at least its a start.