Tree Managers : Public Discussion

Mark

My reply is interwoven with your posting.

Regards

Mike Ellison

------------------------------------------------------
> Whilst the system produces statistical data, I
> have yet to see research data to support its
> accuracy. Based on Australian figures, the risk of
> death or serious injury is extreemely low (less
> than 1 in a 1,000,000 of population). Given that
> there are more than 10 trees for every person the
> odds of a tree causing problems is less than 1 in
> 10,000,000. How then do we account for the
> problems we do have? Clearly these are the result
> of trees that have a lot higher risk of failure
> that are high trafic areas!

ME: The data are published as refered to below. Without going into the figures that you have quoted, what you are discussing is risk of death or serious harm (RoH) from tree failure in the population as a whole, which is likely to be very low (in the region of 1/5,000,000 in the UK based on historical data). When we look at individual land ownership, one land owner may have a RoH far lower than 1/5,000,000 and anther land owner will have a RoH that is much higher. Whilst 1/5,000,000 may be the mean, there are huge peaks and troughs, which accounts for the fact that some tree owners have problems with high RoH

> Given that it is these trees that are the subject
> of risk it would seem that the data on these would
> need to be extreemely accurate. The concern I have
> is that the assesment is subjective and not
> quantataive at this point. Yes the tree does have
> girdling roots but can I say to what extent that
> this has impacted on the stability of the tree and
> the thigmomorphogenic responses. As a result we
> generally err on the side of caution and assume
> the risk is higher than it is. (Hence the
> Wessolley Mattheck debate)

ME: The starting point is the question "are there any significant targets?" because if there are not, there cannot be a significant risk of significant harm. Targets can be accurately quantified by frequency of occupation or by repair or replacement value. I believe that you are correct in suggesting that as aresult of the difficulty in interpreting the mechanical effects of decay and defects, we err on the side of caution. If your assessment of probability of failure (PoF) is cautious, this will be reflected in the qtra RoH outcome. However, qtra has substantially improved my assessment of PoF and therefore improved the outcome of the qtra assessments. And yes we can quantify and improve the assumptions that we are already basing our decisions upon.
>

> Clearly any program that considers the size of the
> falling parts, the probable frequency and the
> target must pick out the more hazardous trees (eg
> the ISA hazard assesment program) but how is this
> subjectivity converted into statistical accuracy
> without reseach data?

We use our best estimates and as we use the qtra procedures and observe trees that have been the subject of our assesments, our assessments become more accurate. Research data could improve our understanding of tree failures as a proportion of the opulation and this is something that I am hoping to assist with by encouraging cooperation with the ITFD and through other avenues. Have you had a look at the articles under the 'Resources for Tree Managers' section of the qtra website?


> If any arborist can tell me the liklihood of a
> 100mm dead branch falling I would like to know how
> they do so.

Based on species, size and condition, it is not unreasonable to expect s skilled arborist to be able to estimate how likely it is that the dead branch will fail. Currently, this skill requires improvement and use of qtra can only help advance our observational techniques.


Instead we assume that it can fall at
> any time and it will most certainly fail so we
> might as well be smart and remove or at least
> reduce it if we want to keep the habitat.

This approach would fail to take due consideration of targets so I assume that this would be in the context of the target. In terms of safety, the majority of dead branches are irrelvant because either the target value is so low or the usage of the site is substantially reduced during the weather conditionsthat are most likely to result in failure of the branch. Judging by the substantially degraded state of most fallen dead branches, it is fairly safe to assume that they degrade over a period of years and that the failure of any particular branch in the coming year is by no means certain.


On the
> other hand what are the actual statistical figures
> on death or serious injury due to the failure of a
> dead branch of this size. Again here in Austarlia
> the odds are very very small.

To the best of my knowledge, there are none.


> If 1 in 10,000 is an acceptable risk is there data
> for 10,000 trees at that risk level to prove or
> falcify the statistical assertion. Throwing the
> average tree with a risk factor of less than 1 in
> 10,000,000 into the data soop is surely poor
> ststistics

You may have misunderstood our use of the 1/10,000 risk of significant harm as a threshhold of acceptable risk. Other sources suggest that people generally would not spend their own resources to reduce an annual risk of death that was already as low as 1/10,000 and that a 1/10,000 risk of death is broadly acceptable where the risk is imposed on the public in the wider interest. I am uncear as to how data for 10.000 trees would prove or falsify this advised limit (which is not a not statistical assertion). Where does your second point relate to qtra?
>
> In short what is the error factor and does the
> data consider the mean, mode, maximum, minimum or
> average liklihood of any event. What is the
> statistical difference between the minimum and
> maximum (the error factor) and how have these
> figures been proved? I certainly would hate to
> find myself in a court case because I acted too
> quickly or because I did not act quickly enough.

The issue of error is currently being considered, but we do not operate to an absolute prescriptive threshold and any outcome that contradicts intuition can be reassessed and refined if necessary. The system calculates the likelihood that a significant loss will occur as a result of tree failure. The data used are set out in the articles mentioned above and there have been various modifications to the process since their publication.