See below ...

greg.gorbett Wrote:
-------------------------------------------------------
> This study evaluates the ability of 500+
> investigators to determine the area of origin
> through the use of photographs and measureable
> data. It would be wrong to say that this study is
> an error rate for an overall origin determination,
> as it does not include witness statements, arc
> mapping, or fire dynamics – the other three areas
> of information for origin determination were not
> included.

The paper I was refering to was by Campanell and Avato. I do not believe it included these aspects in this particular paper.

I do not think anyone is arguing the point that the determinations were based on anything other than the post-fire physical evidence.

>
> I do find it interesting that there are always
> naysayers to others work and testing (knowing full
> well that there are thousands of variables that
> cannot be accounted for in every fire test), yet
> they have not done any testing themselves to help
> solve the problems. I hear a lot of complaints
> and dismissive tone about the testing from our
> article and others not being long enough in
> post-flashover duration…my question to them - is
> where is your data and tests that support your
> assertions. I would love to see more
> testing...let me know when you have published your
> results, I would love to see how it helps the
> profession.

I think most would agree that this work falls under the category of "applied science." The discuss was started because the paper as written (and reviewed) did not include any discussion of the limitation of the application of this applied science by the authors. The comments were directed to the potential "users" of the applied work as opposed to the authors of the work. When evaluated in the context of the growth and development of fires, the time to detection of the fire, the fire department response, and the time to fire suppression, the time to fire suppression in the tests in most fire incidents is not substantially similar to the conditions in the tests. This is a significant and important issue within the fire investigation community with the increasing frequency of reliability challenges.

One of the Carman's papers shows the error rate in the origin determination versus the duration of post-flashover conditions. As would be expected, the error rate increases as the duration of post-flashover conditions increases. So believe this issue is relevant to a discussion of the application and limitations of the work.

Applied science also has to address the issue of the control of variables. Stating that there are thousands of variables that can not be accounted for in every fire test does goes against the concept of scientific reliability and applied science. There is an art to designing scientifically reliable tests and experiments. In his paper "The Needed Fire Science," Howard Emmons of Harvard University, Division of Applied Sciences ("Grandfather of Fire Modeling") explains "When we solve technical problems, it is only sensible to keep all phenomena (terms) down to some minimum size and to discard the rest. It takes a high level of scientific understanding of fire to know exactly what aspects of the fire environment are important and which are not." The work done here seems to fit this concept and not the condition of thousands of variables unaccounted for.

My data for addressing the issue of reliable application is not derived from my own testing, but from previous available data from the available scientific literature associated with the time frames of the growth and development of compartment fires and the time frames of detection and fire suppression response.

So where is my test data associated with fire patterns and origin determination? There is none. As was put fourth in my previous posts, I would characterize the use of "fire patterns" into two (2) regimes. 1) Interpreting fire patterns to reach a determination of the origin, and 2) using fire patterns as a means to test one's origin and fire cause hypothesis. There is a difference.

In my estimation, the first regime has significant problems that may not be overcome with an infinite number of additional testing as discussed in my previous posts. Thus, given the scarcity of funding for such testing, there is a much higher technical risk with respect to developing a scientifically reliable methodology.

The second regime is already embedded in the 921 and the SM. The "predictive validity" of the SM where one tests their hypothesis to try and disprove it by showing that the outcome is not substantially similar to the outcome in a specific fire incident. For example, in a fatal fire where the fire origin and fire cause would not produce a fatal outcome. An actual occurrence in a case I worked on where fire modeling was used by an opposing fire expert.

Testing associated with the second regime would be more fundamentally based with respect to the variables associated with how specific patterns are formed and not formed. This (once characterized by Vyto Babrauskas as an "encyclopedia of fire patterns") is still a daunting task. This additional fundamental knowledge would allow more detailed analysis and testing of hypotheses using the SM than may be currently available.

There is certainly a lot one can do to further the knowledge base with respect to compartment fires and applied science. There has not been a comprehensive effort to characterize the measurement uncertainty of replicate compartment fires. Such grant monies have never been awarded, although there is general agreement among reviewers that this has not been done reliably and would be a worthy technical effort.

Douglas J. Carpenter, MScFPE, CFEI, PE, FSFPE
Vice President & Principal Engineer
Combustion Science & Engineering, Inc.
8940 Old Annapolis Road, Suite L
Columbia, MD 21045
(410) 884-3266
(410) 884-3267 (fax)
www.csefire.com