Pub Quiz

[Daryl]

I am with the Radial clan!

[kim]

Well done to all who said 'Radial', Robert got it first so over to you.

Gunther, Not sure what you said but it sounded good so kudos to you as well.

Kim

[Robert Bradley]

Hooray ... my question is;

Visiting the ERT-SAR website I noticed the following;

ERT Search and Rescue have been developing their Canine section since late 2006. In 2007 - Various SAR Dog groups, European SAR handlers and the Police Dog Section, assisted ERT In establishing their K-9 / Dog Unit. It officially starts in February 2008.

Can anyone give me either of the two other names that this SAR group has gone by?

[Pete]

OSAR and UKSAR I think......

Pete.

[Robert Bradley]

Or both .....

Well done Pete, off you go...

[Pete]

Here's a great controversial question for you

Is it mathematically possible to create a probability of detection when using search dogs? Plus explain your answer in full.

Pete.

[Robert Bradley]

Oh yes......I must answer this one.

Yes, it is mathematically possible to predict a probability of detection for a search dog (and indeed people are working on it).

That said, it is extremely difficult to do so because of the multitude of factors that influence the effectiveness of a dog search.

Starting at the beginning, probability of detection is itself calculated from a number of factors.
Koopman stated that POD=1-e(to the power of -Coverage).
Coverage equals the area effectively swept over the actual area to be search.
Area effectively swept equals the distance travelled by the search resource(s) times by the sweep width for that search resource in that environment.

And it is here that the problem with a dog search team starts.

Sweep width is an empirically calculated figure that represents an "area" where 50% of all possible detections are detected. Hard to explain simply (and off the top of my head as the rules dictate) but essentially the area under the graph of possible detections at this distance is the same as outside . In order to calculate this distance sweep width experiments are undertaken and sweep width distances are calculated (along with factors which affect this - for aerial search these factors include speed, height, size of object being searched for and so on.)

It is the difficulty of calculating all these factors which is causing problems in calculating sweep width figures for search dogs. Imagine trying to quantify the difference an extra mph of wind makes to the detection ability of a search dog, or a gradient has, or the time of day etc. etc. etc.

So whilst it is mathematically possible to calculate this figure, it is a bit of a way off at the moment.

That answered, going further...

Attempts to subjectively predict POD have been proven not to be effective - in fact the research states that you might as well pull a figure from a hat. One assumes that the same holds for dog search.

In land field search at the moment, however, the Average Maximum Detection Range (AMDR) is being used to estimate sweep width in the field for the purposes of calculating POD. One possibility for dog search is to try to calculate an AMDR for a dog for the time and environment of the search and use this to predict POD (although this is very much an idea rather than theory, I must add!)

AMDR could perhaps be calculated by putting someone 50 metres or so off to one side, then walking a dog up and down parallel to them getting 5 metres closer at a time until the dog "detects" that person. (Some sort of dog equivalent to doing a Northumbrian Rain Dance) In this way POD could be estimated using dog AMDR, distance dog travelled, search AMDR, distance searcher travelled and so on.

Of course, the effectiveness of this couldn't be proven without experimentation.

Anyway, am I right?

[Pete]

Yes your right, Robert.

As stated it would never be feasable to use as there are far to many variables.

Pete.

[Robert Bradley]

I wouldn't go so far as to say never feasible.

I'm not great at predictions but give it ten to fifteen years and dog SAR will, I'm sure, have workable sweep width tables.
The reason; now that the mathematics of search theory are being taken up around the world for land SAR, a workable model for dog search must be produced otherwise the figures produced by the search management team will essentially be meaningless - either ignoring the contribution made by dog search, or using subjective figures (which as I've stated are likely to be meaningless.)

Any volunteers to start working on it?

[Pete]

Actually I've just looked at this from a statistical point of view and as every dog is different the calculations would have to be made for every different dog.....

Anyway we digress....... Your question.

Pete.