Are Victim Detection and Forensic Evidence Search Dogs reliable?
Well, not always as the case of Thomas Quick proves. He was someone who confessed to a string of murders leading detectives to the places where he said the murders were committed. A police dog called Zampo was used to build a case against him.
"During the course of the investigation, Quick mentioned at least 24 different places in Sweden and Norway where he had committed murders, handled dead bodies or left body parts," says Leyla Belle Drake, who was Hannes Råstam's literary agent. "Zampo marked for human remains 45 times at those 24 locations. Not a single trace of blood or body parts was ever found."
http://www.theguardian.com/world/2012/oct/20/thomas-quick-bergwall-sweden-murder
The reason why no trace of blood or body parts was found in those locations despite the positive alerts was because Thomas Quick didn't actually commit any of the murders.
Why, I wonder, did the dog alert 45 times at those 24 locations if no murders had taken place there? Perhaps one of the dog experts on this forum could explain...
There are few credible explanations for this behaviour of Zampo.
One might be that there was a scent relating to an unknown cadaver in the locations alerted at but which has never been found despite human intervention. This seems highly unlikely as an explanation given the high number of locations where such a coincidental scent would have to have existed.
Another might be that the dog's training was flawed and that some other scent in the environment was, unbeknown to the handler, also capable of triggering an alert by this dog.
The third possibility also relates to possible flaws in the dog's training or handling. The training is in the form of a game with rewards. Perhaps the dog was too keen to receive rewards, perhaps the handler was not sufficiently aware of the potential for cueing of the dog.
We can only speculate, of course, as to possible reasons. There is no right answer to the question you posed.
But what the whole episode does re-inforce very strongly is that simply telling people that "dogs don't lie" or that "the dog alerts are evidence" without further careful and serious investigation and without being open-minded enough to accept that there are many examples of such flawed alerts is pointless. Such behaviour does not help anyone find the truth.