LSTM’s Seminar Series continued this week with a presentation by Dr Steve Le Comber, Senior Lecturer at the School of Biological and Chemical Sciences, Queen Mary University of London. His Seminar “From Jack the Ripper to Malaria: geographic profiling in biology” was introduced by LSTM’s Dr David Weetman.
Dr Le Comber began by looking at the background and origins of geographic profiling (GP), which was the brainchild of Canadian police officer, and now Professor of Criminology, Kim Rossmo. It is used by many law enforcement agencies around the world in serious serial offences and uses the location of the crimes to reduce a pool of suspects by likely area in which the offender either lives or works. While it can’t provide the name of the offender it gives those investigating the opportunity to prioritise their search strategy using the spatial locations of linked crimes.
Steve used several high profile case studies to demonstrate the use of GP including the case of serial sexual offender who committed four crimes between 1982 and 1995 in the Leeds and Bradford areas and the tracking of serial killer Peter Sutcliffe.
He went on to explain the mathematics behind GP and the fact that it was driven by a practical need to find a solution. He pointed out that a without using the model, it would be expected that on average 50% of an area would be searched before the offender was found, but when using the model it would be expected that the location of the offender should be found within the first 5% of the search – which is described as it’s hit score. He looked at the different kinds of GP that are used: Criminal Geographic Targeting (CGT) is used mainly by law enforcement and is highly effective but difficult to adapt owing to intractability of the underlying model. A newer, more mathematically elegant model has been proposed by O’ Leary, but is practically limited by assumption of a single crime source. This is where innovation from Steve’s team has proved valuable by developing a novel Bayesian model in the open source software R which can work with multiple sources and critically has the capacity to incorporate additional information a priori. Steve demonstrated the performance of this model through a series of simulations and additional case studies including the Jack the Ripper Case, though lack of a proven culprit limited proof of the model here. The Bayesian model easily outperformed the O’ Leary model and slightly but consistently outperformed the CGT model across a range of conditions.
So far so interesting, but what about biology? Steve described three types of biological information to which the model has now been successfully applied. The first was in finding animals roosts based on their foraging information, with details of successful studies including both bats and bees. The second was to find the source of invasive species, with the model being more successful in finding the source of 52 out of the 53 invasive species looked at than any other method. The third application described was in epidemiology and here is where tropical medicine enters the fray, with a study which looked for the sources of a past outbreak of malaria in Cairo. From survey data water sources that contained Anopheles species capable of transmitting malaria were identified. When the model was applied, the seven water sources containing the vectors where listed one to six and number 21 (which contained a minor vector) in the geographic profile.
To conclude Dr Le Comber discussed some of the projects that his team is currently working on, pointing out a number of papers that have been published for further reading. His seminar was very well received and he went on to answer a number of questions from both staff and students in the audience, taking the seminar well over the allotted timeslot, but tellingly few left before the end!
Further information and publication links can be accessed through Steve’s website: