CSIR predictive modelling will boost battle against rhino poaching


Researchers at the Council for Scientific and Industrial Research (CSIR) specialising in data fusion and mathematics have developed a predictive modelling tool that will assist SANParks in decision-making. Tasked with patrolling an area some 19 500 square kilometres in size with limited resources, this newly implemented model will assist counter-poaching units to deploy sensors and manpower at the right time, at the right place.

The predictive modelling tool utilises both a statistical model based on expert knowledge, which is used mainly for reasoning about the rhino poaching problem, and a data-driven model, which is used mainly for making predictions.

Dr Pieter de Villiers, a CSIR principal researcher, says the CSIR predictive modelling tool will assist SANParks with better decision-making in the deployment of technologies.
“The Kruger National Park is vast in terms of surface area with a limited number of rangers available for patrols. This tool will help to better direct rangers and deploy technologies such as radars, perimeter surveillance systems and other sensors used to detect poachers,” he said.

CSIR applied mathematician, Hildegarde Koen, started working on a concept that could potentially tip the balance back into the favour of the rangers.
“When the poaching problem flared up in 2010, I decided to focus my PhD study on developing a model that could theoretically predict poaching incidents. As I started working on it, I quickly realised there were no existing rhino poaching models and very limited information available. Owing to the lack of data available at the time, we decided to build an expert knowledge model based on the experience and information provided by experts working on counter-poaching programmes in the Kruger National Park. We try to consider the variables that would influence poaching events, such as proximity to water, vegetation density and moon cycles,” Koen said.

To improve predictive performance, the team then built an alternative data-driven model based on additional data that became available. The structure of the data-driven model was based on actual poaching data collected over several years. Through a process called data mining, algorithms are deployed to find patterns within poaching data. By recognising patterns in poachers’ behaviour, counter poaching efforts can be directed to specific areas of interest.

A complex problem such as wildlife crime requires innovative solutions, built through collaboration. In this context, CSIR safety and security experts, as well as data mining and data science researchers, are working with the University of Pretoria’s Statistics Department and researchers from the Thales Research and Technology group of laboratories based in the Netherlands to further develop the predictive modelling tool.

Data science researcher and project team member Dr Vukosi Marivate commented on the possible implications of a successful prediction.
“At the moment there are many experts working on the poaching problem. By adding computers that are continuously crunching numbers and providing possible locations to the team, you can give forces on the ground a better chance of catching poachers. There is no single solution to solving the rhino crisis. It is only through collective effort and a range of technologies, such as the predictive modelling tool, that poaching incidents will be reduced.”

If implemented successfully, SANParks will benefit from better decision-making and it will help key role players to better understand the causal contextual factors promoting or reducing poaching. It will also directly impact the arrest rate by directing rangers and air support in hot pursuit of poachers. “It is much easier chasing something if you have a better idea of where they are heading,” is how De Villiers summed it up.

The number of African rhinos killed by poachers has increased for the sixth year in a row with at least 1 338 rhinos killed by poachers across Africa in 2015, according to new data compiled by International Union for Conservation of Nature (IUCN) Species Survival Commission’s African Rhino Specialist Group (AfRSG). This is the highest level since the current crisis began to emerge in 2008. Since then poachers have killed at least 5 940 African rhinos, IUCN said.

Last year South Africa lost 1 175 rhinos to poachers – 40 less than 2014’s all-time high of 1 215. South Africa currently conserves 79% of Africa’s rhinos and has suffered the bulk (85%) of poaching on the continent since 2008. The country’s vast Kruger National Park is home to the world’s largest rhino population and has borne the brunt of the killing. While the margins of error around the Kruger National Park rhino population estimates between 2012 and 2015 overlapped, statistical modelling suggests that in all likelihood the populations of both black and white rhinos have decreased in the Park. This has however been countered by net increases in the numbers of black and white rhino elsewhere in South Africa and other countries, IUCN said.