Missouri Botanical Garden Open Conference Systems, TDWG 2014 ANNUAL CONFERENCE

Font Size: 
Brazilian Information System in Wildlife Health - SISS-Geo
Eduardo Krempser, Douglas Adriano Augusto, Livia Abdalla, Marcia Chame

Building: Elmia Congress Centre, Jönköping
Room: Rydbergsalen
Date: 2014-10-29 10:15 AM – 10:30 AM
Last modified: 2014-10-04


Environmental changes, including climate change and loss of biodiversity, are determinant factors for the emergence of wildlife diseases that might potentially spread to and infect humans. The majority of infectious diseases are shared among wildlife and humans (60.3%) and, of these, 71.8% are caused by pathogens with origins in wildlife. These emergencies are almost always associated with areas that have undergone natural or anthropogenic impacts, composing also a range of parameters that makes social inequalities even stricter.

In this context, rather than looking for efficient answers to crisis situations, it is more important to take preventive actions that anticipate problems and then, if the prevention fails, to react quickly. In Brazil, strategies of surveillance and prevention of occurrences of diseases arising from the biodiversity are still developing. To remedy this scenario, the Brazilian Information System in Wildlife Health (SISS-Geo in Portuguese) was developed to systematize the practices of surveillance and prevention of diseases in wildlife health despite major challenges. These challenges include the country's vast territory, immense biodiversity, and the complex intrinsic relationships among factors that characterize risks to wildlife and human health.

SISS-Geo is part of the Wildlife Health Information Center: http://www.biodiveridade.ciss.fiocruz.br  and works as a service for the Brazilian population and government. We emphasize two fundamental components of the system: data collection and data analysis. The data collection component provides a set of web services that receive and send information via web and mobile interfaces. In addition, a process that collects data from  environmental and climate layers was also implemented. All information handled by SISS-Geo are georeferenced and therefore they can be related across layers.The second component is considered as the core of the system, analyzing all collected data by data-driven modeling techniques in order to generate models and alerts through the application of machine learning algorithms. These algorithms take into account information such as territorial distances of the occurrences, time intervals among them, taxonomic similarity, and the physical conditions of the animals involved. The workflow of the analysis can be described as: (i) a new register is added to SISS-Geo; (ii) an unsupervised learning technique defines if the register belongs to a previous event/group or if it will form a new event; (iii) a feature extraction technique is applied to describe the event; (iii) supervised learning is applied to classify the event as requiring an alert or not; (iv) if the event is classified as an alert, the information is sent to a network of laboratories and the Wildlife Health Diagnosis Network, which will then confirm whether the event is actually an alert, therefore validating the classification process and augmenting the training data set.

The alerts will be useful to health and environmental services established in the country, allowing new data collection and analysis in the field and the improvement of preventive measures.