Missouri Botanical Garden Open Conference Systems, TDWG 2014 ANNUAL CONFERENCE

Font Size: 
Mitigation of data positioning errors in the modelling of ecological opportunity for the occurrence of diseases in the Wildlife Health Information System – SISS-Geo
Marcia Chame, Livia Abdalla, Eduardo Krempser, Edmar Moretti

Last modified: 2014-09-29

Abstract


Evaluating the influence of data geopositioning errors in the process of modelling is not a simple task, a fact that is confirmed by the lack of research in the literature related to the sensitivity of models to errors in inputting positioning data. Studies that compare the performance of models do not classify data according to the accuracy of positioning.

The challenge faced in the modelling of ecological opportunity for the occurrence of diseases, which is being development by the Wildlife Health Information System – SISS-Geo, designed by the Oswaldo Cruz Foundation – Fiocruz in cooperation with the National Scientific Computation Laboratory - LNCC, is to integrate a dense mass of data with varied cartographic scales and characteristics in the construction of models designed to alert officials of potential disease outbreaks. It is therefore essential to minimize the effect of positioning errors in these models when they will be used to determine public health policies.

The SISS-Geo database will be made up of data that can be grouped into three large sets: Records, Cartography and Analyses. The Records are the observations of wildlife that are entered into the system. The Cartography set is a thematic geographical database (rivers, highways, borders, biomes, types of vegetation), on a national scale, obtained through different institutions that are data providers. The set referred to as Analyses are the results of the modelling by SISS-Geo, performed with the use of machine learning techniques that relate both the records and the environmental data to the occurrence of diseases, a task that essentially involves spatial queries between the data in the records and the cartography.

In an attempt to minimize the effects of positioning errors on the SISS-Geo models, a  tolerance in the spatial intersections is allowed, using the precision of the positioning of each mapping and the observation point collected. This is because the spatial queries return not only a verification of the incidence, but also give for each one a weight, considering factors such as how close the point is to the areas of interest. This strategy does not eliminate positioning errors, but it allows the creation of solutions that aid in the construction of more realistic and reliable models to be used by decision makers for public health.