Students' Research Circle    
» 2019
Call for papers
The conference
Veterinary Session
Veterinary Jury
Biology Session
Home » Archive » 2019

TDK conference 2019

QGIS-plugin development for geographical sampling
Papp Márton János - year 6
University of Veterinary Medicine Budapest, Center for Bioinformatics
Supervisor: Dr. Norbert Solymosi


Gathering new informations in most scientific fields would be impossible without the use of samples. However not only science depends on sampling but economy, politics and everyday life make decisions daily based on samples. We can conclude that extrapolation from samples originate from early humanity although there are explicit differences in the quality, usability and interpretation of samples.

A sample is considered good when we can make proper assumptions about the population from it originates, based on it. Mostly it is called a representative sample, which is basically the miniature of the population.

To make sure that the sample has the above characteristics we need methods for obtaining it that keeps the original distribution among the elements. These are mostly take elements in a random way. The most well-knowns are the simple random sampling and the systematic random sampling. Despite they are representative techniques they not always result in representative samples. Sometimes different types of elements are overrepresented or aggregated points can be observed in the two-dimensional space.

We can state that the connection between sample elements is not negligible when sampling in two-dimensions. A sample that is well distributed over the sampling area is more likely representative as it is minimizing autocorrelation.

Sometimes however it is favourable to use different distribution: for example when some types of the elements take a small ratio of the whole population but are important for the study. It is especially important in cases, when limited sample numbers are accessible.

Technical difficulties can also occur especially when geographical sampling is presented. Sampling locations can easily be placed in private property or can be physically inaccessible. It is helpful to use methods that can deal with these situations.

These features are representing the method described by Stevens and Olsen in 2004. They named it Generalised Random-Tesselation Stratified (GRTS) model. We made an application of the areal equal and stratified sampling part of this method. We have constructed it as a plugin for QGIS, a free and open source geographical information system. Our goal was to make an easy to use graphical user interface for the program and to make it accessible and useful for everyone, especially in the field of veterinary epidemiology, parasitology and ecology. These are the main reasons we have chosen QGIS. The current version of the plugin can be downloaded at the official QGIS plugin repository in the name VetSamp.

List of lectures