A TOOL FOR MACHINE LEARNING BASED DASYMETRIC MAPPING APPROACHES IN GRASS GIS
- Department of Geoscience, Environment and Society (DGES), Université Libre de Bruxelles, Avenue Franklin Roosevelt 50, Bruxelles 1050, Belgium
Keywords: GRASS GIS, Dasymetric Mapping, Machine Learning, Random Forest
Abstract. Socio-economic and demographic data is often released at the level of census administrative units. However, there is often a need for data available at a higher spatial resolution. Dasymetric mapping is an approach that can be used to disaggregate such data into finer levels of detail. It relies on the assumption that proxies available at a higher spatial resolution, along with knowledge of an area, can be used to produce weights in order to spatially reallocate the data to a finer scale layer. The power and efficiency of machine learning (ML) approaches can be taken advantage of when producing weighted layers for dasymetric mapping. Less advanced users, however, may find these approaches too complex. To encourage a wider uptake of such approaches, easy-to-use tools are necessary. GRASS GIS is a free and open-source GIS software that contains many modules for processing geographic data. The existing GRASS GIS add-on “v.area.weigh” already makes the dasymetric mapping approach more accessible, however users must provide their own weighted layer. This paper presents the development of a GRASS GIS add-on, “r.area.createweight”, which provides a simple and convenient tool to facilitate the implementation of a ML-based approach to produce weighted layers for dasymetric mapping. The tool will be available on the official GRASS GIS add-on repository to encourage a more widespread uptake of these approaches.