LA SPATIAL CHARACTERIZATION OF LAND COVER FROM SATELLITE INFORMATION, CASE STUDY: AGRO-PRODUCTIVE ZONE OF SANTA FE, ARGENTINA

Abstract

Land cover change has always had a key role in land change science thanks to the
possibilities for mapping and typifying land cover based on remote sensing and
observations. When considering the importance of the functions of land use that provide a
wide range of goods and services, it is essential to carry out more integrated land change
assessments. Methods for mapping and quantifying land cover dynamics improve our
ability to understand and model land system change and adequately inform policy and land
planning. This work aims to develop a model that spatially characterises land covers,
implementing Machine Learning techniques and satellite imagery. The region of study is an
agro-productive area of the General López department, province of Buenos Aires,
Argentina. We implemented two models, Random Forest and Support Vector Machine, for
land cover classification using the Google Earth Engine (GEE) platform. Both models exceed
an overall accuracy of 90% and classify, with high reliability, the different soil classes. The
implementation of GEE facilitated data processing and reduced work time.