Geomorphology, land-use, and hemeroby of foothills in Colombian Orinoquia: classification and correlation at a regional scale.

Abstract

In the management of ecosystem services, it is significant to relate land use with the physical characteristics of the terrain, which allows establishing the conditioning factors of human activities and planning their distribution. These analyzes are based on thematic cartography, usually generated with visual classifications of satellite images. Traditional mapping techniques involve limiting the timely availability of information by taking extended periods for interpretation and integration of multiple data sets. This article presents a methodology to overcome these difficulties, implements big data, machine learning, and cloud computing to generate timely thematic cartography and spatial analysis to support land use planning. The study area was delimited according to altitudinal levels that define braided and anastomosed river systems. Acquisition, processing, and classification of input data for modeling were performed on the Google Earth Engine platform. The spatial correlation between hemeroby and geomorphology was calculated with the odds ratio and its respective confidence interval. Maps of 27 geomorphological units, 11 types of land use, and six hemeroby levels are presented at a scale of 1: 50,000. Confusion matrices of implemented classification models were also reported, allowed evaluating global, user's, and producer's accuracy. Correlations between relict of natural areas with the structural environment and urban infrastructure with alluvial fans stand out. The information generated by these procedures is essential for planning land use and prioritizing the maintenance of ecosystem services.