Item

Abstract

Hydropower dams induce spatial and temporal changes in water and land systems in terms of access and use. Analyzing social-ecological changes triggered by dam construction and operation taking into account the water-energy-food nexus helps to identify the synergies and trade-offs between multiple layers of change. Colombia is moving towards peace agreements in recent years, and has already shown impact in the use of land and water, especially by the agricultural systems. This work focuses on two dams in the Magdalena basin, Betania-Quimbo and Hidrosogamosso. Landsat satellite images were used to build thematically detailed land use and land cover maps for the target years 2009, 2015 and 2021 covering seven classes (rice, palm oil, pasture, forest, water surface, temporary and permanent crops and others). Due to frequently high cloud cover and a complex topography, Colombia is one of the most difficult regions to build these maps. Global, or even local, maps available turn out to be characterized by uncertainties. We trained a random forest model with spectral-temporal metrics (176 in total) and topographic covariates in Google Earth Engine. We collected training and testing data based on a stratified random sample using secondary maps, which were labeled in Google Earth (1196 observations for all years). The model achieved a similar overall accuracy in all years (79%) and for certain agricultural systems a high user’s accuracy. Building these maps is extreme importance to support decisions that will be taken by instruments such as Basin Plans or compensation actions for the dam’s construction.