Metodología para identificar frailejón mediante el uso de imágenes de radar: una herramienta para la conservación ambiental

Abstract

One of the main characteristics of the images captured by synthetic aperture radars (SAR) is the possibility of being generated in almost any climatic condition, thus resulting in extracting information from the earth's surface in areas with a predominant presence of clouds, such as the ecosystem of Paramo. In this work, a methodology for the detection of frailejon was developed, evaluated and validated through the capture of information in the field and the use of Sentinel radar images over an area located in the Sumapaz paramo. Unsupervised and supervised classification methods were used, obtaining the best results in the evaluation phase with the Support Vector Machine (SVM) technique, obtaining a kappa coefficient of 0.76 and an overall accuracy of 88%. Subsequently, as part of the evaluation process, the methodology was replicated on a new area located in the Sumapaz páramo using the same image used in the classification described above and obtaining as a result a kappa coefficient of 0.78 and a global accuracy of 89. % . Lastly, the proposed methodology was validated by applying the same classification method to an image with a different temporality than the one used in the evaluation process, obtaining a kappa coefficient of 0.82 and a global accuracy of 92%. Therefore, the implementation of the methodology reveals the great potential of involving images from active sensors in the identification of frailejon.