Evaluation of Land Cover with Satellite Images from Landsat 8, Sentinel 1 and the Random Forest Algorithm in the Municipality of Villanueva Casanare

SUMMARYSoils are a basic requirement for meeting human needs, but human pressure on soil resources is reaching critical limits. Anthropogenic effects, soil contamination and monocultures pose major challenges for sustainable soil management. In Colombia, the significant increase in oil palm monocultures has generated different changes in soil use, alterations in ecosystems and greater vulnerability in food security. This is the case of the municipality of Villanueva Casanare where oil palm planting has been increasing since 1988, displacing other crops with higher occupancy such as rice, which contributes to improve the food security of the population. This research aims to evaluate land cover with satellite images of Landsat 8, Sentinel 1 and the Random Forest algorithm in Villanueva Casanare for the year 2019. For this purpose, a supervised classification was performed for the Landsat 8 optical sensor and for the Sentinel 1 Synthetic Aperture Radar (SAR), with a training of 7 classes (Forests, Pastures, Water Bodies, Urban Areas, Oil Palm, Other Crops and Bare Soils) that summarize the coverages of the Municipality of Villanueva Casanare, The corresponding reflectance and Spackle corrections were made, then the two supervised classifications were merged and run through the Random forest algorithm, likewise a confusion matrix was performed to corroborate the accuracy, all of the above in the Google Earth Engine platform. As main results of the three methods used with the Random Forest algorithm, the one with the highest accuracy of the seven proposed classes is the combination of SAR and Landsat 8, due to its 99.95% accuracy in the different coverages analyzed in the municipality of Villanueva Casanare.