|dc.description.abstract||The present study was conducted in the Huetar Norte region of Costa Rica, and its purpose was to compare two remote sensing methodologies for the identification of forest plantations. The methodologies consisted of a supervised classification by pixels (methodology 1) and a combination of segmentation-supervised classification (methodology 2), using two Landsat satellite images; one of the year 2005 and another of the year 2016.
It was carried out through a design of six blocks, with two treatments or methodologies. First the images were classified in order to identify four basic coverages: forest, non-forest, clouds and cloud shadows. Then, with these classifications, a comparison was made between the image of the year 2005 and that of 2016 in the six blocks, to determine the changes in coverage in the period 2005-2016. Then the areas that showed change from the non-forest to forest class were selected, they were defined as new forests, and then reclassified using a supervised classification in two classes: plantations and not plantations.
On the other hand, in order to know the consistency of the classification a validation was carried out using control points of the plantation and non-plantation classes. A statistical analysis was applied to these reliability data, using a t-test for paired samples.
The study presented limitations with the images used, since, they presented high percentages of cloudiness, in spite of this, it turned out that there was no significant difference between the methodologies, so it recommends using methodology 1, since it is the simplest to apply.||es