Search

The study of regional-scale land covers is an expensive activity that requires specific equipment. This work aimed to analyze the precision and the accuracy between satellite products and a land cover map obtained usingin situmeasurements in the Sauce Grande river basin (Argentina). In order to do so, we carried out five field trips during 2018 to identify land covers. It was studied the reflectance of each class from georeferenced samples acquired with a spectrometer and a GPS. Besides, spectral indices of vegetation, water, and soil were calculated to analyze the spectral behavior of each class. With this information, we delimited regions of interest to perform the supervised method. Subsequently, we downloaded the MODIS and ESA Landcover products. Then, we established a pixel of 300 meters for the three maps with matrix metrics. The Kappa coefficient, the overall accuracy, as well as the user and producer ones, were calculated to evaluate the estimated land cover against the measured ones. Results showed that MODIS-Landcover presents excellent agreements for determining crops and grasslands, while ESA-Landcover for urban expansion and water bodies. La delimitación de las coberturas del suelo es una actividad costosa que requiere de instrumental específico. En este contexto, el objetivo de este estudio fue analizar la precisión y exactitud de productos satelitales de cobertura del suelo con respecto a un mapa confeccionado con mediciones in situ en la cuenca del río Sauce Grande (Argentina). Para ello, se realizaron cinco viajes de campo durante el año 2018 y se tomaron muestras georeferenciadas y firmas espectrales de las coberturas del suelo. Complementariamente, se calcularon índices de vegetación, suelo y agua a partir del procesamiento de imágenes satelitales Landsat 8. Con la reflectancia adquirida a campo y el comportamiento espectral de cada cobertura, se identificaron muestras espaciales necesarias para realizar una clasificación supervisada. Posteriormente, se descargaron los productos MODIS-Landcover y ESA-Landcover y se diseñaron mapas de coberturas del suelo estimadas. Se estableció un tamaño de píxel de 300 m para los tres productos satelitales y se aplicó con una matriz de confusión del tipo pixel por pixel para compararlos. Los resultados permitieron establecer que el MODIS-Landcover es un producto útil para el monitoreo de cultivos y pastizales, mientras que el ESA-Landcover para la expansión urbana y las variaciones del agua superficial. The study of regional-scale land covers is an expensive activity that requires specific equipment. This work aimed to analyze the precision and the accuracy between satellite products and a land cover map obtained usingin situmeasurements in the Sauce Grande river basin (Argentina). In order to do so, we carried out five field trips during 2018 to identify land covers. It was studied the reflectance of each class from georeferenced samples acquired with a spectrometer and a GPS. Besides, spectral indices of vegetation, water, and soil were calculated to analyze the spectral behavior of each class. With this information, we delimited regions of interest to perform the supervised method. Subsequently, we downloaded the MODIS and ESA Landcover products. Then, we established a pixel of 300 meters for the three maps with matrix metrics. The Kappa coefficient, the overall accuracy, as well as the user and producer ones, were calculated to evaluate the estimated land cover against the measured ones. Results showed that MODIS-Landcover presents excellent agreements for determining crops and grasslands, while ESA-Landcover for urban expansion and water bodies.