Caracterización del uso/cobertura del suelo en Uruguay a partir de series temporales de imágenes MODIS

Published in Agrociencia, v. 18(2):95-105
Authors

Baeza, S., Baldassini, P., Bagnato, C., Pinto, P. and Paruelo, J.M.
Publication year 2014
DOI http://dx.doi.org/10.31285/AGRO.18.470
Affiliations
  • Departamento de Sistemas Ambientales, Facultad de Agronomía, Universidad de la República. Avenida Garzón 780, 12900 Montevideo, Uruguay
  • Laboratorio de Análisis Regional y Teledetección, IFEVA, Facultad de Agronomía, Universidad de Buenos Aires
  • Departamento de Métodos Cuantitativos y Sistemas de Información. Facultad de Agronomía, Universidad de Buenos Aires

 
IAI Program

CRN3
IAI Project CRN3095
Keywords
vegetation phenology
remote sensing
NDVI

Abstract

The spatial distribution of land use/land cover is the main control of energy, carbon and water dynamics is also extremely important in environmental monitoring and decision-making in the agricultural sector. The objective of this work was to develop
rapid and low cost land use/land cover classifications in Uruguay. We performed decision tree classifications using phenological information derived of NDVI-MODIS (Normal Difference Vegetation Index) time series (period: May 2011-March 2012),
field data, and high spatial resolution images (Landsat) to identify «pure» agricultural lots (belonging to a single land use /land cover class). 1.7x107 hectares were classified, discriminating four major categories: Perennial Forage Resources, Afforestation and Forest, Summer Crops, and Winter-Summer Crops they occupied respectively 63.6, 13.1, 14.3 and 7.4 % of the mapped area. Overall map accuracy was high (89.6 %) and the less accurate class was summer crops (85.8 %). This classification is one of the first descriptions of land use/land cover for the entire Uruguayan territory with high levels of accuracy. The results show a very significant decrease in forage resources occurred in recent years mainly due to the advance of the agricultural frontier. The method used is a fast, repeatable, measurable and inexpensive alternative to describe land use/land
cover over large areas and monitor its change over time.