Community Research and Development Information Service - CORDIS

Abstract

Burnt area mapping, from SPOT-VGT imagery, has been performed using a Classification And Regression Trees methodology already successfully applied with satellite images for land cover and burnt area mapping. The algorithm is composed of a set of classification rules extracted from a decision tree.

In order to overcome misclassification errors produced by targets with a spectral signature similar to burnt areas, we introduced, as predictor variables, the temporal differences between ten-day VGT composites.

High resolution satellite imagery (Landsat TM) were used to assess the performance of the algorithm. The introduction of the temporal change detection approach drastically reduced the misclassification errors in dry and sparse vegetation classes. es.

Additional information

Authors: STROPPIANA D, JRC-Space Applications Institute -Global Vegetation Monitoring Unit, (IT);GRGOIRE J.M, JRC-Space Applications Institute -Global Vegetation Monitoring Unit, (IT)
Bibliographic Reference: An oral report given at: MultiTemp-2001. Organized by: University of Trento and JRC-IEEE. Held in: Trento Italy, 13-14 September 2001
Record Number: 200013414 / Last updated on: 2001-06-27
Category: PUBLICATION
Original language: en
Available languages: en