Pre-processing of a sample of multi-scene and multi-date Landsat imagery used to monitor forest cover changes over the tropics

In support to the Remote Sensing Survey of the global Forest Resource Assessment (FRA) 2010 of the United Nations Food and Agricultural Organization (UNFAO), the TREES-3 project of the Joint Research Centre aims at measuring quantitatively forest cover changes for the periods 1990-2000-2005 over the tropics. Landsat data subsets of 20 x 20 km size have been extracted at each integer confluence of the geographic grid resulting in more than 4,000 sample sites distributed systematically over Sub-Saharan Africa, Central America, the Caribbean and the tropical part of South America and of South and Southeast Asia. A single approach was developed to automatically pre-process this large amount of multi-date and multi-scene imagery prior to supervised classification in an operational and robust manner. The paper presents the data selection and the different pre-processing steps applied to more than 12,000 Landsat TM and ETM+ data subsets (one image per reference year: 1990, 2000 and 2005) distributed over the tropics: conversion to top-of-atmosphere reflectance, cloud and cloud shadow detection, haze correction and image radiometric normalization. Initially designed for a few Landsat scenes, haze correction procedure was adjusted and tested on a large data set. The results show that the algorithm has improved the visual appearance of the image and significantly corrected the digital numbers for Landsat visible bands, especially the red band included for later digital classification. The impact of the normalization procedures (forest normalization and relative normalization) was assessed on 210 image pairs where neither land cover change nor seasonal changes were evident between the two dates. In all cases the correlation between the spectral values of the same land cover in both images was improved by the normalization processes. The Jefferies-Matusita distance showed no detrimental effect on spectral separability after the forest normalization. The developed pre-processing chain provided a consistent multi-temporal data set across the tropics that will constitute the basis for an automatic object-based classification.

BODART Catherine;  EVA Hugh;  BEUCHLE Rene';  SIMONETTI Dario;  RASI Rastislav;  STIBIG Hans-Jurgen;  BRINK Andreas;  LINDQUIST Eriik;  ACHARD Frederic; 

2011-10-18

ELSEVIER SCIENCE BV

JRC62843

0924-2716,   

http://www.sciencedirect.com/science/article/pii/S0924271611000475,    https://publications.jrc.ec.europa.eu/repository/handle/JRC62843,   

10.1016/j.isprsjprs.2011.03.003,