Community Research and Development Information Service - CORDIS

Abstract

In this paper we discuss performances and capabilities of two low resolution Earth Observing systems (NOAA-AVHRR and SPOT-VEGETATION) for burnt area mapping in a tropical savanna ecosystem of Northern Australia. Multi-temporal composite images were derived from both the data sets and a set of rules for extracting burnt areas was built using a Classification And Regression Trees (CART) methodology. In the case of VEGETATION data the rules were used to segment the entire composite image; for AVHRR data a seed-growing approach was tested, aiming at reducing as much as possible overestimation errors. The accuracy of the results was assessed using a map derived by visual interpretation of two Landsat-TM frames: proportions of burnt areas were calculated in thirty 14 x 14km windows randomly extracted from the AVHRR, VGT and TM maps. Conclusions are that the better geometry of the VGT sensor, compared to AVHRR, allows a higher accuracy in burnt area estimation, but the lack of a thermal channel is certainly a drawback for both the composing and the burnt area detection processes. The two low resolution systems tend to overestimate burnt areas when these are very fragmented, but they do underestimate errors when burn scars are too small to be detected. Finally, the seed-growing method seems promising and it could be the basis of a combined use of SPOT-VEGETATION and NOAA-AVHRR systems, which show complementary characteristics.

Additional information

Authors: MAGGI M, European Commission, Joint Research Centre, Institute for Environment and Sustainability, Ispra (IT);STROPPIANA D, European Commission, Joint Research Centre, Institute for Environment and Sustainability, Ispra (IT)
Bibliographic Reference: An article published in: Canadian Journal of Remote Sensing, Vol 28, No 2, pp.1-15 (2002)
Availability: Journal website: http://www.casi.ca/index.php?pg=cjrs
Record Number: 200214819 / Last updated on: 2002-08-13
Category: PUBLICATION
Original language: en
Available languages: en