Assessment of AVHRR data for deforestation estimation in Mato Grosso (Amazon Basin)
- 1 July 1994
- journal article
- research article
- Published by Taylor & Francis in Remote Sensing Reviews
- Vol. 10 (1-3) , 35-49
- https://doi.org/10.1080/02757259409532236
Abstract
A large area (∼ 900,000 km2) corresponding to Mato Grosso, a southern most State of the Brazilian Legal Amazon region, was selected for this study. All five channels of selected dates of 1988, 1989, and 1990, full resolution NOAA AVHRR data were investigated to map “forest” and “non‐forest” (“cerrado”) using on the screen visual interpretation. Maximum likelihood classifier (MLC) was used to classify deforestation for the 1988 AVHRR images. The vegetation map obtained by interpretation of AVHRR data agreed well with the available vegetation map used as reference. Results based on cells of 1° by 1.5° showed r2 = 0.94 (significant at α = 0.01) between the AVHRR map of forest and the available vegetation map. Comparisons for 1988 between AVHRR deforestation estimates and two independent estimates using TM data, i.e. Landsat TM band 5 habitat fragmentation and Landsat TM bands 3, 4, and 5 color composite deforestation estimates were within 3% of each other for the entire State (range from 71,128 km2 for the habitat fragmentation to 73,120 km2 for the AVHRR estimate). Results based on 1° by 1.5° cells indicated a good agreement between AVHRR and Landsat TM estimates (r 2 greater than 0.7, significant at α = 0.01). The major source of variation between these estimates are the discrepancies in the “forest” and “non‐forest” boundaries. In areas where the non‐forest boundaries are coincident in the independent estimates, an r 2 = 0.93 was observed. The contribution of AVHRR thermal bands seems to be significant for the discrimination of “cerrado” (“non‐forest”) areas. The sensitivity of MLC to the parameters “threshold” and “bias” was examined. It was observed that if deforestation estimate based on high resolution sensors, such as TM is available for a representative sample, an unbiased estimate of deforestation using AVHRR data (five channels) is possible by adjusting the “bias” parameter of MLC.Keywords
This publication has 13 references indexed in Scilit:
- AVHRR As A Data Source For A Gis: Deforestation In AmazoniaPublished by Institute of Electrical and Electronics Engineers (IEEE) ,2005
- Tropical Deforestation and Habitat Fragmentation in the Amazon: Satellite Data from 1978 to 1988Science, 1993
- NOAA-11 AVHRR visible and near-IR inflight calibrationInternational Journal of Remote Sensing, 1990
- Remote sensing of biomass burning in the tropicsJournal of Geophysical Research: Atmospheres, 1990
- Fire in the Tropical Rain Forest of the Amazon BasinPublished by Springer Nature ,1990
- Determining the rate of forest conversion in Mato Grosso, Brazil, using Landsat MSS and AVHRR dataInternational Journal of Remote Sensing, 1987
- Identifying deforestation in Brazil using multiresolution satellite dataInternational Journal of Remote Sensing, 1986
- Remote Sensing Digital Image AnalysisPublished by Springer Nature ,1986
- Net flux of carbon dioxide from tropical forests in 1980Nature, 1985
- Aerosol emissions from forest and grassland burnings in the southern amazon basin and central BrazilNuclear Instruments and Methods, 1981