Methodology for mapping change in woody landcover over Queensland from 1999 to 2001 using Landsat ETM+

The Statewide Landcover and Trees Study (SLATS) is currently in the process of mapping woody change across the State of Queensland for the era 1999 to 2001. SLATS covers such a large area that Landsat Thematic Mapper (TM) imagery was chosen as the most effective data for this purpose. Since 1995 the woody change in Queensland has been mapped every two years. The 1999 to 2001 change era is the first time that Landsat Enhanced Thematic Mapper Plus (ETM+) imagery has been available for both years of the SLATS change measurement era. A comprehensive review of the SLATS change detection methodology was carried out to determine an optimum methodology for woody change mapping in Queensland. The methodology adopted was based on a threshold decision tree classification of a simple three-band difference image, incorporating a newly developed “woody index”, ETM+ Band 2 and ETM+ Band 5. This paper describes this methodology and the field methodology for verifying the change classification. Introduction The study currently known as the Statewide Landcover and Trees Study (SLATS) commenced in 1995 to provide accurate information on rates of vegetation change and associated greenhouse gas emissions in Queensland. The rate of woody vegetation change has been mapped from 1991 to 1997 using Landsat 5 Thematic Mapper (TM) and from 1999 primarily using Landsat 7 Enhanced Thematic Mapper Plus (ETM+). The current 1999 to 2001 change era is the first time that ETM+ data has been used for both years of the change era. This, together with the desire to improve on and to overcome some of the limitations of the previous change detection methodology, prompted a comprehensive review of the methodology and image rectification techniques. Research has been conducted into: • Selecting the most appropriate Landsat 7 ETM+ Bands, • Developing new vegetation indices, • Adapting recent research into soil colour and brightness indices, • Developing a new Foliage Projective Cover (FPC) Index, • Semi-automated cloud and cloud shadow masking, • Improving the radiometric registration of the imagery both spatially and temporally, • Improving the geometric rectification accuracy of the imagery. This research has generated an improved change detection methodology. It uses a decision tree classification of a simple three-band difference image. It incorporates the new Multiple Regression Vegetation Index (MRVI), a woody FPC index (Goulevitch et al 2002) and ETM+ Bands 2 and 5. The new change detection methodology eliminates the need to geographically stratify the imagery to reduce soil colour and brightness effects and reduces the amount of time taken to manually edit change classifications.