Methodology
The foundations of the SLATS project rely on the proven technology of satellite remote sensing.
Satellite image data is the only practical method at monitoring land cover change given the 1.7 million square kilometre extent of the State of Queensland.
These images are regularly acquired by a number of government and commercially owned and operated satellite platforms orbiting the earth and are widely available for purchase through government and commercial organisations.
The SLATS team has developed and tested computer based methodologies to create data products based on these images that accurately depict vegetation composition and changes in land cover to precise geographic locations.
A brief explanation of the SLATS methodology follows but for more detail see reports and publications.
Acquiring imagery
SLATS obtains its ‘raw’ Level 5 Landsat satellite imagery from the Australian Centre for Remote Sensing (ACRES).
Where possible, cloud-free dry season imagery is chosen to maximise discrimination between grasses and the woody component of the vegetation.
Each image represents approximately 185km by 185km on the ground and a total of 89 individual scenes are required to cover the whole state allowing for overlap.
Image rectification and preprocessing
Prior to use in change detection and mapping procedures, the imagery is radiometrically corrected for variation in sun angle and atmospheric effects (Danaher, 2002). This allows each of the scenes to be spatially and temporally matched regardless of what time of the year they are individually acquired.
The imagery is then geometrically corrected and referenced to the MGA94 ground based coordinate system. This is achieved by registering each scene to a 1999 reference year by using an extensive number of ground control points (GCPs) measured in the field with differential global positioning systems (DGPS) to within 3 metres accuracy.
Cloud, cloud shadow, smoke and water in each image is masked through a combined automatic index and manual threshold classification process.
Image processing and classification
Comprehensive image processing procedures have been developed by the SLATS team for working with the satellite imagery. Most of these procedures have now been automated and standardised into documented macros and scripts to minimise errors.
A huge number of files are generated at the intermediate phases of the image processing process. Approximately 10 Terabytes (1 Terabyte = 1024 Gigabytes) from a 300 Terabyte file store is used to safely store the data (see Technology).
The current method of mapping vegetation change is based on a time-series statistical analysis of the foliage projective cover (FPC) for each given imagery date at a location.
This FPC index is calibrated using multiple regression between transformed Landsat imagery (digital numbers) and stand-scale allometric FPC calculated using SLATS field basal area (BA) measurements (see Reports and Publications).
Manual raster editing (recoding) is then performed to each image supplement the initial classification to adequately isolate the woody vegetation change. This is due to green flush in pastures, new cropping, seasonal variation in overstorey, changes in understorey, fire scars, image misregistration and hill shadows.
This visual interpretation makes use of imagery for the earlier and later date, together with the difference image, the initial change classification, a past series of images, and ancillary information such as aerial photography and photographs from previous field data sites.
Field validation
SLATS is not a desk project. The scientists that process the imagery for a particular satellite scene also personally ground truth and validate the computerised classification of each scene no matter where it is in Queensland.
Once this preliminary change classification is completed for each scene, a field verification program investigates areas identified as having woody vegetation change, particularly those areas where interpretation of change is uncertain.
Extensive fieldwork is undertaken collecting ground control points, checking clearings, and measuring loss of coarse woody debris from clearing to assist in determining greenhouse gas emissions.
This is typically well off the beaten track on rough bush roads in remote areas across all of Queensland - including Cape York and the Gulf, through central Queensland to the remote south-west corner and is conducted using a four-wheel drive vehicle equipped for remote area operations.
Purpose-adapted GIS software is run on a laptop computer receiving a DGPS signal which allows the accurate tracking of the vehicle’s location on the satellite imagery.
Data logged at each site includes:
- Accuracy of the classification
- Method of clearing used
- A visual estimate of the percentage of cleared timber removed or decayed
- Replacement land-cover
- Maturity of timber cleared
- Presence of regrowth
- Original vegetation species type
- Current species type
- Presence of termites or fire
- Soil colour
A digital photograph is additionally taken at each site to assist in future analysis.
Data is also gathered at previously visited field sites which are used to determine coarse woody debris disappearance rates, differing clearing regimes and regrowth rates and regrowth clearing across the State. This data is used to improve the State's greenhouse gas inventories and emission estimates.
Final product
The preliminary change classification created prior to the field program is corrected based on the field observations. This involves recoding possible clearing areas and incorrectly classified change to the field verified status. The final classification is closely peer-reviewed by those with first hand field knowledge to ensure accuracy and consistency across Queensland.
SLATS summary statistics are presented as annual rates of change (ha/year) rather than area change between satellite image dates. This is due to the differences in the start and end-capture dates for scenes across the State.
The full resolution change classification data set together with the dates of the imagery used is then provided to the Queensland Environmental Protection Agency (EPA) to update the Regional Ecosystem mapping and the department's Vegetation Management section to monitor compliance with the Vegetation Management Act and the Land Act.
Information on all SLATS data (metadata) is consistent with the Australian Spatial Industry Standards and is provided with each dataset.
Last updated: 10 March 2009