Scientists at the Department of Environment and Conservation matched data from ground surveys in banksias woodlands of the Gnangara Groundwater system to Landsat satellite images, allowing them to spot changes in the vegetation that could indicate phytophthora infestation.
Field data was taken from 26 one hectare sites using orthophotos to estimate canopy cover, and crown openness (the upwards cover of leaves within the crown outline) at three points—the product of both producing the projected foliage cover-field (PFC-F).
A linear equation is then derived from the regression of the Spectral Image Index of a similar aged Landsat photo of the site, and PFC-F estimates, and creates a calibrated image of projected foliage cover (PFC-I) from 0 to 33.
The DEC computer program RegMachine can then compare the PFC-I for each pixel of images of the same site from different years and assess changes in vegetation by calculating the linear and quadratic regression of a single pixel through time using orthogonal polynomials.
DEC senior research officer Graeme Behn says ground surveys might find pockets of dieback and provide detailed information about the outbreak, but aren’t good at surveying large areas and are likely to miss something.
Trend analysis of satellite imagery would allow scientists to better direct their efforts.
“If we find a trend where there’s a dramatic change and that change isn’t related to any physical event on the ground such as fire, there’s a warning light that comes up that says ‘go and have a look’,” he says.
Mr Behn says if the change occurs in vegetation that’s susceptible to dieback or in a low lying area, in a river course or a river flow area, there’s a strong likelihood they’ve found an outbreak.
He says phytophthora impacts have a pulse of recurring death and recovery that happens gradually, while drought tends to happen very quickly and suddenly.
“We can make an assumption in most cases, but assumptions need to be checked,” he says.
DEC Acting Regional Leader Nature Conservation Barbara Wilson says the calibration would only have to be done once in each kind of environment and this one could be used for banksias woodlands and possibly other kinds of woodlands on the Swan Coastal Plain.
“Having shown that it can be done in this ecosystem, you would be able to use this approach here and possibly other similar ecosystems and it would be quicker,” she says.
Relying more on remote sensing has its financial benefits, with surveys such as a 2006–2009 (212 ha) in the Bell Track Catchment costing $250,000 compared to very little for remote sensing as the DEC already has access to Landsat images.
“When you ground survey for phytophthora we have specialised people who are trained and that’s a very labourious process,” Dr Wilson says.
The DEC will soon supersede Landsat with Urban Monitor, which will offer higher resolution and smaller pixels for even more detailed results.