James Cameron - Mapping Pinus Radiata In Native Sclerophyll Forests In South-eastern Australia For Feral Pine Management

14th Australasian Remote Sensing & Photogrammetry Conference, Darwin, 2008

Department for Environment and Heritage Mapping Pinus radiata in native sclerophyll forests in south-eastern Australia for feral pine management James Cameron & Yuki Tunn




Cats




Rabbits




Lantana




Bridal Creeper




Humans




Radiata Pine

www.environment.sa.gov.au

What’s the problem? P. radiata capable of invading many vegetation communities Seeds can be dispersed over long distances by wind and animal Once established, can out-compete established species resulting in a reduction of native species richness

www.environment.sa.gov.au

The Question? Is there a way to detect feral pines (or pine wildlings) within native vegetation over large areas using remote sensing?

Maybe………

www.environment.sa.gov.au

Presentation outline

Project outline The area of interest How we did it? Did we successfully answer the question? What’s the next step?

www.environment.sa.gov.au

The Why SE Regional Office (DEH), ForestrySA, and Victorian Department of Sustainability & Environment concerned of the spread of Pinus radiata from plantations to native bushland Forestry companies had indicated that harvesting wild pines can be worthwhile – Possess the equipment to harvest pine wildlings with minimal damage to the native vegetation – Costs of the operation offset by the returns from the harvest

They just need to know where to look www.environment.sa.gov.au

Study area Study area includes the lower SE of South Australia and extends into western Victoria – ~1.8 million hectares

“Area 3” chosen due to wide variety of vegetation

www.environment.sa.gov.au

Proposed Methodology Use existing archival aerial and hiresolution satellite imagery – Colour infrared film aerial photography – Jan 2005 – 4 band pan-sharpened Quickbird – May 2004

Trial a series of spectral analyses to increase separability of pines and native vegetation – NDVI – PCA – Supervised vs unsupervised classification

Create a map of hotspots showing locations of Pinus radiata www.environment.sa.gov.au

Success?

X Pines

Aerial photography

www.environment.sa.gov.au

Quickbird

Opportunistic Data Integration Microsoft Vexcel UltracamD 4 band imagery – January 2008 – 90cm spatial resolution – SE NRM region

LiDAR (ALTM3100) – 23-25/7/2007 – 50cm vertical accuracy (1σ) – <1.9m mean point density

Could the combination of Microsoft Vexcel UltracamD imagery and LiDAR detect emergent pines? www.environment.sa.gov.au

Looking for the simple solution…. Subtract the ground data from the nonground data – = height relative to ground (2m grid)

Calculate an average height layer from non-ground points only – = background average canopy height layer

Subtract background canopy layer from non-ground data and threshold – Tall trees?

Combine with classification identifying what we think to be pines using Vexcel imagery – Emergent pines? www.environment.sa.gov.au

Validation At first glance, the Vexcel imagery seemed to better discriminate between pines and native vegetation Inspection of the site indicated the method under-estimated emergent pines – The greater the average height, the taller the emergent pine has to be to be detected – Individual emergent pines more likely to be detected as groups of same age pines often had same heights and weren’t “emergent” compared to surrounding vegetation – Conservative threshold meant only trees greater than 2m higher than the background canopy were selected

Was the integration with the LiDAR data filtering out actual emergent pines? www.environment.sa.gov.au

Third time’s the charm? Original unsupervised classification of the Vexcel imagery was filtered using a routine to remove noise and to amalgamate similar pixels – Groups of less than 4 pixels were removed from the classification – Enhances emergent pines with a more rounded canopy shape

Preliminary validation providing promising results www.environment.sa.gov.au

www.environment.sa.gov.au

www.environment.sa.gov.au

www.environment.sa.gov.au

Conclusions Film based colour infrared aerial photography and Quickbird satellite imagery found it difficult to spectrally separate emergent pines from native vegetation

The integration with LiDAR data “refined” the mapping too much, resulting in errors of omission

Vexcel imagery was a vast improvement, but some errors of commission were still evident www.environment.sa.gov.au

Where to from here CSIRO are investigating the use of a pattern/texture based approach to separate pines from native vegetation – Results tend to agree with the Vexcel classification presented here

Integration with these results with GIS could provide better modelling as to where pre-emergent pines would be more likely to occur – Include information on wind direction, bird populations, proximity to existing plantations www.environment.sa.gov.au

Acknowledgements Nerissa Haby – University of Adelaide

Melissa Herpich & Andrea Lindsay – Department for Environment & Heritage

David Hart – Department for Environment & Heritage

Thank you

www.environment.sa.gov.au