All Content © CRC for Forestry 2007

The Monitor - Issue 5 - Visiting student gets mixed up in remote sensing

Canopy and understorey in blue gum forest
Radiata pine canopy cover

Differences in canopy and understorey in blue gum forests at Wattle Range , SA (top) and radiata pine forest at Green Hills, NSW (bottom)

 One of the most challenging aspects of trying to determine the condition of plantation trees in remotely sensed images is minimising the influence of objects beneath the canopy. A masters student from Katholieke Universiteit (KU) Leuven in Belgium, Ms Eva Ampe, visited the Wattle Range and Green Hills study areas to help determine the best method for unmixing our images to measure forest health.

Differences in the condition, colour and character of soil and understorey vegetation beneath a uniform canopy of trees introduces noise into the images that can make tree condition extremely difficult to identify (see photos). The mixing of the background and canopy signals within pixels causes errors in the interpretation of tree condition from the image data.

Eva Ampe

Visiting student Eva Ampe collecting spectral data for her unmixing study

Images can be processed in a number of ways to ‘unmix’ pixels in order to identify the individual components contained within each pixel. The challenge that Eva recently faced in unmixing images from our study sites was considerable due to large variations in soil and understorey vegetation types, and in the condition of the tree canopies within and between these sites.

The early results from this work look promising. Good correlations have been calculated between canopy cover levels observed during fieldwork and estimated cover levels produced using an unmixing process.  The graph below illustrates the difference between the observed canopy cover levels (green line) and the modelled canopy cover levels (red line).  Remaining challenges in this project include improving the models used for determining tree health in young plantations, as the changing colour of the blue gum foliage with age (young red shoots, to blue juvenile leaves, to green adult leaves) can make interepreting forest health difficult from the images. Furthermore, the structural characteristics of blue gum canopies changes with tree age and this needs to be considered when interpreting the images to give an indication of forest health.

canopy cover graph

Correlation between canopy cover levels observed in the field and estimated using an image unmixing model

Program One has established a close association with KU Leuven, as illustrated by Eva’s visit and the visit of a previous masters student, Mr Kim Calders. The relationship between Program One and the university developed during the tenure of Dr Jan Verbesselt (formerly associated with KU Leuven), a principal research scientist who worked on Project 1.1.2.  Both Eva and Kim have made important contributions to this research project.  Many of the questions being addressed in Eva’s work relate to the principles and processes of remote sensing, and, as such, core scientific issues are being addressed within her project.  Outputs from Eva’s work will also directly influence the interpretation of forest health from the other image sources by providing information about structural aspects of the landscape that can only be assessed remotely using hyperspectral images.


Contact

Dr Neil Sims, CSIRO
Tel: 03 9545 2163
neil.sims@csiro.au