THE SECRET LIFE OF PLANTS

If you’re a geochemist, chances are your next grassroots exploration project will have a better chance of success if you can identify not just rock types in the field but also jack pine, trembling aspen, white cedar, and speckled alder. And infrared images of these trees may put you on to an anomaly right away. Conventional overburden geochemical sampling methods near Katie Lake, in northeastern Ontario, have outlined areas with base metal anomalies in the overburden. And a group of researchers, led by Prof Anthony Beswick, executive director of the Centre in Mining and Mineral Exploration Research at Laurentian University in Sudbury, Ont., has arrived at similar conclusions by performing a series of computer-enhancement techniques on airborne images received from the Ontario Centre for Remote Sensing. They found an anomalous signature in the aspen surrounding Katie Lake over areas of known anomalous soil. This anomaly shows up as a bright red spot on the photograph on page b *

Prof Beswick’s group has developed a preliminary method to analyse imagery, classify it and outline base metal anomalies in the vegetation. The method is based on a number of computerized image-enhancement techniques, namely radiometric enhancement of the data and supervised and unsupervised classification of the data set. The supervised classification has proven to be the most successful in identifying the anomalous vegetation in the image. This technique makes full use of the analyst’s experience and knowledge of the test site and of the remote sensor data.

The image analysis system and software used is a PCI IMAVISION system, operating on an IBM-AT.

Traditional geobotanical research has focused on identifying plants that are more sensitive indicators of bedrock anomalies than samples of the soils in which they grow. When geobotany is coupled with remote sensing, the changes in spectral reflectance of vegetation may not arise simply because of increased metal loading in foliar tissue, but rather from physiological changes in the plant induced by metal toxicity. These spectral reflectance changes in vegetation can be detected by remote sensing techniques and may outline target areas for mineral exploration.

The Katie Lake site was chosen as a blind test site to evaluate the effectiveness of the image analysis method, because it has a small but distinct metal anomaly. Good geochemical data from the Ontario Geological Survey and MEIS data are also available. These factors are necessary in order to determine whether a known metal anomaly, with known soil metal levels, can be detected by remote sensors via some change in the spectral reflectance characteristic of the vegetation cover. Trembling Aspen

MEIS digital imagery for Katie Lake was evaluated in relation to ground truth data obtained from field studies. At six different locations within the over-all site, pure stands of trembling aspen (the most dominant tree species in the area) were identified. Within each of these stands, five such trees were sampled for leaves, twigs, trunk wood and roots. In addition, more than 200 non-vascular plant samples and underlying humus soils were collected over the site for base metal analysis. Samples have been analysed by atomic absorption spectrometry for manganese, copper, nickel, cobalt, lead, zinc and arsenic.

The six aspen stands identified in the field were located on color infra- red air photos supplied by the Canada Centre for Remote Sensing. Distortion was apparent in the photos, so in order to precisely locate these stands, it was necessary to map them on a surveyed 10-m grid. Anomalous sites and non-anomalous, or control, sites (as determined from geochemical data) were outlined.

The image-enhancement techniques were used to classify the MEIS digital imagery for Katie Lake. The intended objective of any so-called radiometric enhancement routine in digital image enhancement is to create an image that can be interpreted visually. The radiometric routine is typically the first stage in the image- enhancement procedure and is done to make the researcher familiar with the various salient characteristics of the data set.

The radiometric enhancement of the Katie Lake MEIS data did not provide accurate delineation of the required categories of information, that is, good separation of signature between the non-anomalous or “control” trembling aspen signature and the “anomalous” site. The enhancement procedure was successful in separating the over-all general aspen signature from the other vegetative cover on the site.

The intended result of an unsupervised classification of a remote sensor data set is to have the image analysis system examine the entire image data set and to group those data into classes based on “natural” breaks in data. The unsupervised classification makes very little use of the researchers experience and operates on the premise that the data values for a specific cover type are closely grouped in spectral number, while data for other cover types will be considerably different and well-separated from other cover- type spectral signatures.

An “unsupervised” classification was conducted for a more precise classification of the data. The results of the unsupervised classification on the MEIS data set for Katie Lake showed, as expected, that the pixel radiance numbers for the control and anomaly training sites were too similar. The narrow and subtle range in radiance number effectively masks the differences between the two classes. On the basis of these results, the supervised classification was chosen as the primary cover classification method.

The intended purpose of using a supervised classification is to make full use of the analyst’s experience and knowledge of the test site and remote sensor data to enable classification of the data to be made in an expeditious and accurate way. The supervised classification requires extensive ground data and field experience. This is used to classify specific ground cover types. Once the cover types have been identified, a complete signature is generated, which contains all of the radiance values for the test site. The visual display then allows certain conclusions about the spatial distribution of that signature to be made.

The results of the supervised classification of the Katie Lake MEIS data show several facets of the aspen distribution pattern and other classes of land cover to be identified: the “control” aspen are delimited on the image, a potentially “anomalous” aspen signature can be identified, the anomalous signature coincides closely with the pattern of anomalous soil and vegetation geochemistry, and other vegetation cover types can be classified and their signature(s) identified.

The resultant supervised classification is shown in Figure 1. The control category of aspen is displayed in a light green color; the anomaly aspen as bright red (the yellow boundary to the red is due to color fringing — an artifact of the reproduction process); other categories of land cover can be identified as well, Jack pine (Pinus banksiana), dark green; alder-cedar- balsam fir swamp, black; and pixels are displayed as white and were unclassified.

The unsupervised classification in effect allows the operator to train the computer to recognize and classify vegetation on the MEIS digital imagery. Preliminary results indicate the ability to identify locations of anomalous reflectance among the aspen vegetation. Geochmeical analysis confirms the presence of anomalous concentrations of base metals in the soil samples.

This method of image analysis suggests the MEIS data can identify other potentially anomalous and non-anomalous vegetation stands within the area for which no ground truth data has yet been obtained. Anomalous aspen, for example, is identified as red on the bottom of Figure 1. The researchers propose to use these potentially anomalous and non-anomalous stands as blind test sites to evaluate the effectiveness of the method.

Field studies at Katie Lake, in parallel with imagery examination, comprised bedrock mapping and sampling for con
firmation of earlier work by the Ontario Geological Survey, as well as sampling of A and B soil horizons and a variety of vascular and non- vascular plant species for chemical analysis. Katie Lake is underlain by a mafic-felsic metavolcanic group associated with ultramafic and minor intrusive felsic rocks, which have been extensively intruded by diabase dykes. Metallic anomalies in the overburden of the area appear to be closely associated with serpentinized komatiites and basaltic komatiites, which contain high nickel and chrome values (650 parts per million and 1,550 ppm respectively) as well as other high base metal concentrations.

During the field work, it was observed that the vegetation cover over the site is dense and diverse with different major plant communities being associated with undulating topographic highs and lows and different moisture regimes. The gentle slopes of the northern section are covered with mixed stands of trembling aspen, white birch, balsam, fir and spruce. The trees are mostly mature. and the trembling aspen reach heights of 30 m. Depressions are filled with swamp species such as speckled alder and white cedar. Prominent rock ridges, especially at the southern end of the site, are covered by jack pine with a close ground cover of abundant lichens. Mosses and lichens also occur in clearings in the mixed aspen stands.

To date, analysis of the digital imagery from Katie Lake, employing image-enhancement techniques, in conjunction with overburden geochemical data for the site obtained thus far (and that taken from the ogs), suggest several of the anomalous aspen stands referred to are located over base metal soil anomalies whereas other stands are located in areas with background base metal levels. These preliminary results indicate the use of remote sensing techniques as an aid to mineral exploration in areas of glacial and residual overburden having heavy vegetation cover.

Prof Beswick’s research is supported by the Ontario Geoscience Research Grant Program. He also works in co-operation with people at RADARSAT, a satellite system at the Department of Energy, Mines and Resources; ERS i, the European Space Agency; and with people at the National Aeronautics and Space Administration (NASA) in the U.S. All research is in relation to the Sudbury area, now being referred to by NASA as the “North American Super Site.” REFERENCES Beswick, A. E., P. J. Beckett, G. M. Courtin, and G. O. Tapper (1986): Remote Sensing and Geobotany as an Aid to Mineral Exploraton in Northern Terrains, Progress Report for Grant 291 of the Ontario Geoscience Research Grant Program, 23 p. Joyce Musial is a Toronto-based geologist and freelance writer. — 30 —