Landsat MSS imagery as an adjunct to aerial survey of Kangaroos

Landsat MSS imagery as an adjunct to aerial survey of Kangaroos

G.J.E. Hill
Australian Key Centre in Land Information studies
Department of Geographical Sciences
The university of Queensland
ST. Lucia, 4067, Australia

Abstract
The larger kangaroos have extensive geographic ranges that favour regional approaches to population survey. Consequently, aerial census techniques form the base for much of the monitoring and management programmes conducted in Australia. However, reliance in existing map products and aerial photography for conducting these surveys and interpreting the results is not wholly satisfactory. Factors such as scale and cost (for photography) and level of detail (fro maps) are severe drawbacks. Recency of data capture is often a severe draw back for both products. Satellite imagery, on the other hand, overcomes most of these problems and is well suited as the map/habitat base for aerial survey of kangaroos.

Introduction
Satellite telemetry is now routinely use in Australia to track animal movement pattern (e.g. Doig and Dyson 1988) which has made remote sensing of more direction relevance to wildlife scientist. However, it is still not possible to conduct censuses of wild life population from space. The use of satellite imagery for habitat mapping, on the other hand,. Has been established for many years (e.g. Falconer 1979, Best 1982). In some cases including (the celebrated) "Wombats detected from space" (Loffler and Magules 1980) this work has been extended to include the specific "homes" of animals as well as land cover in general. As has been the case wit many, if not most, remote sensing applications areas, however acceptance and routine use of remotely sensed data as facet of wildlife monitoring and management programmes has been slow to develop in Australia.

Landsat, multispectral scanner (MSS) imagery has been available since 1972 although Australian coverage was limited until the opening of a receiving facility at Alice Springs in late 1979. Data handling and marketing are managed by the Australian Centre for Remote Sensign (ACRES) located in Canberra. Establishment of the Landsat station, which is currently being upgraded to receive imagery fro the newer satellite sensors, led to a similar growth in the use of remote sensing in Australia. This was matched by a rapid expansion in the availability of digital image processing systems which allow the full potential of the data to be exploited.

ACRS A-2-3-2 The utility of Landsat MSS data for regional habitat inventory in Kangaroo Management was recognized in New South Wales soon after the launch of the first satellite in the Landsat series (Fox 1974). In terms of the scientific literature, though, there has been little reported us in wildlife application is Australia since these early days. By way of contract, satellite remote sensing has been accepted as a basic research tool by applied scientists in many disciplines that overlap or are closely related to wildlife ecology and management.

A research group from the university of Queensland, Commonwealth Department of primary industry and Queensland Department of Geographic information has been making regular use of Landsat MSS imagery as part of a Kangaroo monitoring programme in the Goodiwindi district of Queensland since 1983. Providing an outline of these uses and the potential of Landsat MSS imagery as supporting data for broadscale aerial survey of Kangaroos is the aim of this paper.

Figure 1. Location of the study area and landsat scene

Study area selection
The study area for the overall project is centered on Goondiwindi is southern Queensland. Described in detail elsewhere (eg. Hill et. al. 1985, 1988, Hill and Kelly 1986), the area was selected to incorporate a major land use transition zone between predominantly crop land to the east and well wooded pastoral land to the west. An area such as this was suited to research into pest status of Kangaroos because of the high risk status of agricultural land adjacent to pastoral country. As well, the mosaic of land cover types typical of this region was an appropriate base of studies of habitat usage of kangaroos and favoured high kangaroo densities, facilitating research into aerial survey methodology.

Selection of a suitable region may be aided by a perusal of topographic maps; shire Hadnbooks; land systems reports; r the appreciation of the regional geography of the country that aerial survey practitioners develop for the areas they fly.

To optimize selection of areas with specific attributes, however, requires access to current, readily interpreted land cover and/or land use maps. Scale of the maps is also relevant as aerial survey is a regional tool compatible with small to medium scale cartography (generally 1.20,000 or smaller). In most cases such maps do not exists. Landsat MSS imagery, however, is ideally suited tot eh task, it provides a near real-time portrait of the landscape (16 day overpass cycle) and a single scene covers an area of 185 x 185 km. The segments of the spectrum recorded by the sensor assist in clear delineation of regional cover types although true-color images cannot be produced and this may prove confusing the new users. As long as the regional perspective is of interest the coarse spatial resolution of landsat MSS (80m pixels) is not a significant drawback.

In the work reported here, the non-standard Landsat scene 91/80A incorporates the study are. Even to the uninitiated in image interpretation, the pattern of land use is apparent on this scene. it should also be noted that a black and white, single band image at a scale of around 1/1,600,000 such as the one presented in fig.2, is a poor substitute for a three band, color composite at larger scale. Standard, false color composites at a scale of 1/250,000 produced by ACRES were used to select the study area and aerial survey zones within it that suited the various research topics.

Figure 2. Landsat MSS, band 6, image for scene 91-80A recorded on 18 August 1985.

Aerial survey navigation
As is the case with many Kangaroo survey programmes, 1/250,000 topographic maps have formed the navigational base for the aerial surveys at Goondiwindi. However, at a survey altitude of 76m the raterhspars set of reference features plotted on these maps is not readily visible from the aircraft. Furthermore, the land cover and cadastral information recorded n top maps is often hopelessly out of date and may confuse the navigator and / or survey team. Maps from the study area, for example, are based on aerial photography from the early 1960's. While the sophisticated omega navigation systems now available on survey planes have eased the major problems related to locating specific points and following predetermined flight paths (transects), reliable, readily interpreted maps are still required for wildlife survey work from aircraft.

Landsat MSS colour composites at scales of 1/250,000 and 1/100,000 have proved ideally suited to the navigational tasks involved. These computer generated photographic products are available from ACRES and currently cost AU$ 310 for a 80 c 80 com image ( an additional master generation charge may apply). There is a comparative wealth of navigational cues available from land sat compared with the topographic map sheet. In addition, the clear differentiation between the land cover types and individual paddocks makes it possible to plot the positions of groups of Kangaroos on the imagery where distributional data are required at this level.

Habitat mapping
Jarman (1979) has put the case that aerial census data are not as valuable to the wildlife manager or ecologist as distributional data that relate the animal to its environment. Landsat is capable of providing environmental inventories compatible with the scale at which aerial surveys of Kangaroos are conducted and recorded at a similar time to animal counts.

At the simplest level landsat may be used to manually compile maps showing gross habitat patterns (e.g. Hill et al 1988). The images utilized are normally color composites which combine three of the sensor's four spectral bands. Visual interpretation of imagery, a noted previously, may be used for study into site selection and planning, conducting and interpreting survey selection and planning, conducting and interpreting survey flights. Due to cost and time constraints, visual interpretation is probably the most satisfactory way of incorporating habitat data derived from Landsat with boradscale kangaroo surveys. Monitoring programmes covering hundreds of thousands of square kilometers would be a case in point. Manually derived cover (habitat) maps of extensive areas have been produced from Landsat by organizations such as the Australian Surveying and Land information group (AUSLIG) and by the Ecological survey of south Australia (e.g. Douglas 1979). Used in conjunction with aerial census data products of this sort forma useful base for interpreting gross distribution patterns and changes over time. Where large areas are involved, mosaiking of Landsat imagery is a relatively simple task.

In addition to interpretation of photographic products, Landsat data may also be analyzed using computer based image processing systems. While many studies have established the utility of digital analysis of MSS data for vegetation and land cover mapping in Australia (e.g. Hill and Kelly 1986) little attention has been directed towards habitat mapping per see. Research within the present study area has demonstrated that digital processing of MSS can replicate the standard habitat categories used during aerial survey of kangaroos (hill and Kelly 1978). The accuracy of classification ranged from 80% for open country (<10% projective canopy cover) to 92% for forest. Overall classification accuracy was 91% with most error resulting from the difficulty involved in defining fixed cutoffs along the coverage gradient between different vegetation classes. This approach to habitat mapping and quantification offers substantial improvement over traditional methods based on habitat scoring during survey flights. As the proportion of the area sampled during kangaroo surveys is normally less than 5% and often only 1% the scope for inaccuracy, due to sampling error, is large.

Conclusion
The review provided that Landsat MSS imagery with its long established credentials in wildlife habitat mapping, is a valuable adjunct to aerial survey programmes of the type currently used in Australia to monitor density and distribution of kangaroos. However, there is scant evidence of Australian researchers using Landsat data as a source of habitat fields of study such as forestry, land cover and vegetation mapping.

From the point of view of the non-specialist remote sensing practitioner, Landsat imagery did not receive and auspicious introduction to Australia. While much was promised, the poor quality photographic products that dominated the 1970's did more to discourage than encourage interested on lookers. Likewise, the self taught, part-time remote sensing "specialists" who filled the technology gap during this period were in many cases incapable of providing wise counsel. The same is not true today. With world class receiving and distribution facilities (ACRES), a core of remote sensing specialists, well respected on the international scene, distributed around the country, and an ample supply of image processing systems, Australia is well placed to take advantage of remote sensing technology.

References

Best, R.G (1982) Remote Sensing in fish and wildlife management. Remote sensing Institute, south Dakota State University, Brookings, SD
Doing, F. and Dyson, S. (1988) Satellite tracking: A new direction for research. Aust. Nat. Hist., 22 (10): : 437-441
Douglas, J.S. (1979) : The use o Remote Sensing to map and monitor the natural resources of South Australia, in Landsat 79: proceedings 1st Australasian Remote Sensing Conference. Sudney, 22-25 May, pp. 237-242
Falconer, A. (1979) Landsat data and fauna habitat assessment, in Aerial Sruvey of Fauna Populations. Australian National Parks and Wildlife Service, Canberra, pp. 47-62.
Hill, G.J.E Barnes, A and Wilson, G.R. (1985) Time of day and aerial counts of grey Kangaroos. J. Wildi Manage. 49: 843-49.
Hill, G.J.E Barnes, A and Wilson, G.R. (1988) Wheat crop use by grey kangaroos in southern Queensland. Aust .Wildi. Res. 15: 111-17
Hill, G.J.E Barnes, A and Wilson, G.R. (1986) integrating Landsat and land systems for cover maps in southern inland Queensland. Aust Geogr. Std 24: 235-43.
Hill, G.J.E Barnes, A and Wilson, G.R. (1987) Habitat mapping by Landsat for aerial census of kangaroos. Remote sensing of environ. 21: 53-60.
Jarman, P.J. (1979) The types of information available from aerial surveys and their applicability to management, in aerial survey of fauna population. Aust. Natl. parks & Wildl. Serv. Canberra, pp.107-113.
Loffler, E. and Magules, C. (1980) Wombats detected from space. Remote Sens. of Envrion., 9: 47-56.