Background to rainforest dieback in tropical Queensland:

Dying and dead patches of rainforest associated with the root-rot fungus Phytophthora cinnamomi were first recorded in tropical north Queensland over 20 years ago (Brown 1976) Subsequent soil surveys showed the fungus was widespread and associated with serious disease in two widely separated rainforest areas in north Queensland, one of them in the Wet Tropics World Heritage Area (Brown, 1999). However little else is known of the threat posed by P. cinnamomi to the wet tropics rainforests of Queensland (Goosem and Tucker 1999).

Worldwide P. cinnamomi is regarded as one of the most destructive fungal pathogens of woody plants. In Australia, P. cinnamomi and other related species are responsible for economic losses totalling millions of dollars annually, and dieback caused by P. cinnamomi is listed as a key threatening process under the federal Endangered Species Protection of Biodiversity Act (1999). As a consequence systematic identification and mapping of areas currently affected by dieback and potentially susceptible to it have been given very high priority by the Wet Tropics Management Agency (WTMA) based in Cairns. Their concern and funding provided the basis for this study.

During the earlier soil surveys undertaken by Brown and colleagues (Brown 1999) P. cinnamomi was detected in 645 of the 1,817 sites sampled. Its occurrence was not always associated with dieback but detection rates were significantly higher under patches of dead and dying forest. Interpretation of aerial photography in the Eungella Tableland near Mackay showed that dieback patches occupied some 19% (125 ha) of the study area. Given that this survey occurred prior to the advent of accurate position fixing by GPS, the location data can only be regarded as being within +/- 100m at best. However, at a regional scale the patterns can be instructive, and many sites in the present surveys occur close to those recorded by Brown two decades ago.

Remote sensing and geographic information systems are increasingly becoming an integral part of systems designed for operational sustainable management of worldwide forest resources. These technologies represent the only tools available today which can provide synoptic, objective views of the extent and current management of these natural resources. Most studies of tropical forests have concentrated on mapping cover and deforestation; mapping structural and biomass components; mapping the forest condition; and comparative evaluation of satellite sensor capabilities. This study is the first to integrate remote sensing and GIS for analysis of canopy dieback in an Australian tropical rainforest.

Objectives and Methods

The specific objectives of this project are to:

· Mark-up and interpret observable canopy dieback on aerial photography and satellite imagery;

· Determine the environmental correlates of canopy dieback patches in three study areas;

· Develop a spectral signature of canopy dieback patches for application to remotely sensed data at a variety of spatial scales.

These objectives provide a strategy for enhanced understanding of the spatial extent of canopy dieback and its relationship to environmental variables such as topography, geology and vegetation. In addition, relationships with other variables such as the distribution of roads and drainage can be sought. The patches provide an unbiased sampling frame with which to determine the precise spatial signature of canopy dieback. This will be of great assistance in extending the regional survey across the entire Wet Tropics World Heritage Area. It will also provide a means of scaling up from high-resolution multispectral imagery (at 2m resolution) to commercially available Landsat ETM data with 25m resolution. Three study areas were selected on the basis of historical evidence (Brown 1976, 1999) of rainforest canopy dieback: Tully Falls - Koombooloomba; Lamb Range; Mount Lewis. To date most work has been carried out in the Tully Falls area, and this preliminary study will be reported here. Over one hundred and fifty canopy dieback patches have been mapped from colour aerial photography, and thus provide a reasonable sample from which to infer environmental correlates.

Regional spatial data were sourced from the GIS unit of the Wet Tropics Management Agency. Coverages used in this analysis included:

· Webb & Tracey (1985) vegetation cover (structure and association data);

· Geology based on 1:250,000 map coverage from AGSO;

· Digital elevation models with cell resolution of 80m and 25m (the latter being a provisional compilation without drainage enforcement);

· Rainfall surfaces interpolated by Turton (1995);

· Drainage digitised from 1:50,000 scale topographic mapping;

· Roads digitised from 1:50,000 mapping.

The precise locations of dieback patches were mapped from colour aerial photography at a scale of 1:25,000. Areas of reduced canopy density or canopy senescence (brown) were delineated as polygons on transparent overlays and transferred to topographic maps at a scale of 1,50,000. These polygons were then digitised and stored as AutoCad (.dxf) files, attributed and converted to shapefiles (.shp) in the ArcView version 3.2 GIS software. In addition, remotely sensed data were acquired as follows: Landsat ETM+ multispectral imagery for September 1999 was sourced from ACRES and precision map oriented (to level 9). This imagery comprises seven spectral bands from visible blue to near thermal infra-red, with an additional panchromatic layer in the visible blue to red area of the spectrum. Spatial resolution is 25m and RMS error on rectification is 12m. These data cover the entire Wet Tropics World Heritage Area and potentially provide a valuable resource for dieback mapping and derivation of other spatial data layers in GIS. Airborne multispectral videography was sourced from the Farrer Centre, Charles Sturt University. Four calibrated and filtered video cameras with 12mm focal length lens on each camera give 2m by 2m pixels at 2800m altitude. Each camera gives image of 740 by 576 pixels resulting in ground coverage of 1500 by 1100m or approx. 165ha. This is coupled with differential GPS which gives each image centre's location. The four spectral bands in these data are directly comparable with the Landsat ETM+ imagery, allowing for potential scaling up of spectral signatures.

Environmental correlates evident from the GIS analysis

The digitised dieback patches were overlain on the geology GIS layer, with the parent material attribute being chosen for analysis. Nearly all dieback patches occur on acid igneous rocks, specifically on soils derived from the Mareeba Granite and also on rhyolite and dacite. Within the Tully Falls study area, the distribution of dieback patches by lithology is clearly biased to acid igneous rocks.

Table 1: Distribution of dieback patches by lithology

Lithology	Area (ha)	% of total
Mareeba granite	2042	78.0
Rhyolite and dacite	415	15.8
Tully granite	100	3.8
Basalt	61	2.4
Metamorphics	0	0
Alluvium	0	0
Total	2618	100

This distribution is significantly different from random (c²=6058; p=0.0001) and suggests a correlation between the distribution of dieback patches and that of igneous rocks. This is consistent with the observations of Brown (1997) that dieback patches occur on areas of low nutrient status with poorly drained subsoils.

The digital elevation model provided by WTMA was used to examine the distribution of dieback patches by altitude in the Tully Falls area. Nearly all dieback patches occur at elevations between 750 and 1050m, with some outliers at lower elevations. However it is worth noting that Brown (1999) found dieback at a wide range of elevations from near sea level to over 1000m. An aspect map produced from the DEM showed no clear association between patch occurrence and aspect. This is in contrast to the study of Wills (1993) in the Stirling Ranges, WA, where Phytophthora dieback tends to occur on shaded and moister southerly and easterly slopes. It is likely that in the Tully Falls area there are no sites where soil moisture is a limiting factor. An analysis of a slope map derived from the digital elevation model was carried out. Within the study area, most dieback patches occur on relatively flat areas with slope angle less than 5 degrees. Brown (1999) noted the co-occurrence of dieback patches with moist flat areas on ridge crests at Eungella. Thus this is consistent with the results from the Tully Falls study. Feral pigs also frequently disturb poorly drained detention hollows in such sites.

Most dieback patches occur in simple notophyll vine forest (Webb & Tracey types 8/9). Small numbers of patches also occur in complex notophyll vine forest (type 5a), several mesophyll forest types (types 1a, 1b, 2a), and closed forest with eucalypts (type 13c). The distribution of patch areas by vegetation type (Figure 1) indicates that there is a strong spatial correlation between dieback patches and notophyll forest types (Table 2).

Table 2: Distribution of dieback patches by vegetation type

Vegetation type	Area (ha)	% of total
Mesophyll forest types	125	3.7
Vegetation complexes and mosaics	3	0.1
Eucalyptus, Corymbia and Acacia Closed Forest	76	2.2
Notophyll Forest Types	269	8.0
Acacia Emergent Forest	2	0.1
Notophyll Forest /Microphyll Forest and Thickets	2835	83.9
Tall Open Forest/Open Woodland	68	2.0
Total	3378	100

This distribution is significantly different from random (c2=9935; p=0.0001). The same pattern is also evident from preliminary analysis of mapped dieback patches at Mount Lewis (Rumula 1;100,000 sheet area) and the Lamb Range (Tinaroo 1:100,000 sheet). Further mapping of dieback patches from aerial photography will allow testing of this relationship with a very large sample of patches.

Many dieback patches occur close to roads, implying that road use, or road building and maintenance activities may be implicated in the spread of dieback. The road vector dataset is based on 1:50,000 scale compilations and includes all formed roads from sealed surfaces to minor unsealed roads eg. snigging tracks to former log dumps. It would be difficult to accurately map minor tracks and walking tracks due to the high canopy coverage. Canopy dieback polygon areas (in hectares) were expressed as a fraction of the total area of each road buffer zone (100m wide). This is necessary to avoid bias due to the systematically reducing area of each buffer zone with distance away from the defined road. The regression analysis of fractional area against distance from roads provides an r² value of 0.527, significant at p=0.05 (Figure 2). This was compared with a random sample of polygons with the same total area (930ha), which is poorly correlated (r² = 0.121) with distance from roads. Thus there is a strong suggestion that spread from roads may be a factor in canopy dieback. However, other factors are almost certainly involved. One influence is the normal location of roads on relatively flat areas of land which may also have impeded drainage and thus be susceptible to root pathogens. Another is the predilection of feral pigs for dense regrowth along roads.

Drainage patterns in the Tully Falls area were sourced from AUSLIG topo-250K datasets. These data were compiled at a map scale of 1:50,000 but may omit very small streams, gullies and detention hollows which store water. Many dieback patches occur within or close to a defined drainage line. A histogram of the area distribution of dieback patches within buffer zones of the drainage indicates a strongly skewed distribution which differs significantly (c²=1829.5; p=0.0001) from random. From this analysis, 51.2% of the dieback patches occur within 200m of a defined drainage line. The association of Phytophthora dieback with drainage lines elsewhere is well established (Wills, 1993; Davison, 1994; Peters & Weste 1997). This relates to the transport of zoospores in soil percolation water, providing a ready means for soil fungi to spread.

A linear combination of the three variables most likely to be associated with dieback at the regional scale was used to map country types susceptible to dieback. The three variables were: altitude between 750 and 1050m; notophyll forest types; and acid igneous rocks. The resulting map (Figure 3) indicates that approximately 14% of the World Heritage area may be susceptible on the basis of this combination. This proportion may be reduced with further refinement to incorporate proximity to roads and poorly drained areas, but this would be better done at a local scale.

The next phase of the detection and mapping component of the dieback project will be to develop specific spectral signatures for dieback patches from airborne video (Louis et al, 1995) and Landsat ETM+ data and apply them to the whole study area. The spatial sampling strategy will be based on recorded dieback patches and will extract all pixels within a patch. In addition, spectral signatures from a random sample of equivalent areas of healthy forest canopy will be used to assess canopy variance. It will also be possible to construct linear spectral transects across larger patches where there is significant environmental variation. Comparison of the two spectral signatures should give us a clear idea of the extent to which we can scale up from the high resolution to lower spatial resolution data, making routine monitoring a reality. The development of dieback patches over time will be assessed by comparing the spectral signatures of the same canopy areas for a series of Landsat TM scenes. These scenes will shortly become available through the Rainforest CRC and cover the period 1972 to 1999 at roughly five year intervals.

The prospect of defining environmental correlates looks very good from the preliminary study of the Tully falls area, and we should thus be able to define areas at risk and establish GIS protocols for monitoring using remotely sensed data. By adopting a decision tree analysis we should be able to develop an heuristic which will classify mixed data sets (remotely sensed data combined with mapped GIS variables) into dieback susceptible and non-susceptible areas of the Wet Tropics World Heritage Area. Complementary ground surveys will focus on the nature of the association between the occurrence of dieback and Phytophthora cinnamomi, and the impact of dieback on rainforest diversity.

Acknowledgments

We are grateful to the Wet Tropics Management Authority (WTMA) who provided funding for part of this study through the CRC for Tropical Rainforest Ecology & Management. Mr Terry Webb of WTMA provided several GIS datasets used in this study.

References

Brown, B.N 1976. Phytophthora cinnamomi associated with patch death in tropical rain forests in Queensland. Aust. Plant Path. Soc. Newsl. 5: 1-4.

Brown, B, 1999. Occurrence and impact of Phytophthora cinnamomi and other Phytophthora species in rainforests of the Wet Tropics World Heritage Area. Pages 41-76 in P. Gadek (ed) Patch Deaths in Tropical Queensland Rainforests. Report, Rainforest Cooperative Research Centre, Cairns.

Davison, E.M. 1994. Role of the environment in dieback of jarrah, J. Roy Soc. WA 77:123-126.

Wills, R.T., 1993. The ecological impact of Phytophthora cinnamomi in the Stirling Range National Park, Western Australia, Aust. J.Ecol. 18(2):145-160.

back to Dave's research home page

Dieback mapping & assessment in rainforests of tropical North Queensland

David Gillieson , Jill Landsberg, Paul Gadek and Will Edwards

School of Tropical Environment Studies & Geography and School of Tropical Biology