This type of landcover disturbance defines the ILZ. We summarise our findings for the continent under the following four headings
Aggregate Area
Within the ILZ, we assessed that 1,547,863 km2, or 51.9% of the area was either cleared or thinned.
This in turn translates to approximately 20% of the whole continent.
Table 5: A summary of the five classes of clearing disturbance aggregated for the Intensive landuse Zone (ILZ) and for the continent.
Cover class km2 ILZ (%) Continent (%) _________________________________________________________________________ Uncleared 1059741 35.5 13.8 Thinned 518223 17.4 6.7 Cleared 1029640 34.5 13.4 Indeterminate 365350 12.2 4.7 Other 10953 0.4 0.1 _________________________________________________________________________ Total 2983908 100.0 38.8
How good is this estimate?
There is no single dataset to test the validity of this estimate. There were however two national estimates that allow us to examine its plausibility.
The first of these is the Australian Bureau of Statistics (ABS) estimates for the total area of crops and pastures in 1990; ABS (1994). The ABS figure is 475, 000 km2; roughly one third of the figure derived in this study. However, the areas reported by the ABS have two important qualifications: they do not equate to the total area cleared, and the figures were derived by survey sampling not census sampling as in this study. The figure published by the ABS is the area used for crops in that year. In cereal cropping areas with a 2-3 year crop rotation, it suggests that the total area cleared for cropping may be three times that used in any one year. Since the ABS reports that the area of crops peaked at approximately 220, 000 km2 in 1983-84, then the total area cleared for cropping alone may be 660, 000 km2.
The area cleared for the establishment of pastures was more difficult to separately identify. Undoubtedly, in some agricultural areas there is a reciprocal substitution of pasture and crop depending on commodity prices. Nonetheless, there has been an expansion of clearing for the establishment of sown pastures alone. This clearing and establishment have been most marked in northern NSW, and especially Queensland. Here the second set of figures adds plausibility to the total figure derived in this study.
The estimates of landcover change published by AUSLIG (1990) were based on the manual interpretation of satellite data acquired a decade ago. That survey recorded extensive thinning as well as clearing of woodlands for the establishment or enhancement of pastures. The aggregate figures produced by the AUSLIG survey suggest that in the last 200 years, there has been a net increase by thinning of 400, 000 km2 of open woodlands, and by clearing, a net increase in grassland and pasture areas of 710, 000 km2. The sum of these two AUSLIG estimates is markedly less than the area of thinning and clearing that we measured, indicating an alarming increase of those activities over the last decade.
No doubt there were uncertainties associated with both estimates but it is unlikely that the agreement between the two census assessments based on satellite data was fortuitous.
Where is the clearing?
The distribution of the ILZ and the extent of clearing and thinning within it, is shown in this ILZ disturbance image.
The continental distribution of disturbance by clearing. The classes and colour codes are: uncleared (green); thinned (brown); cleared (white); indeterminate (blue); lakes (black).
Clearing and thinning extends across the southern areas of the continent, including the central area of Tasmania, and northwards through Central Queensland. The most southerly areas of clearing were associated with cereal cropping and improved pastures; that in northern NSW and Central Queensland was mostly for the establishment of pastures.
How to interpret the four disturbance classes displayed in this image?
The three key classes are uncleared, thinned and cleared. These were allocated as progressive steps from uncleared, where the vegetation canopy was intact, to cleared, where the spectral signature of the intact canopy was replaced by that of understorey. The thinned class was an intermediate step within these two extremes based solely on spectral signature; it was not primarily based on the measurements of the degree of thinning, such as tree densities, but the two are closely related. The decision rules for the allocation of each pixel in this image to one of these three classes were formulated after exacting expert interpretation, and consistently applied. The image depicts our assessment of the extent and location of vegetation disturbance by clearing as of approximately 1990.
The indeterminate class displayed in the image illustrates the landcover types within the ILZ for which an assessment of clearing disturbance could not be made.
The largest area was in the Northern Territory where localised clearing, almost entirely for pastures, could be detected but not reliably mapped. The areas involved, including the Ord River irrigation development, were trivial compared with the clearing in southern Australia.
The indeterminate areas outside the Northern Territory were the landcover types without a woody overstorey, the grasslands, for which the classification of clearing was inappropriate.
This is not to say that these landcover types are undisturbed. Cultivation was detectable within the cleared grasslands (xG4) on the west coast of Tasmania and within the most mesic grasslands of the other States and Territories; see, for example, TASMANIA. The grasslands at the drier edges of the ILZ (xG3, xG2) are also disturbed, principally by grazing.
This display of the clearing data-set also illustrates the nature and strength of the uncertainties associated with the assessments reported in this document. The error here is misclassification: uncleared was classified as thinned or cleared, and the converse. As explained earlier, because of the intrinsic nature of the source satellite data-set and the fallibility of the expert process, such errors can never be eliminated.
Rather, in this project, the uncertainty has been managed by minimising and systematising the error by consistently biasing it in one direction. With a focus on the uncleared class, a bias towards the error of commission was preferred. That is, the uncleared class decision rules were set to optimistic values, to err toward a high value. Generally we achieved that so that the actual undisturbed areas with the ILZ were less that the values reported here.
Errors of commission concerning the uncleared class are difficult to demonstrate in the image. However, the converse, error of omission where uncleared vegetation was labelled as thinned or cleared can be illustrated.
Along the clearing boundaries of parts of Western Australia, South Australia, Tasmania and northern Queensland, there are areas labelled as cleared that are not. We encourage you to examine the images of these areas provided in Exploring Your Country. In particular, you will find the images of TOWNSVILLE, CLERMONT, CHARLEVILLE, BOURKE, BROKEN HILL, ADELAIDE, PORT AUGUSTA, TASMANIA, ESPERANCE, ALBANY, PERTH and MEEKATHARRA are of particular interest.
This error could not be avoided, only minimised. There are two principal sources of this error: problems intrinsic to the satellite dataset, and limitations to separating clearing from significant disturbance by grazing. Nevertheless, the nature and size of these errors were small compared with the information content of this dataset that is summarised in the remainder of this section.
Is the clearing and thinning evenly distributed across all landcover types?
Considering the entire ILZ, the distribution of disturbance by clearing between landcover types is summarised in Table 6. A simple index of Loss (thinned and cleared) was also calculated reflecting the view that for the process of biotic erosion, there is little difference between thinned and cleared.
Table 6: A summary of the relative distribution (%) of clearing within each landcover type. The grasslands (xG) and low shrublands (xZ2) were excluded. The index Loss was the sum of the cleared and thinned classes.
Type Uncleared Thinned Cleared Indeterminate Loss
_____________________________________________________________________________
xTML4 73.9 17.2 8.9 0.0 26.1
eTML3 61.6 17.0 20.1 1.3 37.1
wTML3 19.7 36.7 43.6 0.0 80.3
xTML3 29.1 17.0 53.9 0.0 70.9
eM2 26.7 22.5 42.8 8.0 65.3
wM2 19.6 14.2 66.2 0.0 80.4
xM2 43.3 24.7 32.0 0.0 56.7
eL2 18.6 10.3 29.3 41.9 39.6
wL2 13.6 21.3 65.0 0.0 86.4
xL2 90.3 2.0 7.7 0.0 9.7
eM1 7.8 43.3 48.9 0.0 92.2
eL1 20.5 11.0 21.8 46.6 32.9
wL1 96.5 0.0 3.5 0.0 3.5
xL1 60.7 4.1 35.1 0.0 39.3
eS3 1.0 1.5 97.5 0.0 99.0
wS3 37.5 18.5 44.0 0.0 62.5
xS3 43.8 3.1 53.1 0.0 56.2
eS2 17.6 19.1 63.3 0.0 82.4
wS2 97.1 2.2 0.7 0.0 2.9
xS2 38.8 18.4 42.8 0.0 61.2
eS1 14.0 11.6 74.4 0.0 86.0
wS1 100.0 0.0 0.0 0.0 0.0
xS1 100.0 0.0 0.0 0.0 0.0
xZ3 38.7 16.3 45.0 0.0 61.3
wZ1 100.0 0.0 0.0 0.0 0.0
Littoral 100.0 0.0 0.0 0.0 0.0
Overall the average proportion of a landcover type that was cleared or thinned (Loss) was 47% with values ranging between 0-99%.
As might be expected, the highest Loss values are for the smallest landcover types as is illustrated in the following Loss of Area graph.
The relationship between the Loss index (thinned and cleared) and area for the ILZ landcover types. The highest Loss values are for the smallest, and most likely restricted, landcover types.
To appreciate which landcover types have experienced the greatest disturbance in absolute terms, the aggregate contribution of the cleared and thinned areas was plotted in the following descending size rank graph.
Landcover types arranged in descending order of the total absolute area disturbed (thinned and cleared classes).
The landcover type with the largest disturbed area was the open eucalypt forests (eM2), which combined with the next largest landcover type (eS2) account for 45% of the total area.
By far the largest clearing has taken place in the open eucalypt forests (eM2), followed by the open eucalypt shrublands (eS2), forests (eTML3), sparse forests (eM1) and so forth. The area cleared within the first two landcover types accounts for 45% of the total area cleared.
Tenure is a land management tool; a type of landuse licence issued by state governments. Landuse activities are prescribed and proscribed by certain tenures, and clearing is one such activity. Thus, the distribution of the landcover disturbance classes between various tenure types reflects the management of clearing by state governments. All six tenure types defined in Table 4 are represented within the ILZ and their relative sizes are illustrated by the following tenure pie chart.
The relative area of the six tenure types within the ILZ.
Tenure type 1 (Freehold, Mining leasehold) and type 2 (Pastoral leasehold, Defence tenure) between them account for 81% of the ILZ.
What is the proportional area of each tenure type that is either cleared or thinned?
The relative size of the area that was thinned and cleared relative to the total area of that tenure type is displayed as the following tenure-clearing pie chart. For tenure type 1 (Freehold, Mining), this proportion was high (74%) and less so for tenure type 2 (35%) the Pastoral (and Defence) tenures.
The relative area of thinned and cleared landcover within each of the six tenure types. Tenure type 1 (Freehold) and type 2 (Pastoral) dominate; this is where the clearing is mostly found.
In absolute terms, these are the two tenures that contain the cleared and thinned landcovers as would be expected. All of agriculture, a replacement landuse, is held under Freehold tenure while most pastoral activity, a harvesting landuse, is conducted mostly under Pastoral tenure with some under Freehold.
Can we use this finding to check on the overall validity of our assessment of clearing and thinning?
The tenure-clearing pie chart revealed that there was a small occurrence of cleared and thinned classes within the tenure types that offer high protection against clearing; types 4, 5 and 6. Type 6 is dedicated conservation reserve. These occurrences are obviously a misclassification and an error of omission; uncleared landcover classified as cleared.
Table 7: The distribution of absolute area (km2) of the three critical disturbance classes within the six tenure types of the ILZ.
Tenure Grouping Uncleared Thinned Cleared Thinned+Cleared
____________________________________________________________________________________
1 Freehold 278230 280655 822838 1103493
Mining leasehold
2 Pastoral 434700 177134 142970 320104
leasehold
Defence tenure
3 Forestry 146064 12920 10061 22981
Water Supply
Mixed, Other,
Multi-
4 Aboriginal 22153 2001 2639 4640
freehold /
leasehold
national park
Aboriginal
leasehold
national park
5 Unallocated crown 38315 9895 14125 24020
land (unused)
6 Dedicated nature 127778 30228 26585 56813
conservation
Total 1047240 512833 1019218 1532051
In relative terms within each tenure type the error of omission was large; up to 40% for tenure type 5 (Unused). However, in absolute terms, it is less damaging. If we assume that there should be no clearing at all in tenure types 3 to 6, then the proportion of misclassified pixels was 108454 from a total of 549014; an omission error rate of 20%.
This is a worst case error analysis which if continentally applicable, reduces the total area thinned and cleared from 1, 550, 000 km2 to 1, 240, 000 km2.
We are confident that the error detected by this check was not continentally applicable but we are unable to substantiate this opinion. Nevertheless, this calculation is a measure of the uncertainty associated with all aggregate measurements reported in this project.
