Aboriginal sites decision support tool (ASDST)
Aboriginal site features occur across the entire landscape; however some parts of the landscape have a greater capacity to contain certain site features or features of different types. The variation in site feature likelihood across the landscape is useful for planning assessments of potential site impacts. The ASDST has been developed to support the assessment Aboriginal sites issues in NSW at the landscape-scale. The tool extends the Aboriginal Heritage Information Management System (AHIMS) by illustrating the potential distribution of site features recorded in the database.
What are site predictive models?
The maps of site feature predictions made by the ASDST are based on the application of site predictive modelling. This is a technique used to correlate site information in AHIMS with landscape patterns such as proximity to water, vegetation, terrain, soils etc. The maps provide a regional overview about site feature distribution and related issues about the level of accumulated impacts they have experienced, where data gaps in the AHIMS data base remain, and where these gaps can be addressed through further survey.
For more information see:
How can they be used?
The mapping tool provides interactive access to the maps generated by the ASDST, and can be used as a way of visualising site feature potential and related issues across the whole landscape. The ASDST Product Outline Technical Summary (20100956ASDST.pdf, 4.6MB) is a document describing the appropriate interpretation of the ASDST products, their application in regional-scale projects and a technical explanation of how they were derived. Users are encouraged to read this document in order to interpret the ASDST products appropriately and understand how they were derived.
What are the limitations?
The products have been developed to meet the needs of regional planning. For this reason they are designed to be used at scales of 1:100,000 and above. Application at finer scales is possible, but it should be borne in mind that the datasets used to derive the products were themselves derived at a scale of 1:100,000 or coarser, and therefore the inaccuracies of those layers at finer scales will be carried through to the ASDST products.
It is intended that all products will develop over time. They will be periodically reviewed as new data becomes available, parameters are assessed, models are progressively better validated and modelling approaches are refined. For this reason, all products have a version number attached, and for any analysis, mapping or reporting that is performed from the layers, the version that was used should be specified. The current version of all products is version 7.
Contact the ASDST support team
For further information and support in interpreting the ASDST products, please contact:
The ASDST is composed of an integrated suite of spatial GIS layers. The collection of GIS layers are organised around the following products:
- Pre1750 (original) models.
- Current models
- Combined accumulated impacts
- Combined model reliability
- Combined survey priority.
The pre1750 (1) and current models (2) products are comprised of 10 raster (grid) GIS layers, each representing different feature types. Splitting the products into feature types permits the landscape distribution and nature of impacts of each feature to be modelled separately, leading to a more specific product. The last three products are single grid layers which are derivatives of the Pre1750 and current products.
The following layers form the pre1750 and current model suites:
- All feature types combined (ALL)
- Stone artefacts (AFT)
- Rock art (ART)
- Burials (BUR)
- Western earth mounds and shell (ETM)
- Grinding grooves (GDG)
- Hearths (HTH)
- Coastal shell middens (SHL)
- Stone quarries (STQ)
- Scarred trees (TRE)
For each product (and individual feature products) is a raster GIS layer. Raster GIS layers describe a feature that varies continuously over a landscape, and is represented by cells (like pixels in a digital photograph) that represent a part of the landscape as a square. The square in the ASDST products represents 1 Ha on the landscape. Raster GIS layers are different to GIS shapefiles that represent features on a landscape as points, lines or polygons. Raster GIS layers have a regularly spaced arrangement of cells that cover the whole landscape. Each cell is attributed with a value indicating the relative value of what the layer is describing at that place. For the ASDST products, this may be the relative likelihood of a site feature, relative survey priority or the relative reliability of the model.
The pre1750 models are a set of raster GIS layers describing the relative likelihood of Aboriginal site features occurring across the landscape. They have been derived using AHIMS data, and a set of spatial variables describing the landscape as it is estimated to have been prior to European settlement.
The pre1750 models do not take into account the level of destruction of sites in the intervening period (for instance from agriculture, mining or urbanisation); nor the detectability of different site features (ie whether locating them would require excavation), or local conditions that may lead to the lack of a feature being preserved. They are therefore not meant to convey a likelihood profile of the present landscape or how easily sites could be located. When interpreting the pre1750 models, the level of impact in the current landscape should be evaluated by comparing them to the corresponding current extent models.
The pre1750 models describe the relative likelihood of finding a particular type of Aboriginal site feature as it is assumed to have been in the pre1750 landscape (eg scarred tree or stone artefact). They are therefore a baseline for site potential in the landscape. The legend for each layer is scaled from white (low likelihood) to black (high likelihood). When looking at any pre1750 layer, the darker it is, the higher the likelihood that that feature could be located there 200 years ago, according to the predictions made by the model.
It is important to keep in mind when looking at how dark an area of the model is, that the darkness is a relative quality. Black does not guarantee that that feature would have been located there, or would still be there today. It represents an area where the model predicts that the likelihood of that feature at that location is high relative to all other areas of the landscape. Similarly, white areas do not indicate an absolute absence of that feature, but the lowest relative likelihood that was resolved by the model.
The models are not calibrated to the absolute probability of presence or absence of site features, but instead describe relative likelihood as it changes over the landscape
The relative nature of the likelihood values of each model also extends to what each cell value’s likelihood is measured relative to. When looking at any of the models, it should be kept in mind that relative likelihood is relative to the entire extent of the layer. Thus:
The likelihood at any given cell is relative to all other cells in the layer (ie the rest of NSW)
A further complication with the relative nature of the likelihood measure is that it is not directly comparable between site features. So for instance, although one area might indicate high modelled likelihood for both stone quarries and artefacts, that does not mean they are both predicted to have an equal probability of occurrence. In this example, the absolute probability of locating quarries is still less because they are generally less frequently observed than stone artefacts. Thus:
The relative likelihood between different site features is not directly comparable in absolute probability terms
The current models are modifications of the pre1750 models so that they reflect a more realistic likelihood of site features occurring in the present-day landscape. These layers do this by utilising tenure, native vegetation extent, land-use mapping and mining history (eg sand-mining) to place parameters upon the likely survival rates of different features under different types of land-use and land condition. Parameters estimating impacts were derived through consultation and a series of expert workshops. Tenure and tenure history was used to provide an idea for how long parts of the landscape have been managed for conservation (eg as a National Park); native vegetation extent was used to identify those areas of the landscape that have been subjected to clearing; land-use mapping indicates how land is currently used (ie cropping, grazing, roads or urban); and mining reveals which parts of the landscape where site likelihood has been irretrievably degraded (eg even though it might now be in a National Park and covered in native vegetation).
The current models take into account estimated historical impacts on Aboriginal features to describe their potential occurrence in the present-day landscape
The caveats described in the previous section for interpreting likelihood for the pre1750 layers also apply to interpreting the current models (ie their scale is approximately 1:100,00). However, the relative likelihood in the current models takes into account the relative chance of a feature to survive under different land-use conditions in the landscape. It does this by reducing the pre1750 likelihood values using the level of impact a land-use probably had on a feature. So for example, a cleared landscape would be expected to reduce the likelihood of scarred trees to almost zero; whereas the likelihood of stone artefacts surviving in the same landscape would be expect to remain almost as high as its original likelihood prior to the landscape being cleared. Similarly, a landscape that is regularly cropped and has been subjected to laser-levelling for irrigation has probably had a detrimental impact on features like earth mounds and hearths, whereas grinding grooves on rocky outcrops in the same environment might be expected to have a reasonably high likelihood of survival as they would generally be avoided by these types of processes.
The current models are designed to be used in conjunction with the pre1750 models when interpreting the relative likelihood of a feature being present in the present-day landscape. Similarly, the difference between the pre1750 model likelihood for a feature and its current likelihood, gives an indication of the total impact on the feature in that part of the landscape. Comparing the current model to the pre1750 model enables a visual assessment of the level of impact on that feature in that area of the landscape.
The accumulated impacts layer is derived from the difference between the pre1750 models and the current models for each feature type, and then summed together. As such it reflects the combined impacts across all the nine features modelled. Areas with high values in the layer reflect areas where the majority of feature types have been heavily impacted. Areas where the combined impacts are low reflect areas where the land-use has had only a minimum impact on the likely survival of features in that part of the landscape.
The combined accumulated impacts layer indicates the impacts of post-settlement land-use history on Aboriginal site features in the landscape
Examples where impacts are typically high include areas that have been mined, or dense urban areas, or areas that have been cleared and regularly cropped. Low impact areas tend to be places where land-use has had a minimal impact on the landscape, such as pristine environments within long-established National Parks or rangelands where the only agricultural activity has been livestock grazing.
Areas where the accumulated (ie summed) impact has been low, can be interpreted has having a comparatively high chance of preserving features at or close to the relative likelihood indicated in the pre1750 models. Correspondingly, areas where the accumulated impacts are high have little chance of preserving Aboriginal site features, or if they do, they are likely to be in a highly degraded state (ie mounds or middens within areas that are regularly cropped).
The accumulated impacts layer provides a landscape view of the level of impacts on Aboriginal site features, and can be useful in assessing the conservation significance of a site. For instance, a site that has been located within an area of the landscape that has experienced a high level of accumulated impacts could be considered as having higher conservation value because it is likely that many similar sites in that area have either been destroyed or highly impacted.
In contrast, locations where accumulated impacts are low (especially if they occur within a region of high accumulated impact), may be used as areas for ground assessment for the location of un-recorded sites as these areas may preserve the last examples of some features (or features in the best condition) in that area.
The model reliability layer was derived by analysing the pre1750 model with a Survey Gap Analysis (SGA). This algorithm looks at those areas where sites with a given feature type have been recorded in the landscape previously, combined with areas that have been surveyed, and looks to see how much of the total landscape variability has been sampled through that recording and assessment history, taking into account the relatively likelihood of the feature occurring through the landscape. The resulting layer is a guide to interpreting the overall reliability of the pre1750 models. Those areas which have been under-surveyed (high values in the reliability layer) are areas where the models are making predictions on the least amount of data, and therefore, are potentially the least reliable.
The combined reliability layer indicates those parts of the landscape that have been investigated the least. Areas with high values in this layer indicate where the predictions made by the model is based on the least amount of AHIMS data
When assessing how well the landscape has been surveyed, the SGA algorithm compares the AHIMS sample against the same variables that were used to derive the model (eg terrain, proximity to water, soils, vegetation etc). The result of the algorithm is a layer that produces high values in areas where the landscape has been under-surveyed, and low values where the survey coverage has been more comprehensive. The process is applied to all feature types, and then combined by summing them all together.
The survey priority layer is a derivation of the reliability layer, in that it combines the SGA result with the parameters used to derive the current extent models to produce a layer that takes into account impacts on that feature and the relative likelihood of that feature occurring across the landscape. High survey priority is defined as being areas of the landscape where there is an intersection of three conditions:
1. pre1750 likelihood is medium to high
2. accumulated impacts are low
3. site coverage and previous assessment is low
The final step is to combine the survey priority layers of all features into a single combined survey priority layer.
The combined survey priority layer defines survey priority as the combination of low model reliability, high modelled likelihood and low accumulated impacts
The approach to interpreting the priority layer is to use it as a guide to areas that would most benefit from on-ground assessment when the objective is to identify areas that haven’t been adequately investigated previously. Since the combined survey priority layer takes into account the relative likelihood of feature occurrence, and its potential survival, areas flagged as high survey priority can potentially be areas of the landscape where assessment may be most productive in terms of identifying new sites, and improving overall coverage of the landscape.
The ASDST has been developed by the OEH as a tool to support the assessment Aboriginal Sites issues in NSW at the landscape-scale. The tool is an extension of the Aboriginal Heritage Information Management System (AHIMS), and provides predictions about the potential distribution of site features recorded in the database. These predictions are based on the application of site predictive modelling, which correlates site information in AHIMS with landscape patterns. The maps provide a regional overview about site feature distribution and related issues about the level of accumulated impacts they have experienced, where data gaps in the AHIMS data base remain, and where these gaps can be addressed through further survey.
OEH takes no responsibility for decisions informed by the ASDST products. The ASDST maps are provided in good faith, are based on international best-practice in site predictive modelling and were developed exercising all due care and attention.
Page last updated: 29 October 2014