SA Land Cover

Dataset Number2029
Dataset TitleSA Land Cover
CustodianData Authority: Department for Environment and Water
Data Provider: Department for Environment and Water

Description

DescriptionSouth Australian Land Cover provides a consistent through time, whole-of-state, spatial land cover data set for 6 epochs (1987-1990, 1990-1995, 1995-2000, 2000-2005, 2005-2010 and 2010-2015). The dataset uses Landsat satellite imagery and local calibration (or training) data. The most likely layers (one for each epoch) classify each pixel in to one of 17 land cover classes. The dataset provides additional evidence for use alongside other South Australian mapping products to measure and map native and non-native land cover types.
UsageThis dataset can be used to inform spatial and temporal (5-yearly) summaries of the described land cover types for SA. The capture method and general nature of the classes are most useful for landscape and regional scale assessment (1:50,000 scale or smaller). Localised or paddock scale applications may encounter classification errors as a result of the modelling method that characterises spectral similarity. Note: it should not be assumed that this dataset represents a definitive land cover class at any given pixel, rather it aims to provide a consistent spectral description of land covers through time / between epochs. Associated datasets can be obtained by contacting the Data Authority where the class definitions, spatial resolution and five year time intervals are appropriate for a particular type of analysis or application.

Please Note: These layers are provided in a geographic coordinate system and as such should not be used to calculate measured area summaries. Area summaries will differ from those reported by DEW if all grid cells are assumed to have an area of 625 square metres (25 m x 25 m). To calculate area summaries a supplementary GIS layer, SA Scaled Area Grid, available from Data.SA, can be used with GIS techniques to provide area summaries in square metres.

Data Quality

LineageThe SA Land Cover outputs are the product of a research collaboration between DEW, the Arthur Rylah Institute for Environmental Research (ARI) and Geoscience Australia (GA) as described in a Technical summary report. The Most Likely layers have been generated using machine learning and remote sensing techniques to characterise Landsat pixels according to statistical similarity across Landsat spectral bands. The modelling summarises similarity within seasons and then for the six 5-year epochs - (1987-1990, 1990-1995, 1995-2000, 2000-2005, 2005-2010 and 2010-2015). This process was informed and validated by a dataset of 322,330 training points across SA with variations in temporal and spatial context. These were collected/collated in conjunction with departmental and external stakeholders. Some independent variable layers have also been incorporated in the modelling approach to support the classification model including ALOS radar, Landsat band max/min differences and other indices. The six most likely layers containing 17 land cover classes, were generated by ARI from rules based on the combination of primary input layers. These include probability and confidence layers for 55 modelled continuous land cover classes (the continuous layers are not described further in this metadata record). The 17 land cover classes have been develop as part of an iterative process which included feedback from stakeholders, reinterpretation and post- processing after considerable QA/QC through four major ARI iterations. Some land cover types in certain locations have been re-classified (cleaned) to update any gross class-based misclassifications identified as part of the QA /QC process. A DEW technical note that describes the post-processing undertaken is available upon request.
CompletenessDataset may be considered complete for the six 5 year epochs (1987-1990, 1990-1995, 1995-2000, 2000-2005, 2005-2010 and 2010-2015). As data validation and update processes are determined these layers may be superseded by subsequent versions.
Positional AccuracyThe positional accuracy is dependent on the original source of the imagery and the modelling process and where applicable exemplar data has been provided. Although the land cover models layers have not been projected from a Geographical Coordinate System the minimum area of mapping is approximately 25 m x 25 m (1:50,000) in line with the underlying Landsat archive outputs obtained from GA.

Geographic Extent

Extent NameSouth Australia
Min Easting or Longitude129
Min Northing or Latitude-38.25
Max Easting or Longitude141
Max Northing or Latitude-26

Dataset Status

Initially Acquired01/03/2015
Last Updated16/10/17
Update FrequencyAs required
Maintenance MethodDatasets are updated as required - either by the data authority's in-house data validation process analysis or rerunning the remote sensing modelling using adopted methodology created by the collaboration partners. The current version 4.4 was finalised in October 2017 is documented in a Technical summary report.
AusGOALCC BY (Attribution)

Attributes

TableNameField AliasData TypeDescription
1) Raster Attribute TableVALUE IntegerThe raster value of each pixel (1-17)
2) Look Up TableLC_NAME textThe name of the most likely land cover corresponding to the value within each pixel
2) Look Up TableLC_DESC IntegerA short description of the most likely land cover corresponding to the value within each pixel
2) Look Up TableLC_GEN_CDE IntegerA code relating to a generalised DEWNR version of land cover codes. This code relates to the native/ non- native vegetation class determination of each pixel.
2) Look Up TableLC_CODE IntegerThe assigned most likely land cover class code - directly linked to the value of each raster pixel
2) Look Up TableLC_GEN_NME textThe name of the generalised land cover assigned to DEWNR's generalised land cover code.
2) Look Up TableLC_GEN_DSC textThe generalised land cover description assigned to DEWNR's generalised land cover code.

Contact Information

OrganisationDepartment for Environment and Water
Science Monitoring and Knowledge
Postal AddressGPO Box 1047

ADELAIDE
5001
EmailDEWNR.CustomerServiceCentre@sa.gov.au