• Summary
• Downloads
• Explanatory Notes
• Related Information
• Past Releases

Realising the Potential of Satellite Imagery to Estimate Official Crop Area Statistics

The ABS is currently investigating the potential of satellite imagery data to aid in the production of agricultural statistics, such as land use, crop type and crop yield. Satellite data is seen by the ABS as a useful data confrontation source, and in time, may supplement information directly collected in the agricultural program. Therefore scoping out and demonstrating the ability of this big data source to produce comparable and/or more sophisticated agricultural statistics is a key driver behind current research efforts. Success in this space may see significant reductions in provider burden and improve the timeliness of available data and thus extend the array of stakeholder requirements met or exceeded.

The Analytical Services Branch (ASB) has been working with the subject matter and technology areas in ABS to scope out future directions for the project. This has included consideration of external stakeholder needs and any constraints that may be introduced from a technology infrastructure perspective given the magnitude of big data in scope. This has resulted in the formation of three primary future directions:

Establishing partnerships: Nurturing relationships with experts in the field is essential in fast-tracking our own project work. It is also viewed as an ideal opportunity to build the ABS reputation via leading big data discussions and sharing of experiences. ASB has contributed to an international collaboration on big data through the UNECE High Level Group for the Modernisation of Statistical Production and Services. Ongoing engagement has also been established with CSIRO, Australian Bureau of Agricultural and Resource Economics and Sciences (ABARES) and Geoscience Australia, given the wealth of detailed satellite imagery data that can be remotely accessed through their collaborative computing installation - the National Computational Infrastructure (NCI, see http://nci.org.au).

Method development: Gaining an understanding of the sophisticated methods available and the assumptions underpinning them is essential in motivating future research directions. This enables ASB to make an informed assessment of which methods can be appropriately adopted to assist in addressing key research questions and thus meet the needs of stakeholders. Specific methods include state space modelling approaches for leveraging off spatial and temporal trends in the data, machine learning algorithms and parametric modelling approaches. These all seek to classify crop types based on surface spectral reflectance information collected by satellite sensors.

Acquisition of data sources: ABS requires additional data sources in order to confidently test and validate statistical methods and determine the extent to which these may be applicable in practice. There has already been some progress in this space with ground truth data acquired for some regions which enables informative statistical testing. Geoscience Australia has also provided the ABS with satellite surface-corrected reflectance data, which is one of the richest data sources of this nature in Australia. Ultimately ASB views this initial extract of data as a step towards signing up to a more sustainable position of analysing this high quality satellite imagery directly through the NCI. Both sources will see the richness of training datasets improve dramatically in the near future, which will allow for better method assessment.

A Methodology Advisory Committee (MAC) paper was presented in June 2014 appraising potential statistical approaches to extract relevant information from satellite imagery data - see Marley, Elazar & Traeger (2014). It is planned that further method development in this area will see an additional MAC paper focussed on applying state space modelling (SSM) approaches, to be delivered in June 2015. This paper will include results from evaluating the use of SSMs for classification of agricultural crops using the aforementioned test datasets.

References
Marley, J., Elazar, D. & Traeger, K. (2014) ‘Methodological Approaches for Utilising Satellite Imagery to Estimate Official Crop Area Statistics’, cat. no. 1352.0.55.144, Australian Bureau of Statistics, Canberra (not yet released).
Tam, S. & Clarke, F. (2014) ‘Big Data, Official Statistics and Some Initiatives by the Australian Bureau of Statistics’, International Statistical Review (provisionally accepted).

Further Information
For more information, please contact Jennifer Marley (02 6252 5760, jennifer.marley@abs.gov.au) or Ryan Defina (02 6252 7779, ryan.defina@abs.gov.au).

The ABS Privacy Policy outlines how the ABS will handle any personal information that you provide to us.