Forging Close Collaboration Between EO Scientists and Official Statisticians – An Australian Case Study
EO data are increasingly being recognised by official statisticians as an important data source for the production of official statistics. Since the establishment of the UN Global Working Group on Big Data in 2015, the relationship between official statisticians and EO scientists has grown beyond Australian Bureau of Statistics (ABS) and Geoscience Australia (GA) in the use of EO data for environmental accounts, to include the Commonwealth Scientific and Industrial Research Organisation (CSIRO, a member of CEOS), other Australian Federal and State agencies and academics in the development of an EO Handbook for official statistics, and content for capacity-building workshops in Asia and Latin America. The positive experience in this collaboration not only increased the understanding of official statistics production and EO data sources, as well as its strengths and weakness, but also provides a pathway for more creative and productive use of EO data between official statisticians and EO scientists in the years to come.
3.1 Uses of EO data in Australia’s official statistics
Australia’s decision-makers need reliable information about changes in the use, condition and value of land and how this relates to broader economic activity and the state of our environment. This information is used in policy making and spending decisions in land management and economic development. Official statistics are well placed to provide this vital information by integrating geospatial and environmental data with a range of economic indicators.
The ABS uses the UN System of Environmental-Economic Accounting Central Framework (SEEA) to guide the production of data concerning the environment and economy. Land Accounts form the foundation of all environment-economic accounts, and the ABS has produced a series of Experimental Land Accounts across several jurisdictions in Australia. Rather than collecting the required data itself, the ABS sources and integrates data from a number of government organisations to produce these Land Accounts.
An important input to Land Accounts is data about land cover, both the biophysical cover and the built environment. Through the application of geoscientific expertise and capabilities, the raw satellite observations can be used to detect and map land cover.
The ABS currently works in partnership with GA, which is also providing EO services, expert advice and capabilities, and information for decision-makers, to realise the value of geospatial data and EO data to enhance the production of official statistics in Australia. A combination of high level engagement and technical collaboration has strengthened a productive and mutually beneficial inter-agency relationship. This national partnership has been critical in ensuring that GA’s Dynamic Land Cover Data (DLCD), a consistent national dataset of land cover, has been available for the production of Experimental Land Accounts. The use of the DLCD time-series in Land Accounts provides a ‘line of sight’ between EO data and the information available to policy makers via statistical products.
This has also motivated GA to continue the production of DLCD. GA has worked closely with the ABS on the release of a new version of DLCD for Australia. This new data became available in time for the preparation of the Land Account: Queensland, Experimental Estimates, 2011 – 2016 publication by the ABS, which was released in June 2017.
Figure 1: The Data Cube concept
3.2 A global platform for collaboration
The UN Global Working Group (GWG) on Big Data for Official Statistics was created in 2014, as an outcome of the 45th meeting of the UN Statistical Commission (UNSC). In accordance with its terms of reference, the UN GWG provides strategic vision, direction and coordination of a global programme on Big Data for official statistics, including for indicators of the 2030 Agenda for Sustainable Development. It also promotes practical use of Big Data sources, while supporting capability building, training and sharing of experiences. Finally, the UN GWG fosters communication and advocacy of use of Big Data for policy applications and offers advice in building public trust in the use of Big Data from the private sector.
Between 2014 and 2017, ABS chaired the UN GWG and one of its Task Teams (TT), Satellite Imagery (EO data) and Geo-Spatial Statistics. Consistent with the strategic vision and direction of the GWG, the ABS TT decided at its early meetings to establish a work programme to share best practice in using EO data in the production of official statistics on agriculture and environmental accounts, and to host workshops in Asia and Latin America to help build the capability of Asian and Latin American National Statistical Offices in using EO data for official statistics.
In developing the TT Handbook on EO data, and in developing the content of the methodology workshops (a trial run of which was held in Canberra, Australia, in early 2017), official statisticians from ABS worked closely and collaboratively with EO scientists from CSIRO; GA; Australian Bureau of Agriculture and Resource Economics and Sciences (ABARES); and Queensland Department of Science, IT and Innovation (DSITI); and academics from the Queensland University of Technology (QUT) chapter of Australian Research Council Centre of Excellence in Mathematics and Statistical Frontiers (ACEMS).
Throughout the Handbook, many case studies and examples are showcased. In particular, the SDG indicators were often referred to as examples, since they appear as a unique opportunity for EO and NSOs to work in close collaboration. As such, forest monitoring – and especially deforestation tracking to help achieve Goal 15 (Life on Land) – is an example where EO brings a unique contribution. The Global Forest Observations Initiative (GFOI) of GEO, which relies on a few core sub 30-m resolution satellite data sources, enables a continuous, annual and global coverage to monitor such indicators worldwide over the years. It will thus directly help monitoring indicator 15.2.1 (Progress towards sustainable forest management).
Figure 2: Water Observations from Space (WOfS) results for the braided river network of Coopers Creek in south-western Queensland, Australia
3.3 Task Team Handbook
The TT led by ABS produced a Handbook to provide guidance to NSOs about the use of EO data, which include both satellite and in-situ data, for official statistics. This was reviewed by experts including the UN Food and Agriculture Organization (FAO) and the Joint Research Centre of Eurostat before being finalised in June 2017.
CSIRO played a significant role in producing the Handbook as the primary authors of Chapter 2, with some inputs from GA. The content of the report has been driven by the direction given by the UN Statistics Division (UNSD) for TTs to produce a set of guidelines that include:
- Introduction that provides the motivation and makes the business case, including modernisation of official statistics and monitoring the 2030 Sustainable Development Agenda through the SDGs;
- Data sources – description and explanation, as well as scope of the guidelines;
- Statistical methodology and applications;
- Concluding section with further work to be done.
The report also outlines the TT pilot projects and guidelines for practitioners exploring the use of EO data for the first time.
The TT pilot projects described in the report are:
- Pilot Study Proposal for the application of satellite imagery data in the production of agricultural statistics (ABS);
- Preliminary Analysis of Climate Scenarios to Improve the Environmental Characterization of Various Climatic Regions in Mexico (INEGI, Mexico);
- Use of satellite images to calculate statistics on land cover and land use (DANE, Colombia).
The crop yield work by Statistics Canada is also included as an example of how satellite imagery can be used in the production of official statistics in practice. Statistics Canada is an NSO that has used remote-sensing data since the early 1980s for applications such as census and survey validation, crop mapping and area estimation and support program development. In 2016, Statistics Canada became the first NSO to replace a statistical survey with a remote-sensing model-based approach. The goal of the model was to produce a preliminary estimate of the expected harvest yield of the crops in late summer using information from existing data sources.
ABS is also considering EO data for statistical outputs as part of the larger administrative data initiative and has received presentations from Statistics Canada about their methods and experiences. This has been very useful to the ABS agricultural statistics area, the Satellite Imagery TT report and EO workshop development.
Figure 3: Darwin Harbour: representative Landsat images and the time series along a transect through the mangrove forest at Frances Bay. White arrow: area of urban expansion.
3.4 ABS, CSIRO and GA collaboration
Since the inception of the ABS TT, CSIRO has been providing substantial inputs on a number of dimensions of the Handbook, including EO data sources available (besides the most common MODIS and Landsat images), how to get access to it, how to process and analyse it (using new statistical methods and algorithms), and how ABS and CSIRO can share and build on their expertise to collect and use more satellite data to enhance the official statistics for public good.
Through this collaboration, EO concepts like ‘validity’, ‘accuracy’, ‘timeliness’, or ‘coherence’ and ‘analysis-ready data’ (ARD), as well as the different EO sources, their advantages and disadvantages, are adequately explained in the Handbook to help official statisticians make informed decisions on choices of EO data for official statistics.
As mentioned earlier, collaboration between ABS and GA, which started in early 2000s, motivated GA to develop the DCLD and resulted in the incorporation of EO data in the ABS June 2017 release of a Land Account publication.
This national and growing collaboration (future projects includes Data Cube technology for agricultural statistics in Australia) has provided a tremendous opportunity to make EO data more accessible, known and used by official statisticians and hopefully showcases an example for similar partnerships to be forged in other countries.
Following this initial positive collaboration in the context of the TT, Australian agencies including ABS, CSIRO and GA will continue to pursue their efforts in working together and bringing more EO data into the official statistical framework and will continue to share their experience and expertise in official statistics and EO data to provide better information for decision-making in Australia.
3.5 Concluding remarks
As ABS becomes more experienced with EO, the organisation is committed to sharing this knowledge with the international statistical community and helping build the capability of other NSOs in the use of EO data for official statistics.
In this exercise, ABS has found that partnerships between EO scientists and academia are important to progress the use of EO data for official statistics. Through such partnerships, developing and maintaining strong ties with international organisations such as CEOS is essential to successfully harness EO data for official statistics.
Article Contributors
S.M. Tam, J. Holloway, R. Dunsmore, M. Jakab (Australian Bureau of Statistics)
