Sessions at IASSIST 2017 that resonate with food security research

The 2017 gathering in Lawrence brought together data librarians, agricultural economists, climate modellers, and social scientists under a single banner — the idea that research data is a shared language. For food security researchers working in Australia, the program offered a window into how global colleagues were handling the same pressures felt across the Murray-Darling Basin, the northern tropics, and the wheat belt stretching from South Australia into Western Australia. The sessions sketched the scaffolding of a data infrastructure that Australia could plug into without reinventing the wheel.

What made the program particularly useful for Australians was its willingness to address the entire chain of research data, from collection and curation through to analysis and reuse. Papers touched on satellite-derived yield estimates, household consumption surveys, and the messy realities of linking genomic, environmental, and socioeconomic datasets. Several sessions confronted the politics of data — who counts, who is counted, and whose knowledge shapes policy. These are questions that resonate strongly in a country where Aboriginal and Torres Strait Islander data sovereignty has become a live concern in Canberra and in state-level research ethics processes.

For anyone working on food security who could not attend in person, the archived program remains a working roadmap. The sessions provide templates for grant proposals, ideas for data partnerships, and a sober sense of the methodological pitfalls ahead. The themes below draw on that archive, surfacing the talks most likely to inform research practice in an Australian context.

Plenary conversations that set the tone

The opening plenaries argued that data work is not a back-office function but a frontline research practice. Speakers emphasised that decisions about variable definitions, missing-data handling, and provenance ripple through to the conclusions drawn about food systems. For Australian researchers accustomed to working with the Australian Bureau of Statistics' agricultural surveys, this framing was a useful push to treat metadata not as paperwork but as a research output in its own right.

A recurring thread was the tension between data abundance and data quality. Several plenaries pointed to projects drowning in sensor output, drone imagery, and satellite passes while still struggling to answer basic questions about smallholder productivity. This challenge is familiar to anyone working in northern Australia, where monsoon variability, feral animal pressure, and patchy telecommunications create enormous data gaps. The plenaries urged attendees to invest in field-level data infrastructure — a message that maps neatly onto calls from groups like CSIRO for stronger on-ground monitoring across the rangelands.

The plenaries also made space for the human side of data. Presenters from international agricultural research centres described long-term partnerships with farming communities, and how trust determines whether household-level survey data can be collected at all. This is a fair dinkum lesson for Australian researchers engaging with remote Aboriginal communities or with multigenerational farming families in the Mallee — relationships precede datasets, and both are needed to produce credible results.

Big data infrastructure for agricultural research

Several parallel sessions explored the platforms and standards that allow agricultural data to travel between institutions. Talks covered the use of DataCite for persistent identifiers, the role of repository aggregators, and the practical work of harmonising crop yield records from dozens of countries. For Australian researchers, this was a chance to see how their contributions to global datasets are being routed and what standards they need to meet to participate fully.

One session focused on the Global Open Data for Agriculture and Nutrition initiative, which has become a backbone for projects tracking food system shocks. The discussion highlighted how open licensing accelerates response work during droughts or trade disruptions — situations that Australian producers know all too well, whether it is the dry spells hitting southern Queensland or the export tensions affecting barley and wine producers in recent years. The takeaway for Australian institutions was clear: contributing to shared open infrastructure pays dividends when crisis hits.

Another strand of the infrastructure conversation addressed the back-end work of data cleaning and transformation. Presenters demonstrated workflows using R, Python, and open-source ETL tools, showing how raw administrative data from customs records or fertiliser sales can be reshaped into research-ready assets. This kind of applied pipeline work is exactly what university-based food security teams in Brisbane, Adelaide, and Perth have been building, and the IASSIST sessions gave them a chance to compare notes with international peers.

Deep learning applications in crop and land monitoring

Deep learning featured prominently in the Lawrence program, with sessions showing how convolutional neural networks and transformer architectures are being applied to satellite imagery, weed detection, and pest identification. Australian researchers at organisations such as the Australian Centre for Field Robotics have been pushing in similar directions, and the conference offered a useful benchmark for measuring progress against global work.

A standout session examined yield prediction models trained on multi-year Landsat and Sentinel archives. The presenter showed that models trained in one hemisphere transfer poorly to another unless carefully retrained with local phenology calendars. This is a critical insight for Australian researchers, where cropping calendars in the northern tropics are out of phase with those in temperate southern regions. The session also discussed the value of incorporating weather data from the Bureau of Meteorology alongside satellite inputs to improve model robustness.

Another deep learning session addressed livestock monitoring, using video and accelerometer data to track animal welfare indicators. While the talk drew primarily on European dairy systems, the methodology translates well to Australian extensive grazing operations across the outback. Researchers flagged the need for low-bandwidth variants of these systems, since many stations operate with limited connectivity and rely on satellite uplinks for data transfer.

Digital agriculture and the smallholder question

Digital agriculture sessions tackled the gap between precision farming tools designed for large commercial operations and the realities faced by smallholders in low- and middle-income countries. Speakers warned against simply transplanting high-tech solutions into contexts where smartphone penetration is patchy, electricity supply is unreliable, and extension services are stretched thin. For Australian researchers, this is a useful counterweight to the domestic conversation, which often centres on large broadacre and pastoral enterprises.

One paper examined mobile data collection platforms used by agricultural extension workers in sub-Saharan Africa, highlighting how lightweight apps can feed national crop forecasting systems. The approach has clear analogues in northern Australia, where Indigenous ranger groups and pastoral companies are experimenting with similar tools to monitor pasture condition, water points, and feral animal activity. The session emphasised co-design with end users — a principle that aligns with Australian best practice in community-based research.

Another session explored the economics of digital agriculture adoption, asking who benefits when precision tools are rolled out. Presenters found that benefits often accrue to larger operators first, widening inequality unless deliberate distributional policies are in place. This finding matters for Australian food security research because small-scale producers, including Indigenous-owned enterprises, are central to remote food supply chains. Ignoring distributional questions can mask important vulnerabilities in the system.

Data source or platform	Coverage and scale	Strengths for food security research	Key limitations
FAO FAOSTAT	Global, multi-decade	Standardised production, trade, and food balance sheets; widely cited in policy work	Variable reporting quality across countries; limited subnational detail
Global Open Data for Agriculture and Nutrition (GODAN)	Global, federated	Aggregates open datasets; useful for tracking transparency and access	Inconsistent licensing and metadata; curation effort falls on users
Australian Bureau of Statistics Agricultural Census	National, Australia	High-quality, granular data on farms, production, and land use; strong methodological documentation	Conducted infrequently; lags in fast-moving sectors
CSIRO Data Access Portal	National and international	Hosted datasets from climate, water, and agricultural research; supports Australian-led collaborations	Interface can be unfamiliar to international users; some datasets require negotiation
NASA Harvest and similar satellite-derived yield products	Global, near real-time	Frequent updates; supports monitoring in data-sparse regions	Model outputs depend on ground truth that may be thin in remote areas

Climate, water, and environmental data linkages

The Lawrence program devoted significant attention to climate data, reflecting the centrality of weather and water information to food systems. Sessions covered the use of reanalysis products, gridded precipitation datasets, and downscaled climate projections. Australian researchers have a deep tradition of working with the Bureau of Meteorology's datasets, and the IASSIST sessions offered a chance to align local work with international standards and methods.

A particularly lively session examined the challenges of linking household survey data with gridded climate data. The presenter showed how mismatches in spatial resolution, temporal frequency, and data licensing can undermine otherwise promising research designs. Solutions discussed included the use of anonymised village centroids, careful matching of survey and climate reference periods, and clear documentation of every linking decision. These are exactly the methodological questions that Australian food security researchers face when integrating farm survey data with remotely sensed land cover products.

Another strand of the climate conversation turned to the institutional work of maintaining long-term monitoring stations. Speakers from Latin America and Africa described how decades of investment in agrometeorological stations are now paying off, as long time series become available for trend analysis. The implicit message for Australia was that continued support for the Bureau's network, including the more remote stations, is a research asset worth defending — even as technology shifts towards satellite-based proxies. Researchers interested in thermal data standards will find useful parallels in the way international teams are formalising temperature monitoring protocols for agricultural applications.

Indigenous data, sovereignty, and Australian research contexts

A welcome feature of the IASSIST 2017 program was its attention to Indigenous data sovereignty. Sessions covered frameworks like CARE — Collective benefit, Authority to control, Responsibility, and Ethics — alongside the more familiar FAIR principles. Presenters argued that data about Indigenous communities and lands must be governed by those communities, not simply made open for academic convenience.

For Australian food security researchers, this conversation is directly relevant. Work on remote food systems, bush tucker enterprises, and Aboriginal community food security requires engagement with Indigenous data governance principles being developed by bodies such as the Australian Indigenous Governance Institute. The sessions offered concrete examples of data sharing agreements, governance committees, and benefit-sharing arrangements that could be adapted to Australian conditions.

One session highlighted the work of Māori researchers in Aotearoa New Zealand, showing how tikanga-informed data practices can produce richer and more ethical research outcomes. The parallels with Aboriginal and Torres Strait Islander research contexts in Australia are strong, and the IASSIST discussion gave Australian attendees a vocabulary and a set of precedents to draw on. Researchers returning home could use these examples to strengthen ethics proposals and to negotiate more meaningful partnerships with Indigenous organisations.

Reproducibility, documentation, and long-term curation

The closing sessions returned to questions of reproducibility and curation. Presenters argued that reproducibility is not just a statistical concern but a social one — it depends on shared norms, training, and institutional incentives. For food security researchers in Australia, where project funding cycles are short and academic positions are precarious, the discussion was a useful prompt to invest in documentation practices that survive beyond any single grant.

One session offered a practical workflow for reproducible agricultural data analysis, using containerisation tools like Docker and version-controlled code repositories. The presenter showed how the same workflow could be applied by a small research team in a regional university or by a large international consortium. Tools of this kind have become standard in computational research but are still unevenly adopted in applied agricultural work, where time pressures and software fragmentation remain serious obstacles.

Another session examined the afterlife of research data, asking what happens to datasets once a project ends. The discussion pointed to the growing network of domain repositories and to the importance of planning for data deposit at the start of a project, not at the end. Australian researchers heard about funding requirements from the Australian Research Council and the National Health and Medical Research Council that mandate data management plans, and the IASSIST sessions offered practical guidance on writing plans that genuinely support future reuse rather than ticking a compliance box.