Food Security Metrics at the 2017 IASSIST Conference

The archived IASSIST 2017 conference, “Data in the Middle: The Common Language of Research,” brought researchers to the University of Kansas in Lawrence from 23 to 26 May 2017. Its program linked plenary sessions with presentations on big data, deep learning, digital agriculture and global food security. Across these topics, food security metrics connected otherwise separate evidence: satellite imagery, household surveys, environmental records and transaction data could all describe whether people have reliable access to adequate food.

For an Australian audience, the archive is useful less as a collection of settled answers than as a record of how evidence was framed. Researchers had to choose indicators, units, time periods and visual forms that made complex conditions legible. The same choices become consequential in Australia, where drought, bushfires, remoteness, supermarket pricing and sharp regional differences shape both the available data and the meaning of food security.

A shared language for evidence

Food security is not a single condition that one percentage can capture. Its established dimensions are availability, access, utilisation and stability: enough food must exist, people must be able to obtain it, it must be used safely and nutritiously, and access must persist over time. Conference presentations made these dimensions visible by pairing each measure with the question it could answer.

The common-language idea also concerned form. A yield number, a price series and a map of household expenditure answer different questions, even when all are labelled food security. Effective presentations stated the denominator, geography and period, distinguished counts from rates, and exposed uncertainty. That discipline was especially important when big data or deep learning made complex estimates look deceptively exact.

Availability from production to supply

Availability was framed through production and supply: yields, harvested area, livestock numbers, stocks, imports, exports and post-harvest loss. Researchers also had to distinguish national totals from food actually reaching households. A tonne of grain may be exported, reserved as seed or stranded by transport costs, so production statistics became evidence only when paired with distribution and access.

Digital agriculture extended this evidence base. Satellite-derived crop maps, soil sensors, weather records and machine-learning estimates could reveal variation hidden in official averages. The conference’s deep learning strand highlighted how algorithms might support yield forecasts or detect risk. Useful presentations tested model accuracy alongside whether the model’s scale and error matched the policy decision it was intended to inform.

An Australian lens makes these distinctions immediate. Grain and irrigated production across the Murray–Darling Basin sit beside pastoral rangelands, coastal fisheries and food-import dependence in Sydney, Melbourne and other cities. Rainfall variability, drought and bushfires can shift output and prices quickly. An Australian output total may therefore rise while households in a particular regional town face a sharper shortfall.

Access through prices and distance

Access required metrics about purchasing power and proximity. Researchers compared food prices with wages or income, tracked the share of household spending devoted to food, mapped travel time to shops, and recorded coping strategies such as reducing meal size or skipping fresh items. Composite measures were useful for ranking areas, but only if their components and thresholds remained visible.

Conference presentations could turn these measures into maps, time series, dashboards and layered graphics. Spatial displays showed that a citywide median concealed different exposure in outer suburbs, while time plots exposed price spikes between annual surveys. Strong presentations paired each visual with data provenance, uncertainty and a local explanation; otherwise a polished chart could imply precision the underlying survey did not possess.

Australian shopping routines make affordability unusually concrete. Many households plan around a weekly supermarket shop, compare specials at Coles or Woolworths, combine fresh produce markets with takeaway, and use community gardens or food bank services where budgets are tight. In Melbourne and Sydney, delivery and convenience shops reshape access; in regional Victoria or New South Wales, distance, freight costs and a limited local inventory may matter more. For conference material intended for wider reuse, clear research writing guidance can help preserve those distinctions without flattening them into one national score.

Utilisation stability and uncertainty

Utilisation connected food to health and diet. Measures included dietary diversity, meeting nutrient requirements, child growth, illness, breast-feeding and access to safe water and sanitation. These indicators often came from health or nutrition surveys rather than agricultural records, which made the middle of the research process visible: data from clinics, households and governments had to be reconciled before a security claim could be made.

Stability was shown through variation through time rather than a single snapshot. Yield volatility, food-price swings, seasonal availability and exposure to drought, conflict or displacement were common signals. Forecasting models and deep learning approaches could extend early-warning possibilities, while historical comparisons revealed whether an apparent improvement would survive the next shock. Confidence intervals or scenario ranges were essential when the future formed part of the metric.

The presentations also exposed limits hidden by aggregation. Small samples can change a district estimate, national averages can erase distributional inequality, and satellite pixels do not show whether food is affordable. Missing data may follow the same roads and boundaries as the underlying inequality. Good metric design therefore documented error, provenance and ethical consequences rather than presenting a dashboard as neutral fact.

Dimension	Core question	Example metrics	Interpretation caution
Availability	Is enough food present?	Production, yields, stocks, imports and post-harvest loss	National totals can hide seasonal and regional gaps
Access	Can households obtain food?	Food prices, income, food expenditure, travel time and market distance	Spending shares vary with prices and household income
Utilisation	Is food used safely and nutritiously?	Dietary diversity, nutrient adequacy, growth, illness, water and sanitation	Survey delays and small samples weaken local estimates
Stability	Can access be maintained?	Price and yield volatility, seasonal supply and shock exposure	Period averages can conceal sharp shocks and unequal exposure
Equity	Who benefits and who is excluded?	Location, income, gender and disability gaps within the population	Composite scores can conceal the worst outcomes for particular groups

What the presentation format revealed

Conference format shaped what the metrics communicated. Plenary sessions supplied common concepts and political context, while parallel presentations allowed researchers to test detailed methods with peers. The archived schedule shows a deliberate movement between broad framing and specialist evidence on data-intensive agriculture, learning methods and global food systems. That structure encouraged participants to ask whether a measure was comparable before asking whether it was novel.

The same archive contains practical information on check-in, transportation and local recommendations. Those details may appear separate from research, yet they illustrate the principle behind the meeting: shared specifications become useful only when participants can interpret and apply them consistently. In slide terms, the equivalent was a clearly named source, denominator, period and limitation. The archive therefore preserves both the intellectual argument and the infrastructure required to negotiate it.

Australian metrics for practical decisions

Adapting this framework to Australia starts with aligning sources and definitions. The Australian Bureau of Statistics, Australian Bureau of Agricultural and Resource Economics and Research, Australian Institute of Health and Welfare, state agencies and global sources such as the Food and Agriculture Organization offer different levels of detail. Their series must be reconciled before food availability, affordability and nutrition can support one argument.

Regulation and market structure also shape which metrics matter. The Food Act 1988 (Cth) underpins food standards administered by Food Standards Australia New Zealand, while the Australian Consumer Law in Schedule 2 of the Competition and Consumer Act 2010 governs claims and unit pricing. Major supermarket chains sit alongside independent grocers and fresh produce markets, so branded price indexes, local market surveys and household expenditure data answer different questions. Australia’s food environment needs all three, with no single chain serving as a proxy for the whole market.

Measures that transfer well

• Food supply: Combine production, stock and trade measures so national abundance is connected to actual household availability.
• Affordability: Compare food-price changes with income and expenditure rather than presenting prices without purchasing power.
• Nutrition: Pair dietary diversity with health outcomes, water access and sanitation to show how food is used.
• Stability and equity: Track variation over time and disaggregate results by location and population group.

Checks before release

• Scope: Define the population, geographic boundary, unit of measure and reference period for every indicator.
• Uncertainty: Report confidence intervals, model assumptions and the effects of missing or revised data.
• Distribution: Disaggregate results so averages do not conceal food insecurity in smaller or more remote communities.
• Traceability: Preserve source files, processing decisions, model settings and links to the underlying datasets.

For Australian cities, the priority may be transport, convenience and housing costs. For remote communities and regional towns, reliable roads, stocked shops, affordable freight and functioning health services may dominate. First Nations communities may also require metrics shaped by local priorities and community-controlled data. A strong Australian metric set must let those differences remain visible instead of averaging them away.

Keeping metrics credible after the conference

Credibility beyond 2017 depends on reproducibility. A preserved presentation should identify its data release, processing steps, model settings, missing-data treatment and revision history. Archived slides can show how a result was communicated, but they cannot replace the underlying evidence. Citation of the IASSIST archive is therefore a starting point for inquiry rather than proof that a measure remains current.

The conference’s lasting contribution was to make measurement a shared research practice. Researchers working with imagery, models, clinical records and household data could place their results beside a common structure: what was measured, for whom, over what period, and with what uncertainty. When those elements are presented together, food security metrics become tools for comparison and scrutiny while still respecting the local realities that give each number meaning.