Visualising Agricultural Data: Techniques Shown at IASSIST 2017

The IASSIST 2017 conference, hosted at the University of Kansas in Lawrence from May 23 to 26, gathered data specialists, librarians, and applied researchers around a single theme: turning raw numbers into shared understanding. Sessions on digital agriculture and global food security drew attention because they showed how charts, maps, and dashboards can change the way growers, policy advisors, and scientists talk to one another. The techniques on display in Kansas remain relevant for anyone working with farm and landscape data, including researchers in Australia, where drought cycles, mixed cropping systems, and export-oriented supply chains create a strong appetite for clearer visual evidence.

For Australian audiences, the value of these methods stretches well beyond the lecture hall. Producers in the Murray-Darling Basin, viticulturists along the Barossa Valley, and broadacre grain operations across the Western Australian wheatbelt all generate streams of records that are hard to compare without thoughtful design. Visualisation offers a path to combine paddock-level sensor readings, satellite imagery, and market information in ways that respect the scale of the landscape and the everyday decision cycles of the people who work it. What follows draws on the visualisation work presented at IASSIST 2017 and considers how those practices translate into Australian contexts.

Why visualising agricultural data needs more than a spreadsheet

A common thread in the conference programme was that agricultural datasets rarely arrive tidy. Soil moisture probes, weather stations, yield monitors on harvesters, and remote sensing platforms each produce records in their own format and cadence. When researchers join these streams, the resulting tables grow wider than any human eye can scan. The presenters argued that the choice of visual representation is not decorative: it determines which patterns a viewer can perceive. A line chart may hide a localised spike, while a choropleth map may mask a time-bound event. Choosing a chart is also choosing a question.

For Australian research teams this matters because mixed farming systems layer several enterprises onto a single property. A southern New South Wales operation, for example, may carry sheep, winter cereal, and a pulse crop across the same paddock in a single year, generating three overlapping yield records. The Kansas sessions emphasised small multiples, a grid of mini-charts that share scales, as a way to give each enterprise its own panel without losing the shared time axis. The approach was cited in Lawrence as a practical remedy for visual overload and was paired with advice to limit each chart to a single decision point.

Mapping the landscape: geospatial views from the conference

Several IASSIST 2017 contributions focused on geospatial visualisation, demonstrating how satellite indices such as NDVI can be layered with administrative boundaries and farm-level plots to support yield forecasting. Participants walked through animated time series of vegetation greenness that show how regions respond to rainfall pulses across seasons. The animation technique helps users spot lag effects between weather and crop performance that static maps tend to bury.

Australian researchers face similar problems at a national scale. ABARES publishes outlook reports that pair charts with maps showing shifts in winter crop production across the eastern seaboard. The Lawrence presentations offered concrete tips for pairing raster data with vector overlays, using semi-transparent fills so parcel boundaries remain visible without overpowering the underlying signal. Presenters also recommended labelling the colour ramp directly on the legend, since viewers often misread greenness gradients when the scale is implicit. These craft choices matter when the audience includes growers reading the map on a phone in the paddock, often with limited connectivity.

Time series and yield monitoring

Time-based visualisation was another recurring topic in the conference programme. Speakers highlighted the value of running mean charts, anomaly plots, and stacked area charts for monitoring yield against long-term expectations. An anomaly plot shows how a value differs from a reference average, making it easier to spot when a season is performing below or above the historical norm. Stacked area charts place multiple series on top of one another to show total and composition, and were discussed as a way to convey the changing mix of crops across regions. The conference also featured a session on annotating time series with weather events, irrigation turns, and sowing dates, so the chart becomes a conversation between data and the story behind it.

In Australia, the Grains Research and Development Corporation funds long-term monitoring sites across the southern and western cropping zones, generating the kind of multi-decadal records that lend themselves to these methods. Conference attendees from Australian institutions noted that a clear running mean over twenty seasons is often more useful to a grower than a single year on its own. The Kansas discussions emphasised building these charts with open tooling so collaborators can fork the visualisation and adjust thresholds without writing new code, a practice that suits research consortia spread across universities, state agencies, and grower groups.

Dashboards for climate, soil, and crop interaction

Dashboards were a central theme of the digital agriculture track at IASSIST 2017. Multiple presenters argued that an effective dashboard should reduce, not amplify, complexity. The speakers laid out a few design rules: keep the top-left tile focused on the headline indicator, use consistent colour scales across tiles, and provide a clear path from an overview chart to the underlying record. A dashboard, in this framing, is a guided tour of the dataset rather than a wall of independent widgets. Speakers warned against clutter, suggesting that any tile that does not change a decision should be cut.

For Australian audiences, this guidance aligns with the work of CSIRO's agriculture and food unit and with state-level decision support systems used during dry seasons. The Bureau of Meteorology's seasonal outlooks are increasingly paired with farm-level dashboards in pilot programmes across Queensland and northern New South Wales. The conference's design principles help explain why some pilots have struggled: when too many indicators share one screen, growers revert to the single number they trust. The Kansas advice was to design for one decision at a time, then layer additional views behind a deliberate click.

Open data infrastructure and the FAIR principles on display

A thread running through several plenaries was the importance of FAIR principles, an acronym standing for Findable, Accessible, Interoperable, and Reusable data, applied to agricultural research outputs. Visualisation was framed as part of the reuse step, since a chart without its data, metadata, and provenance quickly loses value. Presenters demonstrated repositories where every published figure links back to a dataset, a script, and a license, so downstream researchers can rerun the visualisation with new inputs. This provenance tracking matters for international collaborations spanning institutions and countries, including projects that bring together Australian partners with North American and European counterparts.

Australian researchers operate within a specific legislative backdrop. The Privacy Act 1988 governs personal information handling, while state-level regulations cover water use, land tenure, and livestock identification through the National Livestock Identification System. When designing visualisations, Australian teams must balance openness with these obligations, particularly when farm identifiers, water allocations, or individual enterprise data are involved. The IASSIST 2017 sessions offered practical workflows for documenting consent, aggregating to safe levels, and stripping direct identifiers. For Commonwealth-funded research, these practices align with the Australian Research Data Commons and its guidance on sensitive data.

Practical visualisation choices for Australian researchers

Conference participants left Lawrence with a refined sense of which chart suits which agricultural question. The following list captures the choices most often recommended:

• Choropleth maps for regional yield, rainfall, or disease prevalence across shires, local government areas, or natural resource management regions.
• Small multiples for comparing paddock, variety, or livestock class across the same time window.
• Anomaly plots for showing seasonal performance against a 10- or 20-year reference period.
• Stacked area charts for changes in land use or enterprise mix across regions and years.
• Scatter plots with marginal histograms for pairing soil test values with yield response curves.
• Interactive dashboards, built with reusable components, for ongoing decision support during sowing, in-crop monitoring, and harvest.

These choices are not mutually exclusive. The conference's recurring advice was to start with the question rather than the tool. A grower wondering whether to topdress nitrogen in mid-season needs a different chart than a policy analyst modelling five-year land use change, and both differ from the chart a community group needs to argue for better rural connectivity on the National Broadband Network. For Australian extension officers, the practical translation often happens at field days, where attendees may view a chart on a laptop, a printed handout, or a phone. Presenters recommended designing for the lowest common denominator first, then layering in interactivity, a sequence suited to Australian conditions where some stations still rely on intermittent mobile coverage.

Tools and formats compared for farm and field reporting

The closing sessions of IASSIST 2017 included hands-on demonstrations of open source visualisation stacks alongside commercial platforms. Presenters compared static outputs with interactive web builds, weighed PDF reports against HTML pages, and discussed the long-term sustainability of each approach. The table below summarises the trade-offs that resonated most strongly with Australian practitioners, given the country's uneven connectivity and mixed digital literacy of end users.

Approach	Strengths for agricultural reporting	Weaknesses for Australian contexts	Best fit
Static PDF report with embedded charts	Reliable on low-bandwidth connections; easy to print at field days	Charts cannot be filtered or updated without a new file	Seasonal outlook summaries, end-of-year grower reports
Interactive web dashboard	Allows filtering by region, crop, or enterprise; links back to source data	Requires reliable internet, which is patchy in some rural areas	In-season decision support, multi-stakeholder monitoring
Open data notebook (Jupyter with R or Python)	Fully reproducible; supports FAIR principles and citation of code	Steeper learning curve for non-technical readers	Research publications, postgraduate teaching, ABARES-style analysis
Story map with embedded charts	Strong narrative flow; good for community engagement	Can become long and hard to maintain	Catchment group reports, public consultation on land use plans
Spreadsheet with pivot charts	Familiar to many growers and agronomists	Limited handling of geospatial layers and time series	Small enterprise record keeping, on-farm trials

The table is a starting point, not a verdict. The conference's deeper message was that any visualisation tool should be chosen for the audience and the decision at hand, with maintenance cost weighed alongside visual appeal.

Beyond tool selection, presenters also recommended a short checklist for teams preparing to share agricultural visualisations more widely:

• Document the data source, vintage, and any transformations applied before the chart was drawn.
• Choose a colour scale that remains readable in greyscale printing, in case field-day handouts are photocopied.
• Provide a textual summary of the headline finding so screen readers and low-bandwidth users still receive the message.
• Label units clearly, including whether hectares are national, regional, or paddock-scale.
• Test the visualisation with at least one non-specialist viewer before publishing, and revise where confusion appears.
• Include a license statement that matches the underlying data license, in line with Australian Research Data Commons guidance.

Taken together, the visualisation work presented at IASSIST 2017 points to a craft that combines careful data preparation, considered design, and respect for the audience. For Australian researchers, the conference offered both a vocabulary and tested patterns that travel well across cropping zones, livestock regions, and the mixed enterprises that define much of the continent's agricultural output. Applying these patterns with attention to local realities, from the connectivity constraints of remote stations to the legal frameworks that shape data sharing, can help the sector move from raw numbers to clearer decisions.