Byte by byte: how AI is reshaping agriculture

By 
Smriti Daniel

28 October 2025
8 min learn

The Kakhovka Dam in Ukraine as soon as stood 30 metres excessive and stretched 3.2 kilometres throughout the Dnieper River, creating one of many nation’s largest reservoirs. Holding some 18 cubic kilometres of water, it provided irrigation throughout Kherson Oblast and past, and cooled Europe’s largest nuclear plant at Zaporizhzhia.

When the dam collapsed in June 2023, its loss unleashed catastrophic floods downstream and drained the huge reservoir upstream. Fields had been left parched, canals emptied, ecosystems shattered. Toxic sediments flowed into the Black Sea, algal blooms unfold, and what had been a reservoir, started remodeling right into a uncooked floodplain forest, sprouting willows and poplars over unstable floor.

Amid the upheaval, Dr Sarah Hartman, then a PhD candidate on the University of California, Berkeley, set out along with her group to reply a urgent query.

“We wanted to understand how these regions were responding to disruption. Could we track agricultural resilience, even in crisis, by using AI and satellite imagery?”

With entry on the bottom unsafe or not possible, Sarah and her group turned to satellites and AI. Their mannequin used publicly accessible Landsat imagery through Google Earth Engine and personal knowledge from Planet Labs to observe crop well being earlier than and after the dam’s destruction. Using unsupervised machine studying – a technique that identifies patterns in knowledge with out pre-labelling – they detected shifts in vegetation throughout the area. A choice-tree mannequin then sorted these patterns into actively managed farmland and areas displaying indicators of abandonment or regrowth.

The Ukraine analysis was undertaken as a part of Sarah’s PhD, with revisions to the revealed paper accomplished after she joined CSIRO as an Early Research Career (CERC) Postdoctoral Fellow. What emerged, she mentioned, was “a mosaic of loss and persistence.”

“We were able to identify zones of collapse – but also zones of recovery,” she mentioned.

This potential to see what’s hidden on the bottom, to interpret disruption from afar, is on the coronary heart of AI’s promise in agriculture.

What AI can reveal

Through satellite tv for pc imagery, AI can spotlight the place irrigation methods are drying up, spot greenhouses that weren’t there a decade in the past, or flag fields all of the sudden off-course of their development. In locations the place knowledge is lacking or misplaced, AI can reclaim information, generally going again many years. With one caveat – Sarah stresses that AI’s energy is just pretty much as good as its databank.

“AI is what it eats,” she mentioned. “If we feed it sparse, outdated or biased data, we get misinformed models. But when we train models with rich, local and timely information, we grow AI that’s relevant to the real and pressing needs of agriculture.”

This precept is what she and colleagues name “responsible, application-driven AI” – RAD-AI for brief. The strategy is straightforward however important: AI must be designed with and for the individuals who will use it, grounded in context, and fuelled by the fitting knowledge.

Building belief in a sector underneath stress

“It’s RAD because it reminds us that responsible, application-driven AI should be grounded in context and designed for the real world,” she defined.

At its coronary heart is co-design: working alongside the individuals who will use the know-how – agronomists, growers, agtech firms – to make sure AI displays actual wants and builds belief.

“Building context-aware capabilities responsibly from the start, with a focus on the communities that will be served, is essential,” Sarah mentioned.

Trust, she mentioned, can be about range. Incorporating Indigenous information methods, native knowledge sovereignty and participatory design helps guarantee AI displays Australia’s distinctive farming methods.

“By weaving different agricultural paradigms into development, we can build tools that are not only technically powerful but also socially equitable.”

This is occurring at a second when Australia’s agricultural sector is underneath stress from each route – local weather change, labour shortages, productiveness challenges and risky world markets. But with out focused funding and cautious design, agriculture dangers lacking out on the advantages of synthetic intelligence (AI) at exactly the second it’s most wanted.

“We have an incredible opportunity to harness AI but we need to ensure it is done responsibly and in ways that reflect the realities of Australian farming,” cautioned Sarah.

As innovation gathers tempo, Sarah believes accountability can’t be an afterthought. It should be constructed into the method from the start – not solely to keep away from dangers of bias, misuse or exclusion, however to unlock the total potential of AI as a companion in farming.

The instruments already rising

We’re already seeing thrilling initiatives at work within the area. On distant cattle stations, drones geared up with AI-driven algorithms are being trialled to automate mustering. Developed by the Australian firm SkyKelpie, these methods predict livestock behaviour and optimise flight paths, chopping prices and environmental impacts in contrast with helicopters and quad bikes.

At the identical time, robotics firms comparable to Swarmfarm are growing absolutely autonomous tractors that combine a number of AI applied sciences. These machines maintain potential to remodel area operations, although technical and scaling challenges stay.

AI can be reshaping how farmers take care of the climate. Breakthroughs comparable to Google’s GraphCast and Microsoft’s Aurora are delivering forecasts with unprecedented pace and precision, supporting decision-making from planting to drought administration. And generative AI is starting to enter agriculture, from Microsoft’s AgPilot chatbot to rising platforms that may combine farm-specific knowledge into tailor-made suggestions.

Each of those advances factors to AI’s versatility – from livestock to logistics, forecasts to fieldwork. But in addition they reveal vulnerabilities. Generative AI – methods educated to create new content material comparable to textual content, photographs or code – are liable to “hallucinations”: producing assured however incorrect solutions. Misapplied in agriculture, such errors may result in expensive errors.

Preparing for the longer term

Sarah’s “Trusted AI Agronomist” venture exhibits how AI could make advanced crop science usable on farms. A feed-forward neural community is educated to copy CSIRO’s Agricultural Production Systems Simulator, producing crop development forecasts underneath totally different situations. Crucially, it additionally supplies uncertainty bands – giving farmers not only a prediction, however a measure of confidence to information risk-based choices.

“Most farm decisions are made under risk, not certainty,” she mentioned. “Knowing the confidence around a prediction, and not just the average, allows farmers with different tolerances for risk to choose the action that makes the most sense.”

Seeing not only a single consequence however a spread of doable ones builds each belief and resilience into the decision-making course of.

Meanwhile, as a part of CSIRO’s Ag2050 initiative, Sarah has been examining the three AI breakthroughs with the greatest potential to transform farming: knowledge integration and analytics, climate and local weather forecasting, and generative AI chatbots.

These will not be summary ideas however tangible applied sciences already in movement, pointing in direction of the sort of agricultural methods Australia may construct by mid-century.

“Farming is complex, and AI can revolutionise how decisions are made,” mentioned Dr Rose Roche, CSIRO’s Ag2050 lead.

“To keep farm productivity growing in the face of climate change, shifting markets and resource pressures, our Ag2050 work is showing how AI can help farmers adapt and stay resilient into the future,” Rose mentioned.

Frank Sperling co-edited Ag2050’s special issue of the Farm Policy Journal, a landmark version launched by the Australian Farm Institute in collaboration with CSIRO. He mentioned it’s important to discover numerous futures for Australian agriculture.

“With Australian agriculture having to navigate an increasingly uncertain future, where global and local risks intertwine, we need to consider multiple plausible scenarios and explore what can be done to harness opportunities, while minimising adverse outcomes,” Frank mentioned.

“This is why this work by Sarah and colleagues on responsible AI is so important,” he added.

Choosing what sort of AI takes root

For Sarah, the lesson from Ukraine nonetheless resonates. Out of devastation, AI helped reveal each collapse and persistence – and now, comparable applied sciences are being developed and trialled in Australia. From drone-assisted cattle mustering and superior climate forecasting to generative AI instruments like Microsoft’s AgPilot, AI is steadily transferring from analysis to real-world software throughout the agricultural sector.

The similar can be true in Australia: the applied sciences we select to construct – and the best way we select to construct them – will form whether or not AI turns into a supply of belief and resilience, or a reason behind additional threat.

It’s an perception that underpins Sarah’s current recognition as winner of the Women in AI Asia-Pacificaward for AI in Agribusiness and Rural Development.

“This award is a reflection of my drive to deeply engage when making (high and low) technology solutions and ensuring they work for people,” she mentioned.

But Sarah is fast so as to add a observe of warning: “AI is not a silver bullet. But when responsibly designed and rooted in good data, it can become a powerful tool for transforming agriculture.”