BY Michelle Howie 25 October 2021
Our GovHack 2021 challege explored how AgTech can improve sustainable agribusiness.
The potential of Aussie AgTech
Australia is well-known for its world-leading agriculture exports. Perhaps lesser known, however, is the potential of Australian agtech (agriculture technology ) to bring more sustainable, scalable and relevant food and resources to the world.
As an export, AgTech is scalable and less dependant upon population size or geographical location. It's more sustainable and economic to ship software around than fresh produce or livestock. This puts us Aussies at a huge advantage.
Plus, with the increased prevalence of droughts, natural disasters, harmful pests and climate change, it's never been more important to innovate for sustainable and resilient farming practices. Environmentally-friendly techniques and technology are key to meeting the food demand and offering the right nutrition to consumers.
That’s why Telstra and Microsoft came together at GovHack 2021 to challenge the hackathon participants to think of ways that IoT insights might improve regional agribusiness at scale. See the full challenge breakdown here on GovHack.
The Telstra and Microsoft GovHack IoT challenge
How might we harness open data and IoT insights in near-real-time to enable local agribusiness and farms in our regions to be more productive and sustainable at scale? How can we share private IoT data and regional open data for an overall more productive and sustainable agribusiness?
To set the 2021 GovHack-ers up for this challenge, Michelle from Telstra and Valeria from Microsoft shared a pre-hack workshop introduction to cellular IoT networks and visualisation tools with Azure. Participants learned about how thesev tools can be used to unlock all the useful data that you can pull from your IoT devices. Catch the replay here.
And here are some of our favourite entries from the GovHack 2021 participants -- you can read more about each entry below. You'll notice that most of our entrants focused upon rural farming, which highlights the need for digital inclusion in our vulnerable regional communities. In these areas, coverage across the whole property can be a challenge, which is why the Narrow Band Internet of Things (NB-IoT) network comes in handy. Telstra’s NB-IoT network has been globally recognised for its coverage capability of up to 120km from the nearest mobile tower, covering ~ 4 million square kilometres in Australia.
Cattle fertility predictions
This entry used Artificial Intelligence (AI) to scan video and images of cattle to predict whether or not they were “on heat”, before sending the farmer an SMS with the resulting data.
The system uses a trained model over previous images to recognise the patterns inside the new query image for classification. The cameras are solar powered and connected over Telstra’s NB-IoT network. Telstra Messaging API could be used to send the SMS.
We thought this entry had the potential to seriously help farmers manage the difficult dependencies of cattle fertility. Over the past few years, there's been a lot of research in this area and the use of IoT sensors to infer when and which cows are on heat has become increasingly popular. This solution took the next step by leveraging AI and ML (Machine Learning) so human monitoring is no longer required. This speeds up the process, reduces the risk of human error, and allows farms to utilise labour in other areas.
I think we’ve seen something similar at a previous Telstra Innovation Hackathon, which perhaps means there's still a gap in the market. Going forward, we’d love to see how the video analytics can be paired with other IoT sensors like temperature.
Crop predictions and recommendations
Get more of a yield, from less land, water, energy and pesticides. This solution combined open source weather data, manual sensor input from things like crop yield and site management, as well as static and dynamic data from sensors on site. They proposed using AI to predict crop yield and recommend certain actions for the farm hands to get greater yield with optimised resource allocation.
The Telstra and Microsoft judges liked how this entry demonstrated how such technology would work on a single farm, but could also be expanded to a collection of farms within a local region. This really addressed our challenge brief of sustainable agri-tech at scale.
Digital regenerative farming
This team took the challenge of more sustainable farming into the practice of farming itself. When addressing the impacts of climate change, two key strategies are prevention and mitigation. Carbon capture or sequestration is when CO2 from the atmosphere, usually a gas, can be trapped in solid form as Carbon. This is a form of mitigation, to reduce existing pollution.
We want to see more of this in agriculture, but it’s traditionally hard to actually measure and reward. This team has compared the limitations of existing carbon sensors, and provided a new alternative to measure and reward carbon capture. They proposed partnering with established Aussie AgTech robotics company: Agerris, who was featured on the AgTech session of our monthly AusIoT meetup. The live data collected from the roaming robots would be sent via Telstra’s NB-IoT network to an model in the Azure IoT cloud. If you want to test out this network yourself, you can grab a trial SIM and an Arduino MKR NB 1500 -- I’ve done some tutorials to show you how to get started with Arduino for IoT on TelstraDev.
This project shows a practical solution to support the carbon reduction incentive program, which will be rolled out by an increasing number of countries in the next 5-10 years. They’ve considered public and private partnerships that can make it happen, including using UN open data, and who the end users of that data could be, e.g carbon credit aggregators.
To take it to the next level, we’d love to see how this data can be fed into the Telstra Data Hub. If more institutes have secure access to the carbon capture data, this could be a powerful solution for climate change mitigation.
Environmental health tracker
Combining crowd sourced data on frog sightings from the New South Wales community – along with other open source data and well placed IoT sensors – Frogly is an interactive map that tracks the environmental health in key areas of the city. It infers that an ecosystem that attracts wildlife is a health one. The team hope to use this data to support the development of green spaces.
Many citizens already use local data from environmental sensors to keep track of their local indicators, but this was an interesting extrapolation to citywide IoT by engaging the community in the data collection and creating civic ownership.
HydroSight water usage tracking
This team offered a very relevant solution to tracking an important asset: water. Developed in consultation with farmers, they tackled a problem that Telstra and Microsoft have heard time and time again from public, private and community groups. Here we see a solution that asks farmers to input their water usage. This info is then paird with some existing data sets and the insights derived consider both environmental and regulatory factors, in an easy-to-consume app.
We’d love to see how this can be applied to the Murray-Darling River at scale, to increase the accountability and awareness of up and down stream impacts on water usage in the basin.
Internet Of Ag
Moreton Bay is already very hooked into the Internet of Things, so this group paired the existing sensors with open weather data in an app that tells local farmers what they need to know, from bushfire to flood warnings. This solution also does some predictive modelling on the impact of crop yield and revenue for both domestic and exportation.
This was one of the most mature and achievable solutions we saw over the hackathon weekend, demonstrating not only value to the end user but also the ecosystem of the region more broadly.
SuperCattle vegetation measurements
The Lone Alpaca single-handedly put together this solution to measure the suitability of vegetation in an area for grazing, using computer vision for farmers to work out where to move their cattle next.
They used images from Google to train the model, as well open data on financial performance of livestock farms to get cattle station size and location, which was crucial for defining the project.
This team also proposed an array of sensors, attached to an Arduino board in the field, that use computer vision to evaluate the quality of the grass. Is it good, feedable grass? Too dry? Or just good enough to eat? They don’t just show you image by image, but map that to a readable colour coded map.
We loved that they managed to train a working model over the weekend.