Client: Silver Fern Farms
Silver Fern Farms are New Zealand’s largest red meat producer with exports of grass-fed products to over 60 countries. Earlier this year Silver Fern Farms launched ‘Net Carbon Zero’ which is New Zealand's first certified grass-fed, carbon zero red meat product, where 100% of end-to-end emissions have been removed within the farms where the animals were raised.
Silver Fern Farms’ approach is unique as it utilizes “insetting” which means the carbon credits needed to account for the emissions of the product are found by working with farmers to optimize the role that the farms can play in acting as carbon sinks, rather than having to rely on purchasing carbon credits from other projects.
Image: Silver Fern Farms.
“Taking care of our emissions is our own responsibility, no-one else’s,” says Silver Fern Farms CEO Simon Limmer. “We are not outsourcing our emissions, rather we are recognising and incentivizing our farmers for their efforts to create farm environments that are better able to capture carbon, increase biodiversity and support nature positive food production.”
Recognizing that 96% of beef emissions occur on farm, Silver Fern Farms commenced a program with Lynker Analytics to map and measure the sequestration potential from the many types of vegetation present on their farms.
Working closely with carbon forestry specialists Carbon Forest Services and New Zealand’s most trusted environmental verification body, Toitū Envirocare, Lynker Analytics developed a data processing pipeline using satellite imagery, aerial photography and Artificial Intelligence (AI) to map permanent woody vegetation on farms.
Matt Lythe from Lynker Analytics explains “The system we have developed uses satellite imagery with a high temporal return time combined with high resolution aerial photography in conjunction with Machine Learning to map, classify and age vegetation at a sub-hectare scale. A human supervised data harmonisation method is then used to integrate and blend the machine generated data and create carbon stock assessments.”
The AI based system maps vegetation units larger than 1000 square metres in area, enabling a calculation of each individual farm’s ability to sequester carbon.
The on-farm vegetation includes woodlot forests, shelter belts, regenerating native bush, summer shade and winter animal shelter, erosion, and riparian planting.
Vegetation map produced using the AI System. Each polygon area represents an eligible and permanent carbon sink classified by plant community.
Non-eligible vegetation including vegetation registered with MPI for the government emissions trading scheme (ETS) and vegetation areas less than 0.1 hectares are removed.
Each stand of vegetation is assigned an annual carbon stock change in Tonnes CO₂ based on type, hectares, and age. The system also allows for farmer input to confirm or amend the results and opt to include or exclude each stand of vegetation from the programme.
Carbon stock assessment discriminating vegetation group and age
“The commitment by our farmers to whole-of-farm accountability runs deep,” says Simon Limmer.
“In the last couple of years many farmers have been tracking and managing emissions on their farms. Through our Net Carbon Zero programme we are connecting our hardworking farmers with customers who want to support them to plant, restore, and regenerate vegetation to increase the amount of carbon their farms can naturally remove from the atmosphere.”
This work extends previous AI based vegetation classification work Lynker Analytics has completed in New Zealand and also internationally. We have adapted and tuned existing neural networks to this multi-temporal image data and multi-class output requirements.
The harmonisation process involves several processing stages and data checks and is designed to ensure all vegetation stands are captured, classified correctly and meet the specification for carbon removals. The system will also support future government requirements for greenhouse gas sequestration on New Zealand farms.