Ken Moore is the owner of Gilmay Farm an intensive grazing property in Northcliffe. Ken is an experienced senior research manager having worked for the national R&D organisations Land & Water Australia and the Rural Industries Research and Development Corporation (now Agrifutures Australia) over many years. In this capacity he managed major national research programs and hundreds of research projects. Recently, he has committed to research in soil carbon sequestration on Australian farms particularly based on satellite imaging for predicting soil organic carbon. During 2020 he hosted an international webinar series involving Australian and Chinese Academy of Agricultural Sciences’ experts in soil organic carbon measurement on behalf of the Australian not for profit organization Soils for Life. The webinars were funded under the Australia China Agricultural Cooperation Agreement and have led to negotiations on a collaborative research project on soil carbon measurement, mapping and monitoring using satellite imagery and artificial intelligence.
Project Overview: Predicting soil organic carbon in Western Australian agricultural soils using satellite imagery and machine learning (artificial intelligence).
Western Australia’s agricultural soils generally have low levels of soil organic matter (SOM) and hence the soil organic carbon (SOC) component of SOM. Over time agricultural practices have reduced SOM and SOC stores in our agricultural soils. Increasing SOC can improve soil health and production yields. It can also help to reduce greenhouse gas emissions from farming through sequestration of CO2. Subject to meeting Australian regulatory requirements increasing SOC may be a secondary farm income source with the sale of carbon credits.
SOC can be measured relatively easily, accurately and cheaply at farm paddock level through soil sampling and laboratory analysis of the samples. However, for broadacre farms and agricultural regions the costs of sampling and analysis become prohibitive and results based on coarse-level soil data over large areas are less accurate.
Major advances have been made in SOC measurement using remote sensing and near infrared spectroscopy via satellites. Satellite imagery predictions of SOC centre around the relationship of SOC with the Normalized Difference Vegetation Index (NDVI). NDVI is influenced by the fractional ground cover, vegetation density and vegetation greenness. NDVI is limited by the fact that vegetation particularly pasture is not green all the time. Many Australian pasture systems are based on various perennial species in different growth stages and variable (or no) greenness at any one time.
This project is for the students to trial and conduct field work on case study farms in the WA Central Wheatbelt and south west regions using a newly developed methodology of Australian agricultural data analytics company, Cibo Labs Pty Ltd which has agreed to participate and strongly supports the IPREP WA initiative. The scope of the students’ work is to: consolidate information on pasture mass and SOC across different soil types in the study areas; validate the predictive algorithms used in the methodology; calibrate the satellite and machine learning results against on-ground sampling results.
This new methodology uses very-fine resolution predictions of pasture (or crop) biomass via Sentinel multi-spectral satellite imagery and machine learning to assess the relationship of pasture production with soil types and the soil carbon that may be accumulating under those pastures. Carbon estimating areas (CEAs) are generated and split into strata based on soil types associated with common pasture production. Randomized on-ground soil sampling then takes place in the strata to determine actual soil carbon. Cibo Labs provide farmers with a smartphone app (MyFarmKey) to undertake directed in-paddock sampling. This on-ground data is used to validate and improve the accuracy of predictions using satellite imagery and machine learning.
The key outcome of the project is to demonstrate the use and accuracy of a new and potentially world leading satellite and machine learning methodology for quantifying pasture/crop biomass, vegetation coverage and SOC levels over large areas of varying soil types and climatic conditions. If successfully applied on the wheatbelt and south west case study farms, the methodology can then be applied across other Western Australian farms and agricultural regions.
Desired Disciplines and Skill Sets
Remote sensing using high resolution multi-spectral satellite imaging
Machine learning (artificial intelligence) and big data knowledge to understand the predictive algorithms for pasture biomass and soil organic carbon levels
Soil science with specialisation in soil organic matter and soil carbon measurement and management methods to increase SOC
Project Facilitation Preference
Mixture of online and face-to-face engagement