Project Funders: National Landcare Project Smart Farming Partnerships Round 2 (Commonwealth Department of Agriculture and Water Resources); Australian National University; FarmLink
Project Partners: FarmLink, NSW Department of Industry, Australian National University, Holbrook Landcare Group, Central West Farming Systems
Project Code: 4-CS70YJ8
Project Duration: 2019-2023
Project Location: Southern and Central NSW
Soil acidity and declining Soil Organic Carbon affects over half of agricultural soils in southern and central NSW and threatens the viability and resilience of farming systems. Current acid soil management practices are based on outdated models that are failing to prevent the widespread development of subsurface acidity in contemporary cropping and pasture systems. This urgently needed project brings together an expert team from three farming system groups (grower engagement/extension), the NSW DPI (agronomy/soil science) and ANU (computer modelling/visualisation) to develop a new, accurate acidification model using innovative machine learning methods. Model outputs will be used to develop a free online decision support system and API protocols for integration with existing digital platforms. These tools will provide updated liming recommendations and interactive scenario forecasting which will ultimately result in more sustainable soil management and productive farming enterprises.
There are five key components to this project -
- Data collation and curation: Numerous datasets with volumes of existing historic soil and contextual data (e.g. land use, management, climate) collected for research projects (e.g. replicated field trials) and agronomic purposes (e.g. grower/advisor data) will be collated and consolidated into a standardised database. The majority of this work will be performed in the initial 12 months by a project officer to be appointed within the NSW DPI. Each Farming System Group (FSG) will also perform internal data mining and collate privately collected data contributed from grower members.
- Additional data collection: Resampling of high quality, geolocated historic sites will occur throughout the duration of the project. These sites will be strategically selected to fill key data gaps as they are identified. Resampling will be performed by both the NSW DPI and FSGs.
- Acidification modelling: The ANU Research School of Computer Science (ANU-RSCS), working with the NSW DPI will use machine learning techniques to develop predictive analytics that model changes to soil pH based on a variety of inputs including soil chemistry (pH, pH buffering, SOC) sampled in segmented intervals, as well as management factors including lime rate, cultivation (incl. depth) and crop/pasture history. A number of stages will occur including model conceptualisation, training, validation and testing.
- Decision support systems (DSS): ML outputs will be used to develop updated liming recommendations by the NSW DPI and ANU which will be delivered to growers through either (1) a free online DSS, (2) existing farm data management platforms via the open provision of application programming interface (API) protocols and (3) soft and hard copy charts. These outputs will be built by the ANU, with support from the NSW DPI and guidance from FSG’s and land managers via annual grower/advisor workshops. The online DSS system will enable end-users to input paddock locations (auto-populating climate information), target pH, soil parameters (pH, CEC, SOC) and sample depths, as well as information about the farming system/rotation and management actions (e.g. cultivation) to receive customised lime and management recommendations. In addition, the digital tool will also allow end-users to forecast and visualise future soil condition (e.g. pH, Aluminium, SOC) through the soil profile and explore the economics of adjusting inputs and management actions.
- Focus farms and extension activities: Grower engagement, extension and training will be spearheaded by the three FSGs; FL, CWFS and HLN. Five on-farm replicated farmer engagement trial sites will be established by the FSGs in year 1, which will be assessed for three growing seasons (2020/21/22). Contrasting liming rates and incorporation combinations will be assessed, with one site also examining the impact of lime ± organic matter. Trial sites will be used as two-way learning resources, both for parameterising and validating the model and as key extension/grower engagement sites. Trials will be co-designed by local grower reference groups with scientific oversight from NSW DPI. Land manager/advisor workshops will be held throughout the project (one per FSG annually) to gather feedback from end-users, trial and refine projects outputs.