Monitoring agricultural patterns: new insights from big data

Evaluating new and existing ground, aerial and satellite remote sensing data streams for agricultural monitoring; and creating AI tools for exploring patterns in potentially unstructured data 

The amount of data useable in agricultural systems has increased substantially. On–farm hardware produces much digital data, such as spatially determined values of crop yield, for example. Remote sensing of land cover in the last few years offers real potential for weekly continental–type observations of crop and pasture growth. Recent research on the Soil Moisture Integration & Prediction System (SMIPS), which combines satellite and ground observations, provide for daily estimates of soil moisture at root depth.  

Using crop yield monitors as an example, farmers and farming advisors have sought to make sense of climate, soils, landform and management options data. 

Variations in climate and physical conditions are a major barrier to realising the benefits of measurement technologies. This barrier can only be overcome by a large-scale data assembly and analysis platform which can identify and communicate effects of good and poor management options. 

Further, when a paddock and its management strategy is considered in relative isolation, its ignoring a huge body of information – namely, the strategies that other farmers are using. The platform would allow comparison of management strategies across a large number of paddocks, providing greater insights into what worked (and didn’t) in a given year and how a paddock compared relative to others with similar physical conditions.  

This project developed analysis and analytical tools to compare and analyse management strategies of multiple paddocks. 

This project was led by Dr Peter Caccetta. 

For more information, read the recent paper:

Diakogiannis FI, Waldner F, Caccetta P, Wu C (2020) ResUNet-a: a deep learning framework for semantic segmentation of remotely sensed data. ISPRS Journal of Photogrammetry and Remote Sensing 162, 94-114. doi:10.1016/j.isprsjprs.2020.01.013