Resources
Soil sample data: enhancing soil health and fertility management
6 June 2025
A combined dataset of over 4,000 real-world soil samples and more than 300,000 process-based simulated samples, supporting soil health assessment and AI-enabled fertiliser management.
At Mutus Tech, we have developed a soil dataset that combines more than 4,000 real-world soil samples with over 300,000 simulated samples generated using an agricultural process-based model. Together they form a foundation for understanding soil health and supporting sustainable fertiliser management.
Real-world soil samples
The dataset includes more than 4,000 soil samples collected across a range of agricultural regions and soil types. Each sample is accompanied by detailed metadata — environmental conditions, cropping history and land use — providing a practical empirical base for soil health assessment and model calibration.
Simulated data for model training
To expand on the real-world samples, we generated a large simulation dataset using a process-based model. The model uses weather, soil and management data to simulate fertiliser responses under a range of environmental scenarios. This produces over 300,000 synthetic soil entries, alongside more than one million data points covering soil greenhouse gas fluxes under different fertiliser practices.
Applications and ongoing work
Together, these real and simulated samples support the development of AI models for precision fertiliser management, soil health monitoring and emissions modelling. By combining complementary datasets, the aim is to build decision-support tools that are grounded in real-world data while remaining adaptable to a range of agricultural environments. The dataset continues to grow through partnerships with international collaborators and through ongoing field surveys.
Sample data
A sample dataset illustrating the structure and scope of this integrated soil data archive is available. Please contact us for access.