1. Introduction
Blanket fertiliser recommendations derived from a single composite soil sample per field mask substantial within-field spatial variability in nutrient status, which precision agriculture approaches aim to address through denser, georeferenced sampling and variable-rate application.
2. Methodology
Twenty low-cost NPK and pH sensor nodes were deployed on a grid across a 6-hectare field and interfaced to a central logger via LoRa, with readings combined with historical yield-map data in a random-forest regression model to generate a georeferenced fertiliser prescription map at 20m resolution, evaluated against uniform blanket application on an adjacent control area of equivalent size and crop.
3. Results
Trial plots following the site-specific recommendation map reduced total fertiliser expenditure by 19 percent while achieving a 7 percent yield increase relative to the uniformly fertilised control plots, attributed to correcting localised potassium deficiency identified in the eastern third of the trial field that blanket sampling had not detected.
4. Conclusion
Dense IoT-based soil nutrient sensing combined with site-specific prescription mapping can simultaneously reduce input costs and improve yield relative to blanket fertilisation. Future work will integrate real-time in-season sap-testing sensors to refine top-dressing recommendations.
References
[1] Mulla D. J., Twenty five years of remote sensing in precision agriculture, Biosystems Engineering, 2013. [2] Adamchuk V. I. et al., On-the-go soil sensors for precision agriculture, Computers and Electronics in Agriculture, 2004.