A Sensor-Driven Mobile Decision-Support System for Soil Fertility Advisory and Recommendation

Abstract

Soil fertility assessment is required for field-level agricultural decision-making. Raw readings from soil sensors can be unintuitive or ambiguous to farmers. Nitrogen, phosphorus, potassium, pH, electrical conductivity, and moisture parameters require semantic translation into recommendations for crop selection, fertilizer planning, and soil amendments/management. This paper describes a mobile decision-support system for semantic soil fertility advice and recommendations via a developed Android-based prototype. Soil sensor data from the Internet of Things are coupled with a pre-trained fertility prediction layer to generate a categorical output of low/medium/high fertility conditions. The predicted fertility is then passed onto an advisory layer, which automatically generates recommendations for stakeholders. The prototype graphical user interface includes location selection, displaying soil test information, displaying fertility status, NPK advice, pH advice, EC advice, moisture advice, crop recommendations, fertilizer recommendations, fertilizer quantity recommendations, water quality advice, and government resources. Validation of the study was performed by ensuring model-app output consistency, testing L/M/H advisory cases, and ensuring workflow via screenshots. Results indicate that soil sensor data can be processed within one mobile workflow to generate easily interpretable fertility status, advisory alerts, and recommendations. This study investigated and validated an Android-based decision-support tool that can translate sensor-derived soil information into farmer advice. Limitations of this study include prototype-level validation, not a fully deployed commercial tool.