IoT and AI Based Smart Soil Quality Assessment for Data-Driven Irrigation and Fertilization

Abstract

Purpose: The rapidly growing demand for food due to rapid population growth in East Africa is one of the challenging issues and the sustainable way of tackling it, is to enhance the agriculture activities to satisfy the need of increasing farm productivity. However, the climate change, limited water resources and poor soil fertility reduces crops yields. In attempt to solve these challenges, Internet of thing (IoT) in conjunction with artificial intelligence (AI) techniques is increasingly being used in agriculture sector. This study investigates an integration of IoT and a deep learning (DL) driven solution for smart irrigation and fertigation by assessing soil nutrients and soil water content dynamics in Eastern province of Rwanda for optimization of these scare resources while increasing yields productivity.
 Methodology: The research data for analysis was collected from KABOKU-KAGITUMBA irrigation scheme, and data on soil moisture and soil nutrients was gathered over a six-month period from 36 sensor nodes that were installed in approximately 70 hectares with 6 pivots for irrigation. The collected data in real time by sensors was sent to an IoT platform and incorporated with the forecasted weather information there after a deep learning based model used to predict when to irrigate and when to fertigate and the notification sent to farmer with recommendations. The irrigation valves were automatically actuated based on the predictions. The study's main software tools for gathering, displaying, and analyzing real-time data streams were Things Speak, Tensor Flow Lite, and the Arduino Software (IDE). A prototype was finally implemented effectively.
 Findings: The resulting model showed that can perform well with an accuracy of 91.7% and it can work well when deployed in the remote area with minimum internet connection.
 Unique Contribution to Practice: since the currently technologies used in irrigation and fertilization are manual or based on threshold values for automatic irrigation, we recommend the implementation of this solution since it will guarantee data-driven farming, which will help to protect the environment and ensure the optimization use of water resources. Additionally, this will result in lower operating cost, which will raise earnings.