Scholarlink

Water is essential for life and agriculture, yet its availability is increasingly limited due to climate change and rising global demand. Efficient water management in urban and agricultural irrigation systems is crucial for sustainability and minimizing waste. This study presents a data-driven approach for forecasting soil moisture at 8, 16, and 24-h intervals using Machine Learning(ML) and Data Mining techniques. The research was conducted using sensor data collected from an IoT-equipped public garden in Florence, Italy, between August 2023 and March 2025. The garden is part of a smart urban infrastructure project and is monitored through a network of interconnected devices, including six soil sensors (measuring moisture, temperature, and electrical conductivity), a leaf wetness sensor, a weather station, a water consumption meter, and six remotely controlled irrigation actuators. Following the KDD process, extensive preprocessing and feature engineering were performed and multiple regression models were compared, including linear regression, decision trees, random forest, AdaBoost, and gradient boosting. The best performing models achieved high predictive accuracy, with \(R^2\) scores ranging from 0.94 to 0.99 across different zones and time horizons. These results demonstrate the effectiveness of ML-based methods in modeling soil moisture dynamics and their potential to form the basis for intelligent irrigation strategies. The proposed methodology provides a replicable framework for sustainable water management in various urban and agricultural contexts.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Predicting Soil Moisture in an Urban Garden Using Sensor Data and Machine Learning Techniques

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Lucio Colizzi,
  • Giovanni Dimauro,
  • Nunzia Lomonte,
  • Leo Mastronardi

摘要

Water is essential for life and agriculture, yet its availability is increasingly limited due to climate change and rising global demand. Efficient water management in urban and agricultural irrigation systems is crucial for sustainability and minimizing waste. This study presents a data-driven approach for forecasting soil moisture at 8, 16, and 24-h intervals using Machine Learning(ML) and Data Mining techniques. The research was conducted using sensor data collected from an IoT-equipped public garden in Florence, Italy, between August 2023 and March 2025. The garden is part of a smart urban infrastructure project and is monitored through a network of interconnected devices, including six soil sensors (measuring moisture, temperature, and electrical conductivity), a leaf wetness sensor, a weather station, a water consumption meter, and six remotely controlled irrigation actuators. Following the KDD process, extensive preprocessing and feature engineering were performed and multiple regression models were compared, including linear regression, decision trees, random forest, AdaBoost, and gradient boosting. The best performing models achieved high predictive accuracy, with \(R^2\) scores ranging from 0.94 to 0.99 across different zones and time horizons. These results demonstrate the effectiveness of ML-based methods in modeling soil moisture dynamics and their potential to form the basis for intelligent irrigation strategies. The proposed methodology provides a replicable framework for sustainable water management in various urban and agricultural contexts.