Iot and machine learning for multi-horizon soil moisture forecasting using leakage-free reconstructed sensor data in precision agriculture

Smart agriculture has emerged as an important approach for improving crop productivity, resource efficiency, and climate resilience through continuous environmental monitoring and predictive analytics. This study presents an Internet of Things (IoT) edge computing and machine learning framework for multi-horizon soil moisture forecasting in precision Agriculture. The proposed system integrates a Raspberry Pi 5 edge device with soil moisture, temperature, relative humidity, carbon dioxide (CO $$_2$$ ), and camera sensors to continuously monitor environmental and crop conditions on a cassava farm. Sensor data were processed using an edge-computing architecture and subsequently analyzed using statistical and machine learning techniques. To address prolonged sensor outage, a leakage-free time-series reconstruction methodology was developed that preserves temporal causality by reconstructing missing observations using only historical information available before the outage. Following data preprocessing and feature engineering, four supervised machine learning models including Linear Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Multi-Layer Perceptron (MLP), together with Persistence and ARIMA benchmark models, were evaluated for 1-day, 2-day, 5-day, and 7-day soil moisture forecasting. The nonlinear machine learning models consistently outperformed the benchmark approaches, with XGBoost and MLP achieving the highest predictive performance, reaching R2 score of up to 0.94 while substantially reducing forecasting errors. Comparative experiments further demonstrated that leakage-free reconstruction significantly improved forecasting accuracy relative to the non-reconstructed dataset. Feature importance analysis identified cyclical time-of-day information, historical soil moisture observations, and engineered temporal features as the most influential predictors. The study demonstrates the effectiveness of integrating IoT sensing, leakage-free data reconstruction, and machine learning to provide reliable soil moisture forecasting for intelligent irrigation management and sustainable cassava production.

Authors

Institutions

Publication Details

Journal
Discover Sensors
Published
2026-11-17
DOI
https://doi.org/10.1007/s44397-026-00099-1
Primary Topic
Smart Agriculture and AI
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Iot and machine learning for multi-horizon soil moisture forecasting using leakage-free reconstructed sensor data in precision agriculture

Nwamaka U. Okafor, Laeticia Nneka Onyejegbu, Chiamaka E. Njoku, Dorathy C. Dominic et al.
Discover Sensors
Smart Agriculture and AI
article

Iot and machine learning for multi-horizon soil moisture forecasting using leakage-free reconstructed sensor data in precision agriculture

Nwamaka U. Okafor, Laeticia Nneka Onyejegbu, Chiamaka E. Njoku, Dorathy C. Dominic, Grace O. Audu, Hannah Adebayo, Precious O. Oboatarheyere, Treasure K. Ariaga
article en

Abstract

Smart agriculture has emerged as an important approach for improving crop productivity, resource efficiency, and climate resilience through continuous environmental monitoring and predictive analytics. This study presents an Internet of Things (IoT) edge computing and machine learning framework for multi-horizon soil moisture forecasting in precision Agriculture. The proposed system integrates a Raspberry Pi 5 edge device with soil moisture, temperature, relative humidity, carbon dioxide (CO $$_2$$ ), and camera sensors to continuously monitor environmental and crop conditions on a cassava farm. Sensor data were processed using an edge-computing architecture and subsequently analyzed using statistical and machine learning techniques. To address prolonged sensor outage, a leakage-free time-series reconstruction methodology was developed that preserves temporal causality by reconstructing missing observations using only historical information available before the outage. Following data preprocessing and feature engineering, four supervised machine learning models including Linear Regression (LR), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Multi-Layer Perceptron (MLP), together with Persistence and ARIMA benchmark models, were evaluated for 1-day, 2-day, 5-day, and 7-day soil moisture forecasting. The nonlinear machine learning models consistently outperformed the benchmark approaches, with XGBoost and MLP achieving the highest predictive performance, reaching R2 score of up to 0.94 while substantially reducing forecasting errors. Comparative experiments further demonstrated that leakage-free reconstruction significantly improved forecasting accuracy relative to the non-reconstructed dataset. Feature importance analysis identified cyclical time-of-day information, historical soil moisture observations, and engineered temporal features as the most influential predictors. The study demonstrates the effectiveness of integrating IoT sensing, leakage-free data reconstruction, and machine learning to provide reliable soil moisture forecasting for intelligent irrigation management and sustainable cassava production.

Discover SensorsVol. 2(1)
University of Port Harcourt (NG), Argonne National Laboratory (US), Federal Polytechnic Nekede (NG)
Smart Agriculture and AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon