Real-time IoT monitoring of soil parameters and monthly machine learning-based prediction of soil organic carbon in an urban forest
Authors:
- Raihan Azwar Ibrahim PulunganRPRaihan Azwar Ibrahim Pulungan
- Indra ChandraICIndra Chandra
- Jangkung RaharjoJRJangkung Raharjo
- Lindiasari Martha YustikaLYLindiasari Martha Yustika
- Runik MachfirohRMRunik Machfiroh
- Amaliyah Rohsari Indah UtamiAUAmaliyah Rohsari Indah Utami
- Mia RosmiatiMRMia Rosmiati
- SuwandiSUSuwandi
- Adith Adryan DwianggaADAdith Adryan Dwiangga
- Tosy Bima Asadulloh MustofaTMTosy Bima Asadulloh Mustofa
- Eka SugiartoESEka Sugiarto
- Abdul Cholik MustafaAMAbdul Cholik Mustafa
Open Access
- 7 August 2026
- journal article
- research article
- Published by Springer Nature in Discover Internet of Things
- Vol. 6 (1) , 143
Abstract
Soil organic carbon (SOC) is central to soil function and to the global carbon cycle, yet most assessments still depend on laboratory analysis that can only be repeated a few times a year. We describe an Internet of Things (IoT) and machine learning framework that estimates SOC dynamics indirectly, through soft sensing of soil parameters that are measured continuously in a campus urban forest. The sensors do not measure SOC itself; they record soil moisture, pH, electrical conductivity (EC), temperature, and the macronutrient indicators nitrogen, phosphorus, and potassium (N, P, K) in the Telkom University Forest, Bandung, Indonesia. Monthly laboratory SOC measurements, collected from January to August 2025, provided the reference values used to calibrate and test the models. Raw readings were filtered against physical ranges and screened for outliers before being averaged to monthly values, which then served as model inputs. Of the algorithms we compared, Random Forest gave the best result, with an R² of 0.829, a mean absolute error (MAE) of 0.221, and a root mean square error (RMSE) of 0.363 for monthly SOC. While the model demonstrates moderate predictive accuracy, it robustly captures the seasonal dynamics of SOC, aligning with laboratory-validated trends. We therefore present the framework as a tool for early monitoring and trend detection that complements laboratory analysis instead of replacing it, and that brings soil-carbon assessment to a temporal resolution conventional sampling cannot reach.Keywords
Funding Information
- Badan Riset dan Inovasi Nasional (61/II.7/HK/2024)
This publication has 33 references indexed in Scilit:
- Fraction-specific soil carbon responses to decadal warming in two contrasting permafrost grasslands of the Qinghai–Tibet PlateauPlant and Soil, 2026
- Energy- and behavior-aware sensory data collection for ASD monitoring over IoT using unequal clustering and reinforcement learningDiscover Internet of Things, 2026
- Radiocarbon and the Transit Time of Carbon in Terrestrial EcosystemsCurrent Climate Change Reports, 2026
- Modern Methods for Anomaly Detection in Enterprise System Logs: Algorithms, Implementations, and Practical Case Studies2025
- Advancing Soil Organic Carbon Prediction: A Comprehensive Review of Technologies, AI, Process‐Based and Hybrid Modelling ApproachesAdvanced Science, 2025
- Role of topography and physico-chemical properties on organic carbon distribution in semi-arid calcareous soilsEuro-Mediterranean Journal for Environmental Integration, 2025
- Soil Organic Carbon Fractionation Assessment in Areas with High Fire Activity Using Diffuse Spectroscopy and Tree-Based Machine Learning AlgorithmsEarth Systems and Environment, 2025
- Assessing the performance of machine learning models for predicting soil organic carbon variability across diverse landformsEnvironmental Earth Sciences, 2024
- A comprehensive review of soil organic carbon estimates: Integrating remote sensing and machine learning technologiesJournal of Soils and Sediments, 2024
- Online boxplot derived outlier detectionInternational Journal of Data Science and Analytics, 2024