Comparing Machine Learning Algorithms to Enhance Volumetric Water Content Prediction in Low-Cost Soil Moisture Sensor - research paper

Aug 5, 2026

Upload by Levi

Volumetric water content (VWC) is crucial for hydrology and for managing drought and flood risks, yet accurate measurement is often costly and difficult over large areas. Indirect sensing enables continuous, remote monitoring but can suffer from accuracy limits due to non-linear relationships between sensor dielectric properties and true soil moisture. This study compares linear regression, KNN, and decision tree models using gravimetric test results as reference targets, evaluating performance with RMSE. Results indicate KNN achieves the smallest RMSE, followed by decision tree and linear regression, supporting improved calibration for low-cost capacitive sensors and better agricultural water management.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages6

Reading Time15 min

AccessFree to download

Introduction

Importance of VWC measurement

Challenges with traditional and sensor-based approaches

Role of machine learning for sensor calibration

Materials and Methods

Data and sensor setup

Gravimetric test procedure

Data collection criteria

Predictive models and evaluation

Comparing Machine Learning Algorithms to Enhance Volumetric Water Content Prediction in Low-Cost Soil Moisture Sensor Iman Setiawan 1*, Mohammad Dahlan Th. Musa 2**, Dini Aprilia Afriza 3*, Siti Nur Hafidah 4* * Statistic Study Program, Tadulako University, Palu, Indonesia ** Geophysical Engineering, Tadulako University, Palu, Indonesia [npl.untad@gmail.com](npl.untad@gmail.com1)[1](npl.untad@gmail.com1), [ochad1969@gmail.com](ochad1969@gmail.com2)[2](ochad1969@gmail.com2) , [diniapriliafriza@gmail.com](diniapriliafriza@gmail.com3)[3](diniapriliafriza@gmail.com3) , [snhafidah07@gmail.com](snhafidah07@gmail.com4)[4](snhafidah07@gmail.com4) | Article history:
Received 2024-11-16 Revised 2024-11-21 Accepted 2024-11-29 | Measuring soil moisture is possible either with directly using gravimetric test or indirectly using soil moisture sensor. Direct measurements offer accuracy but are not efficient in field measurements. On the other hand, indirect measurement offers remote measurement that will facilitate the user but lacks in accuracy. This research aims to compare and identify the best machine learning model that can improve indirect measurement (soil moisture sensor prediction) using direct measurement (gravimetric test) as a response variable. This research uses linear regression, KNearest Neighbours (KNN) and Decision Tree models. The three models were then compared based on Root Mean Square Error (RMSE). The results suggested that KNN (0.02939128) had the smallest RMSE value followed by decision tree (0.05144186) and linear regression model (0.05172371).

This is an open access article under the CC–BY-SA license. | | --- | --- | | Keyword:
Soil Moisture Sensor, Machine Learning,
Regression, KNN, Decision Tree. | | Article Info ABSTRACT I. PENDAHULUAN Soil volumetric water content (VWC) plays a significant role in the hydrological cycle, particularly in triggering droughts and floods. As a result, proper VWC prediction is essential for effective water resources management as it helps in irrigation planning, increasing crop yields and conserving water. However, traditional methods of measuring VWC, such as taking soil samples or using sensors in the field, can be difficult and expensive, especially for large areas. In addition, complexities in structural properties and relationship with various meteorological factors lead to difficulties in building mathematical models that able to predict soil VWC accurately. [1] Machine learning (ML) techniques, which use data to make predictions, are gaining popularity in the sensor calibration process as they can estimate VWC more accurately. Machine learning also offers a new way of overcoming the non-linear pattern between the soil dielectric constant of soil moisture sensors and soil moisture content. Thus, developing a more comprehensive model to describe the complex relationship of actual soil water content and sensor measurements that can be applied to various soil types is essential. [2] Several recent studies have shown that machine learning can help in estimating VWC. First and foremost, [3] calibration of soil moisture sensors using machine learning. The calibration process is then compared between the laboratory and the field by predicting the volumetric water content, the results show that the calibration of the sensor in the field using a linear regression model is better than other machine learning models. Second of all, [4] saw that the decision tree method has a merit benefit to calculate soil moisture since the decision tree is an algorithm that makes a decision tree from given instances. Additionally, utilizing a decision tree offers a chance to save time. Finally, [5] discusses a system that utilizes the KNN algorithm to provide real-time irrigation data to farmers. By considering environmental elements such as soil moisture, temperature, and precipitation forecasts, the system assists farmers in optimizing crop output while minimizin

FAQs

Why is VWC prediction important for agriculture and water management?

VWC influences irrigation planning, crop yield, and water conservation, and it supports hydrological risk management such as drought and flood triggering.

What is the purpose of comparing machine learning models in this study?

To identify the best model for improving indirect VWC predictions from low-cost soil moisture sensors by using gravimetric test measurements as the response variable.

Which model performed best according to RMSE results?

KNN achieved the smallest RMSE, outperforming decision trees and linear regression in the reported comparison.

Aug 5, 2026

Upload by Levi