Mapping and Estimating Forest Stand Volume using Machine Learning Methods and Multi-Spectral Sentinel 2 Data

Aug 3, 2026

Upload by Eliana

Sustainable forest management requires accurate mapping and estimation of forest stand attributes including density, volume, basal area, and aboveground biomass. This study applies GIS, remote sensing, machine learning, and field inventories to estimate forest stand volume for natural and plantation forests in the Abra River Basin. Regression models using random forest (RF), k-nearest neighbors (KNN), and support vector machines (SVM) were validated, with RF performing best (R2=0.42; RMSE=0.40 m3/plot; MAE=0.31 m3/plot). Topographic variables from DEM and the Near Infrared (NIR) band were most influential. Estimated volumes ranged from 33–115 m3/ha.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages13

Reading Time33 min

AccessFree to download

Abstract

Introduction

Mapping and Estimating Forest Stand Volume using Machine Learning Methods and Multi-Spectral Sentinel 2 Data Nover M. Matso 􀀍 Department of Forestry, Abra State Institute of Sciences and Technology, Philippines Heherson B. Ong College of Environmental Management and Forestry, Isabela State University, Philippines Emerson V. Barcellano College of Environmental Management and Forestry, Isabela State University, Philippines | Suggested Citation | | --- | | Matso, N.M., Ong, H.B. & Barcellano, E.V. (2024). Mapping and Estimating Forest Stand Volume using Machine Learning Methods and Multi-Spectral Sentinel 2 Data. European Journal of Theoretical and Applied Sciences, 2(2), 635-647.
DOI: 10.59324/ejtas.2024.2(2).55 | Abstract: Sustainable forest management necessitates the mapping and estimation of forest stand attributes such as density, volume, basal area, and aboveground biomass. This study was conducted to explore the potential of geographic information systems (GIS), remote sensing, machine learning, and field inventories to estimate the forest stand volume of natural and plantation forests within watersheds in the Abra River Basin. The common machine learning regression techniques, which are random forest (RF), k-nearest neighbors (KNN), and support vector machines (SVM), were used to model and predict forest stand volume. The validation of the three machine learning methods showed that the best model to estimate and map forest stand volume is the RF algorithm (R2 = 0.42, RMSE = 0.40 m3/plot, MAE = 0.31 m3/plot) . Topographic variables such asthe Digital Elevation Model (DEM) and the spectral band Near Infrared (NIR) were the most important variables in predicting forest stand volume. The estimated forest stand volume using the RF model ranged from 33 to 115 m3/ha, with a mean of 59 m3/ha. The results of this study revealed that forest volume can be measured using freely available satellite data and machine learning techniques. Keywords: forest stand volume, geographic information system, machine learning, remote sensing, satellite images, sustainable forest management. Introduction Sustainable management of forests is an urgent need because it results in sustainable development, achievement of internationally agreed development goals, poverty eradication (Aerts and Honnay, 2011), food security, biodiversity conservation, and climate change mitigation (Food and Agriculture Organization (FAO), 2020) . The necessity of updated forest resource inventories required for the assessment of forest stand characteristics is needed in support of sustainable forest management (Wulder et al., 2008). Forest resources assessment provides the foundation of forest planning and forest policy (White et al., 2016) and also describes the state and change of forest ecosystems (Gschwantner et al., 2022) . Forest inventory is important for assessment and analysis (Dau and Chukwu, 2018), for monitoring (Henry et al., 2021), as a tool in decision-making (Kangas and Maltamo, 2006), and to obtain information on the multifunction of forests (Fridman et al., 2014) . The lack of reliable and sufficient information on forest resources can lead to poor management decisions, which in turn result in poor outcomes for ecosystem health and human well-being (Modzelewska et al., 2016; Lister et al., 2020). Traditionally, forest inventories involved complete enumeration (Kangas and Maltamo, 2006). However, this method is time-consuming, laborious, and costly (Ronoud et al., 2019) . The use of remotely sensed data has recently revolutionized the way forest resource assessments are conducted. Recent advances in GIS, remote sensing technology, artificial intelligence (AI), and the Internet of Things (IoT) enable the rapid, up-to-date, and dependable extraction of information on forest features (Fan et al., 2018; Ahmadi et al., 2020). Remote sensing can provide consistent, reproducible, and up-to-date data on multiple forest attrib

FAQs

Which machine learning algorithms were used to model forest stand volume?

Random forest (RF), k-nearest neighbors (KNN), and support vector machines (SVM) were used to build regression models for forest stand volume prediction.

Why are topographic variables and the NIR band important in the predictions?

The study reports that DEM-derived topographic variables and the Near Infrared (NIR) spectral band were the most important variables for predicting forest stand volume.

What model performed best for estimating and mapping forest stand volume?

The random forest (RF) algorithm showed the best performance in validation for estimating and mapping forest stand volume (R2=0.42, RMSE=0.40 m3/plot, MAE=0.31 m3/plot).

Aug 3, 2026

Upload by Eliana