A Comparison of Stand-Level Volume Estimates from Image-Based Canopy Height Models of Different Spatial Resolutions

Ivan Balenović,Hrvoje Marjanović,Anita Simic Milas

doi:10.3390/rs9030205

Ivan Balenović, Hrvoje Marjanović + Show 1 more

Open Access

https://doi.org/10.3390/rs9030205

Copy DOI

Abstract

Digital aerial photogrammetry has recently attracted great attention in forest inventory studies, particularly in countries where airborne laser scanning (ALS) technology is not available. Further research, however, is required to prove its practical applicability in deriving three-dimensional (3D) point clouds and canopy surface and height models (CSMs and CHMs, respectively) over different forest types. The primary aim of this study is to investigate the applicability of image-based CHMs at different spatial resolutions (1 m, 2 m, 5 m) for use in stand-level forest inventory, with a special focus on estimation of stand-level merchantable volume of even-aged pedunculate oak (Quercus robur L.) forests. CHMs are generated by subtracting digital terrain models (DTMs), derived from the national digital terrain database, from corresponding digital surface models (DSMs), derived by the process of image matching of digital aerial images. Two types of stand-level volume regression models are developed for each CHM resolution. The first model is based solely on stand-level CHM metrics, whereas in the second model, easily obtainable variables from forest management databases are included in addition to CHM metrics. The estimation accuracies of the stand volume estimates based on stand-level metrics (relative root mean square error RMSE% = 12.53%–13.28%) are similar or slightly higher than those obtained from previous studies in which stand volume estimates were based on plot-level metrics. The inclusion of stand age as an independent variable in addition to CHM metrics improves the accuracy of the stand volume estimates. Improvements are notable for young and middle-aged stands, and negligible for mature and old stands. Results show that CHMs at the three different resolutions are capable of providing reasonably accurate volume estimates at the stand level.

Highlights

As the most widely distributed terrestrial ecosystem on Earth, forests provide many direct and indirect benefits to human well-being [1,2]
Sens.canopy, 2017, 9, 205 the smaller gaps cannot be determined by CHM5 (Figure 2a–c)
This research confirmed the great potential of digital aerial photogrammetry for stand-level forest inventory when fast, simple and low-cost approach based on existing data is needed

Summary

Introduction

As the most widely distributed terrestrial ecosystem on Earth, forests provide many direct and indirect benefits to human well-being [1,2]. In combination with field reference data and established prediction models, ALS metrics could serve for the estimation of various tree and forest inventory attributes (e.g., height, basal area, volume, biomass, etc.) [10,11,12,13,14,15,16]. There are two main approaches to derive forest information from ALS data: the area-based approach (ABA) and the individual-tree-based approach (ITBA). The ABA uses point cloud or CHM metrics of the sampled area (e.g., plot) as input to statistical models for forest attributes estimation [10,11,12,13,14,15,16]. Unlike the ABA, the ITBA is not used operationally yet, primarily due to difficulties in individual tree detection [7,18]

Objectives

Methods

Results

Discussion

Conclusion