Comparison of Coniferous Plantation Heights Using Unmanned Aerial Vehicle (UAV) Laser Scanning and Stereo Photogrammetry

Mei Li,Zengyuan Li,Erxue Chen,Qingwang Liu

doi:10.3390/rs13152885

Abstract

Plantation forests play a critical role in forest products and ecosystems. Unmanned aerial vehicle (UAV) remote sensing has become a promising technology in forest related applications. The stand heights will reflect the growth and competition of individual trees in plantation. UAV laser scanning (ULS) and UAV stereo photogrammetry (USP) can both be used to estimate stand heights using different algorithms. Thus, this study aimed to deeply explore the variations of four kinds of stand heights including mean height, Lorey’s height, dominated height, and median height of coniferous plantations using different models based on ULS and USP data. In addition, the impacts of thinned point density of 30 pts to 10 pts, 5 pts, 1 pts, and 0.8 pts/m2 were also analyzed. Forest stand heights were estimated from ULS and USP data metrics by linear regression and the prediction accuracy was assessed by 10-fold cross validation. The results showed that the prediction accuracy of the stand heights using metrics from USP was basically as good as that of ULS. Lorey’s height had the highest prediction accuracy, followed by dominated height, mean height, and median height. The correlation between height percentiles metrics from ULS and USP increased with the increased height. Different stand heights had their corresponding best height percentiles as variables based on stand height characteristics. Furthermore, canopy height model (CHM)-based metrics performed slightly better than normalized point cloud (NPC)-based metrics. The USP was not able to extract exact terrain information in a continuous coniferous plantation for forest canopy cover (CC) over 0.49. The combination of USP and terrain from ULS can be used to estimate forest stand heights with high accuracy. In addition, the estimation accuracy of each forest stand height was slightly affected by point density, which can also be ignored.

Highlights

Plantation forests, accounting for 7.3% (290 million ha) of the world’s forest cover (3999 million ha) [1], play a critical role in the restoration and reconstruction of forest ecosystems, the provision of forest products, the increase in forest carbon sink, the improvement of ecological environment, meeting demands for wood and other forest products [2,3].The covering area of planted forests in China is the largest worldwide and is about approximately 36% of the total national forest coverage [4]
It can be found that the estimation accuracy of each forest stand height was little affected by point density, slightly increased with the point density changing from 0.8 pts to 30 pts/m2, and became more stable, which could be found from the corresponding RMSE results
These heights can be estimated with high accuracy by simple linear regression based on the plot-level metrics

Summary

Introduction

Plantation forests, accounting for 7.3% (290 million ha) of the world’s forest cover (3999 million ha) [1], play a critical role in the restoration and reconstruction of forest ecosystems, the provision of forest products, the increase in forest carbon sink, the improvement of ecological environment, meeting demands for wood and other forest products [2,3]. The covering area of planted forests in China is the largest worldwide and is about approximately 36% of the total national forest coverage [4]. As is widely-known, regardless of the purpose of plantation forest management, ecological benefits, or economic benefits, plantation forests are inseparable from forest management decision-making. Forest structural attributes and biophysical properties reflect the growth and health of the forests to a certain extent. Efficient and accurate estimations of the properties of plantation 4.0/).

Objectives

Results

Discussion

Conclusion