Abstract

Airborne lidar scanner (ALS) technology is used in a variety of applications, including forestry. ALS has enormous potential for the estimation of relevant biometric parameters in forest plantations. This study investigates the use of an object-oriented semi-automated segmentation algorithm for stands delineation, based on modeling ALS data, in plantations of Eucalyptus grandis and E. dunnii in Uruguay. The results show that non-parametric methods delivered more accurate and less biased results for total volume (TV) with R2 0.93, RMSE 20.04 m3 h−1 for E. grandis and R2 0.93, RMSE 18.43 m3 h−1 for E.dunnii; and above ground biomass (AGB) with R2 0.95, RMSE 70.2 kg h−1 for E. grandis and R2 0.96, RMSE: 71.2 Kg h−1 for E. dunnii. Parametric methods performed better for dominant height (Ho) with R2 0.98, RMSE 0.67 m and R2: 0.96, RMSE: 0.8 m for E. grandis and E. dunnii, respectively. The most informative ALS metrics for the estimation of AGB and TV were metrics related to the elevation in parametric models (Elev.70 and Elev.75), while for the non-parametric models (k-NN) they were Elev.75 and canopy density. For Ho, the ALS metrics selected were also related to elevation both in the parametric (Elev.90 and Elev.99) and random forest models (Elev.max and Elev.75). The segmentation methodology proposed here matched closely the segments delineated by human operators, and provides a low-cost, cost-effective, easy to apply and update model aimed at generating AGB or TV maps for harvest tasks, based on rasters derived from ALS metrics. The present research shows the capacity of ALS metrics to improve extensive strategic inventories; validating and promoting the adoption of ALS technology for inventory forest stands of Eucalyptus spp. in Uruguay.

Highlights

  • The quantification of forest stock is of key importance to forest management operations, such as habitat assessment, timber harvest, timber extraction, replanting, ecosystem modeling and stand delineation

  • In all cases, predicted values were in near-perfect agreement with the observed measurements

  • This study provides tools that allow the improvement of precision in two fundamental steps of forest stock quantification: stand variables modeling, using LIDAR and the delimitation of stand structure based on unsupervised evaluation

Read more

Summary

Introduction

The quantification of forest stock is of key importance to forest management operations, such as habitat assessment, timber harvest, timber extraction, replanting, ecosystem modeling and stand delineation. The most important variables affecting forest planning are: tree basal area, tree height, and stand-level volume [3]. Airborne lidar scanner (ALS) can support traditional forest inventories by providing precise measurements of forest attributes (e.g., stand density, mean basal area and dominant height) [4]. Discret ALS data has been employed to describe stands forest structure, through two approaches according to the scale of analysis; single tree identification and data aggregation [1]. Aggregation is known as the area-based approach (ABA) and is more efficient. This is because, by reducing the volume of discrete ALS returns to manageable grids, the data can be summarized within fixed-area units such as pixels, irregularly shaped segments, or forest stands [1]. In the ABA approach, stand attribute estimations are calculated from the statistical relationships between plot-level metrics from LIDAR data and plot-stand attributes [6]

Objectives
Methods
Results
Discussion
Conclusion
Full Text
Paper version not known

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call