Use of close-range LiDAR devices and statistical inference approaches in operational stand-level forest inventories

Abstract

Close-range LiDAR devices are considered to have great potential for enhancing forest inventory (FI) estimates. However, this potential is still being explored in the case of ground-based LiDAR devices, especially when the target is focused on relatively large spatial scales, such as stand level. This study explored the performance of close-range LiDAR devices in terms of bias and error, particularly terrestrial laser scanning (TLS) instruments, as measurement tools in stand-level FIs. The main premise of the research is that close-range LiDAR devices provide auxiliary information that can be used to accurately and precisely predict the dependent variable of the target population, thereby reducing errors. To this end, this study compared the performance of different statistical inference approaches that can be implemented with these technologies, such as the simple expansion estimator (EXP), two-stage model-assisted regression (REG), conventional model-based (CMB) and three-phase hierarchical model-based (3pHMB) approaches. These approaches were used to compare the following types of data: field measurements and TLS single-scan data (EXP, REG); field measurements and unmanned aerial vehicles (UAV)-LiDAR data (CMB); and field measurements, TLS single-scan data and UAV-LiDAR data (3pHMB). The case study was carried out in a 16 ha experimental plot dominated by Pinus radiata and Pinus pinaster in northwest Spain, focusing on stand volume ($V$, m³ ha⁻¹) estimates. The findings showed that the use of close-range remote sensing devices as a source of auxiliary data provided lower error in $V$ estimates than the EXP approach using a single data source. The findings also suggest that close-range LiDAR devices can potentially be used as FI instruments. Therefore, the transfer of these sampling techniques may play an important role in operationalizing the use of close-range LiDAR devices in FIs.