An investigation into lidar scan angle impacts on stand attribute predictions in different forest environments

Dayal, Karun; Durrieu, S.; Lahssini, Kamel; Alleaume, Samuel; Bouvier, M.; Monnet, Jean‐Matthieu; Renaud, Jean-Pierre; Revers, Frédéric

doi:10.1016/j.isprsjprs.2022.08.013

“…Volumes were then computed using the allometric > REPLACE THIS LINE WITH YOUR MANUSCRIPT ID NUMBER (DOUBLE-CLICK HERE TO EDIT) < equations available in [29]. [30] and [13] followed the same protocol. A summary of field measurements is given in Table I.…”

Section: A Study Area and Field Measurementsmentioning

confidence: 99%

Enhancing forest attribute prediction by considering terrain and scan angles from Lidar point clouds: a neural network approach

Dayal¹,

Durrieu²,

Lahssini³

et al. 2023

Preprint

0

View full text Add to dashboard Cite

Sensitivity of lidar metrics to scan angle can affect the robustness of area-based approach (ABA) models, and modelling the interplay of scan geometry and terrain properties can be complex. The study hypothesises that neural networks can manage the interplay of lidar acquisition parameters, terrain properties, and vegetation characteristics to improve ABA models. The study area is in Massif des Bauges Natural Regional Park, eastern France, comprising 291 field plots in a mountainous environment with broadleaf, coniferous, and mixed forest types. Field plots were scanned with a high overlap from multiple flight lines and the corresponding point clouds were considered independently to expand the standard ABA dataset (291 observations) to create a dataset containing 1095 independent observations. Computation of lidar, terrain and scan metrics for each point cloud associated each observation in the expanded dataset with the scan information in addition to the lidar and terrain information. A multilayer perceptron (MLP) was used to model basal area and total volume to compare the predictions resulting from standard and expanded ABA datasets. With expanded datasets containing lidar, terrain and scan information, the R² for the median predictions per plot were higher (R² of 0.83 and 0.85 for BA and Vtot) than predictions with standard datasets(R² of 0.66(BA) and 0.71(Vtot)) containing only lidar metrics. It also outperformed an MLP model for a dataset with lidar and terrain information (R² of 0.77 (BA and Vtot)). The MLP performed better than RF regression, which could not sufficiently exploit additional terrain and scan information.

show abstract

“…However, there are few studies exploring the differences of canopy height loss in the horizontal direction perpendicular to flight routes and the influence factors contributing to this discrepancy. Scan angle proves to be a factor severely affecting the prediction of structural traits in forests ( Keränen et al., 2016 ; Liu et al., 2018a ; Dayal et al., 2022 ) and narrower scan angle usually leads to a more accurate estimation of the mean height. But now, the effects of scan angle on estimating mean height of grassland are still unclear, which is essential for developing robust LiDAR-based models for regional grassland height estimation with various types.…”

Section: Introductionmentioning

confidence: 99%

Correction of UAV LiDAR-derived grassland canopy height based on scan angle

Xu

¹

,

Zhao

²

,

Zheng

³

et al. 2023

View full text Add to dashboard Cite

Grassland canopy height is a crucial trait for indicating functional diversity or monitoring species diversity. Compared with traditional field sampling, light detection and ranging (LiDAR) provides new technology for mapping the regional grassland canopy height in a time-saving and cost-effective way. However, the grassland canopy height based on unmanned aerial vehicle (UAV) LiDAR is usually underestimated with height information loss due to the complex structure of grassland and the relatively small size of individual plants. We developed canopy height correction methods based on scan angle to improve the accuracy of height estimation by compensating the loss of grassland height. Our method established the relationships between scan angle and two height loss indicators (height loss and height loss ratio) using the ground-measured canopy height of sample plots with 1×1m and LiDAR-derived heigh. We found that the height loss ratio considering the plant own height had a better performance (R2 = 0.71). We further compared the relationships between scan angle and height loss ratio according to holistic (25–65cm) and segmented (25–40cm, 40–50cm and 50–65cm) height ranges, and applied to correct the estimated grassland canopy height, respectively. Our results showed that the accuracy of grassland height estimation based on UAV LiDAR was significantly improved with R2 from 0.23 to 0.68 for holistic correction and from 0.23 to 0.82 for segmented correction. We highlight the importance of considering the effects of scan angle in LiDAR data preprocessing for estimating grassland canopy height with high accuracy, which also help for monitoring height-related grassland structural and functional parameters by remote sensing.

show abstract

“…So far, most studies involving lidar remote sensing for forestry applications have followed this convention [3]- [9]. Recently, some studies [10]- [13] have tried to assess the impact of scan angles greater than 20°, and many studies involving UAVbased lidar data routinely used much higher scan angles [14], [15]. It may be difficult for highly inclined lidar pulses to reach the ground surface owing to the increased occlusions.…”

mentioning

confidence: 99%

“…Nonetheless, it is also true that probing lidar canopies with inclined pulses may also lend newer insights or different perspectives [16]. A related study [13] observed that datasets comprising nadir point clouds did not always result in better ABA models, thereby emphasising that forest canopies are not a homogenous medium, and the lidar-derived information (lidar metrics) depends on how the lidar pulses sample the canopy. Furthermore, two lidar acquisitions may not have identical properties, and the lidar metrics could be affected by the overall acquisition geometry as characterised by the acquisition properties (sensor properties, scan angle, scan azimuth, flying height), terrain properties, and vegetation structural characteristics.…”

mentioning

confidence: 99%

“…A previous study over mountainous terrain [29] demonstrated that neural networks are well-suited to exploit terrain-related information to improve standard ABA predictions. In another study over the same area [13], single flight line datasets resulted in variable ABA predictions with parametric models comprising widely used metrics sensitive to scan angle. The objective of this study was to evaluate the benefit of using individual point clouds obtained from multiple flight lines independently to a) increase the number of observations and b) retain acquisition properties of each flight line to improve ABA predictions.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Enhancing forest attribute prediction by considering terrain and scan angles from Lidar point clouds: a neural network approach

Dayal¹,

Durrieu²,

Lahssini³

et al. 2023

Preprint

0

View full text Add to dashboard Cite

Sensitivity of lidar metrics to scan angle can affect the robustness of area-based approach (ABA) models, and modelling the interplay of scan geometry and terrain properties can be complex. The study hypothesises that neural networks can manage the interplay of lidar acquisition parameters, terrain properties, and vegetation characteristics to improve ABA models. The study area is in Massif des Bauges Natural Regional Park, eastern France, comprising 291 field plots in a mountainous environment with broadleaf, coniferous, and mixed forest types. Field plots were scanned with a high overlap from multiple flight lines and the corresponding point clouds were considered independently to expand the standard ABA dataset (291 observations) to create a dataset containing 1095 independent observations. Computation of lidar, terrain and scan metrics for each point cloud associated each observation in the expanded dataset with the scan information in addition to the lidar and terrain information. A multilayer perceptron (MLP) was used to model basal area and total volume to compare the predictions resulting from standard and expanded ABA datasets. With expanded datasets containing lidar, terrain and scan information, the R² for the median predictions per plot were higher (R² of 0.83 and 0.85 for BA and Vtot) than predictions with standard datasets(R² of 0.66(BA) and 0.71(Vtot)) containing only lidar metrics. It also outperformed an MLP model for a dataset with lidar and terrain information (R² of 0.77 (BA and Vtot)). The MLP performed better than RF regression, which could not sufficiently exploit additional terrain and scan information.

show abstract

An investigation into lidar scan angle impacts on stand attribute predictions in different forest environments

Cited by 10 publications

References 54 publications

Enhancing forest attribute prediction by considering terrain and scan angles from Lidar point clouds: a neural network approach

Enhancing forest attribute prediction by considering terrain and scan angles from Lidar point clouds: a neural network approach

Correction of UAV LiDAR-derived grassland canopy height based on scan angle

Enhancing forest attribute prediction by considering terrain and scan angles from Lidar point clouds: a neural network approach

Contact Info

Product

Resources

About