Forest in situ observations using unmanned aerial vehicle as an alternative of terrestrial measurements

Test area

The experimental site was established in 2014 and located in a boreal forest in Evo, Finland (61.19° N, 25.11° E). The main tree species were Scots pine (Pinus sylvestris L.), Norway spruce (Picea abies (H. Karst.) L.), Silver (Betula pendula Roth) and Downy (Betula pubescens Ehrh.) birches. On the site, 24 forest sample plots, 32 m by 32 m for each, were selected by foresters to test various in situ measurement approaches.

The sample plots were classified into three stand complexity categories, i.e., “easy”, “medium” and “difficult”, according to the amount of occlusion at the ground level, the spatial stem density and the distribution of the DBH. The category “easy” represented clear visibility with minimal understory vegetation and low stem density (~ 700 trees/ha); “medium” represented sample plots with moderate stem densities (~1000 trees/ha) and sparse understory vegetation; the “difficult” category represented plots with high stem densities (~2000 trees/ha) and dense understory vegetation.

Figure 1 illustrates tree maps of three example plots from the three different complexity categories. The mean DBH, mean tree height and mean basal areas are 20.5 cm, 18.3 m and 23.1 m²∙ha^− 1, respectively, for the easy plot (Fig. 1a); 17.3 cm, 16.3 m and 30.2 m²∙ha^− 1 for the medium plot (Fig. 1b), respectively; and 12.3 cm, 13.2 m and 29.3 m²∙ha^− 1 for the difficult plot (Fig. 1c), respectively. In general, the mean DBH and tree height decrease and the basal area increases along with the growing complexity of stands.

Figure 2 shows the distribution of the DBH in each stand complexity category by the time of the UAV flight, which illustrates the variation of tree size within each complexity category. As shown by the DBH distribution, the higher the stand complexity, the higher is the population of small trees (e.g., DBH < 15 cm).

In situ observations using conventional field methods

Conventional forest field measurements were carried out between May and August of 2014 (Liang et al. 2018a). All trees with DBH larger than 5 cm were measured in the sample plots. The tree height and the DBH were measured using conventional field measurement methods, i.e., using calipers and inclinometer. Tree maps were produced for each plot by combining measurements in the field and in the TLS data. Preliminary tree positions were firstly mapped from TLS point clouds. The preliminary tree positions were then verified and updated during in situ investigations, and trees missed on the preliminary tree maps were added. The stem curves were manually digitized from the multi-scan TLS point clouds starting at the height of 0.65 m above the ground, continuing at the DBH height and then every meter above until the maximum measurable height in the point cloud. The stem volume was calculated directly using the tree height and stem curve measurements similarly as in Liang et al. (2018a). The total tree biomass was calculated using Finnish national allometric models (Repola, 2009). The DTM was generated through rasterization of the classified ground points using TerraScan software (TerraSolid Oy, Helsinki, Finland), where the ground points were automatically classified and manually edited when needed.

The forest plots were revisited in December 2017 to update the field reference. Possible changes since 2014 were first visually interpreted from the ULS data with respect to the existing reference tree maps. Among the 24 plots, 22 plots remained almost unchanged. One plot was completely cut. For another plot, the ULS data mis-matched the plot location. Therefore, 22 plots were employed in this study. In the 22 plots, altogether 72 trees had been felled in 18 plots during 2014–2017. The maximum number of felled trees in a sample plot was 10, which accounted for only a small proportion of the plot tree population. Thus, the classification of complexity categories did not change for the sample plots. The results of the ULS evaluations are therefore comparable with the results from TLS and MLS/PLS previously reported in other studies (i.e., Liang et al. 2018a, 2018b).

In situ measurements using TLS and MLS/PLS

TLS data were collected in 2014 using Leica HDS6100 (Leica Geosystems AG, Heerbrugg, Switzerland) with a multi-scan approach, that is, one scan at the plot center and four scans at the four quadrant directions. No pre-scan preparations were implemented in the field measurements. The data were registered using artificial spheres. The mutual scan registration accuracy was at a 2-mm level. The point spacing was 15.7 mm at a 25-m distance to the scanner in both horizontal and vertical directions.

The kinematic in situ measurements, namely, MLS from an all-terrain vehicle and PLS from a backpack, were also collected in 2014 (Liang et al. 2018b). The core measuring system for both platforms was identical, namely, AkhkaR2 (Finnish Geospatial Research Institute, FGI, Finland). Both platforms used the same scanning parameters: scanning frequency of 95 Hz, which resulted in an approximate 4-cm on ground point spacing along the profile at a range of 35 m and an on ground profile spacing of 1.0–1.4 cm at a typical platform moving speed of 1.0–1.45 m/s .

ULS measurements

The ULS data were collected in September 2017 using a Riegl RiCOPTER with VUX-1UAV (RIEGL, Horn, Lower Austria, Austria), as shown in Fig. 3. The UAV campaign lasted for 3 days and the average time spent on UAV flight per plot was 10–20 min. The flight altitude was approximately 50 m above the ground. Each plot was covered by 4–5 flight lines. The overlap of all flight lines at the plot area was typically high, e.g., 100%.

The GNSS-IMU system installed on the UAV was Applanix AP20​ GNSS-Inertial System (Trimble Applanix, Ontario, Canada). The position measurement accuracy was better than 0.1 and 0.2 m in horizontal and vertical directions, respectively. The roll, pitch and heading were measured at an accuracy of better than 0.015, 0.015 and 0.035 degrees, respectively. These give a point location accuracy of better than 3.6 cm at the nadir and 7.1 cm at the far end of the field-of-view at 50 m above the ground, if the positioning and ranging errors are left out of consideration.

The applied laser sensor was a Riegl VUX-1UAV. The scanner was mounted to scan nadir profiles and it operated at the 1550-nm wavelength. The maximum measurement range of the scanner was 300 m. The beam divergence was 0.5 mrad, providing 2.5-cm and 5.0-cm footprints at 50-m and 100-m distances from the scanner, respectively. Since the targeted flight altitude was 50 m, the 50-m and 100-m distances corresponded to the scanning distances at nadir and at far end of the field of view. In order to gain better stem visibility, the data were collected with a 120-degrees field-of-view and a 550-kHz laser pulse rate, resulting in 106 scan lines per second and in a 0.07-degrees (1.2-mrad) measurement resolution along each scan line. Alone each scan line, the on ground point spacing was 6.1 cm at nadir and 24.4 cm at the far end of the scan line. The typical flight speed was 2.0–4.0 m∙s^− 1, resulting in a 2.0–4.0 cm on ground spacing between scan lines. The point density was around 100–800 points∙m^− 2 on the horizontal ground surface if only one echo per pulse was considered. In practice, due to the high overlaps between flight lines, the point density increased vastly, resulting in 4000–18,000 points∙m^− 2 at the sample plot areas. Two Sony ILCE-6000 cameras were also mounted on the UAV for colorizing the point cloud. Figure 4 illustrates the examples of colored ULS point cloud data in the three stand complexity categories. Color information was not used in the following processing.

The visibility of trees varied significantly according to tree species and forest structure, as shown in Fig. 5. For example, pine trees in easy forest stands typically had an excellent visibility and the stem was visible for most of its length. For spruce trees, the visibility of the stem decreased significantly as the tree stem can be either partly visible or totally occluded by the tree’s own canopy and/or by canopies of the surrounding trees.

Figure 6 illustrates a pine tree under moderate occlusion in the ULS data. ULS points at five different stem heights are shown in the subfigures. The stem structure is visually identifiable in two lower slices, but is totally missing in other three upper slices, indicating the impact of tree’s own crown on its stem digitization.

Geometric inconsistency among different flight lines was visible in the delivered point clouds. Possible reasons for such errors include registration errors, influences of wind, and/or varying geometric accuracies from different flight lines at identical locations. Figure 7 shows four examples of mismatches between the flight lines, which indicate the challenge of tree parameter estimation brought by geometric inconsistencies.

Automated retrieval of tree and plot parameters

The point cloud from stationary TLS, MLS/PLS and ULS, were processed through the same processing chain as described in Liang et al. (2018b). The specialty here is that the ULS data were understood equal to the terrestrial point cloud and processed with a stem detection and modeling algorithm that was identical to the other terrestrial point clouds.

In a preprocessing stage, the original point cloud was first sampled through an equivalent sampling method. The point cloud was digitized into a voxel space, and in each voxel the point closest to the center of gravity was selected as the representative point for the point distribution within the voxel. The voxel size was 1 cm for both the TLS and MLS and 0.5 cm for the ULS considering the given density of raw point clouds. The DTM was reconstructed using a morphological filter and linear interpolation. The point cloud was firstly rasterized in 2D. The lowest point in each pixel (20 cm by 20 cm) of a 2D raster space was selected as a seed point and the largest connected group was interpreted to be part of the ground. Detached groups were accepted as ground if they were smoothly connected with the accepted ground. Due to the above canopy viewing position, ULS collects tremendous amount of canopy points that are less important for the stem analysis. Therefore, a canopy filtering was applied to remove the topmost canopy layer (i.e., 20% topmost) before the stem detection and modeling. After the stem detection, all canopy points were used in the tree height estimation.

Stem points were identified through a point-based approach. Points on stems were identified by analyzing the structure in their immediate neighborhood using principal component analysis. Tree stem models were built from the recognized stem points as a series of 3D cylinders representing the changes in the growth direction of stems. The DBH and location of a stem were then estimated from the cylinder element at breast height (1.3 m above the ground). The stem curve was estimated from the cylinder elements at predefined heights. The tree height was estimated differently for big and small trees, which were separated according to a DBH threshold of 15 cm. Big trees were assumed mostly to be dominant or co-dominant trees that are exposed directly to the sunlight and no trees are above them (Wang et al. 2016). The tree top was assumed to be the highest point around the stem. The small trees were mostly intermediate and suppressed trees, and the treetop was therefore found from the largest connected point group around the tree stem. For both big and small trees, the elevation difference between the tree top and the DTM beneath it was used as a height estimate for the tree.

The MLS/PLS data were processed with one additional step. Since a tree may be observed several times from different trajectories and spatial inconsistency from different observations prevails in the mobile data, the tree mapping followed the multi-single-scan type of processing (Liang et al. 2018a).

Retrieval of tree and plot parameters from ULS data by manual measurements

In addition to the automated processing, 7 plots, 3 from easy, 2 from medium and 2 from difficult categories, were randomly selected and manually measured in order to benchmark a data-level accuracy, thus, to enhance understandings of the automatically derived tree parameter estimates. The manually measured tree parameters include the tree position, the DBH and the tree height of individual trees that are visually identifiable from the ULS data. A circle at the breast height was fitted to each visually identified tree. If only few or none points were recorded for a stem at the DBH height, a circle at a higher position was searched for and fitted when possible. The tree position and the DBH were therefore the center of the fitted circle and the diameter of the fitted circle, respectively. The treetop and the stump were also visually identified, and the elevation difference between the tree stump and top was recorded as the tree height.

Evaluating the accuracy of individual tree parameters

The performance of TLS, MLS/PLS, ULS was evaluated by comparing the accuracies of the estimated tree-level parameters with respect to the field measurements.

The evaluation followed the procedure in Liang et al. (2018a). The detected trees were matched with the reference trees based on the horizontal stem locations and the DBHs of the trees. The search distance was 50 cm for TLS and ULS, and 150 cm for the MLS/PLS. The mapping accuracy was evaluated using the completeness, which indicates the proportion of reference trees that were automatically detected. The accuracy of the tree position, tree height, DBH, stem curve, stem volume and AGB were all evaluated using the relative Root Mean Squared Error (RMSE) and relative bias, both in percentages, with an exception of the tree location where only the absolute RMSE was calculated. The evaluation was carried out at an individual-tree-level, i.e., the estimated and reference parameters were compared, and reported at a plot-level, i.e., relative RMSE and bias were calculated at a plot level and further averaged at the plot-complexity-category-level. The results are reported separately for each stand complexity category.

Stem mapping

The stem mapping results from ULS, MLS/PLS and TLS are reported in Fig. 8. Similar to the MLS/PLS and TLS, ULS had a steady declining trend in the completeness of stem mapping when the stand conditions became more complicated. The manual stem mapping from ULS provided very similar results as the automated stem mapping from MLS/PLS and TLS. However, when processed with the automated algorithm, the completeness of ULS-based stem mapping was remarkably lower (approximately 50% lower) than that of the MLS/PLS- and TLS-based results. Moreover, the decrease of ULS-based automated stem mapping completeness with respect to the increase of the stand complexity was more significant than that of the MLS/PLS- and TLS- based results.

These results suggest that, firstly, the state-of-the-art ULS data are capable to digitize similar amount of stems in forest plots as the other terrestrial point clouds; secondly, the stand complexity has more significant influence to the ULS-based stem digitization than to the other terrestrial point clouds. Even though the stem point cloud in ULS data (e.g., the density and the geometrical accuracy) is sufficient for human recognition, it is not yet comparable with the other terrestrial point clouds to support the automated stem detection, at least with the applied algorithm.

Stem position

The stem positioning accuracy from ULS, MLS/PLS and TLS are reported in Fig. 9. In general, the average RMSE of stem locations for digitized stems in ULS data was between 6.0 cm and 15.0 cm considering the stand complexity. The ULS-based stem positioning error is about two times as much as that of TLS-based results as shown in the Fig. 9. Compared to ULS, the MLS-based stem positioning accuracy was significantly lower due to the geometric inconsistency brought by the unstable GNSS signals under canopies.

Diameter at breast height

The relative RMSE (RMSE%) and the relative bias (bias%) of the DBH estimates from the three platforms are reported in Fig. 10.

The columns of “UAV” and “UAV (gross errors removed)” in Fig. 10 represent the evaluation results of the automated DBH estimates from ULS data. The differences between these two results were caused by gross errors, which mainly occurred at the stems of small trees that were digitized with low quality in the ULS point clouds. That is, the stems were digitized in the ULS data to an extent that is not enough for a reliable delineation by the automated algorithm. The stem modeling from ULS data produced large errors, for example, due to the low density, completeness, and geometric accuracy of stem points. Errors larger than 20 cm with respect to the reference were considered as gross errors. Gross errors accounted for only a small proportion of detected stems, that is, maximum three in a plot. Once the gross errors were removed, the RMSE% of automated ULS-based DBH estimates became comparable to that of the manual recognition from ULS data, as well as to that of the automated DBH estimates from TLS and MLS/PLS data.

The RMSE% of ULS-automated, ULS-manual, and automated estimates from MLS were at a same level (15%–30% in different stand complexities), and they were twice as much as the RMSE% of the TLS-based estimates. The bias% of ULS and MLS suggested that both aerial and terrestrial mobile platforms lead to overestimated DBH. The overestimation from ULS is more significant than that from MLS, indicating the geometric accuracy at the stems is higher in the MLS compared to the ULS point clouds.

The tree height

The advantage of an aerial observation perspective for tree height measurements is clearly shown in Fig. 11. ULS presented the best performances across different stands among all point clouds due to a better visibility of the upper parts of crowns. On the other hand, the tree height is the only parameter, for which the ULS-based estimates outperformed the MLS- and TLS-based estimates.

The automated tree height estimates from ULS performed similarly as the manual mensuration in easy plots and better in medium and difficult plots. This however does not mean that the automated algorithm outperforms the manual identification in the treetop recognition. The difference is due to the amount of recognized trees, which is smaller in the automatic detection. Automatic algorithm detected only trees whose trunk was clearly visible in the data. This underlines the claim in the TLS benchmarking (Liang et al. 2018a) that the accuracy of individual tree parameter estimates should be analyzed together with the detection rate (i.e., completeness). Treetop identification in point clouds is a challenging task even for human eyes especially in medium and difficult stands where species are mixed and tree crowns are entangled with each other. Thus, treetops may be misinterpreted, leading to a larger error than that of the automated recognition where only the trees clearly recorded were measured. These results also indicate that the manual and the automated interpretation of point clouds perform very similarly in treetop recognition for trees that are clearly digitized.

The stem curve

The evaluation of stem curve estimates are reported in Fig. 12. While the RMSE% of the stem curve estimates from ULS was similar to that of the DBH, the bias% was almost doubled. This was caused by more gross errors and less accurate estimates due to, e.g., occlusions in the data. Similar to DBH and tree height estimates, the RMSE% and bias% increased along the increased stand complexity.

The percentage of tree height covered (PHC, the extracted stem length covered by the stem curve divided by the tree height) of ULS-based stem curve estimates was in general higher than that of MLS and lower than that of TLS. The PHC value of the stem curve estimates based on ULS was expected to be higher than that of the terrestrial point clouds from TLS and MLS, because ULS has a favorable viewing position to digitize upper part of tree stems. However, the results suggest that the stem curve modeling from ULS can be disturbed by the missing parts in the middle of the stems, therefore, the clearly digitized upper stem parts can hardly be included in the stem curve model. While outperforming the MLS, ULS still cannot compete with TLS in the stem curve modeling in the sense of completeness, i.e., PHC. More discussion about stem curve estimates is given in the Section of Discussion.

Stem volume and total tree biomass

The results of the stem volume estimates are reported in Fig. 13. The stem volume is a function of stem curve, but it is proportional to the square of the radius of the stem. Therefore, the increase of ULS-based volume error with respect to the stand complexity is higher compared to the increase of ULS-based stem curve error (Fig. 12).

The results of the total tree biomass estimation are reported in Fig. 14. The RMSE% values in easy plots increases from 11.1% in TLS to 24.6% in MLS and to 37.6% in ULS. In the medium and difficult plots, the differences between different systems become larger. Since the stem curve and DBH accuracies play a more significant role in the volume and biomass estimates (Liang et al. 2018a), the more accurate tree height estimates of the ULS do not improve the ULS-based stem volume and total tree biomass estimates observably.

Pros and cons of the above-canopy view perspective in forest field inventories

The main motivation of applying mobile platforms (e.g., MLS/PLS/ULS) in forest inventories is three-fold. Firstly, mobile platforms improve the efficiency of data collection significantly. Secondly, the area of interest is more thoroughly covered by the kinematic data collection compared to stationary scans. Thirdly, the occlusion effects can be mitigated by moving the observation viewpoint around the targets.

Uplifting the viewpoint from the ground to the air using a UAV platform indeed raises the mobility to a level that has never been seen before. Due to the absence of obstacles along the path, ULS’s mobility is even higher than what MLS can provide. In comparison with TLS and MLS, the forest plots are covered more comprehensively in the ULS data, that is, close to 100% plot coverage in each and all flight lines. However, currently, such coverage improvement is present mainly in the horizontal dimension. In the vertical dimension, ULS is able to digitize trees comprehensively if the viewing geometry is ideal. Complete coverage of a tree from the stump to the top can be achieved if occlusions, e.g., from the canopies and other nearby trees, are absent, e.g., for an isolated tree. However, occlusions still exists which remarkably reduce the ULS data coverage of individual trees in the vertical direction, specifically at the lower or middle parts of the stems. As such, recent application of UAV platform does not yet reach the objective of mitigating the occlusion effects present in terrestrial technologies. Results in this study suggest that the mobile platforms on the ground lessen the occlusion much more successfully, even though the ground mobile platforms suffer from the geometric inconsistency due to GNSS signal losses.

Unlike the terrestrial point clouds where the occlusion effects are mainly caused by the bushes, small trees and lower parts of trees, in the ULS point clouds, the occlusions are caused significantly by the upper parts of the canopies. Data collection in leave-off season might be able to mitigate certain occlusions of deciduous trees, but cannot solve the problems in coniferous forests. As illustrated in Figs. 6 and 7, tree stems may be fully, partially or barely digitized by ULS, depending on the scanning geometry, species and forest stand conditions. The applied ULS system in this study is a high-end product. The point clouds from the survey-grade laser scanner and the fairly good navigation system represent a high quality point cloud data in the ULS domain in terms of the geometric accuracy and point density. With the applied high-end ULS, the capability of stem digitization from above canopies seems to be, however, still limited.

Data acquisition configurations

The density and geometric consistency of the point cloud data from ULS are influenced by the forest stand conditions, e.g., the structure and species of the forest, and by the data acquisition configurations, e.g., the setups of the scanner, positing system and flight lines.

In general, the point density of a scanner with the rotating mirror scanning mechanism is a function of the Pulse Repetition Rate (PRR), the Field of View (FOV) and the flight height and speed, which determines the point spacing within and between the scan line(s). A high PPR, low flight height, slow flight speed and small FOV lead to a high point density, and vice versa. Also, high overlap rate between flight lines over the forest site increases the point density. Higher point density typically achieve better canopy penetration capacity. But this is not guaranteed. Forest conditions have significant impacts on the point penetration capability.

Using the sensor and platform configurations applied in this study as described in the Section of ULS measurements, the on ground point spacing of the ULS point cloud is 6–15 cm within a scan line and 2–4 cm between the scan lines, which produced 4000–18,000 points∙m^− 2 point density. However, such data are still not enough to record trees comprehensively, as shown in Figs. 5 and 6. In addition, the stems of small trees (e.g., 5–10 cm DBH) are recorded vaguely in the ULS point cloud, taking into account the occlusion effects under forest canopies.

Multiple flight lines further increase the point density, and consequently increase point coverage on the targets. However, the geometrical consistency inside the point cloud may decrease because of the inaccuracy of the registration between flight lines, as shown in Fig. 7, which may dwarf the benefit of multiple flight lines.

Generally speaking, the best practice of the ULS measurement, e.g., the FoV, flight height, speed and overlap rate between flight trajectories, are not yet established. These research questions deserve more research in future.

Point geometrical accuracy matters

The geometrical accuracy of a point in a mobile point cloud from a linear laser scanner is determined by the scanner settings, such as the beam divergence and FoV, and platform status, such as rotations and sensor-target distances.

In the applied ULS system, the angle measurement accuracy of the platform is better than 0.015 and 0.035 degrees for roll/pitch and heading, respectively. If the GNSS positioning and laser ranging errors are not considered, this angle measurement accuracy means that, from a 50-m altitude, a laser point’s position accuracy of a laser point on the ground is better than 3.6 and 7.1 cm, at the nadir and at the far end of the field-of-view, respectively. Such data accuracy is insufficient considering the required 1–2 cm accuracy of DBH estimates in NFIs.

In addition, the survey-grade laser scanner applied in the ULS system has a 0.5 mrad laser beam divergence, which corresponds to a 2.5-cm footprint at nadir and a 5-cm footprint at the far end of FOV for a 50-m flight altitute. While negligible in TLS data due to the generally smaller beam divergence (e.g., 0.2 mrad) and the shorter scanner to target distances in forests (e.g. typically under 40 m), the influence of the beam divergence should be taken into account when using ULS data since it negatively affects the positioning accuracy of the points in the point cloud.

Moreover, the geometrical accuracy of points is heterogeneous through out a point cloud, namely, higher accuracy at nadir and lower accuracy at far end of FOV, which brings further challenges on point cloud registration between different flight lines. The relatively large uncertainties in ULS- and MLS/PLS-based tree parameter estimates can thus be understood as a result of error propagation from the platform positioning and orientation, and the geometric accuracy of points from the laser sensors. A slightly better performance can be observed from MLS/PLS than ULS, probably because MLS/PLS can achieve a similar performance as TLS in short periods of time. When the time span is short, the positioning drift is small and the relative positions between scanner and target are close to stable. Furthermore, the multi-single-scan principle for stem modeling applied in the MLS/PLS data processing shortens the time span and thus also reduces positioning errors that propagate to the point cloud data.

Outlook of ULS for in situ measurements of forest

ULS is the first aerial technique that achieves a stem digitization at a similar level of detail as TLS. Despite the recent technical challenges, the amount of recognizable stems in ULS data with human interpretation is similar to that with automated detection from TLS data.

So far, the geometrical accuracy of the ULS point clouds has not yet reached the level achieved through terrestrial LS systems. Even with the reasonably good UAV positioning system, the absolute 3D accuracy of a ULS point (3.6–7.1 cm on ground) is not yet accurate enough to estimate DBH and stem curve with a required accuracy. However, ULS outperforms terrestrial systems on the tree height estimations. Moreover, ULS provides a new in situ observation possibility that the observations are performed above the forest site. ULS frees the platform from any on-ground mobility constrains and enables fast and vast observations. Therefore, ULS may have a fundamental impact on the forest inventories.

The geometric inconsistencies is ubiquitous in point clouds, as results of winds, measurement inaccuracy and registration errors (Liang et al. 2018b; Pyörälä et al. 2018). While the wind impacts can hardly be eliminated, accuracy in measurement and registration can be improved. Smaller beam divergence, more accurate range measurement and higher angular resolution should provide higher measurement accuracy. Small FOV and slow flight speed can also reduce point spacing on the ground in ULS data.

All mobile systems in the air and on the ground face the same challenges of accurate registration between trajectories. The inconsistencies brought by overlapping trajectories in the ULS data appear less significant than that in the MLS/PLS point cloud if measurements from all trajectories are merged. The reason for this is that the GNSS signal degenerates under the forest canopies significantly, which hinders the accurate registration in terrestrial systems. In ULS, higher-quality GNSS-IMU system improves the positioning accuracy and vision-aided positioning (e.g., from structure-from-motion) can improve the orientation accuracy (e.g., Suomalainen et al. 2014). Actually, the registration problem in ULS may be slightly easier to solve than that in terrestrial platforms since the initial location parameter from GNSS is fairly accurate.

Even though the completeness and the geometric accuracy of stem digitization in ULS data is not as high as in other terrestrial systems, it records treetops in high detail. This is shown by the accurate tree height estimates in this study, which confirms the conclusion in a previous study (Wang et al. 2019). This hints that ULS most probably measures other canopy features accurately as well. Tree height and canopy features with higher resolution than those from ALS may serve as effective explanatory variables in the estimation of forest biomass, age and productivity. Meanwhile, ULS is most probably a suitable option in particular applications. ULS’s areal point of view has the potential to mitigate the occlusion effects within canopy, thus the structure information within canopies may be studied at a higher level of detail than what is presently achievable. Point clouds collected from terrestrial platforms typically have a limited spatial coverage of tree structures inside the canopy. Consequently, the performance of the quantitative structure delineation decreases in modelling the stems and branches above the live canopy base, even with high-resolution TLS point clouds (Liang et al. 2018a; Pyörälä et al. 2019). In addition, considering time and cost efficiency, ULS is also a very competitive candidate for the applications that emphasize tree count and location rather than tree parameter mensuration, such as the inventory of young plots.