Next Article in Journal
GPR-Based Leakage Reconstruction of Shallow-Buried Water Supply Pipelines Using an Improved UNet++ Network
Previous Article in Journal
Multi-Resolution UAV Remote Sensing for Anthropogenic Debris Detection in Complex River Environments
Previous Article in Special Issue
Estimating Leaf Chlorophyll Fluorescence Parameters Using Partial Least Squares Regression with Fractional-Order Derivative Spectra and Effective Feature Selection
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Remote Sensing
Volume 17
Issue 13
10.3390/rs17132173
remotesensing-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Comparison of Advanced Terrestrial and Aerial Remote Sensing Methods for Above-Ground Carbon Stock Estimation—A Comparative Case Study for a Hungarian Temperate Forest

by
Botond Szász
1,
Bálint Heil
2,*,
Gábor Kovács
2,
Diána Mészáros
2 and
Kornél Czimber
1
1
Institute of Geomatics and Civil Engineering, Faculty of Forestry, University of Sopron, 9400 Sopron, Hungary
2
Institute of Environmental Protection and Nature Conservation, Faculty of Forestry, University of Sopron, 9400 Sopron, Hungary
*
Author to whom correspondence should be addressed.
Remote Sens. 2025, 17(13), 2173; https://doi.org/10.3390/rs17132173
Submission received: 11 May 2025 / Revised: 20 June 2025 / Accepted: 23 June 2025 / Published: 25 June 2025
(This article belongs to the Collection Feature Paper Special Issue on Forest Remote Sensing)
Browse Figures
Versions Notes

Abstract

The increasing pace of climate-driven changes in forest ecosystems calls for reliable remote sensing techniques for quantifying above-ground carbon storage. In this article, we compare the methodology and results of traditional field surveys, mobile laser scanning, optical drone imaging and photogrammetry, and both drone-based and light aircraft-based aerial laser scanning to determine forest stand parameters, which are suitable to estimate carbon stock. Measurements were conducted at four designated sampling points established during a large-scale project in deciduous and coniferous tree stands of the Dudles Forest, Hungary. The results of the surveys were first compared spatially and quantitatively, followed by a summary of the advantages and disadvantages of each method. The mobile laser scanner proved to be the most accurate, while optical surveying—enhanced with a new diameter measurement methodology based on detecting stem positions from the photogrammetric point cloud and measuring the diameter directly on the orthorectified images—also delivered promising results. Aerial laser scanning was the least accurate but provided coverage over large areas. Based on the results, we recommend adapting our carbon stock estimation methodology primarily to mobile laser scanning surveys combined with aerial laser scanned data.

1. Introduction

The accelerated natural processes caused by climate change are making it increasingly urgent to quantify the carbon stocks of forest ecosystems and monitor their temporal dynamics. This process is of critical importance for achieving sustainable forest management and mitigating the adverse impacts of environmental changes. Due to the accelerating pace of these transformations, traditional carbon stock estimation methods are no longer able to provide timely and adequate assessments. Consequently, the integration of modern technologies into new estimation approaches has become indispensable.
In response to this need, the SoilSense project was launched in 2023 as part of the Climate Action Plan, supported by Hungary’s Ministry of Agriculture. At its current stage, the project has established a model area where topographic and derived thematic analyses have been conducted, exploring both the interrelationships between terrain variables and their links to forest stand characteristics—particularly height growth [1]. This was followed by the development of an integrated method that combines satellite imagery, airborne laser scanning (ALS) data, and field measurements, aimed at estimating both above- and below-ground carbon stocks [2]. In the latter study, the above-ground carbon stock was estimated by deriving critical forest structural and individual tree parameters (e.g., stem position, tree height, diameter at breast height, and crown dimensions) from low-density airborne LiDAR data.
In this article, as a part of the aforementioned methodology [2], we are focusing on the estimation of the abovementioned parameters using various remote sensing techniques. It evaluates the accuracy and applicability of these methods—or the combination of two methods—thereby assessing the potential for expanding or refining the integrated carbon estimation approach.

1.1. Terrestrial Laser Scanning

Terrestrial laser scanning (TLS) can fundamentally be divided into two subcategories: static and mobile scanning. In static TLS, the point cloud is generated from one or more fixed positions, while in mobile laser scanning (MLS), the operator moves continuously with the sensor during data acquisition. Since MLS data is captured while in motion, the resulting point clouds generally have lower density and accuracy, making fine details and small features better captured by static scanning methods [3]. However, mobile scanning is clearly more time-efficient [4]. One key advantage of mobile scanning in forest environments is the reduced occlusion of trees compared to static setups. With the increasing adoption of both methods, numerous studies have investigated their applicability in forestry [5].
According to Wardius and Hein [6], TLS-derived individual tree parameters show a tendency to underestimate tree height (especially at trees above 30 m, where the bias was more than 4 m) and a slight positive bias (1 cm) in DBH (diameter at breast height) measurements. A similar pattern was observed with the voxel-based automatic detection algorithm presented by Brolly et al. [7]. Brolly and Király [8] introduced a “coarse-to-fine” method consisting of two steps: pre-processing and clustering (point cloud segmentation), followed by classification (shape fitting). Brolly et al. [9] also developed an algorithm specifically aimed at detecting young trees with lower heights between 3 and 6 m. Compeán-Aguirre et al. [10] compared five DBH estimation algorithms (Nelder-Mead, least squares, Hough transform, RANSAC, and convex hull), of which the first two yielded the most accurate results. However, the outcomes were significantly influenced by point density, occlusion, vegetation cover, and tree structural characteristics. You et al. [11] presented a tree skeletonization method using TLS point clouds, improving visualization and the precision of tree measurements. Capalbo et al. [12] compared parametric (regression-based) and non-parametric (random forest) approaches for volume estimation, finding that the non-parametric model produced more accurate results. Yang et al. [13] proposed a multiscale framework to explore tree structural features from the whole tree down to small branches and their bifurcations. A notable group of approaches is those based on Quantitative Structure Models (QSMs), which approximate tree branch architecture, geometry, and volume using hierarchical primitive geometric shapes [14,15,16]. In addition to quantitative parameters, species classification methods have also been developed. Meng et al. [17], using random forest classifiers, achieved 94.17% accuracy by analyzing tree structure, bark texture (enhanced with kriging interpolation), and bark color.
The efficiency of mobile laser scanners is largely enabled by Simultaneous Localization and Mapping (SLAM) technology, which allows the scanner’s position to be tracked solely through onboard sensors—making it suitable for GNSS-free mapping and navigation of unmanned systems [18]. These systems can also be deployed in forested environments. Pan et al. [19] designed a disassembly-free dual-scanner backpack laser scanning (BLS) system capable of complete and accurate forest mapping, even in dense understory. Spadavecchia et al. [20], using the Forest Structural Complexity Tool (FSCT) [21], also confirmed that MLS-derived point clouds are suitable for forestry applications. Beyond forest inventory, MLS technology has been successfully integrated into other forestry sectors. Faitli et al. [22], for instance, embedded an MLS system into a harvester, enabling the real-time and automated estimation of tree diameter and curvature. This serves dual purposes: the surveyed point cloud allows for later forest modeling, and the operator benefits from immediate access to critical data.

1.2. Aerial Photogrammetry

Among the listed methods, forest inventory based on aerial imagery is considered the most cost-effective in terms of equipment expenses. During data acquisition, the aerial platform captures overlapping images, which can be used to generate 3D point clouds based on the principles of stereophotogrammetry [23]. To generate a point cloud, image matching must be performed first. In recent years, numerous image-matching algorithms have been developed, including SIFT (Scale-Invariant Feature Transform), SURF (Speeded-Up Robust Features), ORB (Oriented FAST and Rotated BRIEF) [24], SGM (Semi-Global Matching) [25], MVS (Multi-View Stereo) [26], and SfM (Structure-from-Motion) [27].
Due to the development and widespread availability of easy-to-operate drone technology, SfM has become the most widely adopted algorithm in forestry due to its simplicity and relatively low cost per unit area [28]. This algorithm is capable of simultaneously calculating both internal and external orientation parameters while also generating an initial 3D point cloud. Its greatest advantage is that it does not require pre-calibrated cameras, making it suitable for processing images captured by non-professional cameras as well [29].
Among the classical dendrometric parameters, tree height [30,31,32] and crown diameter [32,33] are the most commonly measured directly using aerial photogrammetry. Tree volume [33] and diameter at breast height (DBH) can also be estimated through statistical relationships and regression analysis [34,35]. The latter can only be measured directly in leaf-off conditions in deciduous forests; however, this measurement is also affected by crown architecture, branch density, and the presence of understory vegetation. Mao et al. [36] found that DBH values derived from 2D imagery were more suitable for practical forestry applications than those extracted from 3D photogrammetric point clouds.

1.3. Airborne Laser Scanning

Airborne laser scanning (ALS) refers to data collection using LiDAR sensors mounted on aerial platforms. Traditionally, this has involved fixed-wing light aircraft or helicopters; however, in recent years, the increasing use of drones and compatible sensors has led to the emergence of a separate category (such as the separation of TLS and MLS)—UAV-borne laser scanning (ULS). The two primary differences between ALS and ULS lie in point density, and survey efficiency and cost per unit area: ULS typically yields higher point densities, while ALS conducted by aircraft remains more efficient and cost-effective for covering larger areas [37].
The primary objective of point cloud processing algorithms is to generate digital terrain models (DTMs) and Canopy Height Models (CHMs). Various approaches have been developed for this purpose. Terrain classification and modeling are performed using algorithms such as bandpass filtering and averaging [38], cloth simulation [39,40], data-driven TIN surface fitting [41,42], and morphological filtering [43,44]. Canopy modeling methods include adaptive k-means clustering [45], layer stacking recognition [46], graph-based segmentation [47], and voxel aggregation-based individual tree detection algorithms [1,2]. Even with low-density point clouds (less than 20 pts/m2), tree volume estimation has been achieved by White et al. [48] using a random forest approach and by Oehmcke et al. [49] through deep learning models. Ullah et al. [50] compared three techniques—Multiple Linear Regression, k-Nearest Neighbour, and Support Vector Machine—finding that Multiple Linear Regression yielded the most accurate estimates for stand-level parameters. Sačkov et al. [51] applied the “Tree Crown Identification” module of the reFLex software, which resulted in a 21% RMSE.
Using ULS, point clouds with high density (over 100 pts/m2) can often allow for the direct estimation of diameter at breast height (DBH), typically through circle fitting along breast-height cross-sections. Kuželka et al. [52] compared three circle fitting methods—Hough transform, Random Sample Consensus (RANSAC), and Robust Least Trimmed Squares (RLTS)—and found that RLTS performed slightly better in complex cases such as incomplete or noisy point groupings. However, Hao et al. [53] noted that in dense forest stands, even ULS might not provide sufficient point density for direct DBH estimation, and thus applied a General Nonlinear Mixed-Effects (NLME) model to infer DBH indirectly. An alternative approach was presented by Chisholm et al. [54], who combined ULS with SLAM techniques to collect data beneath the canopy, enabling direct diameter measurements from low-altitude UAV flights.

2. Materials and Methods

The first study area of the project was selected to be the Dudles Forest, located north of Sopron, in Western Hungary. This forest spans 1067 hectares and is predominantly composed of Turkey oak (Quercus cerris), sessile oak (Quercus petraea), and Scots pine (Pinus sylvestris), as well as black pine (Pinus nigra). Previously, a total of 272 permanent sampling plots were established across the forest using a 200 × 200 m grid system, with minor deviations in certain locations [1,2].
For the present study, 4 sampling plots were selected from the 272 (see Figure 1) to compare different methods for estimating various forest stand parameters. When selecting the sampling areas, care was taken to ensure they were representative of the entire study region. The four selected sites include the four dominant tree species that cover the majority of the area and are located in mid-aged or mature forest stands. Younger stands were excluded due to their dense structure and thin trunks, which limit the accuracy of the applied methods. In terms of topography, the sites are situated on flat terrain, gentle to moderate slopes, and ridges—landforms that collectively represent approximately 90% of the entire study area.
At these four sample locations, five distinct methods were used to determine dendrometric parameters, including stem position, diameter at breast height (DBH), and tree height. The software applied for the estimation of these parameters—introduced in the following sections and developed with our contribution—incorporates procedures for calculating dry biomass and carbon stock. A detailed description of these calculations is provided in [2]. Among these methods, two are terrestrial-based, while three are aerial, including two drone-based techniques and one light aircraft-based approach. The procedural steps of each method are illustrated in Figure 2.

2.1. Field Survey

To enable the comparison and accuracy assessment of various remote sensing methods, reference data were collected through traditional field-based dendrometric surveys. Within a circular plot of 12.61 m in radius (500 m2 area) around each of the four selected sampling points, all trees with a diameter at breast height (DBH) greater than 6 cm were measured, starting from the north and proceeding clockwise. For each individual tree, species, DBH, height, and estimated distance and azimuth from the plot center were recorded for spatial identification. These data allowed for the determination of the number of stems, as well as the mean DBH and tree height by calculating the arithmetic mean of the individual tree measurements. In addition, the total tree volume per plot could also be derived. The forest stand parameters described in the following sections were determined in a similar manner for each method applied. To evaluate the accuracy of these methods, we calculated the RMSE (Root Mean Square Error) by comparing the extracted parameters of each tree to the corresponding reference values. We also performed a correlation analysis between the parameters to reveal potential relationships and trends in the measurements.

2.2. Mobile Laser Scanning

For the survey, a Stonex X120GO handheld SLAM laser scanner (Stonex Ltd., Milan, Italy) was used. The device has a performance of 320,000 points/sec, a range of 120 m, and an accuracy of 1 cm. It is equipped with a Hesai Pandar XT16 sensor (Hesai Technology, Shanghai, China). The instrument is operated with two applications: the GOapp (v2.9.1.292) mobile controller and the GOpost (v2.3.5.0) desktop processing software. Data collection was conducted within a circular plot of 12 m in radius, with the plot center and the north direction marked using a survey pole. Raw data processing was performed in the GOpost software.
Tree detection and dendrometric parameter extraction from the generated point cloud were carried out using the dotXpert v1.25.5.6 software (TopoLynx Ltd., Kőszeg, Hungary) (Figure 3). The processing workflow includes the filtering of outlier points, generation of the DTM via hierarchical ground point filtering (upward) and interpolation (downward), and filtering and segmentation of points between 1.1 and 1.5 m, followed by three types of fitting to the segmented clusters: regression circle, regression cylinder, and RANSAC circle. The best-fitting model is then selected, and the results are exported. Numerous parameters can be configured in the software for each processing phase, including ground filtering options, DTM resolution, height range for trunk detection, segment size, noise filtering, DBH range, fitting method, and fitting threshold.

2.3. UAV Photogrammetry

The study was also conducted using photographs, for which a DJI Mavic 3 drone (SZ DJI Technology Ltd., Shenzhen, China) was employed. Since the four selected sample points are located in close proximity to each other, the survey was completed with a single flight. The flight was performed at an altitude of 100 m, with 90% image overlap both along and across the flight lines. The resulting images had a geometric resolution of 2.7 cm/pixel. Although an RTK connection was available, due to a weak and unstable signal, field control points (a total of 5) were placed, and these were surveyed using a FORGEO Puli GNSS receiver (FORGEO Ltd., Baja, Hungary).
Image orientation and processing, as well as point cloud generation, were carried out in the Agisoft Metashape v2.0.2 (Agisoft LLC., St. Petersburg, Russia) software. Based on the resulting photogrammetric point cloud, a terrain model was created, and trunk positions were determined from clusters of points in the height range of 1.1–1.5 m. Subsequently, an orthophoto was generated on the 1.3 m elevated virtual terrain model, all in the topoXmap v1.23.7.27 (TopoLynx Ltd., Kőszeg, Hungary) software. The detected trunk positions were projected onto the orthophoto, and the diameter at breast height (DBH) was measured directly from the images using the principle of central projection (Figure 4). Initially, this was performed manually, after which the potential for automating the method was explored. To achieve this, image segmentation and classification of the trunk-shaped and colored image segments were performed using the eCognition Developer 9 (Trimble Inc., Westminster, CO, USA) software, with expert classifiers. In most cases, the width of the segments provided the required diameter.

2.4. UAV-Borne Laser Scanning (ULS)

The UAV-based LiDAR point cloud was created using a DJI Matrice 350 RTK drone (SZ DJI Technology Ltd.) and a DJI Zenmuse L1 LiDAR sensor (SZ DJI Technology Ltd.). The sensor’s maximum operational range is 450 m, assuming sufficient target reflectivity (e.g., 10% at 190 m and 80% at 450 m distance), and its maximum performance can reach 480,000 points per second. The horizontal accuracy is 10 cm at a range of 50 m, and the vertical accuracy is 5 cm. The flight was planned using the DJI Pilot 2 software, where the overlap between flight lines was set at 55% to ensure that every terrain feature was surveyed from two directions. We used the sensor with a 70° tilt relative to the nadir position, thereby increasing the chances of detecting tree trunks. The flight was conducted at an altitude of 100 m with a speed of 7 m/s, resulting in an estimated point density of 404 points/m2 by the software. Similarly to UAV optical imaging, the four sample areas were surveyed in a single flight, meaning a continuous flight path was used for data collection. During the scanning, we set the recording of the first and last returns, which ensures the highest point density.
The raw point cloud was processed and projected using the DJI Terra v.4.4 software (SZ DJI Technology Ltd.). Since the RTK connection was not perfect during the flight due to poor mobile signal strength, post-processing was applied, and correction data was obtained from the GNSSnet.hu state provider. The resulting point cloud was also processed in the dotXpert software, starting with the deletion of scattered points. Then, a terrain model was created, and points at heights of 1.1 to 1.5 m above this model were filtered, followed by attempts to detect individual trees.

2.5. Aerial Laser Scanning with Light Aircraft

This survey also used an airborne LiDAR point cloud collected over a larger area at a higher altitude than the previous one, which was provided by Envirosense Ltd. (Debrecen, Hungary). The point cloud was created using a small aircraft and a Riegl V780 II (RIEGL Laser Measurement System GmbH, Horn, Austria) sensor. The average point density of the cloud is 20 points/m2. The point cloud processing was carried out using the TreeDetect v1.25.01.12 (TopoLynx Ltd.) software, which uses a voxel aggregation method for detecting individual trees. The procedure begins with classifying the point cloud by searching for ground and canopy points, producing both the terrain and canopy models. Then, using the voxel aggregation algorithm, it segments the trunks and crowns of the individual trees. The iterative crown building occurs on the voxel surface, and the output includes crown apex points and polygons, along with related attribute data such as height, crown diameter, DBH calculated using a regression method, tree volume, and carbon stock. The model used for DBH estimation is based on yield tables. To ensure its applicability under the specific conditions of the sample area, a correction factor is required. This factor is determined by comparing field data with the general model. The model has already been calibrated for the Dudles Forest area, as described in [2], and the same version was used in the present study.

3. Results and Discussion

3.1. Field Survey

The results of the ground survey serve as a reference for evaluating the other methods. The results consist of a data set for each test plot, containing the diameter, height, and species of each tree. The average and total values obtained by aggregating the data are presented in Table 1, while Figure 5 shows a DBH-based density chart for each sampling point. It is important to emphasize that in our case, the measurement of the distance from the center and the angle with the northern direction used to determine the trunk positions was not performed with geodetic-grade equipment but was estimated, with an accuracy of a few meters. Therefore, some deviations may be observed, but despite this, the individual trees can be paired with the results of the remote sensing methods. This is illustrated in Figure 6, where we compare these positions with those recorded by the MLS.
The advantage of the ground survey is that it provides accurate, reference-grade results, and in the case of species identification, it is considered the most reliable method, with accuracy only influenced by human factors. The disadvantage is that it is time-consuming and requires manual labor, as at least two people are needed to complete the survey.

3.2. Mobile Laser Scanning

The mobile laser scanning provided convincing results. Thanks to the SLAM technology, there is no occlusion, so the trunks were scanned from all directions within the sample area in every case (Figure 6). In the 12 m radius survey area, the handheld laser scanner can survey tree specimens within a much larger area, almost a 25 m radius. This area is nearly four times the size of the sample point area. This should be considered in future carbon stock surveys. By increasing the filter, the software was able to recognize most of the trees on the younger and denser southwestern plot 20I.
Among the methods examined, this was one of the most accurate, as the trunk shapes clearly appeared on the cross-section, and due to the nearly perfect circle fitting, the deviation of the average diameter from the reference is within 1 cm. Furthermore, it is very time-efficient, taking only 2–3 min to survey an area of this size. Another advantage is that only one person is needed to conduct the survey; the point cloud processing is fast; does not require specialized knowledge; and, as mentioned above, thanks to SLAM technology, there is no occlusion, and the stems are scanned from all directions.

3.3. UAV Photogrammetry

A dense photogrammetric point cloud was generated from 1400 high-overlapping drone images through image matching (Figure 7). From the point cloud, a high-resolution digital terrain model can be created. On the special DEM+1.3m, the direct measurement of trunk diameters was possible both manually and automatically, by identifying similar image objects. Although the images were taken during the leafless season with high overlap, some trunks were not visible due to the branches of denser canopies.
Because of this occlusion, not all trunks were represented with the correct number of points, which is particularly noticeable in thinner trunks. This is illustrated in Figure 8, where for the 19I sampling point, 12 trunks are clearly visible in the photogrammetric point cloud compared to the MLS-derived trunks, while 7 thinner trunks are not. This can be improved by increasing the point cloud filtering height from 1.1–1.5 m to a 1–5 m strip.
A clear advantage of this method, compared to the others, is that it requires the least financial investment and can now be performed even with a simple hobby drone. Due to the geometric resolution of the images, measurements with an accuracy of 2–3 cm can be made. Only one person is required to carry out the survey. A disadvantage is that it is highly sensitive to weather conditions. The lighting is also a critical factor: diffuse light is ideal. Adequate contrast is necessary (e.g., dark trunk and light ground cover, or illuminated trunk and dark ground cover) for automatic identification. Another drawback is that not all trunks are visible in the case of dense branching structures, and in coniferous forests, the method is nearly unusable. Since successful image matching requires high (at least 80%) overlap, this significantly reduces time efficiency. With a 20 min flight, approximately 10 hectares can be surveyed. However, if image matching (and thus automatic trunk position recognition) is not the goal, much lower overlap is sufficient, allowing at least five times larger areas to be surveyed in the same amount of time.

3.4. UAV-Borne Laser Scanning

The point cloud produced by UAV laser scanning, with points filtered at heights above 1.1–1.5 m, clearly outlines the trunks. The positions of the points obtained match the positions of the points in the same height band from the TLS survey, with trunk positions nearly perfectly aligned. Unfortunately, the point dispersion is around 8 cm, which prevents fitting a circle or cylinder to the trunk points. Considering the sensor’s specifications, the observed high dispersion is not entirely surprising. To mitigate this variability, the sensor was tilted from the conventional nadir position to capture more trunk points; however, this approach proved ineffective. We suppose that by significantly reducing the flight altitude, possibly even halving it, the points around the trunk may become more defined and less scattered. This assumption is based on the premise that a lower flight altitude, and thus a reduced distance between the sensor and the target object, should lead to decreased point dispersion, thereby making accurate fitting more achievable. However, with a relative flight altitude of 100 m, it seems this is not feasible (Figure 9).
The advantage of this method is that thanks to RTK (or PPK) technology, accurate trunk positions can be obtained. From the point cloud, height and crown dimensions can also be extracted. The measurement method is relatively fast, as with a 20 min flight, up to 100 hectares can be surveyed, and only one person is required to conduct the survey. The disadvantage is the high investment cost of the ALS sensor. Lower flight altitude might be required for measuring diameters, but this significantly reduces the size of the area that can be surveyed in a single flight.

3.5. Aerial Laser Scanning with Light Aircraft

The comparison of the crowns detected from the aerial laser-scanned point cloud with the trunks obtained from the MLS yields good results. For example, at the 19I sampling point, all trunk points, except for two thinner ones, could be paired with a crown. The ALS method detected four additional crowns, which actually belong to existing trunks, meaning that in these cases, one trunk corresponds to two crown segments (Figure 10). Generally, the recognition accuracy is around 85%.
Table 2 shows a comparison of the DBH (diameter at breast height) and height values calculated from the MLS and ALS point clouds at each sampling point, along with the RMSE of the ALS-derived values relative to the MLS-based measurements. It can be seen that the algorithm processing the ALS point cloud detected more trunks everywhere, meaning in several cases, a crown was divided into at least two parts. This is common in deciduous stands, where branching could occur closer to the ground, resulting in a two-part crown.
Regarding DBH, it is observed that the smaller the number of trunks in the sample area, the greater the discrepancy between the two methods. Since the ALS trunk diameter is calculated using regression functions based on the crown and height, it is recommended to redefine this value for every area. It is important to mention that the trunk count obtained from the MLS point cloud is the closest to the real value, and for DBH, in all four areas, the deviation from the reference value is less than 1 cm. In terms of height, the data obtained from ALS shows smaller average values compared to those from MLS. This is due to the fact that the algorithm processing the ALS point cloud generates the CHM using a smoothing function based on a quartic kernel; furthermore, the measurement took place one and a half years earlier than the MLS survey.
One of the main advantages of small aircraft-based aerial laser scanning is the ability to capture large areas wall-to-wall, with a daily capacity of up to 50,000 hectares, making it highly cost-effective. Furthermore, both DEM (Digital Elevation Model) and CHM (Canopy Height Model) can be generated from the resulting point cloud, with the latter also being segmentable using voxel aggregation. The disadvantage is that it is only economical to apply on large areas, and in younger and middle-aged stands, the trunks are less visible. Therefore, diameter cannot be directly extracted from the point cloud but must be derived indirectly, using crown size and height.

3.6. Comparison of the Results

After the discussion of every method’s advantages and disadvantages, we summarized the results in Table 3. For sample points 19I and 19J, the MLS method produced near-perfect results. In contrast, on the other two sites, fewer individual trees were detected compared to the actual number. This discrepancy is likely due to thinner trunks and a dense, low-lying branch and shrub layer. The other methods also yielded their best performance in the first two sample areas. Notably, the trunk detection rate and the average DBH error achieved from the photogrammetric point cloud clearly exceeded our expectations.
Tree height estimation showed notably high RMSE values, which can be attributed to the subjective errors associated with field-based height measurements—locating the treetop in the field is often challenging. As a result, these reference measurements yielded systematically lower values compared to all the other methods. Although they are considered reference values, height estimation is presumably more accurate when derived from the generated models. Therefore, in Table 3, we used the MLS-derived heights as the basis for RMSE calculation, and the heights obtained by the other three methods were compared against them. The resulting low RMSE values clearly indicate that the four methods yield similar results.
In terms of correlation with field data, the DBH values extracted from MLS show a clearly strong relationship (R2 = 0.96). A similarly high correlation was observed for DBH values derived from the photogrammetric point cloud (R2 = 0.94). In contrast, the correlation for the ALS-based DBH values is significantly lower (R2 = 0.55). Regarding tree height, when compared to the MLS-derived values, the photogrammetric and ALS-based heights showed strong correlations (R2 = 0.88), while the correlation for ULS-derived heights was slightly lower (R2 = 0.84).
For a more detailed comparison of the methods, Table 4 summarizes the detection rates and measured parameters for each approach, alongside the corresponding reference values for sampling point 19J. The average values presented here may differ from those shown in Table 2, as the latter includes averages calculated from all the detected trees, whereas the current table (as well as Table 3) only includes values associated with matched reference trees. Thus, overdetections and false positives have been excluded. The slight overestimation of diameters derived from MLS point clouds reported by Wardius and Hein [6] was not observed in our study; even when considering the average across all the sample areas, the results closely matched the reference values. Likewise, no significant underestimation trend was observed for tree heights using this method. Major underestimation was only apparent in the ALS-derived data, which, as discussed in Section 3.5, can be attributed to specific methodological factors and observation time difference.

4. Conclusions

In our article, we examined and compared five different survey methods (classic dendrometric field survey, MLS, ULS, UAV-photogrammetry, and ALS) on four sample areas, each 500 m2 in size, with a total of 133 trees. These methods can be used to determine forest stand parameters, and through these, the above-ground carbon stock can be estimated. Among the methods, mobile laser scanning (MLS) proved to be one of the most efficient and accurate. With MLS, both circular and strip sampling, as well as full-area surveys, can be conducted; however, the latter is quite time-consuming (10 min per hectare). This measurement can be effectively complemented by wall-to-wall airborne laser scanning, which is the most cost-effective per unit area and provides homogeneous, automatically processable data from large areas.
On the other hand, the processing of ULS-based point clouds did not yield the expected results. Unfortunately, the 100 m relative flight height is not able to adequately represent the geometry of the tree trunks, even with a point density of around 400 points/m2. This is due to the sensor’s accuracy, which becomes significant at 100 m. It is highly likely that lowering the flight height and speed would improve it, but this would significantly increase survey time and reduce area coverage.
Photogrammetric surveying and processing stand out due to their lower costs compared to the above methods, but they are much more sensitive and capable of detecting fewer stems. However, the accuracy of the method increases as the forest stand becomes older and sparser. By improving the segmentation and classification of trunks in images and advancing automated processing, relatively good temporal efficiency can be achieved with this method as well.
Based on the results, we consider the combination of MLS and ALS to be the most effective method for determining above-ground carbon stock. We are planning our future research on carbon stock determination with this conclusion. Using the combined method, the production of tree and forest stand parameters, complemented by tree species information obtained from satellite image classification, allows for the estimation of above-ground carbon stock with adequate accuracy. Assessing and monitoring this stock over time is essential for forest management planning and for emphasizing the role of forests in mitigating climate change.

Author Contributions

Conceptualization, B.H. and K.C.; methodology, B.S. and K.C.; software, B.S. and K.C.; validation, B.S. and K.C.; formal analysis, B.S.; investigation, B.S.; resources, D.M.; data curation, D.M.; writing—original draft preparation, B.S. and K.C.; writing—review and editing, B.S., K.C., B.H. and G.K.; visualization, B.S. and K.C.; supervision, B.H.; project administration, D.M.; funding acquisition, G.K. All authors have read and agreed to the published version of the manuscript.

Funding

This research is supported by the Hungarian Ministry of Agriculture, State Secretariat for Forests and Land Affairs, in the frame of the Climate Change Action Plan, in the project “Digital Sitemapping and Carbon Stock Assessment in the Forest Soils of Hungary” (EGF/365/2023).

Data Availability Statement

The data that support the findings of this study are available from the corresponding authors upon reasonable request.

Acknowledgments

We thank the colleagues who participated directly and indirectly in this research for their support, especially the forest engineering students in the field measurements, and the university colleagues in the procurement and administrative tasks.

Conflicts of Interest

The authors declare no conflicts of interest.

Abbreviations

The following abbreviations are used in this manuscript:
TLSTerrestrial Laser Scanning
MLSMobil Laser Scanning
UAVUnmanned Aerial Vehicle (drone)
ULSUAV-based Laser Scanning
SLAMSimultaneous Localization and Mapping
DBHDiameter at Breast Height
SfMStructure-from-Motion
LiDARLight Detection and Ranging
RMSERoot Mean Square Error

References

  1. Szász, B.; Heil, B.; Kovács, G.; Heilig, D.; Veperdi, G.; Mészáros, D.; Illés, G.; Czimber, K. Investigation of the Relationship between Topographic and Forest Stand Characteristics Using Aerial Laser Scanning and Field Survey Data. Forests 2024, 15, 1546. [Google Scholar] [CrossRef]
  2. Czimber, K.; Szász, B.; Ács, N.; Heilig, D.; Illés, G.; Mészáros, D.; Veperdi, G.; Heil, B.; Kovács, G. Estimation of the Total Carbon Stock of Dudles Forest Based on Satellite Imagery, Airborne Laser Scanning, and Field Surveys. Forests 2025, 16, 512. [Google Scholar] [CrossRef]
  3. Coops, N.C.; Irwin, L.A.; Seely, H.S.; Hardy, S.J. Advances in Laser Scanning to Assess Carbon in Forests: From Ground-Based to Space-Based Sensors. Curr. For. Rep. 2025, 11, 11. [Google Scholar] [CrossRef]
  4. Conti, A.; Pagliaricci, G.; Bonora, V.; Tucci, G. A comparison between Terrestrial Laser Scanning and Hand-held Mobile Mapping for the Documentation of Built Heritage. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. 2024, 48, 141–147. [Google Scholar] [CrossRef]
  5. Liang, X.; Hyyppä, J.; Kaartinen, H.; Lehtomäki, M.; Pyörälä, J.; Pfeifer, N.; Holopainen, M.; Brolly, G.; Francesco, P.; Hackenberg, J.; et al. International benchmarking of terrestrial laser scanning approaches for forest inventories. ISPRS J. Photogramm. Remote Sens. 2018, 144, 137–179. [Google Scholar] [CrossRef]
  6. Wardius, Y.; Hein, S. Terrestrial laser scanning vs. manual methods for assessing complex forest stand structure: A comparative analysis on plenter forests. Eur. J. Forest Res. 2024, 143, 635–649. [Google Scholar] [CrossRef]
  7. Brolly, G.; Király, G.; Lehtomäki, M.; Liang, X. Voxel-Based Automatic Tree Detection and Parameter Retrieval from Terrestrial Laser Scans for Plot-Wise Forest Inventory. Remote Sens. 2021, 13, 542. [Google Scholar] [CrossRef]
  8. Brolly, G.; Király, G. Means of Terrestrial Laser Scanning. Acta Silv. Lignaria Hung. 2009, 5, 119–130. [Google Scholar] [CrossRef]
  9. Brolly, G.; Király, G.; Czimber, K. Mapping Forest Regeneration from Terrestrial Laser Scans. Acta Silv. Lignaria Hung. 2013, 9, 135–146. [Google Scholar] [CrossRef]
  10. Compeán-Aguirre, J.L.; López-Serrano, P.M.; Silván-Cárdenas, J.L.; Martínez-García-Moreno, C.A.; Vega-Nieva, D.J.; Corral-Rivas, J.J.; Pompa-García, M. Evaluation of Two-Dimensional DBH Estimation Algorithms Using TLS. Forests 2024, 15, 1964. [Google Scholar] [CrossRef]
  11. You, L.; Sun, Y.; Liu, Y.; Chang, X.; Jiang, J.; Feng, Y.; Song, X. Tree Skeletonization with DBSCAN Clustering Using Terrestrial Laser Scanning Data. Forests 2023, 14, 1525. [Google Scholar] [CrossRef]
  12. Capalb, F.; Apostol, B.; Lorenț, A.; Petrila, M.; Marcu, C.; Badea, N.O. Integration of Terrestrial Laser Scanning and field measurements data for tree stem volume estimation: Exploring parametric and non-parametric modeling approaches. Ann. For. Res. 2024, 67, 77–94. [Google Scholar] [CrossRef]
  13. Yang, W.; Wilkes, P.; Vicari, M.B.; Hand, K.; Calders, K.; Disney, M. Treegraph: Tree Architecture from Terrestrial Laser Scanning Point Clouds. Remote Sens. Ecol. Conserv. 2024, 10, 755–774. [Google Scholar] [CrossRef]
  14. Åkerblom, M.; Raumonen, P.; Mäkipää, R.; Kaasalainen, M. Automatic tree species recognition with quantitative structure models. Remote Sens. Environ. 2017, 191, 1–12. [Google Scholar] [CrossRef]
  15. Brede, B.; Calders, K.; Lau, A.; Raumonen, P.; Bartholomeus, H.M.; Herold, M.; Kooistra, L. Non-destructive tree volume estimation through quantitative structure modelling: Comparing UAV laser scanning with terrestrial LIDAR. Remote Sens. Environ. 2019, 233, 111355. [Google Scholar] [CrossRef]
  16. Markku, Å.; Raumonen, P.; Kaasalainen, M.; Casella, E. Analysis of Geometric Primitives in Quantitative Structure Models of Tree Stems. Remote Sens. 2015, 7, 4581–4603. [Google Scholar] [CrossRef]
  17. Meng, Y.; Dong, X.; Han, K.; Liu, H.; Qu, H.; Gao, T. Classification of Tree Species Using Point Cloud Features from Terrestrial Laser Scanning. Forests 2024, 15, 2110. [Google Scholar] [CrossRef]
  18. Chen, P.; Zhao, X.; Zeng, L.; Liu, L.; Liu, S.; Sun, L.; Li, Z.; Chen, H.; Liu, G.; Qiao, Z.; et al. A Review of Research on SLAM Technology Based on the Fusion of LiDAR and Vision. Sensors 2025, 25, 1447. [Google Scholar] [CrossRef]
  19. Pan, D.; Shao, J.; Zhang, S.; Zhang, S.; Chang, B.; Xiong, H.; Zhang, W. SLAM-Based Forest Plot Mapping by Integrating IMU and Self-Calibrated Dual 3-D Laser Scanners. IEEE Trans. Geosci. Remote Sens. 2023, 61, 1–13. [Google Scholar] [CrossRef]
  20. Spadavecchia, C.; Belcore, E.; Grasso, N.; Piras, M. A fully automatic forest paramteres extraction at single-tree level: A comparison of MLS and TLS applications. Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. 2023, 48, 457–463. [Google Scholar] [CrossRef]
  21. Krisanski, S.; Taskhiri, M.S.; Gonzalez Aracil, S.; Herries, D.; Muneri, A.; Gurung, M.B.; Montgomery, J.; Turner, P. Forest Structural Complexity Tool—An Open Source, Fully-Automated Tool for Measuring Forest Point Clouds. Remote Sens. 2021, 13, 4677. [Google Scholar] [CrossRef]
  22. Faitli, T.; Hyyppä, E.; Hyyti, H.; Hakala, T.; Kaartinen, H.; Kukko, A.; Muhojoki, J.; Hyyppä, J. Integration of a Mobile Laser Scanning System with a Forest Harvester for Accurate Localization and Tree Stem Measurements. Remote Sens. 2024, 16, 3292. [Google Scholar] [CrossRef]
  23. Woltring, H.J.; Huiskes, R. Stereophotogrammetry. In Biomechanics of Human Movement: Applications in Rehabilitation, Sport and Ergonomics; Bertec: Columbus, OH, USA, 1990; pp. 108–127. Available online: https://pure.tue.nl/ws/portalfiles/portal/2808839/585701.pdf (accessed on 6 April 2025).
  24. Karami, E.; Prasad, S.; Shehata, M. Image Matching Using SIFT, SURF, BRIEF and ORB: Performance Comparison for Distorted Images. In Proceedings of the 2015 Newfoundland Electrical and Computer Engineering Conference, St. Johns, NS, Canada, 3–6 May 2015. [Google Scholar] [CrossRef]
  25. Hirschmuller, H. Accurate and efficient stereo processing by semi-global matching and mutual information. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. 2005, 2, 807–814. [Google Scholar] [CrossRef]
  26. Furukawa, Y.; Hernández, C. Multi-View Stereo: A Tutorial. Found. Trends Comput. Graph. Vis. 2015, 9, 1–148. [Google Scholar] [CrossRef]
  27. Westoby, M.J.; Brasington, J.; Glasser, N.F.; Hambrey, M.J.; Reynolds, J.M. ‘Structure-from-Motion’ photogrammetry: A low-cost, effective tool for geoscience applications. Geomorphology 2012, 179, 300–314. [Google Scholar] [CrossRef]
  28. Iglhaut, J.; Cabo, C.; Puliti, S.; Piermattei, L.; O’Connor, J.; Rosette, J. Structure from Motion Photogrammetry in Forestry: A Review. Curr. For. Rep. 2019, 5, 155–168. [Google Scholar] [CrossRef]
  29. Verhoeven, G.; Doneus, M.; Briese, C.; Vermeulen, F. Mapping by matching: A computer vision-based approach to fast and accurate georeferencing of archaeological aerial photographs. J. Archaeol. Sci. 2012, 39, 2060–2070. [Google Scholar] [CrossRef]
  30. Mielcarek, M.; Kamińska, A.; Stereńczak, K. Digital Aerial Photogrammetry (DAP) and Airborne Laser Scanning (ALS) as Sources of Information about Tree Height: Comparisons of the Accuracy of Remote Sensing Methods for Tree Height Estimation. Remote Sens. 2020, 12, 1808. [Google Scholar] [CrossRef]
  31. Krause, S.; Sanders, T.G.M.; Mund, J.-P.; Greve, K. UAV-Based Photogrammetric Tree Height Measurement for Intensive Forest Monitoring. Remote Sens. 2019, 11, 758. [Google Scholar] [CrossRef]
  32. Panagiotidis, D.; Abdollahnejad, A.; Surový, P.; Chiteculo, V. Determining tree height and crown diameter from high-resolution UAV imagery. Int. J. Remote Sens. 2016, 38, 2392–2410. [Google Scholar] [CrossRef]
  33. Bohlin, J.; Wallerman, J.; Fransson, J.E.S. Forest variable estimation using photogrammetric matching of digital aerial images in combination with a high-resolution DEM. Scand. J. For. Res. 2012, 27, 692–699. [Google Scholar] [CrossRef]
  34. Balenović, I.; Marjanović, H.; Indir, K.; Vuletić, D.; Ostrogovic, Z.M.; Benko, M. Estimation of the stands’ arithmetic mean diameter using manual method of digital photogrammetry. Period. Biol. 2013, 115, 399–407. [Google Scholar]
  35. Cao, L.; Liu, H.; Fu, X.; Zhang, Z.; Shen, X.; Ruan, H. Comparison of UAV LiDAR and Digital Aerial Photogrammetry Point Clouds for Estimating Forest Structural Attributes in Subtropical Planted Forests. Forests 2019, 10, 145. [Google Scholar] [CrossRef]
  36. Mao, Z.; Lu, Z.; Wu, Y.; Deng, L. DBH Estimation for Individual Tree: Two-Dimensional Images or Three-Dimensional Point Clouds? Remote Sens. 2023, 15, 4116. [Google Scholar] [CrossRef]
  37. Bruggisser, M.; Hollaus, M.; Kükenbrink, D.; Pfeifer, N. Comparison of forest structure metrics derived from UAV LiDAR and ALS data. ISPRS Ann. Photogramm. Remote Sens. Spat. Inf. Sci. 2019, 4, 325–332. [Google Scholar] [CrossRef]
  38. Kraus, K.; Pfeifer, N. A new method for surface reconstruction from laser scanner data. Int. Arch. Photogramm. Remote Sens. 1997, 32, 80–86. [Google Scholar]
  39. Li, F.; Zhu, H.; Luo, Z.; Shen, H.; Li, L. An Adaptive Surface Interpolation Filter Using Cloth Simulation and Relief Amplitude for Airborne Laser Scanning Data. Remote Sens. 2021, 13, 2938. [Google Scholar] [CrossRef]
  40. Cai, S.; Zhang, W.; Liang, X.; Wan, P.; Qi, J.; Yu, S.; Yan, G.; Shao, J. Filtering Airborne LiDAR Data Through Complementary Cloth Simulation and Progressive TIN Densification Filters. Remote Sens. 2019, 11, 1037. [Google Scholar] [CrossRef]
  41. Dong, Y.; Cui, X.; Zhang, L.; Ai, H. An Improved Progressive TIN Densification Filtering Method Considering the Density and Standard Variance of Point Clouds. ISPRS Int. J. Geo-Inf. 2018, 7, 409. [Google Scholar] [CrossRef]
  42. Axelsson, P. DEM Generation from Laser Scanner Data Using Adaptive TIN Models. ISPRS—Int. Arch. Photogramm. Remote Sens. Spat. Inf. Sci. 2000, 33, 110–117. [Google Scholar]
  43. Pingel, T.J.; Clarke, K.C.; McBride, W.A. An Improved Simple Morphological Filter for the Terrain Classification of AirborneLIDAR Data. ISPRS J. Photogramm. Remote Sens. 2013, 77, 21–30. [Google Scholar] [CrossRef]
  44. Zhang, K.; Chen, S.-C.; Whitman, D.; Shyu, M.-L.; Yan, J.; Zhang, C. A Progressive Morphological Filter for Removing Nonground Measurements from Airborne LIDAR Data. IEEE Trans. Geosci. Remote Sens. 2003, 41, 872–882. [Google Scholar] [CrossRef]
  45. Lee, H.; Slatton, K.C.; Roth, B.E.; Cropper JR, W.P. Adaptive Clustering of Airborne LiDAR Data to Segment Individual Tree Crowns in Managed Pine Forests. Int. J. Remote Sens. 2010, 31, 117–139. [Google Scholar] [CrossRef]
  46. Ayrey, E.; Fraver, S.; Kershaw, J.A., Jr.; Kenefic, L.S.; Hayes, D.; Weiskittel, A.R.; Roth, B.E. Layer Stacking: A Novel Algorithm for Individual Forest Tree Segmentation from LiDAR Point Clouds. Can. J. Remote Sens. 2017, 43, 16–27. [Google Scholar] [CrossRef]
  47. Strîmbu, V.F.; Strîmbu, B.M. A Graph-Based Segmentation Algorithm for Tree Crown Extraction Using Airborne LiDAR Data. ISPRS J. Photogramm. Remote Sens. 2015, 104, 30–43. [Google Scholar] [CrossRef]
  48. White, J.C.; Wulder, M.A.; Buckmaster, G. Validating Estimates of Merchantable Volume from Airborne Laser Scanning (ALS)Data Using Weight Scale Data. For. Chron. 2014, 90, 378–385. [Google Scholar] [CrossRef]
  49. Oehmcke, S.; Li, L.; Revenga, J.C.; Nord-Larsen, T.; Trepekli, K.; Gieseke, F.; Igel, C. Deep Learning Based 3D Point CloudRegression for Estimating Forest Biomass. In Proceedings of the 30th International Conference on Advances in GeographicInformation Systems, Seattle, WA, USA, 1–4 November 2022; Association for Computing Machinery: New York, NY, USA, 2022; pp. 1–4. [Google Scholar]
  50. Ullah, S.; Dees, M.; Datta, P.; Adler, P.; Koch, B. Comparing Airborne Laser Scanning, and Image-Based Point Clouds by Semi-Global Matching and Enhanced Automatic Terrain Extraction to Estimate Forest Timber Volume. Forests 2017, 8, 215. [Google Scholar] [CrossRef]
  51. Sačkov, I.; Bucha, T.; Király, G.; Brolly, G.; Raši, R. Individual tree and crown identification in the Danube floodplain forests based on airborne laser scanning data. In Proceedings of the EARSeL 34th Symposium Proceedings, Warsaw, Poland, 16–20 June 2014; pp. 16–20. [Google Scholar] [CrossRef]
  52. Kuželka, K.; Slavík, M.; Surový, P. Very High Density Point Clouds from UAV Laser Scanning for Automatic Tree Stem Detection and Direct Diameter Measurement. Remote Sens. 2020, 12, 1236. [Google Scholar] [CrossRef]
  53. Hao, Y.; Widagdo, F.R.A.; Liu, X.; Quan, Y.; Dong, L.; Li, F. Individual Tree Diameter Estimation in Small-Scale Forest Inventory Using UAV Laser Scanning. Remote Sens. 2021, 13, 24. [Google Scholar] [CrossRef]
  54. Chisholm, R.A.; Rodríguez-Ronderos, M.E.; Lin, F. Estimating Tree Diameters from an Autonomous Below-Canopy UAV with Mounted LiDAR. Remote Sens. 2021, 13, 2576. [Google Scholar] [CrossRef]
Remotesensing 17 02173 g001
Figure 1. Location of the four selected sampling plots within the Dudles Forest.
Figure 1. Location of the four selected sampling plots within the Dudles Forest.
Remotesensing 17 02173 g001
Remotesensing 17 02173 g002
Figure 2. Flowchart of the five presented methods.
Figure 2. Flowchart of the five presented methods.
Remotesensing 17 02173 g002
Remotesensing 17 02173 g003
Figure 3. Point cloud processing and stem detection on dotXpert software.
Figure 3. Point cloud processing and stem detection on dotXpert software.
Remotesensing 17 02173 g003
Remotesensing 17 02173 g004
Figure 4. Measurement of DBH on the DEM+1.3m-based orthophoto.
Figure 4. Measurement of DBH on the DEM+1.3m-based orthophoto.
Remotesensing 17 02173 g004
Remotesensing 17 02173 g005
Figure 5. Tree density by DBH in each sampling point.
Figure 5. Tree density by DBH in each sampling point.
Remotesensing 17 02173 g005
Remotesensing 17 02173 g006
Figure 6. Approximate (field survey) and MLS-derived stem positions (large left) and circle fitting (small right) at 19I sampling point.
Figure 6. Approximate (field survey) and MLS-derived stem positions (large left) and circle fitting (small right) at 19I sampling point.
Remotesensing 17 02173 g006
Remotesensing 17 02173 g007
Figure 7. Point cloud derived from image matching at 19J sample plot.
Figure 7. Point cloud derived from image matching at 19J sample plot.
Remotesensing 17 02173 g007
Remotesensing 17 02173 g008
Figure 8. Stem positions derived from MLS and photogrammetric point cloud with the narrower filtering strip.
Figure 8. Stem positions derived from MLS and photogrammetric point cloud with the narrower filtering strip.
Remotesensing 17 02173 g008
Remotesensing 17 02173 g009
Figure 9. Stem positions derived from MLS and ULS survey (large left) and the dispersion of ULS points (small right).
Figure 9. Stem positions derived from MLS and ULS survey (large left) and the dispersion of ULS points (small right).
Remotesensing 17 02173 g009
Remotesensing 17 02173 g010
Figure 10. Comparison of MLS-derived stem positions and ALS-derived crown segments.
Figure 10. Comparison of MLS-derived stem positions and ALS-derived crown segments.
Remotesensing 17 02173 g010
Table 1. Field survey results (reference values).
Table 1. Field survey results (reference values).
Sampling PointNumber of StemsAverage DBH [cm]Average Height [m]Total Volume [m3]
19I2725.417.715.3
19J2129.220.116.5
20I4919.712.19.4
20J3618.515.910.8
Table 2. Comparison of data derived from MLS and ALS surveys.
Table 2. Comparison of data derived from MLS and ALS surveys.
Sampling PointNumber of StemsAverage DBH [cm]Average Height [m]
MLSALSMLSALSRMSEMLSALSRMSE
19I273225.020.65.520.318.81.3
19J213029.223.05.522.520.71.6
20I425418.718.25.214.013.81.5
20J364319.419.47.818.117.63.0
Table 3. Comparison of hit rate and RMSE values for each sampling point.
Table 3. Comparison of hit rate and RMSE values for each sampling point.
Sampling PointHit Rate [%]DBH RMSE [cm]Height RMSE [m]
MLSULSPhot.ALSMLSPhot.ALSULSPhot.ALS
19I10096.284.61001.02.35.40.70.71.3
19J10066.781.095.21.02.65.50.50.71.6
20I83.769.467.373.52.72.76.03.01.71.5
20J83.338.950.066.71.52.67.73.43.43.0
Merged89.465.968.280.31.92.56.22.21.91.9
Table 4. Comparison of data derived from each method for sampling point 19J.
Table 4. Comparison of data derived from each method for sampling point 19J.
Tree NumberField SurveyMLSULSPhotogrammetryALS
d [cm]h [m]d [cm]h [m]h [m]d [cm]h [m]d [cm]h [m]
145.821.845.121.722.14222.336.520.9
235.922.236.221.821.83322.127.221.1
327.122.026.721.9 20.820.6
424.319.024.322.622.72223.130.218.6
527.018.026.721.621.6232223.820.8
637.320.037.223.723.83424.137.723.2
732.221.131.123.322.22822.836.021.2
818.517.618.023.8 1924.4
930.220.829.923.424.5 37.723.5
1032.222.229.423.724.52824.425.023.2
1131.322.830.321.421.4322331.219.6
1229.121.029.223.622.92723.421.321.8
1328.420.929.921.421.52821.526.420.4
1429.322.229.222.021.8 21.519.5
1529.223.629.821.3 3122.726.420.6
1628.318.630.022.6 2722.424.721.5
1719.012.220.221 17.419.7
1823.518.423.523.623.32123.619.521.4
1924.517.623.522.519.92522.423.121.4
2025.718.025.422.0 2423.222.319.6
2134.122.135.023.924.13624.227.023
Mean29.220.129.122.522.728.223.026.821.1
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Szász, B.; Heil, B.; Kovács, G.; Mészáros, D.; Czimber, K. Comparison of Advanced Terrestrial and Aerial Remote Sensing Methods for Above-Ground Carbon Stock Estimation—A Comparative Case Study for a Hungarian Temperate Forest. Remote Sens. 2025, 17, 2173. https://doi.org/10.3390/rs17132173

AMA Style

Szász B, Heil B, Kovács G, Mészáros D, Czimber K. Comparison of Advanced Terrestrial and Aerial Remote Sensing Methods for Above-Ground Carbon Stock Estimation—A Comparative Case Study for a Hungarian Temperate Forest. Remote Sensing. 2025; 17(13):2173. https://doi.org/10.3390/rs17132173

Chicago/Turabian Style

Szász, Botond, Bálint Heil, Gábor Kovács, Diána Mészáros, and Kornél Czimber. 2025. "Comparison of Advanced Terrestrial and Aerial Remote Sensing Methods for Above-Ground Carbon Stock Estimation—A Comparative Case Study for a Hungarian Temperate Forest" Remote Sensing 17, no. 13: 2173. https://doi.org/10.3390/rs17132173

APA Style

Szász, B., Heil, B., Kovács, G., Mészáros, D., & Czimber, K. (2025). Comparison of Advanced Terrestrial and Aerial Remote Sensing Methods for Above-Ground Carbon Stock Estimation—A Comparative Case Study for a Hungarian Temperate Forest. Remote Sensing, 17(13), 2173. https://doi.org/10.3390/rs17132173

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop