1. Introduction
The ocean has recently emerged as a critical domain for addressing global issues such as climate change mitigation, carbon neutrality, ecosystem conservation and restoration, and maritime domain awareness (Gattuso et al., 2018; De Pryck and Boettcher, 2024; Şengül et al., 2023; Ward et al., 2022). In line with these shifts, the purpose of oceanographic surveys has expanded beyond traditional chart production to encompass investigations aimed at understanding the marine environment across diverse spatial and temporal scales, as well as assessing regional environmental changes and the impacts of human activities (Bastos et al., 2016; Lin and Yang, 2020). Consequently, the demand for marine surveys is increasing, and survey environments such as polar regions and deep-sea areas are becoming more complex, further emphasizing the need for long-term and repeated observations (Stanghellini et al., 2022; Lin et al., 2023; Cavanaugh et al., 2025).
Traditional marine surveys have primarily relied on manned platforms such as merchant vessels and research ships. However, their dependence on human resources, high operational costs, and limitations imposed by weather and navigational conditions restrict the spatial and temporal expansion of survey activities. As a result, large portions of the ocean remain insufficiently explored (Liu et al., 2016; Smith et al., 2019; Kubacka et al., 2022). These constraints make it increasingly difficult for conventional manned survey methods to meet the growing demand for observations and the challenges posed by increasingly complex survey environments. Consequently, unmanned platforms survey technologies are rapidly emerging as a viable alternative. Advances in artificial intelligence (AI) based data processing, autonomous navigation, high precision sensor fusion, real time communication, and remote control technologies have significantly enhanced the performance and operational efficiency of various unmanned platforms, including unmanned aerial vehicle (UAV), unmanned surface vehicle (USV), and autonomous underwater vehicle (AUV) (Bae and Hong, 2023).
With these technological advancements, countries around the world are actively adopting unmanned platforms to improve the efficiency and safety of marine surveys. Numerous field demonstrations have confirmed that unmanned survey methods offer higher operational efficiency and greater stability compared to traditional manned approaches (Leder and Duplančić Leder, 2018). Yaney-Keller et al. (2021) reported that in a survey of sea turtle nesting sites around Heron Island, UAVs achieved more than twice the detection efficiency per unit time compared to manned vessels.
In addition, UAV demonstrated superior operational performance, including improved individual identification accuracy, enhanced accessibility to hazardous or narrow waterways, and reduced operational costs. Angliss et al. (2018) conducted concurrent surveys using manned aircraft and UAVs in waters near the Arctic and found that UAVs captured comparable areas at approximately five times higher spatial resolution. They also demonstrated superior operational flexibility and cost efficiency as UAVs could autonomously take off and land with minimal personnel. Furthermore, UAVs exhibited high accuracy in detailed ecological monitoring tasks, including species identification, group size estimation, and detection of juvenile individuals. Schmidt et al. (2023) reported that during operational trials involving the National Oceanic and Atmospheric Administration (NOAA)’s Thomas Jefferson and the Drix-12 USV, continuous overnight operations enabled nearly double the survey coverage compared to traditional NOAA manned vessels.
Moreover, fuel consumption per nautical mile was reduced by approximately 86%, demonstrating significantly superior survey efficiency and operational performance relative to conventional manned survey methods. According to Saildrone (2025), saildrone and meta conducted a 26 day deep ocean seafloor mapping mission in the North Atlantic using a 20 m class USV, surveying approximately 4,500 km of seafloor with a multibeam echosounder (MBES) capable of mapping depths up to 5,500 m with a 10 km swath. The results demonstrated that the USV achieved bathymetric accuracy comparable to conventional manned cable route surveys, while also providing substantial operational advantages including fully autonomous 24 hour operations, an estimated reduction of 243 tons of carbon dioxide (CO₂) emissions, and minimized weather and sea state constraints through real time satellite communication.
According to Terradepth (2023), seafloor mapping conducted with an AUV in rough sea conditions with wave heights exceeding 3 m demonstrated approximately 55% faster survey completion compared to traditional manned survey vessels and diver-based methods. The unmanned approach also offered lower operational costs, higher data precision, and improved safety and flexibility by eliminating the need for direct diver deployment and reducing weather-related operational constraints. Eiler et al. (2013) conducted concurrent marine fauna surveys using an AUV and a manned vessel in waters near Alaska. Their results showed that while both platforms exhibited similar detection efficiency at a depth of 20 m, the unmanned underwater vehicle achieved approximately 1.6 times higher detection rates at 100 m, and at 500 m, it uniquely detected sablefish that the manned vessel failed to observe. For king crab, the AUV also demonstrated an average of 31% higher detection performance. These findings highlight the superior efficiency of unmanned underwater platforms in high risk, deep water environments in terms of time, cost, and detection accuracy.
Various international demonstrations have shown that unmanned platforms can replace or complement traditional manned survey methods. In line with this trend, the use of unmanned systems for seafloor mapping, shoreline surveys, and oceanographic observations is gradually expanding in Korea as well. However, to enhance the applicability of unmanned survey data, ensuring their reliability is essential, which requires rigorous validation of unmanned survey performance. In particular, bathymetric data are a fundamental component for navigational safety, route design, and nautical chart production, and therefore demand more stringent accuracy verification compared to other survey parameters (International Hydrographic Organization, 2022).
Accordingly, it is necessary to compare the bathymetric results obtained from unmanned surveys with those from manned survey methods that have been verified to meet international accuracy standards. Constantinoiu et al. (2023) compared MBES bathymetric data acquired by a manned survey vessel and a USV in the coastal waters between the Sesimbra and Sines harbors in Portugal using a raster-based analysis. Confirming the applicability of USV in coastal environments and the efficiency of the combined uncertainty and bathymetry estimator (CUBE) algorithm for rapid surface generation.
Additionally, Park et al. (2024) compared MBES bathymetric survey results collected by a manned vessel and a USV along mainlines and checklines near Yeondo, Gunsan, South Korea. Demonstrating that USV-based surveys achieve a comparable level of accuracy and spatial consistency to traditional manned survey methods. Although these previous studies demonstrated the applicability of USV-based bathymetric surveys in coastal environments, they were limited in terms of geographic scope and evaluation methodology. Therefore, to ensure the reliability and broader applicability of unmanned hydrographic surveys, systematic research that incorporates diverse environmental conditions and comprehensive validation techniques is required.
This study aims to address these limitations by comparing and analyzing MBES-based bathymetric survey results acquired by a manned vessel and a USV in critical navigation areas, including harbor environments where navigation safety is essential and island regions where access by manned vessels is restricted. In addition to conventional crossline quality control analysis and depth difference analysis, quantitative evaluation methods such as shallow water area analysis were applied to comprehensively assess performance differences between the two platforms. Through these analyses, the reliability of USV-based hydrographic survey data was verified, demonstrating the potential applicability of USV across diverse marine environments.
2. Materials
2.1. Study Area
In this study, MBES based bathymetric survey data collected from two 1:5,000 scale map sheets on the eastern side of Yongho Harbor in Nam-gu, Busan, and three 1:5,000 scale map sheets around Dokdo in Ulleung-eup, Ulleung-gun, Gyeongsangbuk-do, were used to compare the survey performance of manned vessels and USV (Fig. 1). The Suyeong bay area, where the Yongho Harbor is located, gradually deepens with increasing distance from the coast, reaching depths greater than 30 m near its opening toward the Korea Strait (Lee et al., 2009). The study site encompasses an area within approximately 2.5 km east of the Yongho Harbor, with maximum depths of about 19.5 m. The manned survey was conducted in April 2023, and the USV-based survey was carried out in September 2023 using MBES.
Fig. 1. Bathymetric survey tracks acquired by manned and unmanned platforms in the study areas: (a) Dokdo and (b) Yongho Harbor.
Dokdo consists of the East Islet, the West Islet, and several surrounding small islets. In the nearshore waters shallower than approximately 50 m, the irregular topographic structure of the islands extends continuously into the seafloor, creating complex geomorphological characteristics (Kim et al., 2023; Lee et al., 2017). This irregular morphology persists to depths of around 100 m. Between 100 and 200 m, the seabed forms a gentle slope, but beyond roughly 180 m, the terrain transitions into a steep, cliff-like structure with rapid depth changes. The spatial extent of the Dokdo study area is approximately 2.5 km to the east, 3.3 km to the west, 1.8 km to the south, and 0.63 km to the north, with a maximum depth of around 150 m. The USV surveys conducted in 2020, 2022, and the first survey of 2023 covered the entire area. In contrast, the 2021 survey covered only the western side of Dokdo, and the second survey in 2023 excluded the eastern side and some coastal sections. Meanwhile, the 2023 manned survey was conducted only in parts of the eastern and western areas and selected nearshore regions.
2.2. MBES Bathymetric Data
Bathymetric data were acquired using both a manned vessel and a USV in the Yongho Harbor. The manned vessel was equipped with an R2Sonic 2024 MBES, while the USV carried a Seabat T50-P. In the Dokdo area, the manned surveys also utilized the R2Sonic 2024, whereas the equipment mounted on the MBES varied depending on the survey year. In 2020, three USVs were deployed, each equipped with a different MBES. In 2021, two USVs were used, one equipped with an R2Sonic 2020 and the other with an R2Sonic 2024. The 2022 survey utilized the R2Sonic 2024, while both the first and second surveys in 2023 employed the R2Sonic 2020. Although the MBES models differed between platforms, their key specifications, including frequency range, sounding depth capability, swath width, and number of soundings, were sufficiently similar to allow performance comparisons under comparable survey conditions (Table 1).
Table 1. Specifications of MBES systems used in manned and unmanned surveys
The bathymetric data acquired from both the manned vessel and the USV were post-processed using CARIS HIPS & SIPS following the same workflow. The post-processing procedure included raw data conversion, navigation and attitude correction, sound velocity correction, tide correction, swath editing, and subset editing (Fig. 2). Raw data conversion involves transforming the original data files acquired in *.XTF and *.PDS formats using MBES acquisition software such as QINSy and PDS2000 into a structure compatible with processing in CARIS HIPS & SIPS (National Oceanic and Atmospheric Administration, 2014). During this process, the positional offsets between the MBSE, the global navigation satellite system (GNSS) antenna, and the vessel reference point or inertial measurement unit (IMU), as well as draft information, are entered to establish the vertical reference for the depth measurements. In addition, calibration-derived corrections for pitch, roll, yaw, and time latency are applied to minimize depth errors caused by sensor misalignment and temporal discrepancies. After raw data conversion, the analysis and correction consist of navigation correction and attitude correction.
Fig. 2. Post-processing workflow of MBES bathymetric data (Teledyne CARIS, 2016).
Navigation correction involves identifying segments where anomalous speed, distance, or heading values occur for the manned vessel or USV and removing the corresponding depth soundings. Attitude correction detects abnormal variations in roll, pitch, and heave typically caused by sharp turns or unfavorable sea conditions to identify and eliminate attitude related distortions in the raw data, thereby minimizing depth errors (Canadian Hydrographic Service, 2021). Sound velocity correction compensates for ray bending caused by variations in sound speed due to changes in temperature and salinity (Grządziel, 2021). For this purpose, a sound velocity profile (SVP) is measured in the field to obtain the vertical sound speed structure, and the acoustic ray path is recalculated to correct refraction-induced depth errors. Tide correction is performed to minimize vertical inconsistencies by harmonically decomposing tide gauge observations and adjusting depth values to a consistent vertical datum. Swath editing is performed after navigation, sound velocity, and tidal corrections, and involves reviewing the continuity between adjacent profiles to identify and remove outliers along each survey line. Subset editing represents the final stage of data cleaning, in which the processed bathymetric data are examined from multiple perspectives in both two-dimensional and three-dimensional views. Any remaining outliers are manually removed, and areas that were excessively filtered are restored to improve the overall accuracy of the final dataset. After completing all post-processing steps, the bathymetric data acquired from the manned vessel and the USV were exported in .xyz format and subsequently used for comparative analysis of survey performance over the same areas.
3. Methods
3.1. Crossline Quality Control Analysis
Errors in bathymetric surveys arise from a combination of various sources, including positioning system uncertainty, sensor location uncertainty, beam angle uncertainty, seabed echo detection uncertainty, vertical motion sensor uncertainty, vessel motion sensor uncertainty, vessel heading uncertainty, and ray-path model uncertainty (General Bathymetric Chart of the Oceans, 2019). Although these uncertainties are mitigated through post-processing, they cannot be completely eliminated. Therefore, crossline verification must be conducted to validate depth measurements obtained along the main survey lines using independent check lines (Hydrographic Surveys Division, 2025).
This crossline quality control analysis is performed by first calculating the depth differences between the mainline and checkline measurements at individual sounding points, and then generating a difference surface by converting these values into grids of identical resolution (Teledyne CARIS, 2016). The resulting fdepth deviations are evaluated against the total vertical uncertainty (TVU) standards specified in the International Hydrographic Organization (2020) S-44 guidelines to determine compliance with the appropriate hydrographic survey order (Fig. 3a). Depending on the characteristics and intended use of the survey area, hydrographic survey orders are classified into five levels: Order 2, Order 1b, Order 1a, Special Order, and Exclusive Order (Table 2). Order 2 is applied to areas deeper than 200 m where general seafloor morphology is sufficient for the intended use. Order 1b and order 1a apply to shallow waters less than 2 m deep or to depths between 50 m and 200 m, and are also used in coastal areas (2–50 m), national harbors, and recommended routes where seabed features may pose a potential navigational hazard. Special order is assigned to areas requiring a high level of navigational safety such as commercial and coastal ports, designated channels, and traffic safety control zones with depths shallower than 200 m.
Table 2. Minimum bathymetry standards for the safety of navigation hydrographic surveys
Fig. 3. Flowchart of bathymetric comparison between manned and unmanned surveys: (a) crossline quality control analysis, (b) shallow water area analysis, and (c) depth difference analysis.
Finally, an exclusive order represents the highest level of hydrographic surveying, conducted in very shallow waters where precise navigation and berthing safety must be ensured. The TVU is calculated by substituting the parameters a and b and the depth d into Eq. (1), and its allowable value varies depending on the hydrographic survey order. (Table 2). The parameter a represents the portion of the uncertainty that does not vary with depth, while b is a coefficient representing the component of uncertainty that increases as a function of depth. The variable d denotes the measured depth. The allowable TVU values vary by survey order and depth range, with the minimum required thresholds provided at the end of this section (Table 3). Through this crossline analysis procedure, the accuracy and consistency of bathymetric data acquired from each platform can be evaluated and validated (Hydrographic Surveys Division, 2025).
\(\begin{align}TVU=\sqrt{a^2+(b\times\;d)^2}\end{align}\) (1)
Table 3. Minimum TVU requirements by survey order and depth
3.2. Shallow Water Area Analysis
Shallow water areas are defined as zones where the seabed is noticeably shallower than the surrounding waters (Korea Hydrographic and Oceanographic Agency, 2022). Such areas often contain natural features (e.g., shoals, boulders, reefs, rocks, outcroppings) as well as anthropogenic features (e.g., wrecks, obstructions, piles), which pose potential navigational hazards and therefore require careful examination (Hydrographic Surveys Division, 2025). Therefore, precise bathymetric surveying of shallow water features is essential for accurately identifying dangers to navigation (DTON) and ensuring navigational safety and chart reliability (Masetti et al., 2018). Comparing the shallow water sounding performance of USV with that of manned vessels also serves as an important measure for assessing the operational applicability of USV in real survey environments. Shallow water features are generally defined as those that rise more than 1 m above the surrounding seabed in areas with depths of 0–20 m, or features that are located at depths more than 10% shallower than the surrounding area when the water depth exceeds 20 m (Hydrographic Surveys Division, 2025). Using these criteria, shallow water points are identified from the point-based bathymetric datasets of both the manned vessel and the USV, and the depth differences between the two platforms are calculated (Eq. 2).
Based on the computed depth differences, the mean error (ME), mean absolute error (MAE), standard deviation (STD), and root mean square error (RMSE) are derived to compare the survey performance of the manned vessel and the USV (Eqs. 3–6; Fig. 3b). ME represents the mean value of the depth differences, including their sign, and is used to assess the degree of bias between the two platforms (Moknatian et al., 2019). MAE is the average of the absolute errors, providing an intuitive indication of the magnitude of the actual depth differences (Willmott and Matsuura, 2005). STD describes the spread of the error distribution and is used to evaluate the precision of the survey results (Paul et al., 2017). Finally, RMSE which is calculated from the squared errors, responds more sensitively to larger deviations and provides an overall assessment of the accuracy and reliability of the bathymetric data (Chai and Draxler, 2014; U. S. Geological Survey, 2020). Values of these metrics approaching zero indicate that the shallow water depth measurements obtained by the USV are by the manned vessel (Chen et al., 2016; Erdoǧan, 2010). consistent with, and largely unbiased relative to, those collected by the manned vessel (Chen et al., 2016; Erdoǧan, 2010).
∆D = DepthManned – DepthUnmanned (2)
\(\begin{align}M E=\frac{1}{n} \sum_{i=1}^{n} \Delta D_{i}\end{align}\) (3)
\(\begin{align}M A E=\frac{1}{n} \sum_{i=1}^{n}\left|\Delta D_{i}\right|\end{align}\) (4)
\(\begin{align}S T D=\sqrt{\frac{1}{n-1} \sum_{i=1}^{n}\left(\Delta D_{i}-M E\right)^{2}}\end{align}\) (5)
\(\begin{align}R M S E=\sqrt{\frac{1}{n} \sum_{i=1}^{n} \Delta D_{i}^{2}}\end{align}\) (6)
3.3. Depth Difference Analysis
Depth difference analysis is used to compare bathymetric datasets acquired by different platforms over the same area under equivalent conditions. This method quantifies differences in raster based digital terrain model (DTM), providing an effective means of evaluating consistency in equipment performance. Slinning (2016) rasterized multibeam bathymetric datasets collected at different times, calculated cell-wise deviations, and assessed accuracy levels by comparing the results with the standards outlined by the International Hydrographic Organization (2020). Pimentel et al. (2020) compared bathymetric sidescan data with multibeam echosounder DTM acquired in the same area and demonstrated the consistency between the two sensors as well as their potential complementarity. Pasandeh et al. (2025) converted multibeam and single-beam echosounder (SBES) datasets collected in the same region into respective DTM, computed their depth deviations, and performed statistical analyses to quantitatively assess accuracy differences between the instruments.
These previous studies demonstrate that depth difference analysis is a highly effective method for validating performance differences between platforms and sensors. Accordingly, the same analytical approach was applied in this study to compare MBES-derived bathymetry obtained from the manned vessel and the USV. To conduct the analysis, raster conversion was performed using identical spatial resolution and coordinate systems, and spatial alignment between datasets was ensured. Depth deviations were then computed on a cell-by-cell basis from the constructed DTM, and statistical indicators such as ME, MAE, and STD were used to quantify differences in survey performance (Fig. 3c). This approach enables evaluation of overall errors across the survey area and examination of spatial patterns, thereby providing a comprehensive assessment of performance differences between the manned vessel and the USV.
4. Results and Discussion
4.1. Crossline Quality Control Analysis result
In the Yongho Harbor, crossline validation showed that both the manned vessel and the USV satisfied 100% of the special order, order 1a/1b, and order 2 requirements (Table 4). This indicates that, in relatively stable coastal environments such as port areas where seafloor variability is minimal, the USV is capable of achieving data quality equivalent to that required by the International Hydrographic Organization (2020). In the Dokdo area, both the manned survey conducted in 2023 and the USV surveys carried out from 2020 onward satisfied the special order, order 1a/1b, and order 2 criteria at levels ranging from 99.6% to 100% (Table 4). This demonstrates that the USV consistently provides high-quality bathymetric data even in island environments where depth variability is relatively high. Notably, the consistent accuracy observed across multiple survey years indicates that the USV ensures reliable operational stability and data production performance.
Table 4. Crossline quality control analysis results in Yongho Harbor and Dokdo
4.2. Shallow Water Area Analysis Result
A total of 61 shallow water points near the Yongho Harbor were analyzed to compare the bathymetric performance of the manned vessel and the USV (Fig. 4a). The selected shallow water points ranged from approximately 3 m to 14 m in depth, and the survey results from both platforms are shown (Fig. 5a). The comparison revealed that depth differences between the manned vessel and the USV were generally very small. The ME was 0.01 m, indicating that the USV did not exhibit overestimation or underestimation tendencies relative to the manned survey. The MAE was 0.06 m, and both the STD and RMSE were 0.07 m (Table 5). This level of error remains well below the Special Order depth tolerance of approximately 0.25 m for depths of 3–15 m specified by the International Hydrographic Organization (2020). Consequently, the magnitude of these discrepancies of only a few centimeters indicates that both platforms provide nearly identical depth measurements in coastal harbor environments.
Fig. 4. Locations of shallow water points: (a) Yongho Harbor and (b) Dokdo.
Fig. 5. Comparison of shallow water depths between manned and unmanned surveys: (a) Yongho Harbor, (b) Dokdo (2020), (c) Dokdo (2021), (d) Dokdo (2022), (e) Dokdo (2023-1st), and (f) Dokdo (2023-2nd).
Table 5. Statistical comparison of manned and unmanned survey results in shallow water areas of Yongho Harbor and Dokdo
The Dokdo area is characterized by rocky terrain around Dongdo and Seodo, and due to the difference in survey periods between the manned vessel and the USV, areas with minimal morphological change were selected to compare the shallow water bathymetric performance of the two platforms. From 2020 to the 2023-1st survey, a total of 19 identical points were analyzed against the manned survey results, whereas in the 2023-2nd survey, only 8 of the 19 points were included due to incomplete survey coverage or the absence of supplementary shoal checking in some areas (Fig. 4b). The depth range of the selected survey points was approximately 3–28 m. The annual comparison results between the manned vessel and the USV are presented (Figs. 5b–f), and the corresponding statistical analysis is summarized (Table 5). In 2020, the errors were the largest among all survey years, with ME of 0.14 m, MAE of 0.19 m, STD of 0.19 m, and RMSE of 0.23 m. In 2021, the MAE decreased to 0.11 m, with a STD of 0.09 m and an RMSE of 0.13 m, indicating an improvement compared with 2020. The 2022 results showed the most stable performance, with ME of 0.01 m, MAE of 0.09 m, and RMSE of 0.12 m. Although the RMSE slightly increased to 0.15 m in the 2023‐1st survey, the ME (-0.09 m) and MAE (0.12 m) remained relatively consistent. In the 2023‐2nd survey, despite the analysis being limited to only eight points, the ME (0.00 m) and RMSE (0.13 m) still showed stable agreement between platforms.
In 2020, when the largest errors occurred, several shallow water points exhibited notably greater depth differences compared with other years. At depths of 10.44 m and 13.74 m, deviations exceeding 0.39 m were observed only in the 2020 dataset, whereas most other years showed discrepancies of only a few centimeters (Table 6). Considering that the USV size and MBES specifications were similar across survey years, these discrepancies are attributed more to environmental factors during the 2020 survey than to structural or system-related issues with the USV.
Table 6. Depth deviations at specific shallow water points for manned and unmanned surveys
In addition, the average RMSE in the Dokdo area was 0.16 m, which may appear slightly higher than that observed in the harbor survey. However, despite the strong wave and tidal influences characteristic of offshore island environments, this level of error remains well below the Special Order depth tolerance of 0.25–0.34 m for depths of 3–30 m specified by the International Hydrographic Organization (2020). This indicates that the USV-derived shallow water soundings achieve accuracy and reliability comparable to those of the manned vessel, even in island environments.
4.3. Depth Difference Analysis Result
The depth difference analysis was conducted by converting the point-based bathymetric soundings into raster surfaces and computing cell-by-cell deviations between the manned-vessel DTM and the USV-derived raster data. For the raster conversion, the conventional inverse distance weighting (IDW) interpolation method, widely used for DTM construction due to its balance of computational efficiency and accuracy, was applied (Maleika, 2020). Using this approach, bathymetric datasets from both the Yongho Harbor and the Dokdo region were generated at an identical spatial resolution, and the cell-based depth differences between platforms were subsequently calculated and analyzed (Fig. 6).
Fig. 6. Fitting the normal distribution function to the difference between manned and unmanned data: (a) Yongho Harbor, (b) Dokdo (2020), (c) Dokdo (2021), (d) Dokdo (2022), (e) Dokdo (2023-1st), and (f) Dokdo (2023-2nd).
The bathymetric data collected at the Yongho Harbor ranged from approximately 1.3 m to 19.5 m in depth. The cell-based depth difference analysis between the manned vessel and the USV was carried out across most of the study area, except for several locations, such as near the Gwangan Bridge and other very shallow zones, where the manned vessel could not operate, and no survey data were available. The results showed that the overall bias was negligible, with an ME of 0.01 m. In addition, the MAE was 0.08 m and the standard deviation was 0.12 m, confirming a very high level of spatial agreement in depth measurements between the two platforms (Table 7). In particular, the standard deviation of 0.12 m indicates that more than 95% of the depth differences fall within ±0.25 m, which meets the minimum allowable depth-error thresholds specified by the International Hydrographic Organization (2020). This confirms that the USV can reliably provide bathymetric data of a quality comparable to that of the manned vessel in harbor environments. In addition, shallow areas where deviation analysis could not be performed correspond to regions inaccessible to the manned vessel, requiring supplementary SBES surveys using small boats.
Table 7. Statistical comparison of the depth difference analysis results of Yongho Harbor and Dokdo
In contrast, the USV, with its smaller hull size and shallow draft, can conduct continuous multibeam surveys in such extremely shallow and obstacle-dense areas without the need for platform changes (Fig. 7). This demonstrates the USV’s clear advantages over manned vessels in terms of operational efficiency, accessibility, and survey safety in shallow water environments. However, relatively large deviations were observed in some steep bottom slopes. This is likely attributable to the fact that the manned vessel performed multiple repeat surveys in these problematic areas to reduce uncertainty and validate depth consistency, whereas the USV operated with a single-pass acquisition strategy. In complex seafloor conditions, including steep gradients, rocky outcrops, repeated observations, and multi-directional coverage are essential for minimizing bottom detection ambiguities and acoustic artifacts. Therefore, additional repeat surveying or enhanced overlap strategies are required in future USV operations to achieve comparable performance in these challenging environments.
Fig. 7. Survey tracks for manned MBSE/SBES and unmanned MBES.
For the Dokdo area, the depth difference analysis was conducted using bathymetric data acquired between 47 m and 151 m, excluding regions near Dongdo and Seodo, where the manned vessel could not perform crossline surveys due to rugged rocky terrain. Approximately 88% of the data fell within the 90–140 m depth range. According to IHO standards, the allowable TVU for Special Order surveys increases from about 0.78 m at 90 m depth to approximately 1.079 m at 140 m, reflecting the natural increase in sounding uncertainty with greater depth. The deviation analysis showed ME values ranging from –0.13 m to 0.27 m, MAE values between 0.29 and 0.33 m, and standard deviations of approximately 0.34–0.44 m (Table 7). The ME values mostly ranged within ±0.1 m, indicating that the overall systematic bias remained minimal despite variations in environmental conditions across survey years. In 2022, however, the ME increased to +0.27 m, which was notably higher than in other years; this appears to have resulted from temporary environmental factors rather than instrument or system related errors. While the MAE remained around 0.3 m, the STD values indicate that approximately 95% of all deviations fell within 0.68–0.88 m. Considering the Special Order allowance of 0.78–1.079 m, these deviation ranges satisfy the required thresholds and demonstrate that consistent agreement between manned-vessel and USV derived depths was maintained even in offshore island conditions at depths greater than 50 m.
However, relatively large deviations were observed in areas with pronounced seafloor gradients, which is believed to result from a combination of the terrain characteristics and the inherent limitations of GNSS based positioning. The accuracy of GNSS positioning can vary significantly depending on satellite signal conditions, and in remote island or offshore regions such as Dokdo, positioning performance may degrade due to reference-station signal blockage and multipath interference (Makar, 2023). In addition, characteristics of the USV platform itself appear to have influenced the results. Compared with manned survey vessels, the USV used in this study is lighter and has lower hull stability, making it more susceptible to substantial roll, pitch, and yaw motion under offshore wave and surf-zone conditions (Constantinoiu et al., 2023). These motion-induced errors directly affect the accuracy of MBES bottom detection and beam angle estimation, and the resulting depth discrepancies tend to increase particularly over steep slopes, rocky bottoms, and irregular seabed terrain. Therefore, to mitigate these limitations, the application of precision post-processing positioning techniques such as post-processed positioning (PPP) or post-processed kinematic (PPK) should be considered. Furthermore, employing larger USV platforms in offshore environments is expected to improve motion stability and thereby enhance the consistency and reliability of hydrographic survey results.
5. Conclusions
This study evaluated the reliability of USV equipped with MBES by quantitatively comparing their bathymetric survey results with those acquired from a manned vessel in the same areas. Three analytical approaches, crossline quality control analysis, shallow water area analysis, and depth difference analysis, were applied to assess performance differences between the two platforms in both harbor and island environments.
The results showed that, in the crossline analysis, USV data met the uncertainty requirements of all applicable International Hydrographic Organization (2020) survey orders in both harbor and island areas, demonstrating reliability comparable to that of the manned vessel. The shallow water comparison, conducted at 61 points in the harbor and 19 points around Dokdo, yielded RMSE values of 0.07 m and 0.16 m, respectively. These findings indicate that USV-derived shallow water depths closely match those obtained from the manned vessel, confirming similar levels of accuracy and consistency. In the depth difference analysis, point-based bathymetric data were converted into raster surfaces, and spatial deviations between platforms were examined. The results showed high spatial agreement overall, indicating that USV-based bathymetry provides stable and consistent performance across a range of environmental conditions.
Taken together, these results demonstrate that USV-based MBES bathymetric surveys achieve accuracy and reliability equivalent to traditional manned-vessel surveys, while also offering advantages in shallow, complex, or obstacle-rich environments where manned operations are limited. Therefore, USV holds significant potential as an effective platform capable of complementing or in some cases replacing manned hydrographic surveys in future ocean mapping operations.
Future work should include further investigation into areas where relatively large deviations occurred, as well as expanded comparative analyses across more diverse marine environments. Additionally, to broaden the scope of unmanned surveys, the potential application of other unmanned platforms, such as AUVs and sensors, including marine magnetometers and sub-bottom profilers, should be examined. By validating USV-based MBES performance against manned surveys, this study contributes to improving the reliability of unmanned hydrographic data.
Author Contributions
Conceptualization: All authors; Data curation, Formal analysis: Bae JG, Yang JS; Investigation, Methodology: Bae JG, Lee SH, Yang JS; Project administration: Sin DG, Kim JS; Supervision: Kim ED, Kim JS; Validation: All authors; Writing–original draft: Bae JG, Lee SH; Writing–review & editing: All authors.
Conflicts of Interest
No potential conflict of interest relevant to this article was reported.
Funding
None.
Data Availability Statement
The data presented in this study are not publicly available due to security and usage restrictions.
Acknowledgments
The authors express their gratitude to the Korea Hydrographic and Oceanographic Agency for supporting this research through the “2025 Unmanned Survey Project” and providing the manned vessel and USV MBES bathymetric survey datasets used in this study.
Supplementary Materials
None.
References
- Angliss, R.P., Ferguson, M. C., Hall,P., Helker, V., Kennedy, A., and Sformo, T., 2018. Comparing manned to unmanned aerial surveys for cetacean monitoring in the Arctic: Methods and operational results. Journal of Unmanned Vehicle Systems, 6(3), 109-127. https://doi.org/10.1139/juvs-2018-0001
- Bae, I., and Hong, J., 2023. Survey on the developments of unmanned marine vehicles:Intelligence and cooperation. Sensors, 23(10),4643. https://doi.org/10.3390/s23104643
- Bastos, L., Bio, A., and Iglesias, I., 2016. The importance of marine observatories and of RAIA in particular. Frontier sin Marine Science, 3, 140. https://doi.org/10.3389/fmars.2016.00140
- Canadian Hydrographic Service, 2021. Hydrographic survey management guidelines. Fisheries and Oceans Canada. https://www.charts.gc.ca/documents/data-gestion/guidelines-directrices/sg-ld-2021-eng.pdf
- Cavanaugh, K. C., Bell, T. W., Aerni, K. E., Byrnes, J. E.K., McCammon, S ,and Smith, M.M., 2025. New technologies for monitoring coastal ecosystem dynamics. Annual Review of Marine Science, 17, 409-433. https://doi.org10.1146/annurev-marine-040523-020221
- Chai, T., and Draxler, R.R., 2014. Root mean square error(RMSE) or mean absolute error(MAE)?–Arguments against avoiding RMSE in the literature. Geoscientific Model Development, 7(3), 1247–1250. https://doi.org/10.5194/gmd-7-1247-2014
- Chen,Y., Shan, X., Jin, X.,Yang, T., Dai, F., and Yang, D, 2016. A comparative study of spatial interpolation methods for determining fishery resources density in the Yellow Sea. Acta Oceanologica Sinica, 35, 65-72. https://doi.org/10.1007/s13131-016-0966-y
- Constantinoiu, L-F, Bernardino, M., and Rusu, E. 2023. Autonomous shallow water hydrographic survey using a proto-type USV. Journal of Marine Science and Engineering, 11(4), 799. https://doi.org/10.3390/jmse11040799
- DePryck, K., and Boettcher, M.,2024. The rise, fall and rebirth of ocean carbon sequestration as a climate 'solution. Global Envirotumental Change, 85, 102820. https://doi.org/10.1016/j.gloenvcha.2024.102820
- Eiler, J. H., Grothues, T. M., Dobarro, J. A., and Masuda, M. M., 2013.Comparing autonomous underwater vehicle(AUV) and vessel-based tracking performance for locating acoustically tagged fish. Marine Fisheries Review 75(4), 27-42. https://doi.org/10.7755/MFR.75.4.2
- Erdoǧan, S., 2010. Modelling the spatial distribution of DEM error with geographically weighted regression:An experimental study.Computers & Geosciences, 36(1), 34–43. https://doi.org/10.1016/j.cageo.2009.06.005
- Gattuso, J.-P, Magnan, A. K., Bopp, L., Cheung, W. W. L., Duarte, C. M., Hinkel, J., et al., 2018. Ocean solutions to address climate change and its effects on marine ecosystems. Frontiers in Marine Science, 5, 337. https://doi.org/10.3389/fmars.2018.00337
- General Bathymetric Chart of the Oceans, 2019. The IHO-IOC GEBCO cook book. International Hydrographic Organization, Intergovernmental Oceanographic Commission. https://www.gebco.net/data_and_products/gebco_cook_book/
- Grzadziel, A., 2021. The importance of under-keel sound velocity sensor in measuring water depth with multibeam echosounder: Energies, 14(17), 5267. https://doi.org/10.3390/enl4175267
- Hydrographic Surveys Division, 2025. Hydrographic survey specifications and deliverables. National Oceanic and Atmospheric Administration, Office of Coast Survey. https://www.nauticalcharts.noaa.gov/publications/documents/HSSD_2025-0-00.pdf
- International Hydrographic Organization, 2020. IHO standards for hydrographic surveys S-44 edition 6.1.0. Available online: https://iho.int/uploads/user/pubs/standards/s-44/S-44_1Edition_6.1.0.pdf(accessed on Nov. 11, 2025).
- International Hydrographic Organization, 2022. Importance of hydrography. Availableonline: https://ihoint/en/importance-of-hydrography (accessedon Nov. 11, 2025).
- Kim,C.H., Choi,S. Y.,Kim, W.H., Choi, H. O., Park, C.H.,Kim, Y.B., et al., 2023. Monitoring of bathymetry changes in the coastal area of Dokdo, East Sea. Economic and Environmental Geology, 56(5), 589-601. https://doi.org/10.9719/EEG.2023.56.5.589
- Korea Hydrographic and Oceanogiaphic Agency, 2022. Hydrographic survey glossary. Available online: https://www.khoa.gokr/khoa/pgmctrl/selectDictionaryList.do (accessed on Nov.11, 2025).
- Kubacka, M., Gajewski, L., Burchacz, M., Matczak, M., Janowski, P. and Piotrowicz, J., 2022. Many faces of survey equipment failures during marine research at sea-Risk analysis. PLOS one, 17(8), e0272960. https://doi.org/101371/joumal.pone.0272960 101371/joumal.pone.0272960
- Leder, N., and Duplančić Leder, T., 2018. Unmanned vehicle systemsin hydrographic survey - Newopportunities and challenges. In Proceedings of the 2018 18th International Conference on Transport Science, Portrož, Slovenia, June 14–15, pp. 184–188.
- Lee, G. H., Kim, H. J., Kim, D. C., Yi, B. Y., Nam, S. M., Khim, B. K., et al., 2009. The acoustic diversity of the seabed based on the similarity index computed from Chirp seismic data. ICES Journal of Marine Science, 66(2), 227–236. https://doi.org/10.1093/icesjms/fsn142
- Lee, M. H., Kim, C. H., Park, C. H., Rho, H. S., and Kim, D. C., 2017.Comparative analysis of bathymetry in the Dongdo and the Seodo, Dokdo using multibeam echosounder system.Economic and Environmental Geology, 50(6), 477-486. https://doi.org/10.9719/EEG.2017.50.6.477
- Lin, M., and Yang, C., 2020. Ocean observation technologies: A review. Chinese Journal of Mechanical Engineering, 33, 32. https://doi.org/10.1186/s10033-020-00449-z
- Lin, Y., Chen, J., Sun, Z., and Phan-Thien, N., 2023. Editorial: Deep-sea sampling technology. Frontiers in Marine Science, 10, 1182211.https://doi.org/103389/fmars.2023.1182211 103389/fmars.2023.1182211
- Liu, Z., Zhang, Y., Yu, X., and Yuan, C., 2016. Unmanned surface vehicles: An overview of developments and challenges. Annual Reviews in Control, 41, 71–93. https://doi.org/10.1016/j.arcontrol.2016.04.018
- Makar, A., 2023. Limitations of multi-GNSS positioning of USV in area with high harbour infrastructure. Electronics, 12(3), 697. https://doi.org/10.3390/electronics12030697
- Maleika,W., 2020. Inverse distance weighting method optimization in the process of digital terrain model creation based on data collected from a multibeam echosounder. Applied Geomatics, 12, 397–407. https://doi.org/10.1007/s12518-020-00307-6
- Masetti, G., Faulkes, T., and Kastrisios, C., 2018. Automated identification of discrepancies between nautical charts and survey soundings. ISPRS International Journal of GeoInformation, 7(10), 392.https://doi.org/10.3390/ijgi7100392
- Moknatian, M., Piasecki, M., Moshary, F., and Gonzalez, J., 2019. Development of digital bathymetry maps for Lakes Azuei and Enriquillo using sonar and remote sensing techniques. Transactions in GIS, 23(4), 841–859. https://doi.org/10.1111/tgis.12532
- National Oceanic and Atmospheric Administration, 2014. Data acouisition and processing report(U.S. GOV. PRINTING OFFICE: 1985-566-054). US. Department of Commerce, National Oceanic and Atmospheric Administration National Ocean Service. https://data.ngdc.noaa.gov/platforms/ocean/nos/DAPRs/M-L906-KR-08_DAPR_Amend-2.pdf
- Park, M. S., Kim, Y. G., Park, C.H., and Jang, M. S., 2024. Survey research using manned and unmanned platforms in the coastal area. Korean Journal of Hydrography and Marine Spatial Information, 13(1),43-52. https://doi.org/10.23330/KJHMSI.2024.13.2.043
- Pasandeh, N., Mosavat, S. S., Abadpour, S., Niya, A. K., Tajfirooz, B., Zarei, S. M., et al., 2025. Investigating the precision of hydrographic data by comparing the differences between multi-beam and single-beam echo-sounders(case study: Bushehr port in the Persian Gulf). The International Hydrographic Review, 31(1). 184–194. https://doi.org/10. 58440/ihr-31-1-n02 https://doi.org/10.58440/ihr-31-1-n02
- Paul, F., Bolch, T., Briggs, K. Kaab, A., McMillan, M., McNabb, R., et al., 2017. Error sources and guidelines for quality assessment of glacier area, elevation change, and velocity products derived from satellite data in the Glaciers_CCI Project. Remote Sensing of Environment, 203, 256-275. https://doi.org/10.1016/j.ss.2017.08.038
- Pimentel, V. B., Florentino, C., and Neto, A. A., 2020. Evaluation of the precision of phase-measuring bathymetric side scan sonar relative to multibeam echosounders. The International Hydrographic Review, 24, 61–83.
- Saildrone, 2025. Saildrone and meta complete first unmanned, autonomous deep-water route survey in the North Atlantic. Available online: https://www.saildrone.com/media-room/ press-releases/saildrone-meta-first-autonomous-deep-water-route-survey (accessed on Nov. 14, 2025).
- Schmidt, V., Downs, R., Doroba, J., and Levano, M., 2023. Integration and operation of Drix-12 aboard NOAA ship Thomas Jefferson. In Proceedings of the 2023 US Hydro Conference, Mobile, AL,USA, Mar. 13-17, p. 1463. https://scholars.unh.edu/ccom/1463
- Şengül, B., Yılmaz, F., and Uğurlu, Ö., 2023. Safety–security analysis of maritime surveillance systems in critical marine areas. Sustainability, 15(23), 16381. https://doi.org/10.3390/su152316381
- Slinning, T. O., 2016. Analytic comparison of multibeam echo soundings. Master's thesis, Degree- Lund University, Lund, Sweden.
- Smith,S.R., Alory, G., Andersson, A., Asher, W., Baker, A., Berry, D. I., et al.,2019.Ship-based contributions to global ocean, weather, and climate observing systems. Frontiers in Marine Science, 6, 434. https://doi.org/10.3389/fnars.2019.00434
- Stanghellini, G., Bidini,C., Romagnoli, C., Archetti, R., Ponti, M., Turicchia, E., et al., 2022.Repeated(4D)marine geophysical surveys as a tool for studying the coastal environment and ground-truthing remote-sensing observations and modeling. Remote Sensing, 14(22), 5901. https://doi.org/10.3390/rs14225901
- Teledyne CARIS, 2016. CARIS HIPS and SIPS 8.1 user guide. Available online: https://www.scribd.com/document/300265523/Caris-Hips-and-Sips-User-Guide-v81 (accessed on Nov. 12, 2025).
- Terradepth, 2023. Autonomous underwater vehicle (AUV)survey services: Safer and faster operations (55% more efficient) in challenging conditions compared to traditional surveys. Available online: https://www.terradepth.com/uploads/downloads/AUVCaseStudy.pdf (accessed on Nov. 12, 2025).
- U. S. Geological Survey, 2020. 3DEP lidar base specification 2020 rev.A.Available online: https://d9-wret.s3.us-west-2.amazonaws.com/assets/palladium/production/s3fs-public/ atoms/files/Lidar-Base-Specification-2020-rev-A.pdf (accessed on Nov. 13, 2025).
- Ward, D., Melbourne-Thomas, J., Pecl, G. T., Evans, K., Green, M., McCormack, P.C., et al., 2022. Safeguarding marine life: Conservation of biodiversity and ecosystems. Reviews in Fish Biology and Fisheries, 32, 65-100. https://doi.org10.1007/s11160-022-09700-3
- Willmott, C. J., and Matsuura, K., 2005. Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance. Climate Research, 30, 79–82. https://doi.org/10.3354/cr030079
- Yaney-Keller, A., San Martin, R., and Reina, R. D., 2021. Comparison of UAV and boat surveys for detecting changes in breeding population dynamics of sea turtles. Remote Sensing, 13(15), 2857. https://doi.org/10.3390/rsl3152857