Application of MLS and UAS-SfM for Beach Management at the North Padre Island Seawall

Highlights

What are the main findings?

• Mobile lidar scanning (MLS)-derived digital elevation models (DEMs) were used to monitor beach geomorphology, finding significant seasonal and post-nourishment changes in beach slope and width, shoreline position, and beach volume.
• A comparative and operational analysis assessed MLS and uncrewed aircraft system (UAS) structure-from-motion (SfM)/multi-view stereo (MVS) photogrammetry for beach management, finding DEM RMSE differences were similar, averaging up to 3 cm, leading to volume differences of up to 3%.

What is the implication of the main finding?

• MLS and UAS-SfM offer efficient, scalable tools for routine beach monitoring, each with unique operational considerations that can inform coastal policy and management.
• Highest astronomical tide (HAT) shoreline position monitoring helped to identify optimal seasonal bollard placement to restrict vehicular access, thus increasing pedestrian safety.

Abstract

Collecting accurate and reliable beach morphology data is essential for informed coastal management. The beach adjacent to the seawall on North Padre Island, Texas, USA has experienced increased erosion and disrupted natural processes. City ordinance mandates the placement of bollards to restrict vehicular traffic when the beach width from the seawall toe to mean high water (MHW) is less than 45.7 m. To aid the City of Corpus Christi’s understanding of seasonal beach changes, mobile lidar scanning (MLS) surveys with a mapping-grade system were conducted in February, June, September, and November 2023, and post-nourishment in March 2024. Concurrent uncrewed aircraft system (UAS) photogrammetry surveys were performed in February and November 2023, and March 2024 to aid beach monitoring analysis and for comparative assessment to the MLS data. MLS-derived digital elevation models (DEMs) were used to evaluate seasonal geomorphology, including beach slope, width, shoreline position, and volume change. Because MHW was submerged during all surveys, highest astronomical tide (HAT) was used for shoreline analyses. HAT-based results indicated that bollards should be placed from approximately 390 to 560 m from the northern end of the seawall, varying seasonally. The March 2024 post-nourishment survey showed 102,462 m³ of sand was placed on the beach, extending the shoreline by more than 40 m in some locations. UAS photogrammetry-derived DEMs were compared to the MLS-derived DEMs, revealing mean HAT position differences of 0.02 m in February 2023 and 0.98 m in November 2023. Elevation and volume assessments showed variability between the MLS and UAS-SfM DEMs, with neither indicating consistently higher or lower values.

1. Introduction

Mobile lidar scanning (MLS) systems have been increasingly used to map and monitor sandy beach coastal corridor environments. These systems allow for the rapid collection of precise, high-resolution three-dimensional (3D) point cloud data that can capture detailed surface information of the beach and shoreward-facing dune structure [1,2]. The high spatial resolution, combined with ease of deployment for repeatable surveys make MLS an optimal method for monitoring seasonal and yearly erosion and deposition, post-storm damage, and change detection in the beach and lower foredune.

Several examples in the literature demonstrate the application of MLS for beach monitoring. Lim et al. [3] assessed the horizontal and vertical accuracy of MLS systems in sandy beach environments, comparing it to paved areas. Before boresight and lever arm adjustments, the mean errors were 0.22 m (x), 0.036 m (y), and 0.10 m (z), and after adjustments, they were improved to 0.06 m (x), 0.09 m (y), and 0.05 m (z). Donker et al. [4] used MLS systems to perform 13 repeat surveys over 2.5 years to analyze changes in foredune structure and volume. After generating 1 m digital elevation models (DEMs), the morphology of the study site was analyzed. It was found that the vertical root mean square error (RMSE) of the surveys was about 0.01 m, and the volume error estimates were about 0.25 m³/m. Gong [5] used MLS to perform rapid post-storm damage assessments of New York and New Jersey after Hurricane Sandy. Results showed that MLS data can enhance airborne remote sensing data and typical ground survey data. It was also deemed to be useful in assessing sediment change, damaged structures, and reconstructing flood scenarios. Bitenc et al. [1] used an MLS system to survey a 6 km beach on the Dutch coast to assess the accuracy of the resulting point cloud DEM. They found that their system could achieve a vertical RMSE of 0.05 m, and a DEM was available within two days of acquisition.

Uncrewed aircraft system (UAS) photogrammetric surveys are generally processed using structure-from-motion (SfM)/multi-view stereo (MVS) photogrammetry techniques (collectively referred to in this paper as SfM). SfM derives 3D structure from two-dimensional (2D) image sequences through movement of the camera, thereby providing different perspective views of the scene. By using the UAS as the moving platform, SfM can be implemented with an onboard camera by acquiring images with sufficient overlap. UAS is a relatively lower- cost survey method compared to traditional airborne remote sensing techniques for collecting high resolution 3D data of topography and land cover.

In the coastal zone, UAS-SfM has been applied to assess post-storm impacts and study geomorphologic processes such as shoreline change, erosion and deposition, and changes in beach-dune morphology [6,7]. This surveying method can provide high spatial and temporal resolution data of the beach and dune environment from a nadir view. Enwright et al. [8] employed UAS-SfM to monitor a barrier island fronted by the Gulf of Mexico in Mississippi. They tested different methods of deriving digital surface models (DSMs), finding that a minimum bin algorithm resulted in higher vertical accuracy DSMs than interpolation. The final vertical RMSEs ranged from 0.095 m to 0.164 m. Che Mat and Tahar [9] conducted four repeat surveys of a sandy beach with a DJI Phantom 3 (Nanshan, Shenzhen, China) on the east coast of Malaysia to create a DEM for modeling surf zone changes over time. The average final vertical accuracy was 0.419 m, and it was observed that the vertical error was higher in sandy areas. Rotnicka et al. [7] used a fixed-wing UAS to survey a sandy beach and vegetated foredune system, comparing the UAS-SfM derived DEMs to a series of natural ground transects covering both vegetated and non-vegetated terrain. They found that the vertical RMSE of the survey was 0.06 m, but grass and other vegetation significantly affected the UAS-SfM-derived DEM accuracies, with only areas of sparse vegetation revealing accurate sediment budgets. Leal-Alves [10] applied UAS-SfM to assess vulnerability of a beach in southern Brazil to inundation. The final DEM resolution was 0.12 m, while the vertical RMSE was 0.06 m. A bathtub model identified susceptible areas to sea level rise and inundation. Jeyaraj et al. [6] implemented UAS-SfM to conduct two surveys to monitor sediment changes on a sandy beach, collecting a series of ground control points (GCPs) for model validation. They found that the two surveys had vertical RMSEs of 0.02 m and 0.01 m. By analyzing alongshore cross-shore transects. They quantified the impact of monsoon season on beach and foredune morphology, identifying significant volume and elevation changes. Van Alphen et al. [11] applied UAS-SfM to assess volume change in a sandy beach after Hurricane Michael. They computed the pre-and post-volume of the study sites and the volume change after the hurricane by transect and by a raster-based method. Lastly, Nahon et al. [2] combined MLS and UAS-SfM to monitor sandy beaches. They used the UAS data to address occluded features and serve as a form of ground truth validation to reduce the number of GCPs needed during a survey.

The goal of this study was to use a mapping-grade MLS system to examine variability of a sandy beach adjacent to a seawall on North Padre Island, Texas, USA, to inform management decisions. Beach width, shoreline change, erosion, and seasonal patterns were analyzed to inform policy and management decisions. A key focus was monitoring the shoreline to guide bollard placement. The initial objective was to monitor the MHW line relative to the toe of the seawall (the bottom of the bottom stair of the seawall). Due to MHW being submerged for every survey, an alternative shoreline proxy, highest astronomical tide (HAT) was chosen. UAS-SfM surveys were also conducted to compare with MLS surveys in terms of data acquisition; beach elevation and shoreline position measurement; and their utility for beach monitoring and management.

Study Site

The North Padre Island (NPI) Seawall (Figure 1) was built in 1967 using corporate funds, without permit authority, to protect a resort community, condominiums, and a series of hotels. The structure is approximately 1280 m long and 3.5 m tall, consisting of a concrete wall with steps running along its seaward face. It was not built at an exact parallel to the shoreline, with the southern end sitting closer to the waterline. After its construction, the seawall withstood two hurricanes: Hurricane Celia, a category 3 storm in 1970 and Hurricane Fern, as a Category 1 in 1971, with neither storm causing noticeable damage [12]. Hurricane Allen made landfall in 1980 about 130 km south of the seawall as a Category 5 hurricane, with a storm surge that reached an elevation of 2.6 m and lasted about 48 h [13]. The storm caused the seawalls’ foundation to fail due to long periods of inundation, wave runup and sediment loss in many parts of the structure [14]. It was rebuilt shortly after and continues to be managed under shared, private ownership. Management of the surrounding beach is as follows: the portion seaward of the seawall toe is overseen by the City of Corpus Christi; the beach north of the seawall is managed by both the City of Corpus Christi and Nueces County Coastal Parks; and the section south of the structure falls under Nueces County jurisdiction [12,14].

The beach shoreward of the seawall (Figure 1), extending north to Packery Channel and south to Bob Hall Pier, is heavily used by the public. It is popular with both locals and tourists, offers handicap access, and is significantly impacted by development. The construction of the seawall and adjacent jetties disrupted natural sediment dynamics, contributing to increased erosion along the southern portion and directly seaward of the structure [12]. Williams [14] noted that the Packery Channel jetties reduce northward sediment transport and promote sand accumulation south of the inlet. However, the stretch of beach from the mid-seawall southward lies beyond the zone protected by the jetties and has undergone consistent erosion. In fact, at both ends of the seawall, access roads are prone to flooding during storms and elevated water levels. Washover events at the southern access road promote erosion at the southern end of the seawall during these high-water episodes [15]. To maintain beach access, especially for vehicles, sand is regularly mechanically redistributed to counter erosion [14].

According to the Corpus Christi, TX [16] Ordinance No. S10-77, bollards must be placed perpendicular to the seawall to temporarily restrict vehicular access and maintain a safe beach environment for pedestrians if the mean high tide line (defined by mean high water (MHW), an elevation of 0.317 m North American Vertical Datum of 1988 (NAVD88) in this area) is less than 45.7 m (150 ft) [17]. This ordinance does not account for tides, driving winds and other forces influencing wave runup. Because beach width fluctuates seasonally and annually, maintaining beach compliance can be challenging [14]. However, the Texas Open Beaches Act states that the public has free and unrestricted right of ingress and egress for access to all state-owned beaches from the line of vegetation to the line of mean low tide [18]. Once bollards are placed, the City of Corpus Christi must begin planning for beach restoration to the 45.72 m threshold to avoid violating this Act.

In addition to regular mechanical sand redistribution, occasional beach nourishment projects have been conducted in front of the NPI Seawall. Williams [14] reported that two nourishments occurred prior to 2015 using sediment dredged from Packery Channel under a Beneficial Use of Dredge Material (BUDM) program. The first was completed during the construction of Packery Channel between 2004 and 2006, and the second was during the winters of 2011 to 2013. Following the channel’s construction, erosion rates decreased along the north end of the Seawall but remained high along the southern end. The most recent nourishment at the time of this study began on 25 November 2023 and was completed on 22 January 2024.

NPI is described as a “microtidal, low-energy coastline” by both [12,19]. The astronomical tides are primarily diurnal, with a mean range of tide of 40 cm (NOAA, 2004 [17]), with June and July classified as low tidal months. However, tidal fluctuations in this area are largely wind-driven, inducing larger water change levels than meteorological tides, with mean annual wind speeds around 19 km/h [19]. According to [15,20], wind direction along NPI varies seasonally, with bimodal wind patterns generally originating from the northeast during winter and from the southeast in summer, driving sediment transport in their respective directions. The predominant wind direction throughout most of the year is from the south-southeast. During the winter months, typically from about October/November through February, the winds shift from southeast to the northeast, with the arrival of cold fronts, locally known as “northers”. A spring transition occurs from approximately March through May, during which winds gradually shift from northeast to east and southeast. In the summer months, from about June through September, winds are persistent from the southeast. Lastly, a fall transition occurs from about September to October/November, during which southeasterly winds give way once again to northeasterly winds as frontal activity increases. These seasonal periods are approximate and may vary slightly from year to year depending on seasonal weather variability. Tissot and Dell, [21] indicate that this area experiences a median significant wave height, with a range of 0.1 to 2.0 m and a typical wave period of 5.9 s, with a range of 2.1 to 11.5 s. The average water levels vary seasonally with highs in May and October and lows in July and January [22]. Additionally, periods of high, onshore winds occur in the spring, further contributing to elevated water levels [19,20]. Extreme events, such as very high winds and the impact of tropical storms and hurricanes also lead to high water levels.

NPI is also oriented at an angle to the oncoming wave approach, with waves approaching and breaking diagonally along the shoreline from the southeast direction from north of NPI southward to a convergence zone near Big Shell beach on Padre Island National Seashore (PAIS) [20]. This wave angled wave action, combined with wind forcing, drives longshore sediment transport parallel to the shoreline [19]. Williams and Turner [23] note that sediment transport is unobstructed along the open beach south of NPI, while the southern jetty of Packery Channel obstructs sediment flow, resulting in localized sediment accretion. Although seasonal reversals occur, the dominant direction of sediment transport at NPI is from south to north due to the prevailing south-southeasterly winds. In the winter, this pattern reverses, with sediment moving from north to south in response to north-northeasterly winds. According to Morton [24], prior calculations estimated that gross littoral transport along Mustang Island and NPI ranges from approximately 725,000 and 925,000 yd³/yr, while net southwesterly transport lies between 66,000 and 80,000 yd³/yr. The sediment itself is made up of fine, brown sand as described in Morton [12] and predominantly consists of quartz, with traces of rock fragments, feldspar, and minerals such as garnet, zircon, tourmaline, hornblende, and others [19].

2. Materials and Methods

2.1. Data Acquisition

2.1.1. MLS

Data were collected using a MLS system called the HiWay mapper, integrated by LidarUSA (Hartsell, AL, USA) and comprising a Velodyne HDL-32E (San Jose, CA, USA) lidar sensor with a NovAtel position and orientation system (POS) (Calgary, Alberta, Canada) (Figure 2a). The Velodyne unit includes 32 class 1 laser channels operating at a wavelength of 903 nm, housed within a spinning head. It captures approximately 700,000 points per second in single return mode, spins at a user-defined rate of 5 to 20 Hz (set at 20 Hz for this study), and has a maximum effective range of 100 m with an accuracy of ±2 cm (1-σ at 25 m). The NovAtel POS incorporates a 702 gg global navigation satellite system (GNSS) receiver and a Synchronous Position, Attitude, and Navigation (SPAN) IGM inertial navigation system (INS). The INS includes a Sensonor STIM300 inertial measurement unit (IMU) (Vestfold, Norway) and a NovAtel OEM615 receiver. The IMU has 3 Micro-Electro-Mechanical System (MEMS) based gyroscopes, 3 accelerometers, 3 stability inclinometers, and collects at 125 Hz.

Four seasonal MLS system surveys of the beach adjacent to the 1280 m seawall were conducted in February, June, September, and November 2023, followed by a post-nourishment survey in March 2024. MLS data were systematically collected and processed using an optimized workflow developed by [25] to ensure consistency and data quality. The MLS drive paths maintained a minimum of 50% overlap to ensure coverage and alignment. Two passes were completed at the southern end of the seawall and three passes at the northern end. The vehicle was driven at speeds no greater than 4.5 km to enhance point density and maintain pedestrian safety.

Figure 2. (a) MLS system used in this study mounted on the roof of a 4WD truck at the NPI seawall and (b) DJI Phantom 4 RTK used to collect UAS-SfM data [26].

Figure 2. (a) MLS system used in this study mounted on the roof of a 4WD truck at the NPI seawall and (b) DJI Phantom 4 RTK used to collect UAS-SfM data [26].

Survey planning during the winter and fall transitional periods was challenging, as low tides often coincided with onshore winds that produced significant wave runup. As a result, surveys were scheduled during low tide and/or during periods of offshore or cross-shore winds. Most surveys were conducted during low tide conditions. However, the February 2023 survey was instead timed during offshore winds to expose more of the beach. Subsequent surveys were conducted near low tide, although tides began to rise toward the end of each survey.

2.1.2. UAS-SfM

The February 2023, November 2023, and March 2024 surveys were each accompanied by a simultaneously collected UAS survey. The UAS used in this study was a DJI Phantom 4 RTK, (Figure 2b) which is a rotary quadcopter equipped with a multi-frequency GNSS receiver and a 20 MP CMOS camera with a focal length of 8.8 mm [27]. Each flight was conducted concurrently with the MLS data collection using the following UAS settings: 75 m flying height above ground level (AGL) which yielded ~2 cm ground sample distance (GSD), 30° forward of nadir camera angle, and 80% sidelap and endlap in a double grid pattern.

2.1.3. Ground Truth Data

A local, geodetic-grade base station was established for every survey and occupied on the same benchmark atop the seawall. The base station recorded observations during the entire duration of the MLS and UAS surveys, for a minimum of two hours [28]. The GNSS data was then used for post processed kinematic (PPK) trajectory correction of the MLS and UAS surveys during data processing. The benchmark atop the seawall was established using a static GNSS occupation of over four hours, with precise coordinates computed via the National Geodetic Survey (NGS) Online Positioning User Service (OPUS) [28].

GCPs were made of 0.3 m by 0.3 m commercially printed polyvinyl chloride panels, each 1.3 cm thick. Each target was double-sided, featuring a vinyl overlay. One side displayed a high-contrast checkerboard pattern with two matte black and two reflective white squares. The other side had a matte black background with a central reflective white circle approximately 12 cm in diameter. The checkerboard pattern was used to align both the MLS and UAS-SfM surveys.

The GCPs were distributed in intervals every 300 m alongshore, starting from the southern end of the seawall. For the first six intervals, two targets were placed: one near the wet/dry line (where the transition between wet and dry sand is visible), and one near the seawall. In the final 2 intervals, three targets were placed: one near the seawall, one in the center of the beach, and one near the wet/dry line. In each pair of GCPs, one target was used to constrain the adjustment, while the other served as a targeted checkpoint for accuracy assessment. These GCP types alternated alongshore to enhance the control geometry. For the intervals with three targets, the two outer GCPs were used for adjustment, and the central target was used as a checkpoint. The same GCPs were used for both the MLS strip adjustment and the UAS-SfM PPK image geotag corrections to maintain alignment and allow for direct comparison between the two surveys.

Additionally, natural ground topographic checkpoint transects were collected in the cross-shore direction from the wet/dry line to the first exposed stair of the seawall. These were used to assess vertical accuracy in both the MLS and UAS-SfM surveys. The GCPs and topographic checkpoints were georeferenced using real time kinematic (RTK) surveying using coordinate averages from 5 s epochs and 3 s epochs, respectively. The RTK surveys used the same local GNSS base station and benchmark coordinates to ensure consistent alignment to the same datum point across all surveys.

2.2. Data Processing

2.2.1. MLS

MLS system trajectories were processed within Inertial Explorer software v9.00 using post processed kinematic (PPK) techniques, incorporating the static observation data from the local GNSS base station. Using observations recorded by the MLS system’s POS, inertial trajectory processing within Inertial Explorer employed a tightly coupled solution with combined forward and reverse solutions. Point clouds were then created in ScanLook PC, applying a minimum and maximum range filter of 2.5 m and 70 m, respectively. Although the system has a maximum effective range of 100 m, the range was cut off at 70 m to reduce measurement noise and error observed at longer ranges with this MLS system.

Following point cloud generation, noise filtering was performed using the lasnoise module from LAStools. This removed isolated point clusters through a gridded filtering process. Ground classification was completed with LASTools’ lasground, which uses a modified progressive triangulated irregular network (TIN) densification algorithm. This method, based on user defined parameters, selects the lowest elevation point in each grid cell, builds an initial TIN, and progressively densifies it to distinguish ground from non-ground points [29]. To classify water, the discrete attribute analysis tool in Quick Terrain Modeler was used. Points associated with breaking waves or strong absorption were filtered by applying upper and lower intensity thresholds, leaving behind a clear shoreline which was then manually delineated. A strip adjustment using the ground classified points was carried out in Spatial Explorer, which applied its version of an iterative least squares adjustment to correct for discrepancies in roll, pitch, heading, and elevation between overlapping passes. The GCPs were integrated during this step.

Each point cloud was then transformed from ellipsoidal to orthometric heights using the NAVD88 datum and Geoid18 model via the lasvdatum module in LAStools. Lastly, 10 cm DEMs of the beach and seawall structure were generated using the las2dem module in LAStools. This tool uses natural neighbor interpolation, which assigns values to grid cells based on nearby sample points using weighted contributions and TIN interpolation to calculate elevation values [30]. A 10 cm DEM resolution was chosen to give sufficient detail and resolution, without being too time-intensive to process and analyze. After each DEM was generated, the vertical RMSE was computed, shown in

Table 1. MLS system survey dates, DEM vertical accuracies, and point densities.

Survey Date	Time Period *	RMSEz Check (cm)	RMSEz Topo (cm)	Points per m2
3 February 2023	Spring Transitional	2	2	1334
28 June 2023	Summer	3	-	978
7 September 2023	Fall Transitional	3	5	1253
1 November 2023	Winter	5	6	807
6 March 2024	Post-nourishment	3	4	832

* The time period and corresponding seasonal wind conditions are described in the introduction of this paper.

.

2.2.2. UAS-SfM

The first step in the UAS processing workflow involved correcting the onboard GNSS trajectory data and image positions (or geotags) using PPK processing techniques. This was done in REDtoolbox (REDcatch GmbH, Fulpmes, Austria) software v3.4.1, incorporating the static observations from the local GNSS base station [31]. The images and their corresponding positions were subsequently imported into Agisoft Metashape software v2.2.2 for photogrammetric data processing, which applies SfM/MVS techniques. A basic summary of this process is provided below.

First, the software identified scale-invariant features across the full resolution image set, matching key points between overlapping images. Next, a bundle block adjustment was performed to estimate internal camera parameters (e.g., focal length, principal point, lens distortion) and to refine camera positions and orientations, resulting in a sparse point cloud of the scene. GCPs were then incorporated as constraints into the adjustment to improve georeferencing accuracy and further optimize the sparse point cloud, improving alignment to the control data. Finally, the point cloud was densified at half image scale using MVS techniques. From the dense point cloud, the software generated a digital surface model (DSM) to orthorectify the imagery and output a georeferenced orthomosaic image for high accuracy planimetric mapping. For more details on this process, refer to Starek et al. [32].

Dynamic movement of water alongshore captured in the UAS imagery, such as from wave runup, can negatively impact the feature matching and correspondence algorithms during SfM data processing. To reduce this effect, this study fused image masks to conceal water features. These masks were generated by creating a mesh and delineating the water-covered portions to produce binary masks for individual images. Additionally, the photogrammetry workflow described above adapted a bundle block optimization approach outlined by Over et al. [33] to reduce errors and generate a sparse point cloud comprising high-quality points. These enhancements improved the quality of the final dense point cloud data and derivative mapping products. Using the resulting UAS-SfM point cloud data, the same methods described above for the MLS point cloud data were applied to remove noise and classify ground points for interpolating 10 cm resolution bare-earth DEMs.

The same base station data and set of GCPs were used to process and adjust both the MLS and UAS-SfM surveys. The resulting vertical accuracy for each MLS and UAS-SfM survey, reported as RMSE, are shown in Table 1 and

Table 2. UAS-SfM survey dates, DEM vertical accuracies, point density, and flight specifications.

Survey Date	RMSEz Check (m)	RMSEz Topo (m)	Ground Sample Distance (GSD)	Points per m2	Flying Altitude (m)
3 February 2023	0.03	0.03	2.14	544	72.9
1 November 2023	0.01	0.03	2.23	494	76.2
6 March 2024	0.01	0.02	2.24	451	76.5

, respectively. The RMSE values shown in the tables represent the vertical accuracy of the DEMS generated from the surveys measured relative to the RTK surveyed coordinates of the aerial panel targets (i.e., GCPs used as checkpoints) and topographic checkpoints.

2.3. Feature Extraction

Shorelines were delineated from each DEM using the HAT elevation in ArcGIS Pro 2.5. This tidal datum was selected because the MHW datum was submerged in all surveys, despite efforts to time data collection around low tide and low or offshore wind conditions. This is a common occurrence along this part of the coast as a recent study showed that MHW was continuously submerged or on the wet beach for the period of February 2023 through September 2023 [34]. Although both HAT and MHW are derived from the local tide gauge data, they are computed over different periods from different characteristics. HAT represents the highest predicted astronomical tide, factoring in the gravitational influence of the moon and sun, and is calculated over a 40-year period. In contrast, MHW is defined as the average of all high water observations over a 19-year tidal epoch [35]. The elevation values for both datums were obtained from NOAA Tides and Currents based on data from the nearest tide gauge station, Bob Hall Pier [17]. While this station is no longer operational, it continues to provide reference data from 1 January 1986 to 31 December 2018. At this location, the HAT and MHW elevations corresponded to 0.647 and 0.317 m NAVD88, respectively. For shoreline delineation, a smoothed 0.647 m NAVD88 contour was created to represent the HAT elevation. The landward boundary of the beach was defined by the toe of the seawall.

A seawall-parallel baseline was manually digitized landward of the seawall and was used to generate 150 m-long cross shore transects at 3 m intervals (approximately 10 ft) alongshore using ET Geowizards v12.0, an ArcGIS add-on. This resulted in a total of 439 transects spanning the length of the seawall (see Figure 3). The shoreline and seawall toe were used to segment each transect to isolate the beach zone. Points were generated every 1 m along each segmented transect, from the seawall to the HAT shoreline, with the last segment being a length shorter than 1 m. Then, elevation values from the corresponding DEM grid cells were assigned to each point. These elevation points were used to construct elevation-attributed polylines. From each polyline segment, the 2D length and downhill slope were extracted using ET Geowizards create z characteristics tool. Downhill slope for each polyline was calculated using:

$${Slope}_n={\tan }^{-1}\left({\displaystyle \frac{\sum _{i=1}^m{∆z}_i}{\sum _{i=1}^m{l}_i}}\right)$$

(1)

where ${∆z}_i=\left|z_i-z_{i+1}\right|$ is the elevation change between the start and end point of each polyline segment in the transect going from seawall to shoreline, $l_i$ is the length of each polyline segment, $m$ is the maximum number of points along each polyline, and $n$ is the transect number.

MATLAB R2025b was used to plot alongshore beach width and slope, along with corresponding standard deviations. HAT shoreline change was assessed using the United States Geological Surveys (USGS) Digital Shoreline Analysis System (DSAS) v6.0 software, which calculated linear regression shoreline change rates for the 2023 surveys and net shoreline movement between November 2023 and March 2024 (post-nourishment) [36]. The output tables were imported into MATLAB R2025b to visualize shoreline change. Figure 4 presents a summary of the feature extraction workflow.

Volume Change

For all volume computations (Figure 4), each DEM was clipped using the end transects at the northern and southern ends of the seawall, the toe of the seawall, and the HAT shoreline delineated for each respective DEM. Then, to compute transect-based alongshore volume, the elevation-attributed cross-shore transects explained above were used. The volume beneath each transect was calculated in MATLAB using trapezoidal integration, which takes into account the length of each cross-shore transect, the elevation of each point along the transect, and the spacing between points. This method breaks the area between points into adjacent trapezoids and computes the sum of the areas. A base elevation of 0 m NAVD88 was used as the reference plane. The alongshore volume per transect was plotted using MATLAB and the average, minimum, maximum, and standard deviation of alongshore volume was computed for each respective survey.

Raster-based total volume for each survey was calculated in MATLAB using each DEM raster. First, the height of each DEM grid cell was found above a base elevation of 0 m NAVD88. Then, the elevation value for each cell was multiplied by the cell area. The cells were summed to calculate the total volume. Total volume change was computed by differencing the total volumes for two surveys, with the less recent being subtracted from the more recent.

2.4. MLS/UAS-SfM Comparative Analysis

Comparative analyses were conducted to evaluate the performance and data fidelity of the MLS system relative to UAS-SfM. One key objective was to assess shoreline delineation accuracy by comparing the HAT contours for each method and a manually delineated wet/dry line to HAT in the UAS-SfM DEM. This analysis was performed for the November 2023 MLS and UAS-SfM surveys. Notably, during the February UAS-SfM survey, significant runup occurred during the flight, despite targeting offshore wind conditions. This resulted in an orthomosaic with adjacent alongshore sections showing differing water levels. This made wet/dry delineation nearly impossible. Similarly, HAT could not be applied in the March UAS survey, as water levels rose during the flight, submerging the HAT elevation. As a result, the only complete shoreline analysis was conducted for the November 2023 survey.

The wet/dry line was visually identified using several indicators: the variation in sand color between darker (wet) and drier (light) sand, visible wave runup in the orthomosaic and washover on the MLS vehicle’s tire tracks, which were identified using the MLS system point cloud. The UAS-SfM survey was conducted around the time of low tide, but a rising tide combined with wind-driven wave activity resulted in increased wave runup during the survey. The cross-shore topographic checkpoints were collected during GCP placement and an outgoing tide before the UAS-SfM survey was conducted. As a result, most of the topographic checkpoints ended up within the swash zone by the time of image capture, which was noticeable in the orthomosaic, and could not provide a reliable form of ground truth for this process. Additional inconsistencies appeared in the orthomosaic, where multiple wet/dry boundaries were present in some areas. While some of these aligned with active wave runup, others did not. The upper beach remained wet from the receding tide, creating several ambiguous boundaries left by earlier runup. These conditions made it difficult to interpret the wet/dry line. In these cases, the MLS tire tracks provided a more reliable reference for recent runup extent by tracking areas of washover. These tracks were discernable in the MLS-derived point clouds as areas of lower point density due to the minimum distance filter.

To assess elevation differences between the DEMs, a DEM of difference (DoD) was computed for each survey pair by subtracting the overlapping grid cells of the UAS-derived DEM from the corresponding MLS-derived DEM. The uncertainty ($U$) associated with each DoD was calculated following the method outlined in Wheaton et al. [37] where

$$U_{DoD}=\sqrt{{\left({RMSE}_{DEM1}\right)}^2+{\left({RMSE}_{DEM2}\right)}^2}$$

(2)

The RMSE_z values used in this calculation are listed in Table 1 and Table 2, corresponding to each DEM. Once the DoDs were computed, extreme outliers were removed using the interquartile range (IQR) method which is more robust to non-normally distributed data. The minimize the exclusion of meaningful data, a multiplier of 5 was applied to the IQR based on the 25th percentile (Q1) and 75th percentile (Q3), so that only extreme outliers were removed. The filtered elevation differences (Δz) were analyzed to compute the mean, minimum, maximum, and standard deviation of elevation change across the overlapping DEM areas.

To calculate the raster-based total volume of the UAS-SfM DEMs, each DEM was cropped using the seawall toe, the first and last transects, and the HAT shoreline, represented by a smoothed 0.647 m NAVD88 contour, as previously described. Then, the total volume was calculated using the same methodology outlined earlier in this paper.

For total volume change, elevation differences were computed by subtracting the UAS-SfM-derived DEMs from the MLS-derived DEMs, using only overlapping grid cells. Cells without spatial overlap were excluded from the analysis. The resulting elevation differences were then multiplied by the area of each grid cell to compute volume change per cell. Negative values indicated net loss (erosion), while positive values indicated net gain (deposition). Summing these values yielded the net volume change.

3. Results

3.1. Beach Geomorphology and Shoreline Change

The 2023 beach width results from the seawall toe to the HAT shoreline (Figure 5) revealed seasonal fluctuations, with the narrowest widths occurring during February 2023, coinciding with the winter season. The plot also indicated that the beach widths in June and November were similar, while the widest measurements were recorded in September. The post-nourishment survey conducted in March 2024 indicated a substantial increase in beach width along more than half of the seawall’s length.

The average beach widths for each survey, shown in

Table 3. Beach width, slope, and volume statistics for each MLS system survey, from the seawall toe to HAT. The intersection with the 45.72 m pedestrian safety threshold occurs at the listed distance alongshore from the northern end of the seawall.

	February 2023	June 2023	September 2023	November 2023	March 2024
Mean Beach Width (m)	32.75	42.71	48.28	42.02	64.21
Standard Deviation Beach Width (m)	28.82	25.92	26.99	25.32	6.65
45.72 m Intersect (m)	391.04	475.46	560.87	456.07	-
Mean Beach Slope (°)	3.31	2.11	2.15	2.91	2.17
Standard Deviation Beach Slope (°)	1.08	0.49	0.44	0.66	0.34

, display a variation of about 10 m throughout the 2023 surveys, the most variable of which was in February. The widest average width occurred in September 2023, consistent with seasonal shoreline recession and transitioning winds during that time of year. Following nourishment, the average beach width increased significantly, surpassing the 45.72 intersect alongshore, while the alongshore variability significantly decreased.

Seasonal shoreline change was evaluated using a HAT-based shoreline regression, while post-nourishment change was analyzed using a HAT net shoreline movement calculation, both shown in Figure 6. The shoreline regression indicated a higher rate of change along the southern end of the seawall, consistent with the variations in beach width observed in Figure 5. The highest rates of shoreline change in 2023 occurred from approximately 600 m southward from the northern end of the seawall extending to the southern end of the seawall.

The 2023 beach slope results (Figure 7) showed the steepest slopes occurred during the winter, coinciding with the narrowest beach widths and strong wave runup associated with the arrival of “northers”, as described earlier in this paper. The southern half of the seawall consistently exhibited steeper slopes throughout the year, likely due to its position outside the influence of the southern jetty at Packery Channel. The November 2023 survey also showed increased slope values, likely due to the transition into more frequent periods of onshore winds, driving up water levels. In contrast, the March 2024 post-nourishment survey displayed reduced slope variability, as reflected in the statistics in Table 3. The standard deviation of slope decreased during this survey, likely due to the efforts of beach nourishment and subsequent sand redistribution.

3.2. Volume Change

Alongshore profile-based volume (Figure 8) and corresponding statistics (

Table 4. Volume statistics for each MLS system survey, from the seawall toe to HAT.

	February 2023	June 2023	September 2023	November 2023	March 2024
Mean Volume (m3/m)	129.32	147.32	165.24	152.32	237.35
Minimum Volume (m3/m)	11.17	28.63	33.00	21.16	192.41
Maximum Volume (m3/m)	356.19	361.61	367.06	353.24	287.00
Standard Deviation Volume (m3/m)	108.95	97.68	97.08	100.22	21.16
Total Volume (m3)	51,956	63,415	71,567	68,761	100,453
Total Volume Change (m3)	-	11,459	8152	−2806	31,692

) were computed from the seawall toe to the HAT shoreline proxy. The results reveal seasonal variability in 2023, with increased alongshore volume observed near the northern end of the seawall. The alongshore variability of profile-based volume per survey increased from north to south. Consistent with geomorphic trends of decreased beach slope and increased beach width, volume increased during the summer and the early fall transitional period. However, the post-nourishment profile-based volume shows a decrease in volume at the northern end of the seawall. The post-nourishment standard deviation of profile-based volume also significantly decreased, showing a much more uniform alongshore volume along the entire length of the seawall.

The raster-based volume, computed between the end transects, seawall toe, and HAT shoreline for each respective survey (Table 4), display that beach volume fluctuated throughout 2023, signifying the short-term seasonal variation in the beach. The total volume increased almost 20,000 m³ from February to September 2023. The post-nourishment volume calculation concluded that over 31,000 m³ of sediment were distributed on the beach. This was likely more, but the post-nourishment survey took place in the spring transitional period, almost two months after the nourishment was completed during the winter. Winter winds and wave runup likely eroded parts of the beach and impacted the total volume.

3.3. MLS/UAS-SfM Comparative Analysis

Comparisons between UAS-SfM- and MLS-derived HAT shorelines (

Table 5. Shoreline difference statistics of the February 2023 and November 2023 MLS and UAS-SfM shorelines. In the MLS-UAS-SfM comparisons, negative values indicate that the UAS-SfM HAT was landward of the MLS HAT. In the UAS-SFM HAT-wet/dry comparison, negative values indicate that the wet/dry line was landward of the HAT shoreline.

	February 2023 MLS/UAS HAT	November 2023 MLS/UAS HAT	November 2023 UAS HAT/WetDry
Mean (m)	0.02	0.98	0.36
Minimum (m)	−2.32	2.74	−7.44
Maximum (m)	1.63	−2.96	4.13
Standard Deviation (m)	0.47	−0.84	2.66

) show small differences in February 2023 but larger discrepancies in November 2023, with some areas differing by nearly 3 m. This was likely due to elevation differences in the two datasets, which is discussed later in this paper. In February 2023, the UAS-SfM-derived HAT line was positioned slightly landward of the MLS-derived HAT shoreline, averaging 0.02 m closer to the seawall. Similarly, in November 2023, the UAS-SfM HAT shoreline was also mostly landward of the MLS HAT shoreline, averaging 0.98 m closer to the seawall (Table 5 and Figure 9).

The wet/dry shoreline delineated from the November 2023 UAS-SfM survey was generally aligned with the HAT-based shoreline, as shown in Table 5 and Figure 9. On average, the wet/dry line was located 0.36 m seaward of the HAT contour, suggesting strong overall agreement. However, localized differences were observed. The wet/dry line crossed the HAT shoreline in several locations, was located approximately 7 m landward of the HAT line near the middle of the seawall, and remained shoreward of HAT at the southern end.

Alongshore beach width plots, pictured in Figure 10, highlight the differences between MLS- and UAS-SfM-derived HAT lines as well as between the and UAS-SfM HAT and wet/dry lines for the November 2023 UAS-SfM and MLS surveys. From the northern end of the seawall, the intersection point of the MLS HAT with the 45.72 m line occurred at approximately 495 m, while the UAS-SfM HAT intersected at around 480 m. This indicated that the method of data collection can influence bollard placement, while the difference in this case was small.

The UAS SfM wet/dry line and HAT shoreline proxy, illustrated in Figure 10b produced similar beach width measurements during the November 2023 survey, with minor differences. However, the wet/dry boundary is inherently variable, influenced by tides, wave runup, wind conditions, etc. As mentioned, Vicens-Miquel et al. [34], observed that the wet/dry line was generally positioned landward of the HAT contour, suggesting that HAT may offer a more stable shoreline proxy for these methods of data collection. Nevertheless, it remains unclear whether the wet/dry line is a viable alternative to HAT in this study, because it was only analyzed for one survey. Additional data would be needed to evaluate the seasonal variability of the wet/dry shoreline in comparison to the HAT shoreline. However, in this study the wet/dry shoreline identification was very subjective, which indicated that an elevation-based proxy provided a more reliable measure.

The DoD’s, calculated by subtracting the UAS-SfM DEMs from the MLS DEMs, showed a relatively narrow range of mean elevation differences, shown in

Table 6. Elevation and volume change between MLS and UAS-SfM surveys, where a net loss indicates UAS-SfM volumes were higher than MLS volumes.

	February 2023	November 2023	March 2024 *
Elevation Difference
MeanΔz (m)	0.003	0.024	−0.031
MinΔz (m)	−0.22	−0.25	−0.28
MaxΔz (m)	0.24	0.31	0.22
Standard DeviationΔz (m)	0.04	0.05	0.04
RMSEΔz (m)	0.036	0.058	0.036
Volume Difference
Total Volume (UAS-SfM; m3)	52,782	66,697	102,462
Total Volume (MLS; m3)	51,956	68,761	99,400
Total Volume Difference (m3)	−826	2064	−3062

* The March 2024 UAS-SfM survey did not have an HAT shoreline. Analyses were completed to the shoreward extent of the UAS-SfM survey.

. All mean Δz values ranged under ±0.031 m, with the largest (by magnitude) mean elevation difference observed during the March 2024 post-nourishment survey at -0.031 m. This negative value indicates that, on average, the UAS-SfM DEM was slightly higher in elevation than the MLS DEM for the March 2024 survey, in contrast to the February and November 2023 comparisons. The February MLS DEM was roughly equal in elevation difference, while the November MLS DEM was higher. The minimum and maximum mean Δz values for all surveys also ranged within ±0.31 m. Upon closer evaluation, it was noted that the DEM elevations on the MLS drive paths and between larger passes exhibited greater elevation deviations with the surrounding beach, which could potentially be a factor of scan angle, data occlusion, slight sinking from the vehicle, or another factor. Despite these small differences, the standard deviation of elevation change for all survey pairs remained consistent at 0.058 m or less, suggesting relatively strong agreement between the two datasets.

MLS and UAS-SfM raster-based volume analyses revealed mostly loss in volume, as shown in Table 6. The negative volumes indicate surveys where the UAS-SfM-derived DEMs were generally lower in elevation compared to the MLS-derived DEMs. The results suggest that in February 2023 and March 2024, the UAS-SfM DEMs were generally lower in elevation.

Cross-shore transect profiles of the November 2023 survey, from the toe of the seawall to HAT, pictured in Figure 11, emphasize that the surfaces were similar in overall shape. The northern profile displayed a smoother profile for the UAS-SfM DEM and the two transects overlapped in many places cross-shore. The middle and southern transects clearly show the MLS DEM above the UAS-SfM DEM. However, the UAS-SfM HAT cutoff is shoreward of the MLS HAT cutoff in the middle profile, and landward in the southern profile, showcasing differences in HAT shoreline locations alongshore.

4. Discussion

4.1. Beach Geomorphology and Shoreline Change

The results shown in Table 3 suggest that during 2023, bollards were needed to restrict vehicular access at distances ranging from approximately 390 m to about 560 m from the northern end of the seawall, depending on the time of year. This recommendation was based on using the HAT as a shoreline proxy, since the MHW shoreline was consistently submerged during all surveys, despite being conducted at low tide with favorable wind conditions. For context, bollards were actually placed around 785 m from the northern end of the seawall throughout 2023, indicating that their placement was farther south than necessary to ensure pedestrian safety. However, as mentioned, the original placement was based on the distance from MHW to the seawall toe.

Using HAT instead of MHW for the shoreline can significantly impact the alongshore distance used to determine bollard placement. In this area, HAT is approximately 0.33 m higher than MHW, reducing the measured beach width from the seawall toe and thereby shifts the recommended placement further north, lengthening the alongshore vehicular restriction. If the goal is to ensure a minimum of 45.72 m of dry beach for pedestrian safety, referencing HAT, the wet/dry line, or another measure like the high water line (HWL) may offer a more conservative reference. In a study conducted by Vicens-Miquel et al. [34], the HAT shoreline was almost always shoreward the observed wet/dry line, while the MHW shoreline was consistently submerged, making it extremely difficult to capture with the methods used in this study, and impacts the amount of “safe” area for pedestrians.

Ultimately, the choice of shoreline proxy has significant implications for beach management decisions, including bollard placement. While MHW was the intended shoreline of measure in this study, HAT was exposed in all of the surveys and may be a better reference datum for this method of data collection, as it is usually exposed. In any case, the variation in alongshore distance for bollard placement recommendation in this study varied by 170 m, signifying the substantial short-term changes and supporting the claims in Williams [15] and Williams and Turner [23] which state that seasonal fluctuations emphasize the challenges of beach management in this area.

The shoreline change results showed that the highest rate of change was from about 600 m from the northern end of the seawall, extending south, aligning with the previously recommended area for bollard placement, based on the HAT shoreline proxy. Williams [14] noted that shoreline change in front of the seawall is largely seasonal, and is dependent on wind, surge, waves, etc., and is evident by the rate of change over the course of 2023. Post-nourishment net shoreline movement revealed significant alongshore accretion along with pronounced erosion at the northern end of the seawall. These patterns were likely influenced by sand redistribution during the nourishment and erosion during the winter of 2023–2024 and into spring 2024.

4.2. Volume Change

To further analyze the alongshore seasonal variability and post-nourishment impacts, cross-shore transects at the northern end, midpoint, and southern end of the seawall were plotted, illustrated in Figure 12. These transects extend from the toe of the seawall to the HAT shoreline proxy. At the northern transect (Figure 12a), beach elevation was higher on the berm and backbeach area adjacent to the seawall. The post-nourishment survey at this transect displayed a shorter transect profile and a lower berm, features that are not considered in the total volume and total volume change shown in Table 4. As explained, winter winds and storms can cause strong onshore winds, elevated water levels, and wave-runup [14]. This, combined with mechanical redistribution of sand southward, likely contributed to the observed geomorphology.

The middle transect (Figure 12b) exhibited varying beach profiles, with steeper slopes in February and September of 2023. At this point of the seawall, beach widths were about half the length that they were at the northern transect, showcasing the range of alongshore variability. This plot also presents the post-nourishment transect at that location, where the profile overall was much longer, displaying that the beach width significantly increased after the nourishment, as evidenced in Table 3.

The southern transect (Figure 12c) consistently exhibited steep slopes and narrow beach widths across all of the 2023 surveys, aligning with its documented history of high erosion rates. Williams [14] reported that erosion is most severe from the middle of the seawall southward, with the southern end showing the greatest variability in beach width and elevation. Post-nourishment, this area showed a beach width almost three times greater than pre-nourishment conditions and a higher beach elevation indicating a substantial volume of sand was placed there.

4.3. MLS/UAS-SfM Comparative Analysis

In the UAS-SfM HAT and wet/dry shoreline analysis (Table 5 and Figure 9), it was noted that the UAS-SfM orthomosaic showed multiple wet/dry boundaries, representing various stages of wave runup prior to the survey. The delineated shoreline fell between these visible runup marks, which was above the swash zone visible in the orthomosaic and below the upper limit of sand color change, indicating retreating tidal levels. These complexities, indicating ongoing wave activity, low tide, incoming tide, and increased wave action during the UAS-SfM survey, limited the reliability of comparing the delineated wet/dry shoreline to the ground truth topographic transects. Runup at the time of data acquisition likely shifted the apparent wet/dry shoreland inland, creating inconsistencies and rendering the ground truth data unreliable.

These inconsistencies reinforce the importance of shoreline proxy selection when interpreting results. Moore et al. [38] define the HWL as “the landward extent of the last high tide” and found that it differed from MHW by a horizontal average of 18.8 m, emphasizing how the choice of shoreline proxy can greatly affect beach width estimates. To compare, the HWL shoreline was delineated in the November 2023 UAS-SfM orthomosaic, finding that the HWL and HAT shorelines were an average of 12.27 m apart, with the HWL consistently landward of HAT. Nasir et al. [39] also reported a datum bias of up to 32 m between the HWL and MHW shorelines, further demonstrating how such biases can direct interpretations of shoreline positions. Both studies emphasize that the choice of shoreline proxy directly affects the significance of shoreline analyses.

The November 2023 UAS-SfM wet/dry line had an average elevation of 0.62 m NAVD88, with a standard deviation of 0.11 m. Gibeaut and Caudle [40] reported that the wet/dry line typically occurs at 0.6 m above mean sea level, which converts to approximately 0.73 m NAVD88 at the NPI seawall. They also emphasize that the wet/dry boundary is highly influenced by wave activity and recent water levels, as well as subjectivity in manual delineation methods. Additionally, Moore et al. [38] argue that a contour-based shoreline proxy can provide a more objective and repeatable shoreline reference than visual-based shorelines. Ultimately, the shoreline has a significant impact of beach management depending on which shoreline proxy is chosen. As seen in this study, MHW, HAT, and even the wet/dry line were in different locations and would have produced different results that can influence decisions related to beach access, bollard placement, and coastal planning.

When interpreting the differences in volume between the MLS and UAS-SfM DEMs (Table 6) it was found that the UAS-SfM- and MLS-derived DEMs showed relatively strong agreement, with neither having overall higher or lower elevation values. Visual analyses of the DoDs revealed that the highest level of volume change in the February and November 2023 survey pairs was at the northern end of the study site. The March 2024 surveys indicated areas of both higher and lower elevation change, with the MLS DEM higher in the northern and southern ends of the seawall. It was also noted that, in every MLS DEM, the DoD elevation was lower immediately adjacent to the drive path. This indicates that the MLS DEM elevation was consistently higher in these areas. As discussed above, this could be likely due to the lidar’s scan angle, sand compression from the vehicle, or another factor. Additionally, the February 2023 survey pairs displayed the lowest amount of volume change, indicating the highest agreement of the three survey pairs. The lower volume observed is likely influenced by the RMSE_z of the two DEMs, with the February MLS survey having the lowest RMSE_z for that system. This highlights the importance of consistent data collection and processing methods to minimize vertical error.

It should be mentioned that the accuracy of the DEMs ultimately influenced the resulting measurements and computations discussed above. Although the data used was high-resolution and the same reference base station was used for georeferencing and ground truthing, each survey contained its own uncertainty and error, listed in Table 1 and Table 2. This resulted in various RMSE_z values. This ultimately impacted the HAT elevation proxy, beach width, beach slope, volume, and other measurements.

4.4. MLS/UAS-SfM Operational Analysis

While the MLS and UAS-SfM DEMs displayed variations in elevation, profile, and volume, additional differences stemmed from their deployment, data collection techniques, and processing methods. The MLS system used in this study is an active system that emits its own near-infrared (NIR) light, making it less sensitive to ambient lighting conditions. In contrast, the UAS is a passive system that relies on sunlight to illuminate the ground and other surfaces.

Both systems support rapid deployment, but each has operational limitations related to environmental conditions. The UAS used in this study is more sensitive to wind and weather. High winds can reduce accuracy and increase power consumption, requiring more GCPs and batteries, increasing survey time [5]. Che Mat and Tahar [9] discussed that high winds can affect the accuracy of a UAS survey and more GCPs should be used on windier days. Furthermore, the Federal Aviation Administration (FAA) also prohibits small UAS (under 25 kg) from flying over non-participants who are not under a covered protective structure or in a vehicle [41]. This regulation limits UAS flight opportunities to times when beaches are largely empty, which is often during suboptimal wind and/or lighting conditions. Starek et al. [32] also noted that lighting conditions influence the radiometric quality UAS imagery, and that sandy beaches, with their low surface texture, pose a challenge for UAS-SfM performance.

MLS systems, on the other hand, are better suited for fairly high winds, such as post-storm conditions, and can operate on crowded beaches [25]. However, they require vehicular access and enough space to complete multiple passes, which can be challenging in narrow, rugged, or flooded terrain where UAS platforms may have an advantage [30]. In this study, MLS surveys were faster, typically requiring only 2–3 passes and a single battery.

The UAS-SfM surveys, by comparison, were more time-intensive, requiring several flights and batteries, particularly on windy days, which increased power demands to maintain flight stability. The survey time could have been reduced by flying at a higher elevation and using a less rigorous flight pattern. However, these changes would have negatively impacted data quality. A higher flight altitude would increase the ground sample distance (GSD), reducing spatial resolution. Likewise, a less detailed flight pattern would reduce image overlap and weaken tie point matching, potentially introducing greater error along feature-poor areas [31]. The UAS-SfM surveys also required more time and processing power than the MLS surveys, as confirmed by Starek et al. [32], who discussed the differences between UAS-SfM and UAS-based lidar.

UAS-SfM does offer the advantage of a nadir view and produces: (1) a consistent point density, and (2) red, green, blue (RGB) orthomosaics; both of which depend on flying altitude and camera resolution. In contrast, MLS systems scan radially, resulting in a changing point density based on scan pulse rate, driving speed, and sidelap. Geometric accuracy also declines with increasing distance from the lidar sensor [42]. Additionally, ground-based MLS systems are prone to data occlusion in areas behind foredunes, vehicles, beach tents, or beach structures due to their oblique scan angle [1,2]. Additionally, neither system has the capacity to collect useable data in the swash zone.

The high-energy coastal environment presented unique challenges for UAS-SfM processing. Dynamic elements such as moving vehicles and wave action interfered with tie point detection, lowering the quality of the resulting point cloud and orthomosaic as discussed in Bertin et al. [43]. Zekkos et al. [44] also noted that the lack of texture and identifiable features, such as uniform beach sand which lacks contrast, can make mapping sandy beaches difficult and unreliable.

Water removal was challenging for both methods. It can be improved through algorithms, machine learning, or manual techniques, such as delineation of the wet/dry line [45,46]. In this study, ground truth data aided water removal in the MLS surveys, but the process remained subjective. In addition, the wet/dry line could be delineated in the UAS-SfM orthomosaics, but not in the MLS-derived point clouds and resulting DEM. However, this process is also subjective and can be difficult, as described earlier with the November 2023 UAS-SfM dataset. In the future, the ground truth topographic transect datasets should be collected simultaneously with the MLS and UAS-SfM surveys to limit time differences and reduce the risk of wave runup. In February 2023, the UAS SfM survey was conducted during low tide, allowing time to set up the base station, collect ground truth, and prepare both the MLS system and UAS equipment. However, fast-changing winds and wave runup still altered the wet/dry line during the UAS flight, making it difficult to consistently delineate the shoreline. In some cases, waves were captured mid-runup during rising tide conditions, leading to irregular or blurred shoreline features in the resulting data, an example of which is illustrated in Figure 13. These issues were largely due to the longer duration required for UAS flights, which increased the likelihood of environmental variability. Filtering water from the MLS datasets was also challenging, particularly because the system lacked RGB imagery to support classification during postprocessing. However, elevation and intensity values were used to help identify and classify water points.

4.5. Appliction of This Study

This study can be used to support coastal management by providing the City of Corpus Christi and other local decisionmakers actionable guidance for seasonal bollard placement. By quantifying seasonal shoreline position changes, beach width, beach slope, and volume change, the results suggest a spatial and temporal basis for bollard placement ideal for sustaining compliance with Ordinance No. S10-77 [18]. This study found seasonal placement of 390 m to 560 m from the northern end of the seawall in accordance with the 45.7 m threshold, but this may vary from year-to-year, and more data is needed to support this claim. Still, the methods and recommendations presented in this paper offer a starting point that could eventually shift management from reactive to more anticipatory, adjusting bollards with seasonal transitions rather than after the fact. In addition, the post-nourishment shoreline change of over 40 m in some locations illustrates how engineered intervention can improve vehicular management and pedestrian safety, reinforcing the value of regular monitoring for evaluating project performance and gonotrophic response.

In addition to bollard placement, this study integrated MLS and UAS-SfM, demonstrating high-resolution, repeatable data collection methodologies that stakeholders/decision makers can use for long-term planning and monitoring. The difficulties in collecting reliable data and ensuring low tide/offshore winds discussed earlier highlight the importance of how seasonal environmental conditions and survey methods influence shoreline mapping and monitoring. These insights can guide decisions about how frequently data should be collected. Collectively, the results support data-driven policymaking by linking beach geomorphology to ordinance compliance, infrastructure protection, and prioritization of future nourishment or mitigation effects.

5. Conclusions

In conclusion, MLS is an effective method for mapping and monitoring sandy beach environments. It demonstrated vertical accuracy comparable to UAS-SfM, with RMSE_z values ranging from 0.02 m to 0.05 m in this study, and it can be operated in conditions unsuitable for UAS deployment. During the study period, the beach experienced substantial alongshore and seasonal variability, with beach widths, from the toe of the seawall to the HAT shoreline proxy, varying by nearly 20 m and slope by about 1°. Following the nourishment, approximately three-quarters of the beach showed increased widths and reduced slopes. Volume analysis indicated that the total volume change post-nourishment was 31,692 m³. However, the post-nourishment survey took place over a month after the nourishment concluded and these results do not account for subsequent coastal processes such as tides, wave runup, wind, erosion, etc.

The results suggest that HAT may be a more appropriate reference datum for this method of data collection than MHW for bollard placement, as MHW is submerged for most of the year. Additionally, according to the HAT-based analysis, bollard placement should have been positioned farther north to better restrict vehicle access and increase pedestrian safety. However, this method provided a different base of measure to indicate beach width.

A comparative analysis of MLS system and UAS-SfM DEMs found that shoreline delineation using the HAT datum showed the closest agreement in the February 2023 survey, with an average difference of 0.02 m and a standard deviation of 0.47 m. In contrast, the November 2023 survey showed greater divergence, with an average difference of 0.98 m and a standard deviation of 0.84 m. In both cases, the MLS system-derived HAT shoreline was generally seaward of the UAS-SfM-derived HAT shoreline. A comparison of HAT and the wet/dry line from the February 2023 UAS-SfM survey showed that the wet/dry line was, on average 0.36 m shoreward of HAT. The method of delineating HAT, however, was difficult to perform due to there being multiple wet/dry lines from a receding tide and varying levels of wave runup over time.

Elevation comparisons between the UAS-SfM and MLS system DEMs showed that the MLS system and UAS-SfM DEMs varied elevation, with neither being consistently higher or lower, indicating relatively strong agreement. The largest average Δz value was −0.031 m for all of the survey pairs. Total volume change indicated that all survey pairs displayed variations in volume, with higher total volume differences between the November and March surveys. The February 2023 volume difference was the lowest, suggesting stronger agreement between the surveys.

This study contributed to understanding seasonal changes in beach morphology adjacent to the NPI seawall and identified optimal seasonal bollard placement, which ranged from approximately 430 m to 600 m alongshore from the northern end of the seawall, referenced to HAT. The difficulty in capturing MHW using the methods in this study indicates that, to use these methods of data collection for routine monitoring for management, a higher tidal datum for monitoring beach width for vehicular accesses needed, ensuring pedestrian safety on this beach.