A UAV–RTK Lidar System for Wave and Tide Measurements in Coastal Zones
1. Introduction
Airborne scanning lidar is frequently used for mapping topography in coastal areas and for monitoring shoreline variations (e.g., Stockdon et al. 2002) and beach erosion (e.g., Sallenger et al. 2003). Lidar has been extended and proven to be a useful tool for coastal sea surface (e.g., Reineman et al. 2009; Vrbancich et al. 2011) and wave dissipation rate measurements (e.g., Huang et al. 2012). However, the abovementioned studies rely on massive lidar systems, such as Reigl models carried by manned aircrafts.
In comparison with manned aircraft, unmanned aerial vehicles (UAVs) offer a viable alternative to conventional platforms at lower cost, higher operational flexibility, and greater versatility. UAVs have been widely used as an imagery mapping tool for a variety of applications. Review papers of UAV development and applications in civil engineering and coastal and environmental remote sensing can be found in Colomina and Molina (2014), Liu et al. (2014), and Klemas (2015). In applications to coastal research, UAV imagery techniques provide high-spatial- and high-temporal-resolution images that aid in monitoring coastal environments. Brouwer et al. (2015) demonstrated that UAVs are a useful tool for obtaining surfzone and beach characteristics in response to storms and for estimating the temporal and spatial scales of dispersed pollutants. UAVs have proven to be an efficient and cost-effective survey tool for topographic mapping in coastal zones, including mapping beach erosion resulting from storms (Ierodiaconou et al. 2016; Turner et al. 2016) and changes in a tidal inlet (Long et al. 2016). In addition, Holman et al. (2017) applied a UAV to obtain surfzone images without ground control points. They showed that derived Argus products using a UAV agree well with the same products collected using a traditional fixed Argus station. UAVs have been demonstrated to be another useful alternative to fixed camera systems for coastal monitoring.
Given developments in lightweight lidars and larger UAVs that can carry heavier payloads, it is possible to integrate a lidar with a UAV system. Recently, Bandini et al. (2017) equipped a camera-laser distance sensor, a radar, and a sonar into a UAV for measuring water level in rivers and lakes. With a Global Navigation Satellite System (GNSS) providing a relative vertical accuracy better than 3–5 cm, the accuracy of water-level measurements ranges from 5 to 7 cm. Sankey et al. (2017) integrated a lidar system into an octocopter aircraft (additional payload up to 6.5 kg) for topography measurements over forests. The lidar data produced a highly accurate digital elevation model with a root-mean-square (RMS) error of 0.75 m. See the appendix for a list of acronyms used in this paper (Table A1).
This paper assesses the potential for integrating a multirotor UAV, a real-time kinematic (RTK) positioning module, and a small robotic lidar sensor to measure sea surface displacements in coastal zones. Section 2 describes the components of the system. The data processing and calibration are presented in section 3. The performance of the RTK lidar system and algorithm was first validated using a rod and then validated using a UAV, as presented in section 4. Field measurements of tidal elevation (water depth), wave spectrum, wave height, and wave period were compared with the results measured using a pressure sensor, as shown in section 5. The discussion and summary are presented in the last section.
2. System descriptions and measurements
a. Components
The RTK lidar system included four key components: a robotic scanning lidar, an altitude and heading reference system (AHRS), an RTK GNSS, and a small industrial personal computer (PC), as shown in Fig. 1. The system can be carried by a man using a rod, a UAV (a UAV–RTK lidar), or other vehicles. The components of the system and the UAV are now described.
Photograph of the UAV–RTK lidar system. The system includes a UAV, a scanning lidar instrument, an AHRS, and a PC.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Photograph of the UAV–RTK lidar system. The system includes a UAV, a scanning lidar instrument, an AHRS, and a PC.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Photograph of the UAV–RTK lidar system. The system includes a UAV, a scanning lidar instrument, an AHRS, and a PC.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
1) Robotic scanning lidar
The scanning lidar (Hokuyo, UTM-30LX) is a lightweight type (approximately 300 g) that is usually used in intelligent robot industries for detecting distance, shapes, and surface patterns for objects (e.g., Lee and Ehsani 2008). The lidar uses a class one laser light source of 870-nm wavelength. The motor rotation speed of the lidar is 2400 rpm, which results in one scan being completed in 0.025 s (40 Hz). It measures distances to objects in range, and the coordinates of those points are calculated using the step angles. The front of the device is 0°, and the scanning range is from −135° to 135° in steps of 0.25°. The lidar was operated to scan in the range of ±30°, resulting in 241 distance measurement points in one scan.
The maximum detection range is approximately 30 m, but we found that the lidar can obtain only over 80%–90% useful data returns at an altitude of 20 m over the land surface; our flights are therefore controlled at a height within 20 m for topography measurements over land. However, we further found that the rate of data return is much less for sea surface measurements in the field, and therefore the flights were controlled at heights of 6–10 m over the ocean. According to the user manual, the RMS precision and repeated accuracy of the lidar are 1 cm, and the absolute accuracy is approximately 3 cm when the measuring range is within 10 m (Hokuyo Automatic Co. 2012).
2) AHRS
The attitude and rotation angles of the lidar system are measured using an AHRS (Xsens Technologies, MTi 30) at a sampling rate of 100 Hz. In comparison with traditional inertial measurement units (IMUs), this sensor is based on a microelectromechanical system that provides several advantages, such as being inexpensive, lightweight, low power, and compact. It adopts the extended Xsens Kalman filter, version 3 (XKF3), which is an optimal estimation algorithm that fuses data from gyroscopes, accelerometers, and a magnetometer to estimate three-dimensional orientation in the earth-fixed frame of reference (Xsens Technologies 2016). The accuracies of the tilt (pitch and roll) angles are 0.2° and 0.5° in static and dynamic statuses, respectively, and the heading accuracy is 1° (yaw/heading). XKF3 is a proven sensor fusion algorithm developed by Xsens. It uses signals from rate gyroscopes, accelerometers, and magnetometers to compute a statistically optimal 3D orientation estimate of high accuracy with no drift in both static and dynamic operation. This type of drift compensation is often called attitude and heading referencing, and such a system is referred to as an AHRS (Xsens Technologies 2016).
3) RTK GNSS
An RTK GNSS is used for positioning the lidar system. The system includes two GNSS receivers (NovAtel, OEM 628), two antennas, and two lightweight portable radios (433 MHz) to send the radio technical commission for maritime services message from the base station to the rover for differential correction information. When the RTK GNSS is operated under the RT-2 mode (a trademark of NovAtel), it provides an RMS accuracy of 3D positioning in 1 cm according to the specification (NovAtel 2016). The positioning system is operated to undertake sampling at 20 Hz in the World Geodetic System 1984 (WGS84) coordinate system. In operation, the position of the base station is first determined using a virtual base station (VBS)-RTK technique. This technique allows us to perform a rapid and high-quality positioning measurement with only one receiver for the base station. After verifying the actual position of the base station, the classic RTK technique with two receivers and two radios is used in the UAV lidar system.
There are many techniques that can obtain highly accurate positions for a rover station. For example, the VBS-RTK is useful for positioning a lidar system with only one GNSS receiver. However, it requires stable telecommunications to send the differential correction information; unfortunately, the broadband cellular network is not always stable in the coastal zones of Taiwan. Postprocessing of GNSS data is another choice for accurately positioning the rover station (e.g., Reineman et al. 2009). Because the RTK technique works well for operating our system and possibly provides other future applications such as real-time navigation and advantages of obtaining instant responses from the system, the classic RTK technique is therefore adopted in our system.
4) Multirotor UAV
The RTK lidar systems were carried by a commercial multirotor-UAV system (DJI, S1000). The weight of the UAV with one battery (25.2 V, 23 000 mAh) is approximately 7 kg, and the weight of our instruments is approximately 3 kg. The gross weight of the entire assembly is controlled to be less than the maximum safe takeoff weight of 11 kg. After testing, the safe flight time is approximately 10–12 min when using one battery. The battery has to be replaced on the ground between flights. We prepared six spare batteries and recharged the spare batteries during the experiments. The total flight time can exceed 1 h. The UAV system can be controlled by a PC-based ground station to monitor the flight information and to set the flight routes. The control signals were sent by an additional 900-MHz radio system.
b. Measurements
The UAV supports auto-takeoff and auto-landing functions; however, it was manually operated during takeoff and landing for safety reasons. It was switched to an autopilot mode between takeoff and landing to follow the preset flight routes. In the operation, the lidar was set to scan one profile in 0.25 s (40 Hz). The sampling rates of the AHRS and RTK systems were set at 100 and 20 Hz, respectively. The data were recorded by a small embedded industrial computer (Intel Core i7 processor) using Labview.
All of the data samples were recorded in the computer with time stamps. Synchronization between instruments is a challenge of the system because the robotic lidar cannot be externally triggered. Generally, hardware synchronization between instruments could be achieved by triggering the instruments using transistor–transistor logic (TTL) signals. The time offset of the synchronization could thus be smaller than 1 μs or much smaller. The external trigger helps to adjust the timing of each instrument to sample the data at the same time. However, we could not “actively” trigger the instruments using TTL signals. The synchronization was therefore achieved using a “passive” synchronization by recording the data with synchronously recorded PC time stamps (Fig. 2a). There might be some time offset due to data processing, transmission, and recording between instruments. The accuracy of the passive synchronization relies on the accuracy of the time stamp recording. The accuracy of the passive synchronization was checked by comparing the recorded time stamps to the recorded GNSS time. Because the accuracy of the Labview time stamp is 1 ms, only time offsets larger than 1 ms can be identified. We found that there is no time offset larger than 1 ms in the system over hours, suggesting that the error of the passive synchronization is less than 1 ms in operation.
(a) RTK lidar system data acquisition flowchart. (b) Postprocessing flowchart of the lidar data. (c) Schematic diagram of the geometry of the lidar and GPS antenna, and definition of the three-axis rotation angles used to transform the measured surface elevations into the Earth coordinate system.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
(a) RTK lidar system data acquisition flowchart. (b) Postprocessing flowchart of the lidar data. (c) Schematic diagram of the geometry of the lidar and GPS antenna, and definition of the three-axis rotation angles used to transform the measured surface elevations into the Earth coordinate system.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
(a) RTK lidar system data acquisition flowchart. (b) Postprocessing flowchart of the lidar data. (c) Schematic diagram of the geometry of the lidar and GPS antenna, and definition of the three-axis rotation angles used to transform the measured surface elevations into the Earth coordinate system.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Orientation, position, and range measurements are required to be taken at the same time for accurate 3D positioning. A time offset between the instruments will cause a small effect for a calm flight and largely influences the positioning for a turbulent flight (Baltsavias 1999). The error of time offset may affect the performance of the system. More detailed discussions on the accuracy as a result of the time offset are discussed in the next section.
3. Data processing and calibration
a. Data processing
The data processing procedure is shown in Fig. 2b. All of the recorded data are resampled in a time frame of 20 Hz with linear interpolation. The RTK GNSS is operated in RT-2 mode, which provides the georeferenced position of the rover with an accuracy of approximately 1–2 cm (NovAtel 2016). The measured points 
b. Calibration
The accuracies of the AHRS sensor are 0.25°–0.5° and 1° for tilt and heading data, respectively, which are much larger than those of high-standard IMUs. However, the errors may decrease when the system is operated at low altitudes. The attitude errors can be considered as a sum resulting from mean and random errors. The mean error is the system bias that is caused by the coordinate offset and the boresight angles. The random errors are mainly caused by sensor noise. The system biases are the static alignment offsets between the lidar and the AHRS. The system should be calibrated to reduce the mean errors. A calibration experiment of repeated measurements over two horizontal plates was conducted to determine the static offsets of the sensor attitudes. The parameters needing calibration include 
4. Performance, validation, and error analysis
The RTK lidar system was alternatively carried using a rod and a UAV. The performance and validation are presented in the following sections.
a. Validation of the RTK lidar system
The RTK lidar system was first installed on a rod that was able to move with a maximum height 3 m above the ground. Repeated measurements over an open area and over rectangular boxes were performed to quantify the performance of the system. For testing over flat ground, the system was vertically and horizontally moved over vertical and horizontal distances of approximately 1–2 m. The results of the repeated measurements over the test site are summarized in Table 1. The averaged difference in mean elevation was 1.46 cm. The root-mean-square error (RMSE) was computed using all the samples of the five test runs. The standard deviation (std) of the height measurements is approximately 1 cm, and the difference between the values of 
Repeated RTK lidar measurements over flat ground using a rod for heights between 1 and 2.5 m with vertical and horizontal motions. The averaged difference of the mean elevations is 1.46 cm. The RMSE was computed using all the samples of the five test runs.
Figure 3 shows 11 repeated measurements over four rectangular boxes with horizontal motions (defined as the x direction) at a height of 3 m. The standard deviation errors of the detected boundary in the y and x directions are less than 0.8 and 2 cm, respectively. Indeed, the error in the x direction depends on the speed of the moving system; the speed of the moving system was 6–10 cm−1, and the sampling rate of the system was 20 Hz, which resulted in a possible minimum error exceeding 1 cm. The quantified error is under a reasonable value for speeds less than 10 cm−1 at the lidar scan rate. We should note that the error when detecting boundary edges depends on the spatial resolution of the lidar data and will be much greater when the system is operated at a greater speed. The spatial resolution indeed depends on the speed of motion and the height of the lidar. Spatial resolution is discussed in the next section.
RTK lidar measurements over four rectangular boxes. The system was manually carried by a rod. (a) Sample of the measured elevation map, with the white circles showing the detected object boundaries. (b) Detected object boundaries of 11 repeated measurements. (c),(d) The corresponding standard deviation (std) errors of the detected boundaries from the repeated measurements.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
RTK lidar measurements over four rectangular boxes. The system was manually carried by a rod. (a) Sample of the measured elevation map, with the white circles showing the detected object boundaries. (b) Detected object boundaries of 11 repeated measurements. (c),(d) The corresponding standard deviation (std) errors of the detected boundaries from the repeated measurements.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
RTK lidar measurements over four rectangular boxes. The system was manually carried by a rod. (a) Sample of the measured elevation map, with the white circles showing the detected object boundaries. (b) Detected object boundaries of 11 repeated measurements. (c),(d) The corresponding standard deviation (std) errors of the detected boundaries from the repeated measurements.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
The lidar can be operated at low altitudes, and it is therefore possible to measure small objects with heights of an order of magnitude of a few centimeters. Figure 4 shows the lidar measurements over rectangular boxes with different box heights. Errors similar to those presented in Table 1 were obtained. The height measurement errors found by computing the relative height in each scanning line are quantified in Fig. 5. Small errors below 1 cm are found because the system errors are minimized during each scan. Therefore, the height measurement provides the opportunity to quantify surface roughness in the intertidal zones of gravel beaches and rivers.
Repeated RTK lidar measurements over four rectangular boxes. The RTK lidar was moved up and down at heights between 1.5 and 2.5 m. The widths (W) and heights (H) of the boxes were (a) W = 22.8, H = 9.4; (b) W = 36.3, H = 7.4; (c) W = 22.8, H = 4.9; and (d) W = 10.5, H = 3.4 cm. The true elevations of the boxes (black dashed lines), the 60 individual profiles scanned by the lidar (light gray lines), and the mean values (dark gray lines) and std values (error bars) of the individual profiles.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Repeated RTK lidar measurements over four rectangular boxes. The RTK lidar was moved up and down at heights between 1.5 and 2.5 m. The widths (W) and heights (H) of the boxes were (a) W = 22.8, H = 9.4; (b) W = 36.3, H = 7.4; (c) W = 22.8, H = 4.9; and (d) W = 10.5, H = 3.4 cm. The true elevations of the boxes (black dashed lines), the 60 individual profiles scanned by the lidar (light gray lines), and the mean values (dark gray lines) and std values (error bars) of the individual profiles.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Repeated RTK lidar measurements over four rectangular boxes. The RTK lidar was moved up and down at heights between 1.5 and 2.5 m. The widths (W) and heights (H) of the boxes were (a) W = 22.8, H = 9.4; (b) W = 36.3, H = 7.4; (c) W = 22.8, H = 4.9; and (d) W = 10.5, H = 3.4 cm. The true elevations of the boxes (black dashed lines), the 60 individual profiles scanned by the lidar (light gray lines), and the mean values (dark gray lines) and std values (error bars) of the individual profiles.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison off the relative heights measured in each scanning line by the RTK lidar system against the true height (without using the UAV). The circles and diamonds represent objects at different heights, and the vertical error bars are the standard deviations of the repeated measurements from averaging by the lidar. A 1:1 ratio (solid line) and the 1-cm deviations from the 1:1 ratio (dashed lines) are denoted.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison off the relative heights measured in each scanning line by the RTK lidar system against the true height (without using the UAV). The circles and diamonds represent objects at different heights, and the vertical error bars are the standard deviations of the repeated measurements from averaging by the lidar. A 1:1 ratio (solid line) and the 1-cm deviations from the 1:1 ratio (dashed lines) are denoted.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison off the relative heights measured in each scanning line by the RTK lidar system against the true height (without using the UAV). The circles and diamonds represent objects at different heights, and the vertical error bars are the standard deviations of the repeated measurements from averaging by the lidar. A 1:1 ratio (solid line) and the 1-cm deviations from the 1:1 ratio (dashed lines) are denoted.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
b. Validation of UAV–RTK lidar system
The RTK lidar system was carried by the rotary UAV in a hover mode with up and down movements to test the performance of the system. First, an experiment with repeated measurements over flat ground was conducted. The ground truth values were not recorded in the experiment. The flight height was 6–7.5 m. The difference between the 1% highest (
Other experiments with repeated measurements over horizontal plates were conducted and compared to the ground truth values measured by the manual RTK VBS technique. Figure 6 illustrates the measured data of the position of the GNSS; the 
Validation of the UAV–RTK lidar over a horizontal plate. (a) GNSS altitude; (b) range data measured by the lidar; (c),(d) pitch and roll measured by the AHRS, respectively; and (e) elevations of the processed lidar data (gray line) and ground truths measured by the manual RTK VBS (black line). The RMSE between the georeferenced lidar data and the ground truth values is 3.4 cm.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Validation of the UAV–RTK lidar over a horizontal plate. (a) GNSS altitude; (b) range data measured by the lidar; (c),(d) pitch and roll measured by the AHRS, respectively; and (e) elevations of the processed lidar data (gray line) and ground truths measured by the manual RTK VBS (black line). The RMSE between the georeferenced lidar data and the ground truth values is 3.4 cm.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Validation of the UAV–RTK lidar over a horizontal plate. (a) GNSS altitude; (b) range data measured by the lidar; (c),(d) pitch and roll measured by the AHRS, respectively; and (e) elevations of the processed lidar data (gray line) and ground truths measured by the manual RTK VBS (black line). The RMSE between the georeferenced lidar data and the ground truth values is 3.4 cm.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
The temporal and spatial resolutions of the UAV–RTK lidar system can be found in Baltsavias (1999). The spatial resolution of the system can be divided into along-track and cross-track (scanning direction of the lidar) resolutions. The along-track resolution can be computed as 
c. Error analysis
The error (accuracy) of the 3D positioning was discussed for a flight height of 10 m using the basic relations provided by Baltsavias (1999). The major factors affecting the accuracy of the airborne lidar system include 1) the range accuracy of the lidar; 2) the 3D position accuracy of the GNSS; 3) time offsets between instruments; and 4) the accuracy of the roll, pitch, and heading angles of the AHRS. They are discussed as follows.
The range accuracy of the lidar that we used is approximately 3 cm for a 10-m height measurement (Hokuyo Automatic Co. 2012). The lidar’s motor may not be always precise and accurate, which may affect the lidar’s ranging measurements. It may be expected that the motor’s backlash errors may be amplified during a turbulent flight.
When the RTK GNSS is operated in RT-2 mode, the RMS accuracy of the RTK GNSS 3D positioning will be within 1 cm ± 1 ppm (NovAtel 2016). Quality control of the GNSS data was performed to ensure the quality of the GNSS data. Only data in RT-2 mode were selected for further processing; therefore, the error caused by the RTK GNSS is within 1–2 cm.
The time offset between the instruments is less than 1 ms, that is,
The accuracies of the tilt (pitch and roll) angles are 0.5 ° in a dynamic state and 1° in heading. Assuming that the error of the lidar scanning angle is much smaller than the tilt angles, the influence of the roll and pitch on vertical positioning is approximately
The accuracy of the system depends on the accuracy of the components and how the components are integrated. For a flight height of 10 m, the accuracies of the lidar and the RTK GNSS are approximately 3 and 1–2 cm, respectively, and the accuracy is 4–7 cm when the variations in the tilt angles are 5°–7° (Baltsavias 1999). The final accuracy of the system can be determined by comparison with the ground truth values. The RMSE of the system is within 5 cm for a flight height of 10 m. This is because the system was operated in a hover mode at a low altitude of 10 m. The errors caused by the low-cost lidar and altitude sensor are therefore decreased.
5. Field measurements and validations
a. Field experiments
Validations of the UAV–RTK lidar system in the field were performed on 25 August 2017 in the intertidal zone near YongAn Harbor, northern Taiwan. A strain gauge pressure sensor equipped in an acoustic Doppler velocimetry (SonTek ADV-Oceans) was used to validate the results measured by the UAV system. The pressure sensor was deployed in the intertidal zone during low tide and scanned 19 000 data samples at 16 Hz in every 20-min burst. Surface elevations of one tidal cycle from low to high tide were measured continuously by the pressure sensor.
Figure 7 shows a data sample measured by the UAV–RTK lidar system over the sea. However, the most challenging problem is the performance of the robotic lidar. The return rates for professional and expensive lidars—for example, a Riegl LMS-Q240i—are 0.3–0.4 (Huang et al. 2012; Reineman et al. 2009). In comparison with the Riegl lidar, the data return rate was only 0.1–0.2 when the flight was operated at heights of 10–6 m over the sea (Fig. 7b). A higher return rate for lidar data is likely to occur with a lower flight height. Considering the data quality and the safety of the UAV, the flight was operated at a height of approximately 6 m during the field measurements. Approximately 8 min of sea surface elevations were measured by the system in one flight. Nine flights separated by hours were performed in the field experiment.
Field tests of the UAV–RTK lidar measurements over the sea surface. (a) Sample of the raw lidar data. The elevation is the distance measured by the lidar. (b) Return rate of the lidar of (a). (c) Return rates of the lidar over the sea surface for nine flights. (d) The maximum return rates of (c) vs flight height over the ocean (distance between the lidar and the sea surface).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Field tests of the UAV–RTK lidar measurements over the sea surface. (a) Sample of the raw lidar data. The elevation is the distance measured by the lidar. (b) Return rate of the lidar of (a). (c) Return rates of the lidar over the sea surface for nine flights. (d) The maximum return rates of (c) vs flight height over the ocean (distance between the lidar and the sea surface).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Field tests of the UAV–RTK lidar measurements over the sea surface. (a) Sample of the raw lidar data. The elevation is the distance measured by the lidar. (b) Return rate of the lidar of (a). (c) Return rates of the lidar over the sea surface for nine flights. (d) The maximum return rates of (c) vs flight height over the ocean (distance between the lidar and the sea surface).
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
The return rates of the nine flights are shown in Fig. 7c. Because the water surface is much smoother than the surfaces of solid objects, the return rates of the lidar measurements over the sea were lower than those over land. The maximum return rates versus flight height over the ocean are shown in Fig. 7d. Indeed, there is no clear relationship between the two quantities. This might be because other factors, such as sea surface roughness and wave height, may also affect return rate. However, the present system has been demonstrated to be a useful technique for sea surface measurements even under such a low return rate.
b. Measurements of tidal variations
After the lidar data were converted into the georeferencing coordinates, the average of the time series data provided the value of sea level induced by the tide. Figure 8 shows a comparison of the tidal variations measured by the UAV system and the pressure sensor. Ideally, the comparison of the tide measurements should be performed based on removing the mean tidal elevation, that is, removing the mean sea level. Because the field experiment did not cover tidal cycles, the mean sea level cannot be determined. In the comparison, the tidal elevation measured by the UAV was shifted by a constant value of 0.52 m for referencing. This value is the difference between the first measurement point of the pressure sensor and the lidar data. Good agreement between the pressure and the UAV measurements suggests that the system is qualified for tidal elevation measurements over the sea in the field. The RMS error between the two techniques is 4.9 cm.
Comparison of the tidal variations measured by the UAV–RTK lidar system and a pressure sensor. The data measured by the UAV–RTK lidar have been shifted by a constant value of 0.52 m for reference.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison of the tidal variations measured by the UAV–RTK lidar system and a pressure sensor. The data measured by the UAV–RTK lidar have been shifted by a constant value of 0.52 m for reference.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison of the tidal variations measured by the UAV–RTK lidar system and a pressure sensor. The data measured by the UAV–RTK lidar have been shifted by a constant value of 0.52 m for reference.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
c. Measurements of waves
The measured elevations are similar across the spatial span in one lidar scan because the scanning rate of the lidar was much greater (40 Hz) than the dominant wave frequency (0.23 Hz) and the flight height was low (6 m) above the sea surface. The data in one lidar spatial scan were then averaged to obtain one-dimensional time series data. Samples of the time series data measured by the lidar and by the pressure sensor are presented in Fig. 9. Approximately 8000 samples (UAV lidar, 7-min records) and 19 000 samples (pressure sensor, 19-min records) were collected in each measurement. The data were split into 15 and 37 blocks, respectively, using fast Fourier transforms of 512 samples with 30 and 74 degrees of freedom, respectively. A 50% overlap was used to calculate the individual spectra with Hanning windows after mean removal and linear detrending.
Samples of the time series surface elevations measured by the (a) UAV–RTK lidar system and (b) pressure sensor, and (c) the corresponding spectra of the surface elevations.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Samples of the time series surface elevations measured by the (a) UAV–RTK lidar system and (b) pressure sensor, and (c) the corresponding spectra of the surface elevations.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Samples of the time series surface elevations measured by the (a) UAV–RTK lidar system and (b) pressure sensor, and (c) the corresponding spectra of the surface elevations.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Ensemble-averaged spectra for each hourly burst were computed by averaging the spectra. The raw lidar data contain high-frequency spikes that can also be seen in the spectrum analysis. Linear wave theory was used to convert the pressure data to surface elevations with a cut of spectrum energy higher than 0.6–0.8 Hz caused by the failure of the pressure sensor in observing the high-frequency waves.
The spectra and spectrograms of the sea surface displacements measured using the two techniques were computed and compared, as shown in Figs. 9 and 10, respectively. The lidar spectrum agrees well with that of the pressure sensor in the frequency band from 0.1 to 0.6 Hz, which contains most of the spectrum’s energies. The spectral energy of the surface elevation follows 
Spectrograms of surface displacement (m2 s−1) measured by the (a) UAV–RTK lidar system and (b) pressure sensors.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Spectrograms of surface displacement (m2 s−1) measured by the (a) UAV–RTK lidar system and (b) pressure sensors.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Spectrograms of surface displacement (m2 s−1) measured by the (a) UAV–RTK lidar system and (b) pressure sensors.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison of the (a) significant wave heights and (b) weighted wave frequencies measured by the pressure sensor (solid lines) and UAV–RTK lidar system (circles). The vertical error bars denote the 95% confidence limits of the estimates.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison of the (a) significant wave heights and (b) weighted wave frequencies measured by the pressure sensor (solid lines) and UAV–RTK lidar system (circles). The vertical error bars denote the 95% confidence limits of the estimates.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
Comparison of the (a) significant wave heights and (b) weighted wave frequencies measured by the pressure sensor (solid lines) and UAV–RTK lidar system (circles). The vertical error bars denote the 95% confidence limits of the estimates.
Citation: Journal of Atmospheric and Oceanic Technology 35, 8; 10.1175/JTECH-D-17-0199.1
6. Discussion and summary
We have presented the development of a portable, lightweight, and low-cost UAV–RTK lidar system for observing tidal elevations and wave properties in coastal zones. The system is based on a robotic lidar (Hokuyo, UTM-30LX), an AHRS (Xsens, MTi 30), a NovAtel RTK GNSS, and a rotary UAV. A great advantage of the system, which uses a rotary UAV, is that it can be operated at low altitudes, which provides a spatial resolution on the order of a few centimeters. The temporal resolution of the system is 20 Hz, and the spatial resolution ranges from 0.4 to 8.7 cm for flight heights ranging from 1 to 20 m. For a low-altitude flight, the error caused by the low-cost lidar and altitude sensor can therefore be decreased. When operating the system in a hover mode with a flight height of approximately 10 m and with variations in pitch and roll of approximately 8°, the RMSE of the vertical position measurements was approximately 5 cm in comparison with the ground truth values.
The system was validated in the field with comparisons to the results obtained by a pressure sensor. The RMS errors for the tidal elevation, significant wave height, and wave period measurements between the two techniques are 4.9 cm, 4.8 cm, and 0.028 s, respectively, suggesting that the system is useful for tidal elevation and wave measurements.
One challenge of operating the system is that the useful return data rate is 80% for measurements at 20-m height over the land surface and only 10%–35% even for flights undertaken at a height of 6 m over the sea. The total payload capacity of the UAV also limits the instruments being used. These deficiencies could be improved if an upgraded lidar (such as the RIEGL miniVUX-1UAV) and a high-standard UAV are used.
In future applications, the system could be applied to observe beach erosion, shoreline viability, roughness in intertidal zones, water-level gradients caused by wave setup (Lowe et al. 2009), wave dissipation rates (Huang et al. 2012) in coral reefs, and wave dissipation in surfzones. By combining UAV imagery techniques, it is possible to observe land and sea surface signatures in one flight measurement.
Acknowledgments
The authors thank the three anonymous reviewers for their useful comments. This research was supported by grants from the Ministry of Science and Technology, Taiwan, under Contact MOST 106-2611-M-008-002.
APPENDIX
REFERENCES
Baltsavias, E. P., 1999: Airborne laser scanning: Basic relations and formulas. ISPRS J. Photogramm. Remote Sens., 54, 199–214, https://doi.org/10.1016/S0924-2716(99)00015-5.
Bandini, F., J. Jakobsen, D. Olesen, J. A. Reyna-Gutierrez, and P. Bauer-Gottwein, 2017: Measuring water level in rivers and lakes from lightweight Unmanned Aerial Vehicles. J. Hydrol., 548, 237–250, https://doi.org/10.1016/j.jhydrol.2017.02.038.
Brouwer, R. L., M. A. de Schipper, P. F. Rynne, F. J. Graham, A. Reniers, and J. H. MacMahan, 2015: Surfzone monitoring using rotary wing unmanned aerial vehicles. J. Atmos. Oceanic Technol., 32, 855–863, https://doi.org/10.1175/JTECH-D-14-00122.1.
Colomina, I., and P. Molina, 2014: Unmanned aerial systems for photogrammetry and remote sensing: A review. ISPRS J. Photogramm. Remote Sens., 92, 79–97, https://doi.org/10.1016/j.isprsjprs.2014.02.013.
Glennie, C., 2007: Rigorous 3D error analysis of kinematic scanning LIDAR systems. J. Appl. Geod., 1, 147–157, https://doi.org/10.1515/JAG.2007.017.
Hokuyo Automatic Co., 2012: Scanning laser range finder UTM-30LX: Specification. 6 pp., https://www.hokuyo-aut.jp/search/single.php?serial=169.
Holman, R. A., K. L. Brodie, and N. J. Spore, 2017: Surf zone characterization using a small quadcopter: Technical issues and procedures. IEEE Trans. Geosci. Remote Sens., 55, 2017–2027, https://doi.org/10.1109/TGRS.2016.2635120.
Huang, Z. C., B. D. Reineman, L. Lenain, W. K. Melville, and J. H. Middleton, 2012: Airborne LIDAR measurements of wave energy dissipation in a coral reef lagoon system. J. Geophys. Res., 117, C03016, https://doi.org/10.1029/2012JB009424.
Ierodiaconou, D., A. C. G. Schimel, and D. M. Kennedy, 2016: A new perspective of storm bite on sandy beaches using Unmanned Aerial Vehicles. Z. Geomorphol., 60, 123–137, https://doi.org/10.1127/zfg_suppl/2016/00247.
Klemas, V. V., 2015: Coastal and environmental remote sensing from unmanned aerial vehicles: An overview. J. Coastal Res., 31, 1260–1267, https://doi.org/10.2112/JCOASTRES-D-15-00005.1.
Lee, K. H., and R. Ehsani, 2008: Comparison of two 2D laser scanners for sensing object distances, shapes, and surface patterns. Comput. Electron. Agric., 60, 250–262, https://doi.org/10.1016/j.compag.2007.08.007.
Liu, P., and Coauthors, 2014: A review of rotorcraft Unmanned Aerial Vehicle (UAV) developments and applications in civil engineering. Smart Struct. Syst., 13, 1065–1094, https://doi.org/10.12989/sss.2014.13.6.1065.
Long, N., B. Millescamps, B. Guillot, F. Pouget, and X. Bertin, 2016: Monitoring the topography of a dynamic tidal inlet using UAV imagery. Remote Sens., 8, 387, https://doi.org/10.3390/rs8050387.
Lowe, R. J., J. L. Falter, S. G. Monismith, and M. J. Atkinson, 2009: Wave-driven circulation of a coastal reef–lagoon system. J. Phys. Oceanogr., 39, 873–893, https://doi.org/10.1175/2008JPO3958.1.
Morin, K. W., 2002: Calibration of airborne laser scanners. M.S. thesis, Dept. of Geomatics Engineering, University of Calgary, 125 pp.
NovAtel, 2016: OEM6 family installation and operation user manual. Revision 12, NovAtel Inc. Publ. OM-20000128, 208 pp.
Reineman, B. D., L. Lenain, D. Castel, and W. K. Melville, 2009: A portable airborne scanning lidar system for ocean and coastal applications. J. Atmos. Oceanic Technol., 26, 2626–2641, https://doi.org/10.1175/2009JTECHO703.1.
Sallenger, A. H., and Coauthors, 2003: Evaluation of airborne topographic lidar for quantifying beach changes. J. Coastal Res., 19, 125–133.
Sankey, T., J. Donager, J. McVay, and J. B. Sankey, 2017: UAV lidar and hyperspectral fusion for forest monitoring in the southwestern USA. Remote Sens. Environ., 195, 30–43, https://doi.org/10.1016/j.rse.2017.04.007.
Skaloud, J., and D. Lichti, 2006: Rigorous approach to bore-sight self-calibration in airborne laser scanning. ISPRS J. Photogramm. Remote Sens., 61, 47–59, https://doi.org/10.1016/j.isprsjprs.2006.07.003.
Stockdon, H. F., A. H. Sallenger, J. H. List, and R. A. Holman, 2002: Estimation of shoreline position and change using airborne topographic lidar data. J. Coastal Res., 18, 502–513.
Thornton, E. B., 1979: Energetics of breaking waves within the surf zone. J. Geophys. Res., 84, 4931–4938, https://doi.org/10.1029/JC084iC08p04931.
Turner, I. L., M. D. Harley, and C. D. Drummond, 2016: UAVs for coastal surveying. Coastal Eng., 114, 19–24, https://doi.org/10.1016/j.coastaleng.2016.03.011.
Vrbancich, J., W. Lieff, and J. Hacker, 2011: Demonstration of two portable scanning LiDAR systems flown at low-altitude for investigating coastal sea surface topography. Remote Sens., 3, 1983–2001, https://doi.org/10.3390/rs3091983.
Xsens Technologies, 2016: MTi user manual. 86 pp., https://www.xsens.com/products/mti-100-series/.
Young, I. R., 1999: Wind Generated Ocean Waves. Elsevier Ocean Engineering Series, Vol. 2, Elsevier Science, 287 pp.
