Surface Current Measurement Using Large Scale Particle Image Velocimetry (LSPIV) in Ender Estuary, Cirebon

The Large-Scale Particle Image Velocimetry (LSPIV) method is an image- based technique that uses a non-intrusive approach or avoids direct physical contact with the water being studied to measure the velocity of water flow. The present study aims to identify surface velocity using LSPIV captured by Unmanned Aerial Vehicle (UAV) in Ender estuary, Cirebon. Field observation for surface velocity was conducted at three locations during flood and ebb tides. The surface current results of LSPIV compared with in situ velocity measurement by a current meter. The results of the LSPIV method are strongly influenced by the data acquisition process such as lighting, interrogation area, and video recording angle. Based on comparing the surface flow between results of LSPIV and observation data, the coefficient correlation (R) is 0.84 for velocity magnitude and R = 0.92 for flow direction. The Root Mean Square Error (RMSE) of 0.024 m/s for velocity magnitude and 4.81 degree for flow direction, and the Mean Absolute Error (MAE) range from 0.002 to 0.059 m/s (1 to 12 degree) for the velocity magnitude and the flow direction, respectively. Furthermore, the Mean Absolute Percentage Error (MAPE) value of 10.41% (8.84%) for magnitude and direction, respectively. Therefore, LSPIV was applied successfully to obtain surface velocity measurements.


Introduction
Ocean currents play a vital role in estuaries, facilitating processes such as sediment and nutrient transport, water body development, and supporting various human activities such as transportation and tourism [1].Generally, velocity observations are limited to using intrusive measurement method including drifter, current meter, and Acoustic Doppler Current Profiler (ADCP).However, many studies have used photogrammetric methods to measure the surface velocity of rivers in order to lower the risk of field observations.There are more image-based approaches that can be used for surface velocity measurement as a result of the advancement of technology.This related state-of-the-art technology has been continuously developed over many years [2,3].
River surface velocities can be calculated using photos using photogrammetry.Unmanned aerial vehicles (UAVs), which possess highly maneuverable qualities, can overcome terrain constraints, have orthographic capabilities, and provide high economic benefits, are preferred by an increasing number of researchers [4,5].UAVs have been extensively used in agricultural and ecological surveys [6], environmental monitoring projects [7] and disaster monitoring efforts [8], due to their ability to navigate terrain obstacles and conduct aerial photography operations.Among the optical techniques available, large-scale particle image velocimetry (LSPIV) uses digital images taken from a fixed location to measure remote surface flow [2,9].The field's surface velocity can be obtained by using LSPIV.Additionally, LSPIV offers a flexible and affordable flow measurement technique that can efficiently get around the shortcomings of river flow monitoring networks.The results of the surface velocity data obtained by LSPIV and the results obtained by acoustic Doppler current profilers (ADCPs) are comparable, according to a number of studies based on field experiments [10].As a result, it can be concluded that LSPIV techniques are accurate enough to be relied upon for determining river surface velocities.
The LSPIV method has been successfully used to measure surface current velocity in rivers such as the Houlong River, Taiwan [11], Alpine River, Austria [12], East Central Illinois River, USA [13], and Murg River, Switzerland [5].These previous studies were carried out in the area of uniform current velocity.Meanwhile, in this study the LSPIV method will be used to identify current circulation in river mouths, where the conditions of river mouths are more complex due to the influence of tides.The present study aims to measure the surface velocity of the Ender Estuary in Cirebon, West Java, using UAV and LSPIV instruments and an intrusive current meter observation to verify the accuracy of the LSPIV measurements.

Field site and observation
This research was conducted in the estuary and river area in Ender, Pangenan District, Cirebon Regency, as shown in Figure 1a.The Ender River has a length of approximately 18 km, with its river upstream located in Kuningan, West Java.The mouth of the river has a width of 127 m, and the average width of the river is about 40 m.Ender is a shallow estuary with average depth of 2.5 m.The tide is predominantly semidiurnal with the tidal range varies from 0.2 m during neap tides to 0.8 m during spring tides.The Ender estuary area has land use dominated by salt and fish ponds, as well as residential areas.
In this study, three field campaigns were conducted on 26 January, 6 March, and 8 March in 2023 to measure the surface velocity.Measurements were conducted at three stations during flood and ebb tides condition: the first station was located at the river mouth, the second station was situated in the meandering section of the river representing the river body, and the third station was at the point where the ship was docked representing the furthest distance from the river mouth.At the first station, a transect measuring 24 x 41 m in a square shape was established at the river mouth Figure 1b.For the second station, situated in the river's meandering bend, a transect measuring 24 x 25 m was created Figure 1  (c).Lastly, at the third station, located at the ship docking area, a transect measuring 12 x 17 m was set up Figure 1d.At each of the three transects, three disks were placed along each bank.Real Time Kinematic Global Navigation Satellite System (RTK-GNSS) surveying equipment was used to determine 4 points to serve as georeference control points, measuring and marking the georeference control points x, y, z coordinates and marking the position for the drone.
Observation method used in this study involved capturing video footage of the surface flow using a drone (UAV) with the transect plot method.The model of the unmanned aerial vehicle used in this study is DJI Mavic 2 which consists of a drone body, a remote control, a gimbal camera, and the DJI Mavic 2 program.Footage of the event was collected at 1920 × 1080 pixel resolution at an acquisition rate of 30 frames per second (FPS).In addition to the UAV methods combined with the LSPIV method to measure surface velocity, this study also uses traditional surface velocity measurement methods to verify the measurement results of the LSPIV method.A Valeport Current Meter used to measure surface velocity at the three stations.Instantaneous measurements of current meter were taken simultaneously with the UAV recording.About 6-7 locations of surface velocity were observed at each station.

Large Scale Particle Image Velocimetry
Non-intrusive measurement procedures have rapidly advanced in the last few decades due to the development of scientific technology, including computer vision, Large Scale Particle Image Velocimetry (LSPIV) methods, and image-based measurements.[15] used the same PIV method to measure the flow velocity of rivers.Subsequent applications were carried out in the hydrological sector using the LSPIV method [3,14], which has developed angle correction calculations for image acquisition.
Large Scale Particle Image Velocimetry (LSPIV) is an optical remote sensing with spatiotemporal dimensions.Figure 2 show illustration of LSPIV measurement sequence.LSPIV operates on the fundamental principle of using traceable particles (seeding flow) introduced into the flowing water.These particles are then illuminated and recorded in photos or videos in a sequential manner.The seeding process is a concept in fluid dynamics to explain the influence of foreign objects on fluid flow.The output of LSPIV provides a 2D velocity vector field on the surface of the flow, covering a relatively large area.In simple terms, LSPIV captures fluid flow patterns over time using an optical sensor, such as recording videos with a camera or unnamed area vehicle (UAV) [11].Then, the displacement of patterns in the x and y coordinates within the image matrix is analysed with cross-correlation matrices of paired patterns.Entire matrix in the image is processed to obtain velocity vectors on the measurement plane.By assuming the actual distance and pixel distance in the camera image, along with the time difference between two images, the displacement of patterns in the image sequence is converted into units of distance per time, such as m/s.

Performance of LSPIV method
In order to quantity the accuracy of the LSPIV method, the results of LSPIV for velocity magnitude and direction compared with the observations.LSPIV performance is quantified by the Mean Absolute Error (MAE), the Root Mean Square Error (RMSE), the correlation coefficient (R), the Mean Absolute Percentage Error (MAPE) as shown in Table 1 and 2. The MAE between LSPIV and observation is ranging from 0.002 to 0.059 m/s for magnitude and 1 to 12 degrees for direction.RMSE value is 0.024 m/s for velocity magnitude and 4.81 degree for direction.Furthermore, the correlation between LSPIV and observation is 0.84 for velocity magnitude and 0.96 for velocity direction.Percentage error (MPAE) between LSPIV and observation is 10.41% for magnitude and 8.84% for direction.These imply that the LSPIV method provides a reasonably accurate estimation of current magnitude and direction.

Pattern of surface currents
The surface currents obtained from LSPIV at the station 1 as shown in Figure 3.It can be seen that the surface current towards the river mouth during flood tides (Figure 3a).Meanwhile, at low tide, the direction of surface flow is towards upstream of the river (Figure 3b).At high tide conditions, the magnitude of the current ranges from 0.2 -1 m/s with an average magnitude of 0.73 m/s.At low tides, the surface velocity is ranges from 0.4 -1 m/s with an average magnitude of 0.74 m/s.At station 2 (Figure 4), the direction of surface velocity flows to upstream during flood tides with an average velocity of 0.39 m/s.Whereas, the surface currents toward downstream when ebb tides with an average flow of 0.48 m/s.

Effect of interrogation area
The LSPIV method describes the surface velocity field from the Eulerian perspective by determining flow velocities over a regular grid.Interrogation areas (IAs) are used to describe each cell in the regular grid.Figure 6 show surface current pattern at station 2 during ebb tides for two interval of interrogation area (IA).In order to identify the effect of IA, the results of surface velocity with pre-processing stage uses a resolution grid of 0.125 cm (Figure 6a) compared to the grid of 1.25 cm (Figure 6b).The grid is denser in the estimation results using an IA of 0.125 cm, making it challenging to determine the current The results of the present study is similar with [16].They found that when the IA is too small, image matching from frame to frame becomes poor, resulting in less accurate surface velocity estimation.On the other hand, when the IA is too large, the calculated surface velocities for each IA become uniform.

CONCLUSIONS
The LSPIV method was implemented to determine the surface velocity at the Ender Estuary in Cirebon.Field measurements of surface velocity was conducted at three locations during flood and ebb tides.The surface velocity of the LSPIV method was compared with the velocity observed by a current meter.
According to statistical analysis, the coefficient of correlation (R) for velocity magnitude is 0.84 and 0.92 for flow direction.The velocity magnitude and flow direction had the root mean square error (RMSE) of 0.024 m/s (4.81°) and the mean absolute error (MAE) ranges from 0.002 to 0.059 m/s (1 to 12 degree), respectively.Furthermore, the mean absolute percentage error (MAPE) for magnitude and direction is 10.41% (8.84%).As a result, LSPIV was effectively used to obtain the surface velocity, making it a promising technique for further research and development.

Figure 1 .
Figure 1.Study area of the Ender Estuary, Cirebon, West Java (a).Yellow box is the area of investigation for UAV measurement at three locations (Station 1 (a), Station 2 (b) and Station 3 (c)).Longitude (degree)

Figure 2 .
Figure 2. Flow diagram of the Large-Scale Particle Image Velocimetry (LSPIV) (FOV is the Field of View; GRPs is the Ground Reference Points; IA is the Interrogation Area; SA is the Searching Area) [11].

Figure 3 .Figure 4 .
Figure 3. Surface current patterns obtained by LSPIV method at Station 1 during flood tides (a) and ebb tides (b).

Figure 5
Figure5is the results of surface current at station 3 during flood and ebb tides.The direction of surface currents is towards the estuary during flood and ebb condition.The current speed ranges from 0.3 -1 m/s with an average magnitude of 0.52 m/s during low tides, while during high tides the surface velocity ranging from 0.2 -0.8 m/s, with average velocity of 0.31 m/s.

Figure 5 .
Figure 5. Surface current pattern obtained by LSPIV method at station 3 during flood tides (a) and ebb tides (b).
an IA 1.25 cm produces a sparser grid, making the direction of velocity more obvious.

Table 1 .
Quantification of validation between LSPIV and observation for velocity magnitude.

Table 2 .
Quantification of validation between LSPIV and observation for velocity direction.