Unmanned Aerial Vehicle Derived DEM Using in Accurate Geometric Analysis for Water Harvesting in Small-Scale Depressions

Abstract


Introduction
Digital Elevation Model (DEM) is a set of numbers that is organized in a specific way to show the heights of different locations relative to an arbitrary reference point.Essentially, a DEM provides a digital representation of elevation data for a particular area.The term DEM is commonly used to describe topographical features of the Earth's surface, In general, the term DEM refers to a broad range of surfaces (Gupta, 2003).In the context of DEM-based geometric analysis, the accuracy of the data utilized is of critical importance, as it can significantly influence the resulting outcomes.Geometric analysis plays a pivotal role in the calculation of volumetric and aerial characteristics of topographic depressions.Small depressions, in particular, have long posed a surveying challenge, as their spatial coverage is often not adequate for satellite-based DEMs.As a result, traditional land surveying techniques have typically been employed, albeit with limited accuracy and high error ratios.Parallax is observed in still objects when an observer's position changes.This phenomenon occurs in aerial photography, where changes in the aircraft's location during the photo-taking process can cause apparent shifts in the terrain's features, which can be utilized to generate 3D perspectives and compute elevations (Lillesand et al., 2015).Numerous studies have been conducted employing unmanned aerial vehiclesUAV's to generate DEMs for a wide range of hydrological applications, as well as volumetric analyses more broadly.to obtain digital models of coastal environments, UAV-SfM approach has been tested for creating high-resolution DSMs, offering a fast and cost-effective alternative (Mancini et al., 2013).The UAV generated digital surface models (DSMs) used to monitor river-channel morphology, results show accurate ground elevation capture and reasonable vegetation cover capture, as well as revealing the cause of elevation differences before and after a man-made flood (Watanabe and Kawahara, 2016).UAV imagery is used for monitor changes in dynamics and landscape and can use to simulate water flow and validate flooding simulations in areas of hydric-contribution and inland flooding through orthomosaics and digital elevation models (Moreira Furlan et al., 2021).The combined photogrammetric and spectral-based approach using neural networks achieved higher accuracy and effective spatial coverage in fully-covered bathymetry of clear lakes compared to single bathymetric methods.The shallow neural network-based model had the highest accuracy, while the deep neural network-based model provided more detailed water depth distribution (He et al., 2022).Traditional surveying methods in Geodetics are not efficient for calculating the volume of objects in high-risk areas or when the task need to be done in a short time.Using multiple overlapping images taken from different perspectives is a more effective approach that produces precise 3D models, allowing for faster and more accurate measurements (Ruzgienė et al., 2015).Idrees and Heeto (2020), evaluate the accuracy of volume obtained through DEM derived from UAVs images of a dam body with different flight heights and ground control points for geo-referencing.Volumes calculated with UAV images and those obtained through GPS survey techniques have high compatibility.For the camera configuration properties, the area characterized by sudden changes in elevation of the surface orientations is better represented by both nidar and oblique configuration (Rossi et al., 2017).the using UAV-based DEM for volume estimation is faster and more accurate than traditional total station methods.UAV-based DEM was found to acquire data six times faster than total stations (Arango and Morales, 2015).
The main aims of this study are to acquire high-resolution aerial images of the study area, which will be used to generate a DEM using SFM algorithm.The generated DEM will then be evaluated for accuracy, and subsequently utilized to analyze topographic depressions in the study area.The principal finding is the maximum capacity of the valley when the water level is at the maximum water level.

Study Area
The study area is located in AL-Sherqat, the area was selected for its ability to enable the flow of an Intermittent stream, which in turn leads to rainwater overflow into a larger valley.The location consists of a small valley spanning approximately 70,000 m 2 , chosen to evaluate the technical feasibility of generating a DEM that could assist in geometric analysis for small-scale reservoirs for water harvesting purposes.The area is defined by coordinates between two latitudes 35°32'15", 35°32'22" and between two longitudes 43°8'12", 43°8'27".Fig. 1 shows the study site on aerial map.

Materials and Methods
DJI Mavic Air 2s (is a high-performance drone using for aerial photography and videography) had been used in the field surveying work.This kind of UAV belongs to multi-copters.The multi-copters Submitted better results for the ground features in areas with higher slop gradients (Gómez-Gutiérrez and Gonçalves, 2020).The specifications of the selected platform: the sensor includes 20 million effective pixels, a lens with a 35 mm format equivalent of 22 mm, and an aperture of f/2.8-f/11.The electronic shutter speed ranges from 8-1/8000s, and the still image size is 5472x3648.Additionally, the drone has a max ascent speed of 6m/s and a maximum flight time of 31 minutes in still wind conditions.The UAV's GNSS system used is GPS+GLONASS, and the maximum transmission distance is 10000m (DJI, 2020).The hardware GNSS RTK (is a GNSS receiver designed for high-precision positioning and surveying applications.)capabilities of the E600-N model include an RTK accuracy of 4mm+1ppm RMS for horizontal measurement and 8mm+1ppm RMS for vertical measurement.Enables the capacity to track prevailing satellite constellations and signals, encompassing GPS/GLONASS/BEIDOU/GALILEO/QZSS, etc.
In UAV photogrammetry, flight planning is a crucial step that involves selecting the optimal flight path for the UAV to achieve the desired coverage and capture the required imagery.The flight plan should aim to achieve high-quality imagery while reducing flight time and minimizing potential risks to both the UAV and the people in the surrounding area (Pepe et al., 2018).On February 17th, 2022, the Drone Harmony application was used for an automated mission.To maintain a constant altitude during flight, the forward and side overlap were programmed to be 70% with a double grid plane.The flight altitude was set at 80 meters, and the GSD (Ground Sample Distance) of the images was 1.24 pixels per centimeter.The camera configuration was set at Nidar (90 degrees), and a total of 87 images were captured.Despite clear skies, strong winds were observed during the flight.To ensure accurate results, 19 GCPs (Ground Control Points) were evenly distributed throughout the area and were collected using GNSS RTK, with a fixed error range below 1 cm both horizontally and vertically.Fig. 2 shows the distribution of GCPs.The images taken by the UAV were processed using a Structure-from-Motion (SfM) algorithm by Agisoft Metashape software using a specific workflow that solves the geometric constraints of camera position, orientation, and GCPs from many overlapping images simultaneously (Gindraux et al., 2017).The Agisoft Metashape workflow for generating DEM involves several steps.Firstly, you import the photos of the desired area for the DEM into Metashape.The "Align Photos" function is then used to automatically align and match the overlapping photos, utilizing feature points detected in the photos to create a point cloud.This point cloud serves as the foundation for generating the DEM, with the option to apply Ground Control Points (GCPs) for accurate georeferencing.Next, the "Build Dense Cloud" function is used to generate a more detailed and denser point cloud using the matched feature points from the aligned photos.The points may need to be filtered and classified to remove outliers or noise, ensuring the point cloud is clean and ready for further processing.Finally, the "Build DEM" function in Metashape is employed to generate the DTM and DSM , DTM represents the bare earth surface and is used for terrain analysis, hydrological modeling, and civil engineering design, while DSM represents the earth's surface, including all objects, and is used for 3D modeling, urban planning, and aerial imagery analysis (Krauß et al., 2011).The Agisoft workflow is illustrated in Fig. 3.After generating the DEM and exporting the data, accuracy checking can be performed using the Root Mean Square Error (RMSE) value, a common measure of the discrepancy between the DEM and ground truth data (RTK data).
Once the accuracy has been verified, the DEM can be imported into Global Mapper Pro V.24, a software used for geometric analysis processes.In this study, the "cut and fill volume" function is utilized to calculate both volumetric and areal elements of the depression at each selected level, which represents the water level in the depression and is indicated by contour lines.The contour lines generated at the maximum level can demarcate the water holding capacity of the depression, while cropping the depression downstream can mark the outlet or the end of the depression, (which can be assumed as a dam).This approach results in a closed depression that can be further analyzed and processed.To speed up processing and remove unnecessary data, it's crucial to crop all depressions at their highest points.Once the contour is selected and enclosed, it can be turned into an area feature using the "create area feature from selected line feature" command.This area is saved as a Global Mapper package file for comparison.To estimate the volume and aerial parameters, the "cut and fill calculation" command is used in combination with the selection.The software automatically sets the upper surface elevation to the contour height, and the lower surface corresponds to the DEM.The volume between the two surfaces is then calculated.After the cut and fill calculation, a window displays the results, including information on cut and fill volume, area, enclosed area, and perimeter.This data is essential for geometric analysis and can be exported as an Excel CSV report.The collected data can then be consolidated into a single Excel sheet, which allows mathematical equations and graphical representations to be formulated for each parameter or the relationships between parameters.
Geometric analysis is essential for creating an operational plan for dam reservoirs.This involves extracting positive and negative volumes, positive and negative surface areas, as well as positive and negative planar areas.Before building a dam, it's important to study the area's geometric elements.The data collected can be used by designers to create the dam's body and parts, and decision-makers can evaluate the environmental and social impacts of dam construction (Salih et al., 2018).
The geometric elements related to the reservoir include.The water level is the contour line representing the water level ranges between (181-186.8)meters above sea level (m.a.s.l), while positive volume (m 3 ) is the volume of islands within the reservoir boundary.Negative volume is the water volume at a specific water level, and negative planer area is the area of the water surface.Positive planer area (m 2 ) is the area of the island projection, and negative surface area (m 2 ) is the non-planer area of the reservoir bottom.Positive surface area (m 2 ) is the non-planer area of the islands.Divide the reservoir volume by the negative planer area to get the average reservoir depth (m), and divide the positive volume by the positive planer area to get the average island thickness (m).Calculated by subtracting the negative volume at one level from the negative volume at 186.8 m.a.s.l.

Results and Discussion.
The main output of UAV image processing is the DEM. the evaluation of the accuracy of the DEM is made possible by a root mean square error (RMSE) (Villanueva and Blanco, 2019).Jiménez-Jiménez et al., ) 2021), list the accuracy limits for previous studies.The range of RMSE for the x and y coordinates varied from 0.9 to 7.0 m, while for the elevation coordinate, the range was between 1.1 to 5.4 m.The accuracy assessment results obtained from this study were deemed acceptable in comparison to previous studies.Resulting in 1.93m for the x, 1.76m for the y, and 0.893m for the elevation.The calculation of RMSE includes determining the error between predicted and actual values of ground control points (GCPs).This value has the same units as elevation units, here meters.A low RMSE value indicates high accuracy, whereas a high RMSE value indicates low accuracy.Equation ( 1) is used to calculate RMSE (Beg et al., 2023).
Where: Zobs: is the actual value for the observation Zmodel: is is the predicted value for the observation.n: is the number of observations.The DEM that is derived from the process of SFM is presented in visual form within Fig. 4. The values of these geometric elements are presented in Table .1   Geometric studies are based on fundamental principles that involve investigating the connections between geometric elements, such as the relationship between water level and surface area, and the correlation between positive and negative volumes and plane areas.These relationships allow us to comprehend the potential changes that may occur to the land due to storage processes at various levels (Salih et al., 2021).

The Relationship between Water Level and Positive Volume
A graph was generated to depict the correlation between the water level (x-axis) and positive volume (island volume) (y-axis).The results displayed in Fig. 5 indicate that the positive volume increases when the water level rises above 184.6 meters 1.86 m 3 due to the addition of new land in the form of an island.However, the positive volume gradually declines after it surpasses 185.4 meters 0.81m 3 .Of particular note is a significant increase in positive volume at 186.6 meters 8.30 m 3 , which is linked to the emergence of a small island within the water body due to the addition of new lands to the reservoir

The Relationship between Water Level and Negative Volume
the relationship between water level and negative volume in depression is displayed in Fig 6 .The plot indicates that the negative volume generally increases as the water level rises and can be divided into three stages.The first stage spans from 181m to 183.6m with a slow increase in negative volume.The second stage starts at 183.6m with a more pronounced increase in storage than the first stage and ends at 185m with a negative volume of 5795.71m 3 .The final stage occurs from 185m up to the final water level of 186.8m, with the most significant increase in storage during this phase.The total storage capacity of the depression is 18365.91m 3 , these stages can be attributed to the water leaving the main valley on the left side, where the slope angle is less steep than the right side, resulting in a nonsymmetrical valley in terms of slope.

The Relationship between Water Level and Positive Areal Elements
The relationship between water level and positive elements in a reservoir was studied in two plots, one for positive surface area and another for the the positive planar area, as shown in Fig. 7.Both plots show a similar trend where the positive values increase as the water level rises above 184.6 meters, due to the emergence of new islands in the reservoir.However, both positive surface area and positive planar area gradually decline after reaching 185.4 meters.A significant increase in both positive surface area and positive planar area is observed at 186.6 meters, attributed to the emergence of a small island within the reservoir.The results suggest that the small height values of the islands contribute to the similarity in the behavior of positive volume and positive elements.

The Relationship between Water Level and Negative Areal Elements
The relationship between water level and negative elements in the reservoir is shown in Fig. 8. Negative surface area refers to the non-planar areas of the wetland within the reservoir, and negative planar area refers to the planar areas of the wetland.The figure shows a continuously increasing trend in the curves characterized by a gradual slope at the initial stages, followed by a steep incline after the water level reached 183.6m.This phenomenon is duo to the water outflow from the primary valley on the left-hand side.

The Relationship between Water Level and Average Reservoir Depth
Fig. 9 displays the relationship between the water level and the average depth of the reservoir.The curve is characterized by an increasing pattern, with a relatively horizontal phase observed between 183.6 and 185.Subsequently, the curve continues to ascend until it reaches its maximum value at 186.8 with an average depth equal to (1.9 m).

Topographic Cross-Sections
The investigation of the geometric properties of a reservoir necessitates the utilization of both longitudinal and transverse topographical sections of the riverbed.Specifically, the transverse sections serve the purpose of locating the areas where variations in the reservoir's breadth and depth occur with respect to its banks.Conversely, longitudinal sections offer a graphical representation of the changes in depth and slope between the top and bottom of the river.A comprehensive examination of the reservoir's geometry can be achieved through the acquisition of subsequent sections (Salih et al., 2018).

The Cross-section (A-Aˉ).
Cross-section (A-Aˉ) is located near the south end of the reservoir, 30 meters from the assumed dam site.the cross-section length is 65 meters, and the elevation of the highest point is 186.8 m.a.s.l.The reservoir has a steeper gradient on the right bank. the deepest point is on the right side 17 meters from the bank and 6.8 meters deep.Fig. 10 illustrates the cross-section (A-Aˉ).

The Cross-section (B-Bˉ)
the cross-section (B-Bˉ) is located near the north end of the reservoir, about 110 meters from the assumed dam site.The length of the section is (76)m and reaches an elevation of 186.8 meters above sea level.The reservoir has a steeper gradient on the right bank, similar to cross-section (A-Aˉ), but the deepest point is shallower, measuring 3.8 meters in depth and located near the right bank.Fig. 11 illustrates the cross-section (B-Bˉ).

Longitudinal section (C-Cˉ)
The longitudinal section (C-Cˉ) depicted in Fig. 12 follows the valley stream's deepest points and conforms to its deviations over a length of 168 meters.The slope inclination is initially 5 cm/m towards the outlet but starts to reverse as it approaches the outlet due to a concavity in the valley's bottom.This concavity, located 35 meters from the outlet, causes the reversal of the slope inclination.The DEM was exported as a layout with a fixed scale to facilitate a comprehensive visual comparison of the reservoir extension at each level.Observations revealed an increase in the area of the reservoir as the level increased, as well as a notable change in its shape.The change in the shape and extent of the reservoir as the water level rises is shown in Fig. 13.An additional outcome of UAV photogrammetry is the production of the orthomosaic image.An orthomosaic is a type of aerial image created by mosaic individual images captured by an aerial vehicle (Liba and Berg-Jürgens, 2015).The resulting image has high-resolution and georeferenced and can be used to measure distances and areas accurately.The accuracy of the orthomosaic amounts to (2.08 cm/pix).

Conclusions.
A highly efficient way for surveying small topographic depressions and delivering more accurate data than traditional surveying techniques is the use of UAVs to create DEMs.In terms of centimeters per pixel (cm/pix) with the Mavic Air 2S drone, the resolution of orthomosaics for the AL-sherqat site is 2.08 for 70% overlap and 80m altitude.According to the study, the valley in AL-Sherqat has a total negative volume of 18365.91 m 3 when measured using geometric analysis at the highest water level of 186.8m.a.s.l.Water harvesting may be impacted by depression volume, which is the volume of water that collects in low-lying areas following rainfall.These depressions can act as rainwater collecting regions naturally, supplying vital supplies of water for agriculture and other applications.The careful selection of flight parameters for accurate DEM generation and investigation of the integration of multiple UAV platforms and sensors, and combining UAV-based remote sensing techniques with conventional hydrological monitoring methods to develop a comprehensive approach to water resource management.These suggestions can help hydrological assessments be more precise and successful, which will lead to better water resource management.

Fig. 6 .
Fig. 6.Relationship between water level and negative volume.

Fig. 9 .
Fig. 9. Relationship between water level and average reservoir depth.

Table 1 .
Geometric elements values at each selected level extracted from Global Mapper V24 cut and fill report.