Image velocimetry has proven to be a promising technique for monitoring river flows using remotely operated platforms such as Unmanned Aerial Systems (UAS). However, the application of various image velocimetry algorithms has not been extensively assessed. Therefore, a sensitivity analysis has been conducted on five different image velocimetry algorithms including Large Scale Particle Image Velocimetry (LSPIV), Large-Scale Particle Tracking Velocimetry (LSPTV), Kanade–Lucas Tomasi Image Velocimetry (KLT-IV or KLT), Optical Tracking Velocimetry (OTV) and Surface Structure Image Velocimetry (SSIV), during low river flow conditions (average surface velocities of 0.12–0.14 m s ${}^{-1}$ , Q60) on the River Kolubara, Central Serbia. A DJI Phantom 4 Pro UAS was used to collect two 30-second videos of the surface flow. Artificial seeding material was distributed homogeneously across the rivers surface, to enhance the conditions for image velocimetry techniques. The sensitivity analysis was performed on comparable parameters between the different algorithms, including the particle identification area parameters (such as Interrogation Area (LSPIV, LSPTV and SSIV), Block Size (KLT-IV) and Trajectory Length (OTV)) and the feature extraction rate. Results highlighted that KLT and SSIV were sensitive to changing the feature extraction rate; however, changing the particle identification area did not affect the surface velocity results significantly. OTV and LSPTV, on the other hand, highlighted that changing the particle identification area presented higher variability in the results, while changing the feature extraction rate did not affect the surface velocity outputs. LSPIV proved to be sensitive to changing both the feature extraction rate and the particle identification area. This analysis has led to the conclusions that for surface velocities of approximately 0.12 m s ${}^{-1}$ image velocimetry techniques can provide results comparable to traditional techniques such as ADCPs. However, LSPIV, LSPTV and OTV require additional effort for calibration and selecting the appropriate parameters when compared to KLT-IV and SSIV. Despite the varying levels of sensitivity of each algorithm to changing parameters, all configuration image velocimetry algorithms provided results that were within 0.05 m s ${}^{-1}$ of the ADCP measurements, on average.

Since the turn of the 21st century, image-based velocimetry techniques have become an increasingly popular approach for determining open-channel flow in a range of hydrological settings across Europe and beyond. Simultaneously, a range of large-scale image velocimetry algorithms have been developed that are equipped with differing image pre-processing and analytical capabilities. Yet in operational hydrometry, these techniques are utilised by few competent authorities. Therefore, imagery collected for image velocimetry analysis (along with reference data) is required both to enable inter-comparisons between these differing approaches and to test their overall efficacy. Through benchmarking exercises, it will be possible to assess which approaches are best suited for a range of fluvial settings, and to focus future software developments. Here we collate and describe datasets acquired from seven countries across Europe and North America, consisting of videos that have been subjected to a range of pre-processing and image velocimetry analyses (Perks et al., 2020, https://doi.org/10.4121/uuid:014d56f7-06dd-49ad-a48c-2282ab10428e). Reference data are available for 12 of the 13 case studies presented, enabling these data to be used for reference and accuracy assessment.

Image velocimetry (IV) is a remote technique which calculates surface flow velocities of rivers (or fluids) via a range of cross-correlation and tracking algorithms. IV can be implemented via a range of camera sensors which can be mounted on tri-pods, or Unmanned Aerial Systems (UAS). IV has proven a powerful technique for monitoring river flows during flood conditions, whereby traditional in-situ techniques would be unsafe to deploy. However, little research has focussed upon the application of such techniques during low flow conditions. The applicability of IV to low flow studies could aid data collection at a higher spatial and temporal resolution than is currently available. Many IV techniques are under-development, that utilise different cross-correlation and tracking algorithms, including, Large Scale Particle Image Velocimetry (LSPIV), Large Scale Particle Tracking Velocimetry (LSPTV), Optical Tracking Velocimetry (OTV), Kanade Lucas Tomasi Image Velocimetry (KLT-IV) and Surface Structure Image Velocimetry (SSIV). Nevertheless, the true applications and limitations of such algorithms have yet to be extensively tested. Therefore, this study aimed to conduct a sensitivity analysis on the commonly relatable parameters between the different algorithms, including the particle identification area parameters (such as Interrogation Area (LSPIV, LSPTV and SSIV), Block Size (KLT-IV) and Trajectory Length (OTV)) and the feature extraction rate (or sub sampled frame rate).Fieldwork was carried out on Kolubara River near the city of Obrenovac in Central Serbia. Cross-sectional surface width was relatively constant, varying between 23.30 and 23.45m. During the experiment, low flow conditions were present with a discharge of approx. 3.4m3 s-1 (estimated using a Sontek M9 ADCP), and depths of up to 1.9m. A DJI Phantom 4 Pro UAS was used to collect video data of the surface flow. Artificial seeding material (wood-mulch) was distributed homogenously across the rivers’ surface, in order to improve the conditions for IV techniques during slow flows. Two 30-second videos were utilised for surface velocity analysis.This study highlighted that KLT, SSIV, OTV and LSPIV are the least sensitive algorithms to changing parameters when no pre- or post-processing of results are conducted. On the other hand, LSPTV must undergo post-processing procedures in order to avoid spurious results and only then, results may be reliable. Furthermore, KLT and SSIV highlighted a slight sensitivity to changing the feature extraction rate, however changing the particle identification area did not affect significantly the outputted surface velocity results. OTV and LSPTV, on the other hand, highlighted that changing the particle identification area provided a higher variability in the results, whilst changing the feature extraction rate did not affect the surface velocity outputs. LSPIV proved to be sensitive to changing both the feature extraction rate and the particle identification area.This analysis has led to the conclusions that during the conditions of sampling with surface velocities of approximately 0.12ms-1, and homogeneous seeding on the rivers surface, IV techniques can provide results comparable to traditional techniques such as ADCPs during low flow conditions. All IV algorithms provided results that were, on average, within 0.05ms-1 of the ADCP measurements. 

Since the turn of the 21st Century, image based velocimetry techniques have become an increasingly popular approach for determining open-channel flow in a range of hydrological settings across Europe, and beyond. Simultaneously, a range of large-scale image velocimetry algorithms have been developed, equipped with differing image pre-processing, and analytical capabilities. Yet in operational hydrometry, these techniques are utilised by few competent authorities. Therefore, imagery collected for image velocimetry analysis, along with validation data is required both to enable inter-comparisons between these differing approaches and to test their overall efficacy. Through benchmarking exercises, it will be possible to assess which approaches are best suited for a range of fluvial settings, and to focus future software developments. Here we collate, and describe datasets acquired from six countries across Europe and Asia, consisting of videos that have been subjected to a range of pre-processing, and image velocimetry analysis (Perks et al., 2019, https://doi.org/10.4121/uuid:34764be1-31f9-4626-8b11-705b4f66b95a). Validation data is available for 12 of the 13 case studies presented enabling these data to be used for validation and accuracy assessment.

The use of image velocimetry techniques for river monitoring has been increasing in the last few years, but there are some limitations to be solved related mainly to natural environmental conditions and operative framework (Dal Sasso et al., 2021). Along with these issues, the need for surface tracking features or homogeneously distributed materials across the cross-section represents one of the challenges for outdoor applications. In a natural environment, flows can present low seeding densities or locally distributed tracer clusters. These conditions can introduce a high variance and underestimate the flow velocity field, especially near the riverbanks. In this work, the Farnebäck dense optical flow method (Farnebäck 2003) implemented in SSIMS-Flow software (Ljubicic, 2022) was tested and compared with LSPIV technique (Thielicke et al., 2021) to estimate surface flow velocities under different seeding conditions. The application was carried out on the Arrow River (UK) along two meandering river reaches during low-flow conditions. Four different seeding conditions were experimented from low (natural) to high (artificial) seeding density of tracers . Tracers were manually distributed onto the water surface and videos were acquired from DJI Phantom 4 Pro. Seeding metrics were used to estimate seeding conditions including: mean tracer area, seeding density, spatial tracer distribution, and the SDI index (Pizarro et al., 2020). Conventional velocity measurements were used as benchmark purposes along various transects. This study highlighted the good performances of the two tested image velocimetry methods, with results comparable to traditional techniques. On the one hand, the Farnebäck optical flow method proved to be more sensitive to changing setting parameters (e.g., feature extraction rate) with respect to LSPIV. On the other hand, optical flow showed low sensitivity to seeding density (error reduction 30-40%). This is due to the capacity of the Farnebäck method integrated with an ad-hoc pooling technique for spatial velocity averaging to represent surface velocity under sporadic and uneven seeding (e.g., near the convex bank). References Dal Sasso, S. F., Pizarro, A., Manfreda S. (2021). Recent Advancements and Perspectives in UAS-Based Image Velocimetry. Drones 5, 3: 81. Farnebäck, G. (2003). Two-frame motion estimation based on polynomial expansion, Scandinavian conference on Image analysis. Springer, Berlin, Heidelberg. Ljubicic, R. (2022). SSIMS-Flow: UAV image velocimetry workbench, https://github.com/ljubicicrobert/SSIMS-Flow Pizarro, A., Dal Sasso, S.F., Manfreda, S. (2020). Refining image-velocimetry performances for streamflow monitoring: Seeding metrics to errors minimization. Hydrol. Process. 2020, 34, 5167–5175. Thielicke, W., Sonntag, R. (2021). Particle Image Velocimetry for MATLAB: Accuracy and Enhanced Algorithms in PIVlab. Journal of Open Research Software, 9, Ubiquity Press, 2021.