Oriented Object Detection for Complex Hydrodynamic Features: A Multi-Platform Rip Current Identification System

In their article, “Rip current related drowning deaths and rescues in Australia 2004–2011”, the authors endeavored to estimate the percent of surf beach rescues and drowning deaths that can be attributed to rip currents. In doing so, they misinterpreted data we of the United States Lifesaving Association provided to them and thus grossly understated the proportion of rescues from drowning at surf beaches in the USA that are primarily attributed to rip currents. While they reported 53.7 %, the numbers actually range from 80 to 90 % according to our data. To understand the bases of their errors, one may first consider a very logical step Brighton et al. took with respect to data they gleaned from Australian rescue statistics. Specifically they note that those rescues, “. . . that were described to be at a location other than a surf beach have been removed for this analysis (n = 198), because the locations are unaffected by rip currents”. Unfortunately, they did not apply a similar vetting with respect to the US data, which also includes non-surf beaches that should be excluded. More significantly, in reviewing the Australian rescue data Brighton et al. found that the majority of the rescue reports do not include a primary cause and thus they excluded them. Ultimately they considered only a small subset which they define as “major” rescues (n = 1246) and for which the reports are more detailed, although even those reports do not appear to reliably include a primary cause. In the US, a significant number of reporting agencies (typically government-funded lifeguard providers) routinely tabulate the cause of rescue, including rip currents, for all rescues. However, some reporting agencies in our data do not. Unfortunately, Brighton et al. included all of the US rescue numbers, without excluding the data from agencies which do not report a primary cause (as they had endeavored to do with Australian data). One of these reporting agencies, for example, consistently reports over 10 000 rescues per year, but does not provide a breakdown of causation. We have reviewed rescue reports contained in United States Lifesaving Association Statistics (2013) for a comparable period (2004 to 2011) to that studied by Brighton et al. (2013) at our surf beaches which reported a primary cause of rescue. Of these (n = 343 246), 81 % were reported as being primarily attributable to rip currents. In some areas of the US, this percentage is higher. For example, an analysis of California State Parks rescue data from 2004 to 2011, which relies on a Scantron system to record each rescue, finds that 85 % had “rip current” as the primary cause. This includes an indeterminate number of rescues at inland beaches without rip currents, which skews the rip current percent downward. Studies of Southern California agencies have found even higher percentages. We wonder why surf beach rescues in Australia would be less likely to be attributable to rip currents than surf beach rescues in the US, or if there may be other vulnerabilities in the data analysis of Brighton et al. that found only a 57.4 % attribution to rip currents in Australia.

Rip currents are the most dangerous hazard at surf beaches. Rip currents in South Florida have previously not been studied. Beach profiles for three Florida beaches (Miami Beach, Lido Beach, Sarasota, and Pensacola Beach) and one Georgia beach (South Cumberland Island) were chosen for surveying because of their variable sand bar heights. Rip current hazard at each beach was assessed by lifeguard rip rescue and drowning statistics. A relationship was found between sand bar height, beach slope and rip current hazard. Rip current measurements in South Florida, which involved utilizing GPS drifters, laser rangefinder and drone-imaged fluorescent tracer dye, showed that the speed ranged from 0.1-0.5 m/s, which is fairly slow compared to such measurements undertaken in California and Australia. The effect of rip currents on swimmers was analyzed based on the drag force acting on swimmers and the power they generate to overcome the currents when swimming against them. The drag force and power increase quadratically and cubically, respectively, with the increase of rip current and swimming speeds. Hence, even rip currents of low velocity can be dangerous and swimming against the current should be avoided if possible. Strong rips in California have been shown to exhibit a circulatory pattern, which could bring a floater back to the safety of a shallow sand bar. Field measurements of rip currents in South Florida clearly defined the flow characteristics of a nearly straight-line current, sometimes deflected to the east-southeast. Therefore, the traditional approach of swimming left or right, parallel to the shore is the best escape strategy, but not against the longshore current if present. A logistic regression analysis was conducted to predict the occurrence of rip currents based on beach conditions. The logistic model showed that wave height, wave period and wind speed were statistically significant factors in rip generation. Rips were found to be most commonly generated by relatively small, non-threatening waves (e.g., 0.6 to 0.9m in height). These physical factors, along with social and safety considerations, pose a significant problem for coastal management.

Leatherman, S.P., 2012. Undertow, rip current, and riptide. Undertow, rip current, and riptide are terms used to describe a variety of currents, all of which have different characteristics. However, much of the general public, news media, and even dictionary definitions confuse and misidentify these potential hazards at surf beaches. Many beachgoers use the terms interchangeably, when in fact they are distinctly different hazards. Furthermore, undertow, rip currents, and riptides occur for different reasons at different locations along the beach. Avoiding each of them and escaping their grip requires different strategies. Three-quarters of a century has passed since this issue was last addressed in the scientific literature, and rip currents are recently receiving much more attention by scientists and the general public because of a steady rate of fatalities, making this a timely issue.

Rip currents are the greatest hazard at surf beaches. Beachgoers are often unaware that nearshore currents exist, much less that they can be life threatening. Pro-active beach management needs to be undertaken to reduce the large number of rescues, near-drowning events, and loss of life at U.S. and worldwide beaches.

On approximately planar beaches, rip channel development is often preceded by a period in which jet‐like rip currents develop in apparently random locations, and dissipate after minutes to tens of minutes. The subsequent development of sharp‐edged, trough‐like channels extending across the surf zone suggests that rip currents on planar beaches cause local erosion. Conversely, channels are known to cause localized offshore‐directed flow, and once channels have formed on approximately planar beaches, rip currents no longer occur in apparently random locations, but are restricted to the locations of the channels. Apparently, the excavation of channels by rip currents on planar beaches triggers a positive feedback between the morphological development and the flow. However, theoretical analysis indicates that, when depth increases with distance from shore, and sediment transport is treated as a function only of local flow conditions, channel development in the vicinity of a rip current may not occur. In numerical simulations (using a “cellular” model of nonbathymetrically driven rip currents) in which sediment flux on a planar beach is driven by approximately realistic rip current velocity patterns, deposition occurs under parts of the rip currents (especially in the seaward half of the surf zone), and these “rip ridges” cause a negative feedback. In these simulations, as in most models treating surf zone sediment transport, sediment flux is assumed to be strictly a function of local hydrodynamic conditions. However, Observations of sediment‐laden rip‐current plumes extending well beyond the surf zone suggest that suspended sediment transport is not always in equilibrium with local conditions. Other simulations employ a treatment of suspended‐sediment transport that allows for a lag in the adjustment of fluxes to the changes in local hydrodynamic conditions that the sediment is advected through. With this nonlocal sediment‐transport, the flux of sediment out of the surf zone in a rip current is increased, ridges do not develop under any part of the rip currents, and the morphodynamic feedback becomes positive.

The rip current field resulting from the transformation of surface gravity waves over offshore submarine canyons is studied. Employing a wave transformation model and a wave‐induced circulation model over observed bathymetry we find that wave height variations associated with undulations in the canyon contours cause rip current circulation cells with alongshore spacing of O(100m) even though the nearshore bathymetry displays no variations at these length scales. Further, the predicted rips correspond to observed rip currents during the Nearshore Canyon Experiment (NCEX). Motivated by these results we study the relationship between O(100 m) scale variations in offshore bathymetric contours and the resulting rip current field in the nearshore. To isolate the roles of possible bathymetric features, we construct a series of idealized case studies that include site characteristics found at NCEX that are conducive of rip current development, such as a curved shoreline, an offshore submarine canyon and undulations in the canyon contours. Our results show that the first two components are unable to produce the observed short‐scale circulation systems, while wave refraction over undulations in the canyon walls at length scales of O(100 m) provides a sufficient disturbance to generate alongshore wave height variations that drive multiple rip currents for a variety of incident wave conditions. Rips are not generated when the wave period is short, or when the angle of incidence is large. Analysis of the alongshore momentum balances further demonstrates that the rip current locations are also strongly influenced by inertial effects. Hence, nonlinear processes are important within the rip current circulation cell and we find that nonlinear advective acceleration terms balance a large portion of the driving alongshore gradient in the mean water surface elevation in the vicinity of the rip currents with bottom friction accounting for the remainder. Away from the rips, the balance is between the wave forcing and the pressure gradient outside the surf zone and wave forcing and bottom friction inside the surf zone, as expected.

Morphodynamic variability of surf zones and beaches: A synthesis

Globally rip currents are the primary physical hazard facing swimmers on surf beaches. However, beach swimmers also face other hazards such as large waves, tidal influenced currents, and shorebreak waves. The aim of this study was to investigate factors leading to the increased likelihood of surf lifeguard rescues. Rescue data from Surf Life Saving Queensland’s Lifesaving Incident Management System and Operations Console for 54 wave dominated beaches in South–East Queensland, Australia, from July 1st, 2016 to October 6th, 2021 was linked with wind speed and direction, air temperature, phase of tide, wave height and period, beach type, beach hazard rating, and beach swimmer numbers. Stepwise regression was performed to find independent predictors of rescue. There were 8515 rescues, with 3345 (39.3%) females and 5109 (60.0%) males (61 sex not recorded). There were no independent predictors of surf rescue but swimming outside the lifeguard patrol area was nine times more likely to result in rescue than swimming inside the patrol area. Increased rescues were noted at periods of increased rip activity. Rip currents (2992/6523, 45.8%) were the most frequently recorded contributing factor. Rescues occurred most frequently (5902, 69.3%) during the six hours of lower tide levels and during onshore winds (5463, 64.2%). Surf rescues increased with increasing wave height and period, air temperature, and wind speed but decreased as average values for each variable were surpassed. Beaches protected from the prevailing wave direction by headlands had a stronger relationship between rescues and wave height. Beaches adjacent to inlets with tidal flow had a stronger relationship between rescues and the ebb tide. Beach morphology, and hazard rating did not have a relationship with ratio of rescues per 100,000 swimmers. We found no independent predictors for surf rescue, however this study has, for the first time we believe, quantified the increased risk (× 9) posed by swimming outside the patrol area. Open beaches, beaches protected by headlands and beaches with tidal inlets all had different relationships between rescues, tides and wave size. Our findings suggest that lifeguards may need to adopt new approaches to prevent rescues adjacent to the patrol area, as well as a revision of the general hazard rating being required.

Rip currents in the southern Baltic Sea multi-bar nearshore zone

Abstract. Rip currents (“rips”) are the leading cause of drowning on surf beaches worldwide. A major contributing factor is that many beachgoers are unable to identify rip currents. Previous research has attempted to quantify beachgoers' rip identification ability using photographs of rip currents without identifying whether this usefully translates into an ability to identify a rip current in situ at the beach. This study is the first to compare beachgoers ability to identify rip currents in photographs and in situ at a beach in New Zealand (Muriwai Beach) where a channel rip current was present. Only 22 % of respondents were able to identify the in situ rip current. The highest rates of success were for males (33 %), New Zealand residents (25 %), and local beach users (29 %). Of all respondents who were successful at identifying the rip current in situ, 62 % were active surfers/bodyboarders, and 28 % were active beach swimmers. Of the respondents who were able to identify a rip current in two photographs, only 34 % were able to translate this into a successful in situ rip identification, which suggests that the ability to identify rip currents by beachgoers is worse than reported by previous studies involving photographs. This study highlights the difficulty of successfully identifying a rip current in reality and that photographs are not necessarily a useful means of teaching individuals to identify rip currents. It advocates for the use of more immersive and realistic education strategies, such as the use of virtual reality headsets showing moving imagery (videos) of rip currents in order to improve rip identification ability.

Rip currents are narrow, seaward directed flows in the surf zone that can pose a serious threat to swimmers. This issue has received attention particularly on swell dominated coasts (such as the US, Australia, France and UK) where numerous field experiments have been undertaken. However, the threat of rip currents is less recognised on wind-sea dominated coasts such as the North Sea, even though a consistent number of swimmers drift offshore (in rip currents) and require rescue by surf lifeguards each year (for example at Egmond aan Zee, The Netherlands). In August 2011, a five day field experiment SEAREX (Swimmer Safety at Egmond – A Rip current Experiment) was conducted. Lagrangian velocity measurements were taken with drifter instruments and human drifters that were tracked via GPS. Three flow patterns were observed in the experiment: (1) a locally governed circulation cell, (2) a pattern in which the drifter initially floats offshore and then is advected by a strong tidal alongshore-directed current and (3) a meandering longshore current between the shoreline and the bar. A variety of rip current velocities were measured with the strongest being approximately 0.6 m/s. The field data was hindcasted with the numerical model XBeach. Based on this model the sensitivity of rip currents towards wave height, water level and rip channel depth was investigated. Both field and model data show that offshore velocities in a rip increase with increasing wave height and decreasing water level, but that the rip channel depth imposes an upper limit on the rip current velocity.

Kim, I.; Lee, J.L.; Hwang, J.S.; Lee, S., and Lee, J., 2014. Vertical structure of rip current observed at Haeundae Beach. On 4 and 7 August, 2012, more than 200–300 people were swept away by the fast moving seaward current at Haeundae Beach, which is located on the southeastern corner of the Korean Peninsula, and they then had to be rescued Beach patrons were carried away 50 to 100 meters away from the shore by rip currents, to an area where swimming is restricted due to safety reasons. Rip currents are narrow currents that begin close to the shoreline and strongly flow seaward through the surf zone and beyond. Although rip currents can cause tragic accidents and are of considerable interest to scientists and to the general public, our current understanding of the spatial and temporal behavior of rip currents is limited. As a result, the Korean Meteorological Administration installed monitoring equipment that can assess weather conditions and issue rip current warnings on Haeundae beach starting...

Direct observation measurement of drowning risk exposure for surf beach bathers

Chronic lymphocytic leukemia (CLL), the most common leukemia in adults, presents with a broad morphological spectrum, particularly as the disease progresses. Peripheral blood (PB) smear review is essential for evaluating neoplastic cell morphology in CLL. Manual PB smear review, however, is labor-intensive, time-consuming, and subject to inter-observer variability. These limitations underscore the need for more objective and scalable diagnostic tools. Integrating artificial intelligence (AI) into CLL diagnostics may offer a promising solution for automating and standardizing PB smear evaluation. To develop the Convolutional Neural Network (CNN) models for CLL, we are expanding our current image database for MDA-LeukoLens system to include three more categories: (1) typical and atypical CLL cells, (2) prolymphocytes, and (3) large lymphoma cells. Given the variability in cell size across PB smears, we use the ratio of lymphocyte-to-RBC surface area, calculated as the square of the diameter ratio, for quantitative analysis. Approximately 10,000 CellaVision images per cell type are being collected for CNN model training. In addition to having one WBC, each CellaVision image also contains multiple RBCs. The image analysis model for this CLL project was used to create bounding boxes around all RBCs, and those with similar x- and y-axis diameters were selected to calculate the average RBC area per image. The area of each CNN-classified lymphocyte was then compared to the average RBC area in the same image by computing the ratio of the bounding box of the lymphocyte and the average bounding box area of the RBCs. RBCs in approximately 1,000 CellaVision images were labeled with bounding box to train the model for RBC cropping. An independent 1,001 CellaVision images containing atypical lymphoid cells from CLL patients were used to calculate the size ratio of lymphoid cells to RBCs. In the typical CLL cases (737 images), the size of atypical lymphoid cells was variable with the lymphocyte-to-RBC area ratio ranging from 0.5 to 4.49. Approximately 90% of lymphocytes were within the 1.0–2.5 range (1.0–1.49: 27.4%; 1.5–1.99: 42.9%; 2.0–2.49: 19.9%), with only 2.7% exceeding a ratio of 3.0. In the case of accelerated phase (67 images), the area ratio ranged from 2.0 to 7.99. 1.0–2.5: 6.0%; 2.5–4.0: 44.8%, 4.0–5.99: 40.3%, and 6.0–6.99: 7.5%. In the cases of Richter transformation (197 images), the area ratio ranged from 2.5 to 9.99, specifically, 1.0–2.5: 0%; 2.5–4.0: 10.2%; 4.0–5.99: 47.2%; and 6.0–9.99: 28.4%. In summary, our AI-assisted image analysis has revealed certain morphological variability among neoplastic B cells, even in classical CLL cases. A higher proportion of large lymphoid cells correlates with disease progression. This approach has demonstrated improved performance over existing automated imaging platforms and traditional reporting workflows. Overall, these findings support the utility of AI-assisted image analysis in reducing observer variability and providing a scalable framework for monitoring disease evolution and progression in CLL. Citation Format: J. Matthew. Liu, Yun Gong, Amaris Shi, Jeffrey Liu, Xiaoping Sun, Zhihong Hu. Enhancing the Diagnosis and Monitoring of Chronic Lymphocytic Leukemia with Artificial Intelligence-Driven Peripheral Blood Smear Image Analysis [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Artificial Intelligence and Machine Learning; 2025 Jul 10-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(13_Suppl):Abstract nr B046.

We present a hybrid machine learning and flow analysis feature detection method, RipViz, to extract rip currents from stationary videos. Rip currents are dangerous strong currents that can drag beachgoers out to sea. Most people are either unaware of them or do not know what they look like. In some instances, even trained personnel such as lifeguards have difficulty identifying them. RipViz produces a simple, easy to understand visualization of rip location overlaid on the source video. With RipViz, we first obtain an unsteady 2D vector field from the stationary video using optical flow. Movement at each pixel is analyzed over time. At each seed point, sequences of short pathlines, rather a single long pathline, are traced across the frames of the video to better capture the quasi-periodic flow behavior of wave activity. Because of the motion on the beach, the surf zone, and the surrounding areas, these pathlines may still appear very cluttered and incomprehensible. Furthermore, lay audiences are not familiar with pathlines and may not know how to interpret them. To address this, we treat rip currents as a flow anomaly in an otherwise normal flow. To learn about the normal flow behavior, we train an LSTM autoencoder with pathline sequences from normal ocean, foreground, and background movements. During test time, we use the trained LSTM autoencoder to detect anomalous pathlines (i.e., those in the rip zone). The origination points of such anomalous pathlines, over the course of the video, are then presented as points within the rip zone. RipViz is fully automated and does not require user input. Feedback from domain expert suggests that RipViz has the potential for wider use.