| Rip currents are a common natural hazard with characteristics of suddenness,concealment,and rapid changes in flow velocity.Existing literatures show that rip currents are an important factor in drowning accidents on beaches around the world.The research on rip currents in China has just started,and there is insufficient research on the formation mechanism,distribution characteristics and risk evaluation of rip currents.In this paper,54 embayed beaches on Hainan Island were investigated,the characteristics of rip currents were systematically analyzed,and the hazard level and distribution of rip currents around the island were evaluated,and all the results can be divided into the following parts.(1)There is a serious lack of public understanding about rip currents,despite the fact that they play a significant role in drowning accidents.Statistics on drowning accidents on Hainan Island from 2003 to 2022 revealed that Haikou and Sanya had the highest rates of such accidents.Sanya also had the highest number of drowning fatalities,and the analysis of satellite images revealed that Sanya had the highest proportion of rip current-prone beaches.Through the use of online and field surveys,the awareness of beachgoers regarding rip currents was examined.The findings revealed that the majority of respondents were unaware of rip currents,highlighting the significance of rip current study.(2)Rip current growth and formation are intimately correlated with beach state and bar shape.The 54 beaches’ states were determined and classified by morphodynamic model and hierarchical clustering method respectively,while the beach images captured by UAVs were interpreted to determine the presence of sandbars on the 54 beaches.The findings indicated that there were six groups into which the 54 beaches might be divided: complete reflective(R),non-barred dissipative(NBD),low tide bar/rip(LTBR),barred(B),barred dissipative(BD),and low tide terrace with rip(LTTR).A total of three different sorts of sandbars: no sandbar,single sandbar,and multiple sandbar types.Sandbar-affected beaches are more prone to experience rip currents.(3)A deep learning-based rip current detection algorithm was proposed for identifying rip current from satellite images and UAV images.Firstly,the neck region in the YOLOv5 s model was streamlined,and the 80×80 feature map branches suitable for detecting small targets were removed.Then a joint dilated convolutional(JDC)module was proposed to join the lateral connections of the feature pyramid network(FPN),which expanded the perceptual field,improved feature information utilization,and reduced the number of parameters,while keeping the model compact.Finally,the Sim AM module,which was a parametric-free attention mechanism,was added to optimize the target detection accuracy.The findings indicate that the proposed model has the best detection impact on the same dataset when comparing the detection effects of various models.Compared with YOLOv5 s model,our model increased the m AP value by approximately 4%(to 92.15%),frame rate by 2.18 frames per second.The modified model improved the detection accuracy while keeping the model streamlined,indicating its efficiency and accuracy in the detection of rip currents.(4)Our detection model was used to locate the rip current in satellite images,its temporal and geometric scale features were deciphered,and the relationship between the rip current scale and the wave environment was investigated.According to the findings,there is no significant correlation between rip width and rip spacing or between the number of rip currents and rip spacing over a long time scale,but there is a significant positive correlation between average rip length,average rip width,and rip spacing.There is a significant negative correlation between rip density and significant wave height,wave period,wave power and wave energy.Rip currents were frequently observed in the fall and winter,and the east and south coasts of Hainan Island had a higher likelihood of experiencing rip currents with a high hazard level.The rip current is unstable at short time scales,according to field observations.Circulation and advection alongshore are the two main flow patterns observed as the channel rips at embayed beaches.The advection alongshore on beaches is more likely to be persistent than the circulation.In the vertical shoreline direction,the feeder flow converges to the rip root and forms an offshore flow,which flows to the rip neck and rip head successively.The flow velocity increases initially before decreasing and peaks at the rip neck.The flow velocity increases as the wave height and period increase under various wave conditions.(5)Combining the discrimination of morphodynamic model,satellite image interpretation,and field survey results,the distribution of rip hazard level along the Hainan Island coast was firstly mapped.Comparing the results of the Ω-RTR model and the RI model for the evaluation of rip hazard of 54 beaches,the agreement rate is87.04%.In identifying rip currents,the findings of the satellite image identification agreed with the Ω-RTR model identification results by 89.5%.Meanwhile,the results of field observation and satellite image interpretation verified that both models were applicable to the rip current hazard evaluation of embayed beaches on Hainan Island,and the RI model was more accurate.The Ω-RTR model and satellite image interpretation methods have limitations in identifying flash rips,and in future research,we will try to identify flash rips using deep learning methods.In order to get more information about coastal morphodynamics and hydrodynamics for better understanding the movement and evolution of different types of rip currents,long-term observations will be conducted at the embayed beaches of Hainan Island. |
PDF Full Text Request