We present a method based on deep learning for detecting and localizing abnormal/extreme events in sea surface temperature (SST) of the Red Sea images using training samples of normal events only. The method operates in two stages; the first one involves features extraction from each patch of the SST input image using the first two convolutional layers extracted from a pretrained convolutional neural network. In the second stage, two methods are used for training the model from the normal training data. The first method uses one-class support vector machine (1-SVM) classifier that allows a fast and robust abnormal detection in the presence of outliers in the training dataset. In the second method, a Gaussian model is defined on the Mahalanobis distances between all normal training data. Experimental tests are conducted on satellite-derived SST data of the Red Sea spanning for a period of 31 years (1985-2015). Our results suggest that the Gaussian model of Mahalanobis distances outperformed 1-SVM by providing better performance in terms of sensitivity and specificity.

The interannual variability of the exchange flow between the Red Sea and the Gulf of Aden through the Bab-al-Mandeb strait is examined based on a high-resolution, nonhydrostatic hindcast model simulation covering a 19-year period (1995–2013), using the MITgcm (MIT general circulation model). The model is validated against moored profiles and along-strait cruise observations collected during the period from June 1995 to November 1996 and 19-year sea surface temperature satellite observations. The model well reproduces the properties of the water masses at the strait over a wide range of spatiotemporal scales, including the typical two- and three-layer seasonal patterns and the related intraseasonal-to-interannual cycles. The seasonality of the exchange flow is predominately determined by the time-varying surface winds, with a higher correlation over the Gulf of Aden, reflecting the importance of local Gulf of Aden processes for the exchanges at the strait. The alternation of the two seasonal patterns is driven by a combination of the buoyancy-driven mean circulation with the wind-induced transport. The onset/offset of the two patterns is estimated to take place one-to-two weeks after the respective monsoon-driven wind reversal. Model results indicate that the onset dates and durations of both patterns exhibit a considerable interannual variability. Additionally, the duration of the summer (winter) exchange pattern presents a significant increasing (decreasing) trend of ~1.45 day/year (~1.22 day/year) over the 19-year period. Significant interannual variabilities and trends are observed in terms of the total volume of water, salt mass, and stored heat of the exchanges. Budget analysis of these trends suggests that the duration of the two exchange patterns is more important in determining the interannual variability and the related trends than the intensity of the exchange, or the variations in mean salinity or temperature of the exchanged water masses.

Mesoscale ocean eddies play a major role for both the intermixing of water and the transport of biological mass. This makes the identification and tracking of their shape, location and deformation over time highly important for a number of applications. While eddies maintain a roughly circular shape in the free ocean, the narrow basins of the Red Sea and Gulf of Aden lead to the formation of irregular eddy shapes that existing methods struggle to identify. We propose the following model: Inside an eddy, particles rotate around a common core and thereby remain at a constant distance under a certain parametrization. The transition to the more unpredictable flow on the outside can thus be identified as the eddy boundary. We apply this algorithm on a realistic simulation of the Red Sea circulation, where we are able to identify the shape of irregular eddies robustly and more coherently than previous methods. We visualize the eddies as tubes in space-time to enable the analysis of their movement and deformation over several weeks.

Phytoplankton size structure impacts ocean food-web dynamics and biogeochemical cycling, and is thus an important ecological indicator that can be utilised to quantitatively evaluate the state of marine ecosystems. Potential alterations to size structure are predicted to occur in tropical regions under future scenarios of climate change. Therefore, there is an increasing requirement for the synoptic monitoring of phytoplankton size structure in marine systems. The Red Sea remains a comparatively unexplored tropical marine ecosystem, particularly with regards to its large-scale biological dynamics. Using an in situ pigment dataset acquired in the Red Sea, we parameterise a two-component, abundance-based phytoplankton size model and apply it to remotely-sensed observations of chlorophyll-a (Chl-a) concentration, to infer Chl-a in two size classes of phytoplankton, small cells <2 μm in size (picophytoplankton) and large cells >2 μm in size. Satellite-derived estimates of phytoplankton size structure are in good agreement with corresponding in situ measurements and also capture the spatial variability related to regional mesoscale dynamics. Our analysis reveals that, for the estimation of Chl-a in the two size classes, the model performs comparably or in some cases better, to validations in other oceanic regions. Our model parameterisation will be useful for future studies on the seasonal and interannual variability of phytoplankton size classes in the Red Sea, which may ultimately be relevant for understanding trophic linkages between phytoplankton size structure and fisheries, and the development of marine management strategies.

In this paper, we consider the autonomous underwater vehicle (AUV) trajectory planning problem under the influence of a realistic 3D current as simulated by an ocean general circulation model (OGCM). Attention is focused on the case of a deterministic steady OGCM field, which is used to specify data for both the ocean current and for ocean bathymetry. A general framework for optimal trajectory planning is developed for this setting, accounting for the 3D ocean current and for static obstacle avoidance constraints. A nonlinear programming approach is used for this purpose, which leads to a low complexity discrete-time model that can be efficiently solved. To demonstrate the efficiency of the model, we consider the optimal time trajectory planning of an AUV operating in the Red Sea and Gulf of Aden, with velocity, and bathymetric data provided by an eddy-resolving MITgcm. Different optimal-time trajectory planning scenarios are implemented to demonstrate the capabilities of the model to identify trajectories that adapt to favorable and adverse currents and to avoid obstacles corresponding to a complex bathymetry environment. The simulations are also used to evaluate the performance of the proposed approach, and to illustrate the application of advanced visualization tools to interpret the model predictions.

Phytoplankton biomass and size structure are recognized as key ecological indicators. With the aim to quantify the relationship between these two ecological indicators in tropical waters and understand controlling factors, we analyzed the total chlorophyll-a concentration, a measure of phytoplankton biomass, and its partitioning into three size classes of phytoplankton, using a series of observations collected at coastal sites in the central Red Sea. Over a period of 4 years, measurements of flow cytometry, size-fractionated chlorophyll-a concentration, and physical-chemical variables were collected near Thuwal in Saudi Arabia. We fitted a three-component model to the size-fractionated chlorophyll-a data to quantify the relationship between total chlorophyll and that in three size classes of phytoplankton [pico- (<2 μm), nano- (2-20 μm) and micro-phytoplankton (>20 μm)]. The model has an advantage over other more empirical methods in that its parameters are interpretable, expressed as the maximum chlorophyll-a concentration of small phytoplankton (pico- and combined pico-nanophytoplankton, Cpm and Cp,nm , respectively) and the fractional contribution of these two size classes to total chlorophyll-a as it tends to zero (D p and D p,n ). Residuals between the model and the data (model minus data) were compared with a range of other environmental variables available in the dataset. Residuals in pico- and combined pico-nanophytoplankton fractions of total chlorophyll-a were significantly correlated with water temperature (positively) and picoeukaryote cell number (negatively). We conducted a running fit of the model with increasing temperature and found a negative relationship between temperature and parameters Cpm and Cp,nm and a positive relationship between temperature and parameters D p and D p,n . By harnessing the relative red fluorescence of the flow cytometric data, we show that picoeukaryotes, which are higher in cell number in winter (cold) than summer (warm), contain higher chlorophyll per cell than other picophytoplankton and are slightly larger in size, possibly explaining the temperature shift in model parameters, though further evidence is needed to substantiate this finding. Our results emphasize the importance of knowing the water temperature and taxonomic composition of phytoplankton within each size class when understanding their relative contribution to total chlorophyll. Furthermore, our results have implications for the development of algorithms for inferring size-fractionated chlorophyll from satellite data, and for how the partitioning of total chlorophyll into the three size classes may change in a future ocean.

As the Earth's temperature continues to rise, coral bleaching events become more frequent. Some of the most affected reef ecosystems are located in poorly-monitored waters, and thus, the extent of the damage is unknown. We propose the use of Marine Heatwaves (MHWs) as a new approach for detecting coral reef zones susceptible to bleaching, using the Red Sea as a model system. Red Sea corals are exceptionally heat-resistant, yet bleaching events have increased in frequency. By applying a strict definition of MHWs on >30-year satellite-derived sea surface temperature observations (1985-2015), we provide an atlas of MHW hotspots over the Red Sea coral reef zones, which includes all MHWs that caused major coral bleaching. We found that: 1) if tuned to a specific set of conditions, MHWs identify all areas where coral bleaching has previously been reported; 2) those conditions extended farther and occurred more often than bleaching was reported; and 3) an emergent pattern of extreme warming events is evident in the northern Red Sea (since 1998), a region until now thought to be a thermal refuge for corals. We argue that bleaching in the Red Sea may be vastly underrepresented. Additionally, although northern Red Sea corals exhibit remarkably high thermal resistance, the rapidly rising incidence of MHWs of high intensity indicates this region may not remain a thermal refuge much longer. As our regionally-tuned MHW algorithm was capable of isolating all extreme warming events that have led to documented coral bleaching in the Red Sea, we propose that this approach could be used to reveal bleaching-prone regions in other data-limited tropical regions. It may thus prove a highly valuable tool for policy-makers to optimise the sustainable management of coastal economic zones. This article is protected by copyright. All rights reserved.

Mesoscale eddies are a dominant feature of the Red Sea circulation, yet their three-dimensional characteristics remain largely unexplored. This hinders our understanding of about eddy-induced transport in the basin. This study analyzes 14-year outputs from a high-resolution eddy-resolving model to investigate the three-dimensional signature of the Red Sea eddies, their contribution to the air-sea flux, and the eddy-induced transport of heat and salt. Eddies are mostly active and energetic in the central and northern Red Sea. Their associated variability explains ∼8% of the total variance in the surface heat flux, and particularly, ∼39% in the salt flux. The asymmetric eddy structure and meridional gradient drive significant transport of heat and salt across the basin. A negative feedback mechanism is identified that relates the eddy intensity and the meridional steepness of the mixed layer depth in the basin.