“There exist published research papers on unsupervised regime identification in multivariate oceanic current time series, particularly focusing on coastal regions and methods that infer the number of regimes from data, which are relevant for forecasting applications in areas such as Bahia de Santos, Brazil.”

This unique model produces a single time series of regime shift likelihood, using sequential abundance data of more than 300 plankton species. In our study, we have analyzed multiple time series of plankton abundance simultaneously. We here describe the process of generating a single time series representing the likelihood of a regime shift occurring at any point represented by the time series.

In order to understand the variability of the BC, a statistical classification of the BC is proposed, with the identification of three different events: weak, intermediate, and strong. The OFAS show that the coastal region between the coastline and the western edge of the AX97 transect plays an important role in the mean BC total transport, contributing to up to 23% of its value, and further that this transport is not sampled by the XBT data.

This manuscript presents a novel methodology, the Regime Shift Tool (RST) model, for detecting ecosystem regime shifts, with an application to planktonic time series from the North Sea using Continuous Plankton Recorder (CPR) data. The study addresses a critical gap in the detection of regime shifts, particularly within open-ocean systems, by synthesizing multiple plankton time series into a composite likelihood metric.

This theme issue 'Marine regime shifts around the globe: theory, drivers and impacts' has the goal to make a step change towards a more unified understanding of marine regime shifts. We review observational data sets that have been used to detect regime shifts in the ocean.

Data-Driven Response Regime Exploration and Identification (DR^2EI) is a novel and fully data-driven method for identifying and classifying response regimes of a dynamical system without requiring human intervention. This approach is a valuable tool for exploring and discovering response regimes in complex dynamical systems, especially when the governing equations and the number of response regimes are unknown, and the system is expensive to sample. The method was shown to effectively identify a variety of response regimes with both similar and distinct topological features and frequency content, demonstrating its versatility in capturing a wide range of behaviors.

Ocean boundary currents are complex and highly variable systems that play key roles in connecting the open and coastal ocean through cross-slope circulation and upwelling of nutrient-rich water. Continued monitoring of boundary currents is also central to assessing ocean and climate models, improving the accuracy and reliability of weather forecasts locally and remotely, reducing biases in global climate models, and improving climate change projections.

Two-dimensional (2D) maps and time series are displayed for the selected variables at different time scales to monitor the ocean variability. Multivariate time series analysis for ocean indicators in coastal regions.

The Brazil Current (BC) plays an important role in oceanic variability and related processes, as it significantly influences regional and global climate dynamics. The AX97 data integration into assimilation schemes enhances the accuracy of short-term ocean predictions and long-term reanalyses, benefiting global forecasting centers by improving ocean models at regional, basin, and global scales. Moreover, the ongoing expansion of the AX97 program to include shelf waters and the collection of atmospheric data during cruises ensures the continued enhancement of the dataset, directly benefiting climate science and strengthening the resilience of coastal communities.

Dynamically similar regions of the global ocean are identified using a barotropic vorticity (BV) framework from a 20‐year mean of the Estimating the Circulation and Climate of the Ocean state estimate at 1° resolution. An unsupervised machine learning algorithm, K‐means, objectively clusters the standardized BV equation, identifying five unambiguous regimes. This approach appears to be both interesting and useful and is readily generalized to far more complex oceanic states.

This contribution proposes a methodology for improving operational forcings by correcting them with Artificial Neural Networks (ANNs), leveraging remote sensing data as targets and learning complex spatial patterns. Autoencoders (AEs) are used to capture the spatio-temporal patterns in High-Frequency Radar (HFR) data, which is an unsupervised deep learning technique relevant for predicting surface currents in coastal zones and improving short-term wave forecasts.

Dynamically similar regions of the global ocean are identified using a barotropic vorticity (BV) framework from a twenty-year mean of the ECCO state estimate at 1◦ resolution. An unsupervised learning algorithm, k-means, objectively clusters the standardized BV equation, identifying five unambiguous regimes. Cluster 3 covers 14.6 ± 1.0%, characterized by a 'Quasi-Sverdrupian' regime where the beta effect is balanced by the wind and bottom stress term. The small region of Cluster 4 has baroclinic dynamics covering 6.9 ± 2.9% of the ocean. Cluster 5 occurs primarily in the Southern Ocean. Residual 'dominantly non-linear' regions highlight where the BV approach is inadequate, found in areas of rough topography in the Southern Ocean and along western boundaries.

This study investigates modeling the wind-driven current using observed wind stress and an empirically estimated impulse response function for the wind-driven current response to wind forcing. Convolution of the data-driven impulse response function with the wind forcing gives an estimate of the wind-driven part of the current. A practical application is the release of new total surface current estimates such as the Globcurrent CMEMS MOB-TAC based on the same datasets, but here containing also the unsteady part of the wind driven currents (the inertial currents).

The methods reviewed here are primarily those that are used in atmospheric and oceanic (physical and biological) studies. These methods are divided into four groups, depending on the type of shifts they are designed to detect, and placed in Tables 1-4. The Lanzante method is an iterative procedure designed to search for multiple change-points and involves the application of a non-parametric test. The Rodionov method is similar to STARS, but based on the F-test, and signals a possibility of a regime shift in real time, allowing for automatic detection of multiple change-points.

Oceanographers use data on buoy velocity to predict ocean currents and identify “divergences” where water rises to the surface or sinks deeper. A new model that incorporates machine learning makes more accurate predictions than conventional models do, a new study reports. This new model could help oceanographers make more accurate estimates from buoy data, which would enable them to more effectively monitor the transportation of biomass (such as Sargassum seaweed), carbon, plastics, oil, and nutrients in the ocean.

Scientists have developed GOFLOW (Geostationary Ocean Flow), a new method applying deep learning to thermal images from weather satellites to measure ocean surface currents over large areas in greater detail. This approach tracks how temperature patterns move to infer underlying currents, opening the door to testing long-standing ideas about ocean processes and potentially improving weather forecasts by resolving rapidly evolving currents.

Biogenic and rocky reefs are distributed along approximately 27 degrees of latitude and exposed to different temperature, productivity and salinity regimes, which generates a pronounced regionalization of the biodiversity. We compiled data collected between 2001 and 2023 on fish and benthic organisms from 55 locations and 355 unique localities distributed along the Brazilian coast and oceanic islands, including coastal islands of São Paulo (Santos to Ubatuba).

This study applies a nonlinear neural network analysis based on the self-organizing map (SOM), using time series of moored velocity data that span the interval October 1998–September 2001. Three characteristic spatial patterns are extracted in a 3 x 4 SOM array: spatially coherent southeastward and northwestward flow patterns with strong currents and a transition pattern of weak currents. The spatial patterns extracted by the (nonlinear) SOM method are asymmetric, a feature that is not captured by the (linear) empirical orthogonal function method.

This paper proposes an in-depth evaluation study of recent unsupervised anomaly detection techniques in time-series, taking into account more elaborate performance metrics specifically tailored for time-series, model size, and stability. The analysis aims to assess the maturity of state-of-the-art time-series anomaly detection and provide insights regarding their applicability under real-world setups.

Numerous oceanographic systems are providing data on physical and biogeochemical variables, spanning from global, regional and coastal scale. The regional oceanographic services play a crucial role in measuring the Ocean Essential Variables (EOV) defined by the Global Ocean Observing System (GOOS). This description is mainly focused on short term forecasting products, because most systems do not provide long climatological series of the past to understand how ocean conditions are changing over time.

This work aims to improve the forecast of surface currents in the entrance of the Santos estuary in Brazil by applying Quantile Regression Forests (QRF) to estimate the error of the Santos Operational Forecasting System (SOFS), a physics-based numerical model for the region. This was achieved by using in-situ data, measured between 2019 and 2021, associated with historical forecasted data from the SOFS. The use of QRF to correct the SOFS forecasts led to an increase in skill of 0.332 in Mean Absolute Error (MAE) and almost eliminated the bias error of the predicted currents.

In this study we propose to characterize the variability of the occurrence of oceanic fronts in the Santos Basin. Satellite-derived SST images obtained during 2002-2022 were submitted to an edge detection algorithm to identify oceanic fronts in the region. The results indicate the occurrence of oceanic fronts mainly associated with the presence of the Brazil Current, also possibly related to upwelling phenomena and intrusion of cold waters masses coming from the south during the winter.

The aim of this dissertation was to develop a new algorithm called Core-Shell clustering algorithm, which is an extension of the previously developed Sequential Self Tuning Seeded Expanding Cluster (S-STSEC) algorithm. This new algorithm aims at the automatic recognition, definition and ST characterization of coastal upwelling from SST images. The segmentation and tracking of the dynamics of such structures can be performed by the use of a novelty clustering concept introduced in [20] where dynamic Spatio-Temporal (ST) clusters are studied and identified using consecutive Sea Surface Temperature (SST) images over a period of time.

analyses of these data sets enabled identification and quantification of the processes associated with incident waves and winds. The findings indicated that height and angle of incidence of waves was the main driving force of alongshore currents. A pattern has also been identified, in which waves that come from the south generate northward alongshore currents, and waves from the east generate southward alongshore currents.

This work assess the performance of climate-informed Non-Homogeneous Hidden Markov Models (NHMMs) to forecast annual inflow persistence of Orós reservoir, located in the state of Ceará, Brazil. NHHMs represent the system as a set of hidden climate states that influence the inflows occurrence and allow include climate information to influence the transition between climate states. It is shown that similar to ARX models, NHMMs can forecast annual dry and wet period inflows to aid in water resources management.

Based on a vast set of in situ data, a first comprehensive overview of the oceanographic characteristics of Baía de Todos os Santos (BTS) is provided. BTS is the second largest coastal bay in Brazil (maximum area of 1223 km2 and average depth of 9.8 m), and is located in the northeast Brazil, in the vicinity of Salvador city. The circulation inside the bay is mostly tidally driven and does not vary significantly throughout the year. Tidal currents in the bay are mainly bi-directional, and stronger during ebb tide in most of the bay.

Brazil is helping to chart that course, by using cutting-edge technology in creative ways to oversee its coastline with tools that turn complex data into clear, actionable insights. From energy to monitoring, from distant archipelagos to busy ports, Brazil's agencies are using open data and transparency to inform policy and strengthen protection measures. Since 2024, Brazil has been using Global Fishing Watch data as one of the data sources to assess new offshore blocks considered for oil and gas exploration.

Published research exists on unsupervised methods like Hidden Markov Models (HMM) and Gaussian Mixture Models for regime identification in multivariate oceanic time series, including coastal areas; methods often infer regime numbers via BIC/AIC. Specific applications in South Atlantic coastal regions like Santos Basin use current meter data for forecasting oil spill dispersion and circulation regimes.