Early Perspectives on the Planned Brazilian Program to Address Ship-Sourced Pollution

Abstract

A new integrated multi-user monitoring system for Brazilian Jurisdictional Waters (BJW), developed by Instituto Nacional de Pesquisas Espaciais (INPE) with participation from leading universities and research centers in Brazil, proposes a national approach to address oil spills in the South Atlantic. The system incorporates a range of technologies, such as satellite data, AI algorithms, autonomous sensors, and high-resolution modeling, to detect and respond to oil spills and maritime threats. This initiative not only aims to strengthen Brazil’s readiness to address the oil spills but also contribute to the protection of BJW resources and ecosystems. This opinion paper presents third-party viewpoints on SisMOM, analyzing both the positive and negative aspects of the project. It also explores some expectations for SisMOM, including some main and alternative methodologies. This article only reflects the authors’ perspectives, interpretations, points of view, opinions, and discussions about SisMOM’s propositions. This paper does NOT represent an official communication of the program, nor its methodologies and developments.

1. Introduction

In late August 2019, Brazil experienced one of the most extensive oil spill events to ever impact its coastal environment. Over the course of several months, vast quantities of tar balls reached more than 3000 km of shoreline, affecting eleven states along the Equatorial and Atlantic boards. The event was marked by a unique characteristic: the absence of an identified source in the immediate aftermath. No maritime accidents were reported, and the oil appeared suddenly and persistently, contaminating beaches, mangroves, and marine ecosystems. This “Mysterious” oil spill incident underscored the vulnerability of Brazil’s jurisdictional waters [1,2,3].

This environmental disaster prompted a wide mobilization effort involving local communities, governmental agencies, and independent researchers. The response was hampered by delayed detection, lack of predictive tools, and limited preparedness for an event of such spatial scale and uncertain origin. The incident underscored critical gaps in national oil spill monitoring frameworks and raised broader questions about the surveillance and governance of maritime traffic in Brazilian jurisdictional waters [1,2,3].

The institutional response to the oil spill was widely regarded as delayed and poorly coordinated. In the critical weeks following the first reports of contamination along the coast, the absence of clear leadership and the slow mobilization of federal resources revealed the vulnerability of Brazil’s governmental structure in dealing with large-scale environmental emergencies. Small-scale fishers and traditional populations were particularly affected, facing significant economic losses and health risks without immediate governmental support or protective policies [4,5,6].

In the face of this partial state inaction, civil society took the lead in the initial containment and cleanup efforts. Thousands of volunteers spontaneously organized beach cleanups, often without proper equipment or protective gear. Universities, NGOs, and local collectives contributed by providing technical analyses, mapping efforts, and logistical support. This combination of institutional fragility and civic resilience highlighted the urgent need to strengthen Brazil’s preparedness and response mechanisms for environmental disasters. As a direct consequence of the lessons learned from the 2019 event, new initiatives emerged to develop integrated systems for the forecasting, monitoring, and response to marine oil spills. These ongoing efforts aim to enhance marine environmental governance and reduce the impact of future accidents within Brazilian jurisdictional waters [7,8,9].

Since the 2019 “Mysterious” oil spill, at least two more significant pollution incidents caused by ships have been recorded in ecological sanctuaries along the Ceará coast and within the Fernando de Noronha archipelago [10,11,12]. Moreover, large-scale probabilistic modeling estimated that up to 30% of ships on route through the South Atlantic may be exploiting the lack of international surveillance to dump waste and oil close to BJW [13,14]. Brazil’s history of oil spills shows a significant reduction in accidents following the implementation of a new environmental protection legislation on 28 April 2000. However, the number of intentional spills (dumping) has increased over the last 5 years, highlighting the need for a more effective oil spill monitoring and response system [1]. In response, the investment in a new program for oil spill monitoring and modeling (SisMOM: Sistema Multiusuário de detecção, previsão e monitoramento de derrame de Óleo no Mar; i.e., Multi-user System for Offshore Oil Spill Detection, Forecast, and Monitoring) marks a pivotal development in safeguarding Brazil’s coastline against oil pollution [15,16].

SisMOM is the integrated, multi-user monitoring system for BJW, which utilizes space technologies, autonomous vehicles, moored buoys, atmospheric and oceanic models, and artificial intelligence techniques. Instituto Nacional de Pesquisas Espaciais (INPE, i.e., National Institute for Space Research) coordinates the project’s development with the participation of leading universities and research centers in Brazil [15,16]. The public contract was registered under the Agreement for Research, Development and Innovation number 01.22.0102.00, reference 1576/21, signed between the Financiadora de Estudos e Projetos1 (FINEP), the Fundação de Ciência, Aplicações e Tecnologia Espaciais2 (FUNCATE), the Instituto Nacional de Pesquisas Espaciais (INPE), and the Ministério da Ciência, Tecnologia e Inovação3 (MCTI) [17].

This perspective paper presents third-party viewpoints on SisMOM, analyzing both the positive and negative aspects of the project. It also explores the expectations for SisMOM in the coming years.

2. Project Description

SisMOM is a multidisciplinary and inter-institutional system that integrates operational oceanographic and meteorological forecasting models with remote sensing, data assimilation, and artificial intelligence components. Its architecture allows multiple government agencies, including environmental, defense, and emergency response sectors, to access real-time information and collaborate effectively in managing oil spill incidents. The system combines deterministic and probabilistic modeling tools for oil dispersion and weathering, facilitating scenario analysis and impact assessment. One of its core innovations is the incorporation of high-resolution numerical simulations and adaptive learning algorithms, which enable more accurate predictions of oil drift under complex oceanographic conditions [15,16,17].

The system is structured around a modular and scalable framework that integrates various components for real-time data acquisition, atmospheric and oceanic forecasting, oil spill modeling, and user interaction. The system relies heavily on high-performance computing infrastructure to support the execution of atmosphere and oceanic models, such as the ETA model [18] (http://etamodel.cptec.inpe.br/, accessed on 15 June 2025), widely used by INPE for atmospheric dynamics, and the Modular Ocean Model (MOM6) [19] and TELEMAC for regional and coastal circulation [20].

SisMOM is structured around nine goals that focus mainly on detecting vessels and oil slicks at sea, as well as predicting oil drift and dispersion. The monitoring system will be equipped to identify suspicious vessel behavior using artificial intelligence algorithms that analyze Automatic Identification System (AIS) and high-frequency radio wave data and detect potential oil slicks through high-resolution satellite imagery. High-resolution atmospheric and oceanographic data will be assimilated to run oil spill dispersion models, estimating the probability of oil slicks and tar balls reaching the shore. This data will provide critical information to response teams (Figure 1).

The project also includes a data assimilation layer that integrates satellite remote sensing, AIS (Automatic Identification System) ship tracking data, and in situ observations from buoys and coastal stations. Another key feature is the multi-user decision support interface (Situation Room), designed for interoperability across institutions. The Situation Room integrates live data from satellites, weather stations, and ocean sensors to track environmental changes and support coordinated emergency responses.

Thus, SisMOM is structured in a sequence of integrated levels, ranging from the identification of potential polluters to supporting government actions led by the Instituto Brasileiro do Meio Ambiente e dos Recursos Naturais Renováveis4 (IBAMA), Marinha5 do Brasil (MB), and the Ministério Público6 (MP), following what was established by the National Contingency Plan for Oil Pollution Incidents in Waters under National Jurisdiction (PNC)7, established by Federal Decree 10.950/2022.

The development of the SisMOM is organized around nine distinct goals, and each one is segmented into independent working groups. As a result, not all technical details and methodological choices are fully accessible at this time. Several components of the system involve sensitive information that remains protected for strategic and institutional reasons. Furthermore, many procedures and tools are still under refinement, given the current stage of project implementation. In light of these constraints, some analyses presented in this work are supported primarily by literature reviews and established methodologies available in the scientific domain, rather than by finalized operational data from the system itself.

The project design is subdivided into nine complementary goals.

Goal 1—automatic detection of ships at sea

This goal focuses on training artificial intelligence (AI) algorithms to analyze satellite images and auxiliary data, such as Radio Frequency (RF) signals and AIS signatures, to identify the presence, position, and anomalous movements of vessels.

The integration of AI in detecting unusual ship trajectories is crucial for combating illegal oil spills, especially during nighttime operations when satellite-based oil detection is more challenging [21]. Vessels involved in illicit activities often use tactics such as applying dispersants, disabling AIS, altering routes, and operating under the cover of rain or darkness to evade detection [22,23].

Among the available technologies for maritime surveillance, the multi-channel Synthetic Aperture Radar (SAR) has emerged as one of the most effective and sophisticated techniques for monitoring vessel activities at sea. The usual approaches consist of two steps: (1) selection of the target in SAR images of busy traffic, which corresponds to Automatic Identification System (AIS) signals by the “dead-reckoning (DR) position”, and (2) estimation of the position, size, and speed of the ship from SAR images and comparison these results with the AIS “true” data [24,25].

Despite the effectiveness of this approach, one of its main limitations lies in the massive volume of data generated by multi-channel SAR instruments, especially when operating, or over regions with dense maritime traffic. Processing and interpreting this high-dimensional data in real time is a significant challenge and requires the support of advanced artificial intelligence (AI) techniques, such as Convolutional Neural Networks (CNNs) [26,27]. Although there are other options, CNNs are the most well-established and representative deep learning model [28], they are particularly effective in extracting spatial features from SAR images [29], allowing the automated identification and classification of vessels and oil slicks [30,31].

While the detection of vessels using SAR and CNNs is a relatively well-established and technically mature process, likely, the scientific challenge of this goal lies in identifying and defining patterns of anomalous vessel behavior. Unlike other forms of maritime infractions, illegal oil dumping is often carried out while vessels remain on route, exhibiting no abrupt deviations in course or speed. This subtlety makes anomaly detection particularly complex, as traditional indicators such as sudden changes in trajectory or AIS deactivation may not always be present. As such, advancing this goal requires the development of novel behavioral models capable of discerning less obvious anomalies—potentially through the integration of contextual data, long-term movement patterns, and environmental variables. The successful identification of these nuanced behaviors will be critical to the effectiveness of this objective, as it constitutes the system’s first alert signal. From this initial detection, other components are activated, such as the acquisition of aerial imagery (from satellites, drones, etc.) for visual confirmation of the spill, followed by the deployment of forecasting models and assessment of environmental conditions.

Goal 2—automatic detection of oil at sea

Oil spill detection using SAR and deep learning techniques has become a significantly more established and widely adopted practice compared to vessel detection [32,33]. The physical characteristics of oil slicks—such as their tendency to dampen capillary waves on the ocean surface—result in well-defined dark patterns in SAR images, which can be effectively distinguished from the background sea clutter using machine learning and image classification algorithms [34,35,36]. Over the past decade, several studies have refined this process, improving both accuracy and computational efficiency [37,38,39,40].

The “small satellites, big data” approach offers a cost-effective solution for improving oil slick detection. Microsatellites and CubeSats enable continuous global data collection [41], which can be analyzed using machine learning techniques [21]. This approach enhances the detection of oil slicks in BJW by training AI algorithms to analyze satellite imagery, ocean color data, numerical forecasts, and sea state measurements. This extensive data improves the accuracy of oil slick detection and monitoring, providing timely and precise information for response teams.

Moreover, both goals—ships and oil spill detection—could benefit significantly from a unified methodological framework, as the most widely accepted scientific approaches share considerable overlap, particularly the use of SAR imagery and CNNs, which demonstrated strong performance in extracting spatial data and enabling classification tasks in both contexts [31,32,33,34]. However, it is worth noting that oil spill detection has seen broader experimentation with a variety of machine learning and deep learning techniques beyond CNNs, including transformer-based architectures and U-Net for semantic segmentation, FFDNet for noise reduction, Random Forest classifiers for pattern recognition, and YOLO for real-time object detection [39,40].

Goal 3—oil spill modeling

The development is based on open-source oil spill models such as the Spill, Transport and Fate Model (STFM) [42,43,44]; the Coupled Model for Oil Spill Prediction (CMOP) [45,46]; General NOAA Operational Modeling Environment (PyGNOME) [47,48]; and OpenDrift/OpenOIL [49,50], which can be executed independently or as part of an ensemble approach. This ensemble methodology aims to enhance the reliability of simulations by capturing uncertainties inherent in model structures, initial conditions, and environmental forcing.

To support operational usability, a new user interface for the SISMOM oil modeling platform was developed based on the design and functionalities of the STFM system (FAPESP 2023/01542-3 and 2024/06927-3) [51,52,53]. This interface will (i) improve the setup features for both technical and decision-makers; (ii) streamline the configuration, execution, and visualization of oil spill scenarios; and (iii) enable reproducibility and consistency across simulation outputs (Figure 2).

A critical component of the interface development involves the standardization of input data formats across different modeling frameworks. This ensures interoperability between models and allows the seamless integration of outputs for comparative or ensemble analysis. Moreover, the interface is designed to be fully compatible with the metocean datasets generated under Goal 5, including high-resolution wind, current, and wave fields. This compatibility allows for automatic ingestion and conversion of forecast and hindcast environmental data into a unified format, thereby reducing user burden and minimizing errors associated with manual preprocessing.

New optimized algorithms have been developed to enhance the performance of oil spill modeling using the new high-resolution hydrodynamic grids defined in Goal 5. These algorithms are specifically designed to increase the speed of numerical processing, automatically switch between the input data grids as the particles approach the coast, and assimilate satellite images during the simulation period. This will improve the precision of oil spill modeling and enhance the overall predictive capability of the SisMOM. Moreover, the performance of the models is also being evaluated through well-documented case studies based on oil spill data provided by IBAMA. In these analyses, the spatiotemporal variability of oil trajectories generated by the models is compared with aerial imagery of the oil slicks.

Goal 4—data management and communication

Three interconnected situation rooms, managed by MB, IBAMA, and INPE, will coordinate emergency responses using near real-time satellite data (developed under Goals 1, 2, and 8) to locate oil spills far from the coast and identify ships [1,21,41]. They will also be able to launch unmanned autonomous vehicles (developed under Goal 7) for on-site monitoring and data collection, ensuring faster and more precise responses.

Additionally, the rooms will have access to high-resolution meteo-oceanographic data (Goals 5 and 6). They could run the oil spill models in real time, providing decision-makers with critical insights to determine the best course of action (Goal 9).

This goal emerges as a direct consequence of the organizational challenges experienced during the 2019 “Mysterious” oil spill when the slow implementation of a centralized coordination system [54,55] significantly delayed response efforts [56,57]. The fragmented flow of information between institutions, the lack of integrated situational awareness [7,8], and the limited capacity for real-time analysis exposed critical vulnerabilities in Brazil’s emergency response infrastructure [58,59]. By establishing interconnected situation rooms operated by key federal agencies, Goal 4 addresses these shortcomings and aims to institutionalize a structured, collaborative, and data-driven response framework. This development reflects a strategic evolution in national preparedness, ensuring that future incidents can be addressed with greater speed, coordination, and scientific support [60].

Goal 5—coupled ocean–atmosphere modeling and forecasting

The main objective of this goal is to develop a new high-resolution nesting system using the Modular Ocean Model (MOM6) and a hydrodynamic coastal model (TELEMAC) [19,20] to deliver hydrodynamic and sea state forecasts at a 1 km resolution across the entire BJW, with a 100 m spatial resolution in the target areas close to the coastline. This integrated modeling approach with current global models operated by INPE will improve the accuracy of predictions related to water movement, surface elevation, and ocean current behavior, which are essential for understanding the trajectories and weathering of oil spills.

The Brazilian Earth System Model (BESM) is an initiative to provide Brazil with modeling capability ranging from extended weather (2–4 weeks) to seasonal under global climate change scenarios. The BESM development was based on Brazilian Global Atmospheric Modeling (BAM) [61,62], the global ocean model from the Geophysical Fluid Dynamics Laboratory (GFDL-MOM) [63,64] and the Biosphere model (SSiB/IBIS) [65,66], and observationally validated against the PIRATA moorings buoys data [67].

In the SisMOM project, the regional high-resolution coupled ocean–atmosphere model over the tropical South Atlantic Ocean has been developed using the ETA atmospheric model [18] coupled to the Modular Ocean Model (MOM6) model via the GFDL’s Flexible Modular System (FMS) [19] and the BESM as global boundary conditions [61].

CPTEC-ETA still remains one of the main regional atmospheric models used by INPE [68]. Although the Weather Research and Forecasting (WRF) model offers greater modularity and a larger user community [69], the ETA model has been extensively calibrated for the South American domain over decades of use at INPE [70]. This long-term adaptation has resulted in a well-established infrastructure for data assimilation, initialization, and output handling that is tightly integrated with INPE’s forecasting systems [71] from the global to local scales [72].

The NOAA/GFDL MOM has been one of the most extensively adopted ocean circulation models for global climate and Earth system modeling over the past five decades. Its sixth version (MOM6), represents a significant advancement in the field, offering a highly flexible and accurate framework for simulating the ocean’s physical state across a broad range of spatial and temporal scales. MOM6 is designed to resolve complex interactions among physical, biogeochemical, and ecosystem processes, seamlessly integrating phenomena from deep ocean basins to coastal regions and estuarine environments [19].

In Brazil, MOM6 has been widely utilized in several scientific and operational applications [73] as the oceanic component of the BESM. Most of them are related to climate change analysis [74], but are also used in high-resolution regional studies focused on the South Atlantic and the Brazilian coast [75].

The TELEMAC system (http://www.opentelemac.org, acessed on 15 June 2025) is an advanced suite of hydrodynamic simulation tools developed by the French National Hydraulics and Environment Laboratory (EDF R&D), widely recognized for its flexibility in modeling shallow water flows, sediment transport, wave propagation, and water quality dynamics in coastal, estuarine, and riverine environments [20]. In Brazil, TELEMAC has been in applied research and engineering projects [76], such as morphodynamic studies of estuaries in the Amazon region, flood risk assessments in coastal cities, and hydrodynamic modeling of port areas along the Brazilian coast [77]. Recent Brazilian applications have integrated TELEMAC modules for simulating oil dispersion, salinity intrusion, and storm surge impacts in estuarine systems [78,79]. When combined with MOM6, this coupling enables detailed, high-resolution simulations of coastal processes, bridging the gap between global-scale oceanographic dynamics and local-scale hydrodynamic responses required for operational systems like SisMOM.

Given its historical integration with CPTEC systems and proven performance across South America, the ETA model emerges as the natural choice for the meteorological component of the SisMOM project. Its operational reliability, consistency in data assimilation pipelines, and optimized performance in simulating mesoscale features make it well-suited for supporting real-time emergency response needs. This compatibility not only guarantees a more accurate representation of wind and atmospheric forcing conditions, but also facilitates seamless interoperability with other components of the SisMOM architecture, such as the oceanographic models and oil spill forecast systems. Therefore, prioritizing the ETA model within SisMOM leverages INPE’s existing expertise and infrastructure while maintaining the robustness required for high-stakes environmental monitoring applications.

Goal 6—provision of supercomputing and storage services to the multi-user system

The aim is to provide supercomputing resources for executing high-resolution forecast models and processing satellite data using artificial intelligence algorithms. Leveraging these computational capabilities, the system will deliver critical information to the interconnected situation rooms, facilitating data-driven decision-making in emergency response scenarios [1].

Goal 7—study the feasibility of developing a rapid launch system for autonomous sensors for the detection and qualification of oil at sea

This goal aims to evaluate the feasibility of developing a rapid-response system using intelligent and autonomous sensors to verify potential oil spills identified by the operational modules of SisMOM.

These sensors can be deployed on Remotely Operated Vehicles (ROV) or Unmanned low-altitude Aircraft Systems (UAS) to cover affected areas in detail, complementing satellite detections. Furthermore, autonomous systems must be equipped with robust processing capabilities to perform real-time analysis and decision-making in the field, especially in situations where communication with ground stations is intermittent.

A critical aspect of this goal involves identifying sensor technologies capable of detecting not only the presence of oil but also key parameters such as thickness, chemical composition, and weathering stage. These characteristics are essential for assessing environmental risk and planning cleanup operations. Developing such smart systems, capable of integrating IR, lasers [80], and hyperspectral data [81], is critical for transforming raw sensor outputs into actionable information that can guide emergency response teams [82].

There is no definitive answer regarding which remote sensing technology should be prioritized for marine oil spill monitoring, as each sensor type presents distinct advantages and limitations [83,84] (

Table 1. Comparison of oil spill monitoring capability from some remote sensing technologies. Adopted from [83,84].

Technologies	Advantages	Limitations
SAR	all-day, all-weather	high false alarm rate unable to identify the type of oil unable to estimate oil film thickness
Multispectral remote sensing	wide space coverage recognizes heavy and light oils estimate the oil film thickness	low spectral resolution vulnerable to sunglint interference
Hyperspectral remote sensing	high spectral resolution recognizes different oil products estimate the oil film thickness	not applicable to large-scale marine oil spill monitoring, vulnerable to sunglint interference
Ultraviolet remote sensing	sensitive to thinner oil film	low spatial resolution, vulnerable to interferences unable to identify non-emulsified oil and oil–water emulsions
Thermal infrared remote sensing	all-day sensitive to thicker oil film not disturbed by the sunglint	unable to identify non-emulsified oil vulnerable to interference fromtargets with similar thermal properties
Laser-induced Fluorescence/ Raman	able to identify different oil products reverse oil film thickness	unable to detect very thick oil film
Laser Acoustic	all-day/night oil slick thickness measurements	bulky and expensive blocked by fog or clouds

).

The Synthetic Aperture Radar (SAR) is the most widely employed microwave remote sensing technique for monitoring oil spills at sea due to the advantages of all-day and all-weather operations (including nocturnal). However, SAR is often associated with a high incidence of false positives, making it challenging to accurately determine the type of spilled oil and estimate the thickness of the surface film. In contrast, optical remote sensing provides richer spectral information compared to SAR and has been extensively investigated for oil spill applications. Laser-induced Fluorescence can provide the best information for oil spill surveillance as it can detect oil on several environmental backgrounds including ice and the shoreline. However, the horizontal laser range is limited, making it ineffective for monitoring large areas.

No single sensor has the capability to provide all the information needed for oil spill surveillance. Moreover, the selection of ROV/UAS-mounted instruments must complement satellite sensor choices. While satellites are capable of scanning vast ocean regions to detect potential spill locations, ROV/UAS are better suited for targeted, high-resolution investigations of suspicious areas identified by satellite data [83,84].

Goal 8—feasibility study and developing the intelligent satellite constellation

The microsatellite constellation will be equipped with sensors for ocean color monitoring, radar imaging, RF signal detection, and surface data capture. The ocean color sensors will provide critical information on changes in water quality, aiding in the detection of potential oil spills [85]. Radar imaging capabilities will enable all-weather, day-and-night surveillance, which is crucial for tracking oil slicks [86], vessel activities, and other anomalies within the BJW [87]. The RF signal detection system will seek unauthorized activities or support search and rescue operations [88].

The use of satellite constellations has become a widely adopted strategy around the world to enhance maritime protection against illegal activities, including oil dumping.

Since 2008, the U.S. NESDIS Satellite Analysis Branch (SAB) has begun developing oil detection capabilities to support operational demands from NOAA’s Emergency Response Division (ERD), through continuous analysis of SAR and multispectral satellite imagery within and near the U.S. Exclusive Economic Zone (EEZ) to detect both accidental and intentional discharges of crude oil [89].

The European Maritime Safety Agency (EMSA) operates the CleanSeaNet and Copernicus Maritime Surveillance services. These programs use Synthetic Aperture Radar (SAR) and high-resolution visible-band imagery to detect and monitor oil slicks on the ocean surface, supporting EU member states in identifying and responding to potential marine pollution incidents [90].

More recently, some commercial satellite companies have started offering similar services to the oil and gas industry, with the same strategy (large constellation of CubeSats) and economic viability [91]. In Brazil, Telespazio and Petrobras signed a contract in 2019 to monitor the oil company’s activities in the Campos basin, using the COSMO-SkyMed satellites constellation. The agreement prevised a multi-year collaboration and aims to detect any oil spills from the exploration and production activities carried out by the Brazilian oil company). However, no public records confirm whether this service is still active or regularly used by Petrobras or federal agencies [92].

The use of satellite-based remote sensors to detect oil slicks is considered an integral component of surveillance strategies because such images cover extensive sea areas and they can provide a comprehensive picture of the overall extent of pollution. However, the main limitation of all satellite data comes from the passing frequency over the same area, which can range from hours to days depending on the orbit (latency time) [93]. This deficiency in the system needs to be corrected by increasing the number of satellites until good coverage in space and time is achieved, especially considering the dimensions of the BJW.

Beyond data latency, the overall timeline encompassing data acquisition, processing, and analysis (lead-in time) often exceeds the latency period itself. This extended delay is largely attributable to the increasing volume of data acquired during each satellite overpass, the frequency of revisit cycles, the growing number of operational satellites, and the computational demands associated with handling such datasets. Consequently, while the configuration of satellite constellations is fundamental for ensuring spatial and temporal coverage, the effectiveness of oil spill monitoring efforts equally depends on the capacity of data processing systems. A well-integrated and responsive processing infrastructure is essential to ensure that critical information is extracted in a timely manner and delivered to decision-makers without operational delays [93].

Goal 9—data provision for IBAMA and MB

IBAMA and MB are the two major institutions responsible for environmental protection and oil spill response in the BJW. SisMOM’s situation rooms will support these institutions with capabilities in data mining, high-resolution hydrodynamics, satellite data, and data and oil modeling.

The integration of these diverse data streams represents a significant advancement over previous response frameworks, particularly in addressing limitations observed during the 2019 oil spill incidents in Brazil. In 2019, response efforts were hampered by delays in data acquisition, a lack of real-time forecasting tools to predict spill trajectories, and the absence of satellite images [3,4,5,6,7,8,54,55,56,57,58,59,60]. SisMOM’s data provision under Goal 9 directly targets these deficiencies by enabling near-real-time access to satellite imagery and hydrodynamic forecasts, which enhance situational awareness and facilitate timely and informed operational decisions.

Looking forward, Goal 9’s framework could develop enhanced operational protocols that leverage these data resources and coordinated multi-agency communication platforms to streamline response efforts. Such actions promise to reduce response times, minimize environmental damage, and improve accountability.

3. Perspectives

Increasing Brazil’s satellite capabilities is central to SisMOM’s strategy for protecting the BJW from oil spills and other environmental threats. The microsatellite constellation instruments will be the most critical component of the entire project, as they will collect the essential data and provide continuous, near real-time coverage of the BJW. A failure to achieve Goal 8 could undermine SisMOM’s overall ability to fulfill its objectives, potentially leading to the collapse of the entire project and rendering the integrated response and monitoring system ineffective [85,86,87,88].

The new constellation will generate terabytes of unlabeled data and daily broadcasts that AI algorithms must process to optimize the information for decision-makers [21].

AI algorithms can significantly enhance sea surveillance by detecting potential anomalies in the trajectory, speed, or position of ships, using AIS data. This capability improves SisMOM’s ability to identify behaviors associated with oil bunkering or dumping, piracy, and other illegal activities [94]. Additionally, Radio Frequency data and images from the new satellite constellation (Goal 8) must be matched with AIS data to detect ships that fail to activate their AIS equipment as required [95].

The development and expansion of a microsatellite or CubeSat constellation is, in principle, a relatively straightforward task, as the necessary methodologies and instrumentation are well-established within the current state of aerospace engineering. However, the effective operation of such a system poses significantly greater challenges, particularly when the number of deployed satellites reaches a high magnitude. An increase in the number of satellites inevitably leads to a substantial escalation in the volume of data generated, which, in turn, demands a proportionally greater investment in data processing infrastructure and computational capabilities.

Moreover, the prevalent use of Synthetic Aperture Radar (SAR) imagery in such constellations introduces additional operational complexities. While SAR is highly effective for all-weather, day-and-night imaging, it is also prone to generating a considerable number of false positives in oil spill detection scenarios. This limitation can result in frequent and unnecessary deployments of Remotely Operated Vehicles (ROVs) unless the satellite data is meticulously analyzed and filtered.

To ensure the efficient functioning of a satellite constellation designed for environmental monitoring—particularly for detecting oil spills—robust data processing frameworks must be implemented. These should include advanced algorithms capable of minimizing false positives and maximizing detection accuracy. Furthermore, the integration of artificial intelligence and machine learning techniques is essential for automating image interpretation, prioritizing alerts, and supporting real-time decision-making. Thus, while the physical deployment of the constellation may be technically manageable, its successful application relies heavily on the sophistication and scalability of its data processing and analysis systems.

Accurate discrimination between true oil slicks and natural look-alike phenomena—such as low-wind zones, biogenic films, or rain cells—presents technical challenges for SisMOM. Convolutional Neural Networks (CNNs) and Density-based Spatial Clustering of Applications with Noise (DBSCAN) could be combined to improve oil slick recognition and reduce false positive events.

CNNs are particularly well-suited for extracting high-dimensional spatial features from remote sensing imagery. When applied to SAR data, CNN architectures can autonomously learn to identify textural and structural characteristics associated with oil slicks, thereby reducing the reliance on hand-crafted features or heuristic rule sets. Advanced variants such as U-Net and Fully Convolutional Networks (FCNs) have been adapted to the semantic segmentation of SAR images [95,96], achieving pixel-level classification with high spatial fidelity. However, DBSCAN offers a non-parametric, density-based clustering approach that does not rely on labeled data. It is particularly effective in segmenting spatially contiguous dark formations in SAR imagery, which are potential indicators of oil slicks [97,98].

Both CNNs and DBSCAN constitute valuable methodologies for the analysis of SAR imagery in the context of oil spill detection. While CNNs excel in feature abstraction and high-precision classification through supervised learning, DBSCAN offers a computationally efficient, unsupervised alternative for spatial clustering. Their integration within a unified detection pipeline presents a promising avenue for advancing operational oil spill monitoring, especially under the constraints of limited training data, variable environmental conditions, and the need for real-time decision support.

The 2019 “Mysterious” oil spill revealed critical deficiencies in the government’s coordinated response to oil spill scenarios, highlighting the urgent need for improvement across various sectors [8,54]. In response, the three situation rooms (Goal 4) were created to incorporate the valuable lessons learned from this incident. These rooms aim to compile all necessary information in real time, enabling integrated and cooperative decision-making among all involved parties. This approach is essential, as key decisions must be made within hours to a few days (<2 days) following an event.

SisMOM’s oil spill modeling (Goal 3) has already been extensively discussed in Zacharias et al. [53], indicating the perspective of developing a high-performance, custom-built code designed specifically for the internal systems of the situation rooms (Goal 4).

4. Conclusions

SisMOM is designed to integrate various technologies, including satellite data, AI algorithms, autonomous sensors, and high-resolution modeling, to detect and respond to oil spills and maritime threats. However, the success of SisMOM heavily relies on the microsatellite constellation to provide continuous, near real-time coverage of BJW. These satellites are critical to provide surveillance and high-quality input data for the entire system.

Integration of satellite data with AI algorithms enhances SisMOM’s ability to detect anomalies such as illegal dumping or oil bunkering. AI systems are essential to process vast amounts of satellite, RF, and AIS data to identify ships engaged in illegal activities, supporting more effective enforcement.

An operational high-resolution coupled ocean–atmosphere modeling system has been a long-standing goal of the Brazilian scientific community, and it is essential for accurately forecasting oil spill trajectories and the areas affected.

The real-time integration of the three situation rooms enables immediate, coordinated responses from IBAMA, the Brazilian Navy, and INPE, allowing decision-makers to quickly assess data and deploy resources to mitigate oil spill impacts. However, overlapping capacities raise concerns about conflicting or redundant actions, as all parties will access the same data and modeling tools, potentially leading to differing interpretations and responses during emergencies. To prevent inefficient resource allocation and delays, it is essential to clearly define protocols, coordination, and decision-making hierarchies.

Additionally, the integration of these autonomous systems with the broader SisMOM platform presents both technical opportunities and challenges. The real-time coordination between satellite-based alerts (Goals 1 and 2), ocean forecast models (Goal 5), and sensor deployment requires robust data communication protocols and intelligent tasking algorithms. If successful, this system could significantly reduce response times and improve the precision of oil spill confirmation in remote or difficult-to-access maritime areas.

Ultimately, this goal reflects an emerging trend in environmental monitoring that combines remote sensing, robotics, and artificial intelligence to deliver faster, more reliable, and cost-effective responses to marine pollution events.

SisMOM is expected to face significant scientific and technological challenges in achieving its objectives, particularly due to the complexity of integrating heterogeneous data sources and coordinating operational workflows. Therefore, it is not expected to be operational for at least two years. However, just the process’ initiation already represents a shift in the mindset of how Brazil approaches environmental risk governance in its maritime jurisdiction. By fostering cooperation among scientific institutions, federal agencies, and local authorities, SisMOM tries to promote a unified, science-based response to marine pollution threats and contributes to long-term strategic planning through the identification of vulnerable areas, optimization of response logistics, and development of regulatory frameworks for environmental protection.