Impact of Industrialization on the Evolution of Suspended Particulate Matter from MODIS Data (2002–2022): Case Study of Açu Port, Brazil

Highlights

What are the main findings?

• A 20-year MODIS-Aqua analysis (2002–2022) revealed a significant increase (up to 60%) in SPM concentrations near the Açu Port Industrial Complex.
• The developed OWT-based retrieval approach combining optimized bio-optical algorithms provided robust SPM estimates validated against in situ and satellite data.

What are the implications of the main findings?

• Industrial activities substantially contribute to sediment enrichment in coastal waters, while natural hydrodynamics regulate dispersion and transport.
• The study demonstrates the capability of long-term satellite data and adaptive algorithms to detect anthropogenic signals in sediment dynamics.

Abstract

The present study evaluates the influence of industrialization on suspended particulate matter (SPM) dynamics along the northern coast of Rio de Janeiro, focusing specifically on the Açu Port Industrial Complex (APIC). A 20-year MODIS-Aqua (1 km) dataset (2002–2022) was processed using the OC-SMART atmospheric correction. For SPM estimation, a retrieval approach for coastal turbid waters that integrates two optimized bio-optical algorithms based on Optical Water Types (OWTs) was developed. The validity of this approach was substantiated through the utilization of the GLORIA in situ dataset and satellite matchups, which demonstrated its robust performance across a range of turbidity conditions. Its main innovation lies in the OWT-based fusion of two optimized SPM models, enabling robust retrievals across diverse coastal optical conditions. Statistical analyses based on Census X11 decomposition and the Seasonal Mann–Kendall test revealed strong spatial and temporal variability, with SPM concentrations increasing by up to 60% near the APIC during the study period, coinciding with dredging, port expansion, and sediment disposal. These findings indicate a pronounced anthropogenic signal, while spatial and temporal correlation analyses demonstrated that sediment dispersion is consistently directed northward, primarily controlled by currents and wind forcing. The results indicate that industrial activities augment the supply of sediments, while natural hydrodynamic processes govern their dispersion and transport, emphasizing the impact of human pressures and physical drivers on coastal sediments.

1. Introduction

Coastal zones are characterized by a diverse array of systems, including estuaries, bays, lagoons, and a variety of geomorphological features [1,2,3]. The biogeochemical dynamics of these systems are significantly influenced by natural processes and anthropogenic activities. On the natural side, land–sea interactions, sediment resuspension and deposition, as well as transport processes driven by waves, tides, winds, and currents, play a central role in regulating the biogeochemical dynamics of coastal waters [4,5,6,7]. Concurrently, human-driven alterations in land use and cover, including urban expansion, agricultural practices, and dam construction, have the potential to generate substantial pressures, thereby modifying sediment fluxes and nutrient dynamics [8,9,10]. Suspended particulate matter (SPM) has been identified as a pivotal indicator of these processes, given its direct impact on water quality, turbidity, light penetration, and the transport of nutrients and pollutants [11,12,13]. It is therefore imperative to understand SPM dynamics to improve pollution control, preserve marine ecosystems, and support integrated coastal management [14,15].

In Brazil, coastal ecosystems are particularly vulnerable to both anthropogenic and natural pressures. The accelerated pace of environmental changes can be attributed to a number of factors, including industrialization, deforestation, intensive agriculture, and large-scale infrastructure projects such as dams have accelerated environmental changes [16]. Among the most impactful activities are industrial and port developments, whose construction and operation often involve dredging, land reclamation, and altered sediment fluxes [17,18,19].

An interesting example is the Açu Port, located in São João da Barra, Rio de Janeiro, which is the largest private port industrial complex in Latin America [20,21,22]. A substantial body of research has examined its socioeconomic and environmental impacts, including heavy metal contamination and effects on local communities [20,21,23,24,25]. However, the influence of climate change on sediment transport has received scant attention from the scholarly community. A notable exception is the work of Scheel et al. [26], who assessed the impacts of climate change on port development. Similarly, Lima et al. [27] employed integrated hydro-sedimentological modeling to assess port resilience. To date, no study has provided a long-term evaluation of SPM dynamics in the adjacent coastal zone using ocean color remote sensing observations.

Due to the scarcity of long-term and spatially extensive in situ data in Brazilian marine environments, satellite remote sensing of ocean color radiometry (OCR) has become an indispensable tool. However, the elevated optical complexity of coastal waters demands the implementation of robust algorithms capable of retrieving SPM across a range of water clarity levels, from highly turbid to clear [28,29]. In this context, Optical Water Type (OWT) classification provides an effective approach, allowing the use of water-type-specific algorithms and thus improving retrieval accuracy under variable conditions [30,31,32].

This study utilizes data from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor onboard the Aqua satellite, with a spatial resolution of 1 km, from 2002 to 2022, to investigate the dynamics of SPM along the northern coast of Rio de Janeiro, with a particular focus on the influence of the Açu port. MODIS provides the longest continuous time series of ocean color available. While alternative datasets such as GlobColour (4 km) and the Ocean Color Climate Change Initiative (OC-CCI, 1 km) offer global coverage, the coarser resolution of GlobColour and the reduced performance of both datasets in turbid environments make them less suitable for this region [33]. Furthermore, the absence of near-infrared (NIR) bands in these sensors is a critical limitation, as NIR bands are indispensable for the retrieval of SPM and chlorophyll-a (Chl-a) in optically complex waters.

The main objective of this study is to depict the recent evolution of the SPM dynamics in the coastal areas adjacent to the Açu Port and assess the relative impact of anthropogenic activities and climate on the regional SPM spatio-temporal patterns from MODIS 1 km 2002–2022 data. To do so, an SPM retrieval algorithm based on OWTs was developed. This algorithm was adapted for global coastal waters by integrating existing algorithms according to their performance in specific water types. This algorithm was subsequently implemented on a comprehensive dataset spanning two decades, meticulously processed using the OC-SMART (Ocean Color—Simultaneous Marine and Aerosol Retrieval Tool) atmospheric correction method [34]. The resulting SPM time series were then analyzed using the seasonal Mann–Kendall test and Census X11 decomposition method [35,36] to detect monotonic trends over 2002–2022 and to characterize temporal variability at multiple scales (seasonal, interannual). Finally, the SPM spatiotemporal patterns were interpreted considering the main natural and anthropogenic drivers of the study area.

2. Materials and Methods

2.1. Study Area

The Açu Port Industrial Complex (APIC) is located in the northern region of Rio de Janeiro State, southeastern Brazil (21.8528°S, 41.0206°W), within the municipality of São João da Barra [37], occupying approximately 40% of its territory (Figure 1). The APIC is currently the largest private deep-water industrial port complex in Latin America [20,21]. It covers an area of 130 km² and includes 11 terminals distributed between offshore and onshore zones. Offshore terminals have been found to have a depth range of 20 to 25 m, which is sufficient to accommodate large cargo vessels [22]. The APIC construction commenced in late 2006 and early 2007, with official operations initiated in 2014 [25,38]. The complex is situated in proximity to the Paraíba do Sul River delta, which is the second-largest deltaic system in Brazil [21], and in close proximity to the Itabapoana River (see Figure 1). The construction of dams and changes in land use have led to a decline in the discharge and plume size of the Paraiba do Sul River. This has reduced its influence on adjacent coastal areas near the river mouths [39,40].

Figure 1. Location of the study area. (a) Location of northern coast of Rio de Janeiro state, highlighted in red, (b) Regional map showing bathymetry and major rivers, with a red box indicating the position of Açu Port, (c) Satellite view of Açu Port, the specific area analyzed in this study.

The hydrodynamic conditions in the region are mainly influenced by the southwestward-flowing Brazil Current, which originates from the bifurcation of the South Equatorial Current [16,41]. Coastal winds are predominantly northeasterly, though southern, southeastern, and southwestern winds are also common during austral autumn and winter [42,43,44]. The region’s climate is classified as humid tropical, with an average annual temperature of approximately 22 °C. However, the region experiences a marked seasonal alternance in its rainfall regime, with wet summers from December to March and drier winters from June to August [43].

2.2. Ocean Color Satellite Data

2.2.1. Atmospheric Correction

To monitor water quality along the coastal waters of the northern coast of Rio de Janeiro state from 2002 to 2022, MODIS-Aqua Level 1 (L1) images at 1 km spatial resolution were downloaded from the NASA Ocean Biology Processing Group (OBPG) at the Goddard Space Flight Center (https://oceandata.sci.gsfc.nasa.gov/directdataaccess/Level-1A/Aqua-MODIS/ (accessed on 1 December 2022)). This sensor provides the most extensive continuous time series from a single platform, making it particularly valuable for long-term coastal monitoring. The daily L1B data set was processed using the OC-SMART atmospheric correction method [34]. OC-SMART is an advanced platform based on multilayer neural networks (MLNNs) and extensive radiative transfer simulations based on the AccuRT coupled atmosphere–ocean model to assess the remote sensing reflectance ($R_{rs}$(λ)) of the water surface [34]. The atmospheric correction model was selected based on its demonstrated efficacy in retrieving $R_{rs}$(λ) across a broad spectrum of water types. This selection was made following a recent validation study that revealed the model’s superior performance in comparison to alternative methods, including Polymer, ACOLITE, and C2RCC [45,46]. Additionally, it provides superior spatial coverage, especially in highly turbid waters, due to its advanced cloud screening capabilities. Furthermore, this model demonstrates a diminished sensitivity to contamination from weak to moderate sun glint and cloud edges [34]. OC-SMART has also been validated for coastal applications and MODIS-Aqua data by Tran et al. [47], as well as by a recent IOCCG (2025) evaluation group that assessed atmospheric correction schemes for MODIS using global in situ and synthetic data sets, confirming the method’s good overall performance across contrasted environments [48].

2.2.2. Bio-Optical Model for SPM Estimation

In Situ Data

A bio-optical algorithm for SPM concentration was developed and validated using the Global Reflectance community dataset for Imaging and optical sensing of Aquatic environments (GLORIA) [49]. In practice, this global in situ dataset encompasses 7572 quality-controlled hyperspectral ${ R}_{rs}$(λ) spectra with 1 nm resolution over the 350–900 nm range collected in both inland and marine waters. Each $R_{rs}$(λ) spectrum is paired with at least one co-located water quality parameter, including Chl-a, SPM, absorption by colored dissolved organic matter (aCDOM), or Secchi depth. SPM measurements encompass a total of 4149 samples. Subsequent to the implementation of quality control measures and the exclusion of data that had been flagged, a total of 489 points were retained, with each point corresponding to $R_{rs}$(λ) and SPM measurements. These samples encompassed a range of SPM concentrations, from 0.21 to 844.67 mg·L⁻¹, with a mean value of 25.88 mg·L⁻¹ (Figure 2a).

Figure 2. Distribution and classification of GLORIA in situ dataset. (a) Frequency distribution histograms of in situ SPM validation data, (b) Frequency distribution histograms of in situ SPM matchup data.

A matchup data set was constructed by extracting the corresponding MODIS OC-SMART $R_{rs}$(λ) data (3 × 3 pixel window) from the GLORIA SPM stations, which were collected between 2002 and 2022. According to Werdell et al. [50], classical quality control was performed by retaining only the matchups data that demonstrated the following criteria was set for a pixel to be considered valid: (1) at least five valid pixels within the nine pixel subset, (2) a spatial homogeneity of the $R_{rs}$(551) value within the subset considering a coefficient of variation (CV = standard deviation divided by the mean, multiplied by 100) lower than 30%, and (3) time difference between in situ and satellite observations lower than three hours. Additionally, to remove inaccurate measurements, OC-SMART L2 quality flags were applied, and only valid pixels were retained (flag 0). Pixels with flags 1, 4, 16, 64, 256, and 1024 were excluded (L1 reflectance unavailable; out-of-range geometry; land; cloud; Rayleigh-corrected reflectance (Lrc) out of scope; and negative Lrc). Following the implementation of the matchup protocol on the GLORIA data set, the MODIS-validation data set comprises 260 data points, with SPM ranging from 0.1 and 74 mg·L⁻¹ and an average value of 6.59 mg·L⁻¹ (see Figure 2b).

Optical Water Types Classification

Optical water types offer a practical approach for classifying aquatic environments according to their optical properties [30,51,52]. It is particularly useful in optically complex coastal waters, contributing to water quality monitoring and improving the performance of ocean color algorithms performance [31,53,54,55,56].

In this study, OWTs were defined using the methodology described by Vantrepotte et al. [30], and applied to the GLORIA $R_{rs}$(λ) dataset. Normalized remote sensing reflectance was computed from multispectral $R_{rs}$(λ) data at six wavelengths in the visible spectrum (412, 443, 488, 531, 547, and 667 nm), corresponding to MODIS bands, to focus on the spectral shape. Normalization was performed by dividing each $R_{rs}$(λ) value by the integral of the reflectance spectrum, as follows:

$$R_{rs-N}= {\displaystyle \frac{R_{rs}\left(\lambda \right)}{{\int }_{{\lambda }_1}^{{\lambda }_2}{R}_{rs}\left(\lambda \right)d\lambda }}$$

(1)

where ${ R}_{rs-N}$ is the normalized remote sensing reflectance.

The normalized spectra were then classified into OWTs (see Figure 3a,b) using unsupervised hierarchical clustering using Ward’s method [57], which is less sensitive to noise and outliers than other methods [30,58].

Figure 3. Optical water type classification. (a) Average normalized reflectance spectra corresponding to the five OWTs defined from the GLORIA dataset (N = 489) according to Tran et al. [32]; (b) Geographical distribution of the in situ SPM data.

In practice, five OWTs were considered, based on a recent classification of a global dataset, including GLORIA data, which was used to develop Chl-a inversion models for Sentinel 2 and 3 [32], as well as MODIS data [47]. This classification resulted in five distinct $R_{rs}$(λ) spectral shapes. OWTs 1, 2, and 3 (N = 185) correspond to clear oligotrophic-to-mesotrophic waters, with a mean SPM concentration of 5.25 mg·L⁻¹, a minimum of 0.21 mg·L⁻¹, and a maximum of 34.58 mg·L⁻¹. OWTs 4 and 5 (N = 304) represent turbid to highly turbid waters, with mean, minimum and maximum SPM concentrations of 38.44 mg·L⁻¹, 2.37 mg·L⁻¹, and 844.67 mg·L⁻¹, respectively.

Principle of Algorithms Combination Based on OWTs

Five predefined OWTs were used to classify satellite $R_{rs}$(λ) spectra. This process involves associating each satellite $R_{rs}$(λ) spectrum with the identified OWTs from the in situ dataset. This association was performed using a detection technique based on the Mahalanobis distance [59], which has previously been shown to be effective for classifying ocean color data [60,61]. Each OWT is characterized by a specific mean vector ($\mu$) and covariance matrix ($\sum$), computed from normalized, log-transformed $R_{rs}$(λ) spectra [28,30,61]. The Mahalanobis distance ($∆$²) quantifies the similarity between a given satellite spectrum x and each OWT $ic$, and is defined as follows:

$${∆}_{ic}^2\left(x\right)={\left(x- {\mu }_{ic}\right)}^T{\sum }_{ic}^{-1}(x- {\mu }_{ic})$$

(2)

where T indicates the transpose of the matrix.

Using these distances, the class membership probability of spectrum $x$ for OWT $ic$ was computed as [28]:

$$P_{ic}\left(x\right)= 1-P({∆}_{ic}^2\left(x\right),n)$$

(3)

where $P$ is the cumulative χ² distribution function with $n$ degrees of freedom (corresponding to the number of spectral bands). In practice, $P$ is defined as follows:

$$\left({∆}_{ic}^2\left(x\right),n\right)= {\displaystyle \frac{\gamma \left(\frac{n}{2}, \frac{{∆}_{ic}^2\left(x\right)}{\Gamma \left(\frac{n}{2}\right)}\right)}{\Gamma \left(\frac{n}{2}\right)}}$$

(4)

where $\Gamma$ is the Gamma function and $\gamma$ is the lower incomplete Gamma function [62].

The probabilities were then normalized to obtain membership values ($P$*), ensuring that the total sum across all OWTs equaled 1 [28]:

$$P_{ic}^{*}= {\displaystyle \frac{P_{ic}}{{\sum }_{ic=1}^{Nc}{P}_{ic}}}$$

(5)

These OWT membership values were used to combine several bio-optical algorithms to estimate biogeochemical products using a probability-weighted approach [30,61]. When two algorithms are combined to estimate SPM, the final blended product is derived using the following weighted average calculation:

$$SPM= {\displaystyle \frac{W_1{SPM}_1+W_2{SPM}_2}{W_1+W_2}}$$

(6)

where $W_1$ and $W_2$ correspond to the probability that each pixel belongs to the OWTs for which models ${SPM}_1$ and ${SPM}_2$ should be applied, respectively. Depending on the local water optical properties, this approach enables a smooth and adaptative transition between algorithms, ensuring optimal retrieval performance across the full range of water types present in the study area [32].

2.3. Ancillary Data

In order to support the study of SPM dynamics in the coastal area of the APIC, as well as the observed changes in water quality between 2002 and 2022, several ancillary datasets have been considered, including land use and land cover, river discharge, precipitation, winds, currents, and waves to depict the evolution of marine and land conditions over the MODIS time period.

2.3.1. Land Use and Land Cover

Information on land use and cover was gathered from MapBiomas Project Collection 9.0 (https://brasil.mapbiomas.org/en/ (accessed on 1 January 2025)), which provides annual classifications of land cover and use across Brazil from 1985 to the present day. The area of interest corresponded to the land occupied by the Açu Port infrastructure. The annual land use and cover data, expressed in hectares and categorized into classes such as forest, farming, and urban areas (industrialization), were linearly interpolated to produce a monthly time series compatible with the temporal resolution of the MODIS datasets used in this study.

2.3.2. River Discharge and Precipitation

Monthly river discharge data (in m³·s⁻¹) for the Paraíba do Sul River were downloaded from the HIDROWEB platform, which is maintained by the Brazilian National Water Agency (Agência Nacional de Águas—ANA) and provides consistent hydrological time series for major river systems in Brazil (https://www.snirh.gov.br/hidroweb/mapa (accessed on 1 January 2025), station name, 21.6453°S, 41.7522°W). Although this station is situated inland, it was the only one available that provided continuous data covering the complete study period (2002–2022).

Monthly precipitation estimates were downloaded from NASA’s Integrated Multi-satellitE Retrievals for GPM (IMERG) product (https://gpm.nasa.gov/data/directory (accessed on 1 January 2025)), which provides global precipitation data with a spatial resolution of 0.1° and a temporal frequency of 30 min, available from 2000 to the present. Precipitation data for the entire Brazilian territory were extracted, and a regional average was computed over the study area using a 5 × 5 pixel window centered on the red box shown in Figure 1. This process resulted in a monthly precipitation time series (in mm·month⁻¹), which was used to evaluate the influence of rainfall on hydrological variability within the area of interest.

2.3.3. Currents and Winds

We used data from the Global Total Surface Currents product provided by the Copernicus GlobCurrent service (https://data.marine.copernicus.eu/product/MULTIOBS_GLO_PHY_MYNRT_015_003/ (accessed on 1 January 2025)), to analyze surface circulation in the study area. This dataset combines satellite altimetry and gravimetry (GOCE project), along with in situ observations [63]. It offers consistent estimates of total surface currents, as well as their geostrophic and Ekman components, at the surface and a depth of 15 m. The product has a spatial resolution of 1/4° (~25 km at the equator).

Wind data were obtained from the Cross-Calibrated Multi-Platform (CCMP) version 3.0 product [64], available from Remote Sensing Systems (https://www.remss.com/measurements/ccmp/ (accessed on 1 January 2025)). The CCMP v3.0 dataset provides high-resolution, gap-free, global wind vector fields at a spatial resolution of 1/4° based on satellite microwave observations blended with numerical weather prediction model outputs.

2.3.4. Significant Wave Height and Wave Period

Significant wave height (SWH) and mean wave period (MWP) were obtained from the ERA5 monthly reanalysis [65], produced by the European Centre for Medium-Range Weather Forecasts (ECMWF). The dataset has a spatial resolution of 0.25° × 0.25°, covers the global ocean from 1979 to the present, and is available through the Copernicus Climate Data Store (https://cds.climate.copernicus.eu (accessed on 1 September 2025)). Wave power $P$ (W·m⁻¹) was estimated using the finite-depth formulation, expressed as:

$$P= {\displaystyle \frac{1}{8}}{\rho gH}_s^2{C}_g$$

(7)

where $C_g$ is the group velocity derived from the linear dispersion relation, $H_s$ is the significant wave height (m), $\rho$ the water density (1025 kg·m⁻³), and $g$ the gravitational acceleration (9.81 m·s⁻²). This parameter combines wave height and period into a single measure of wave energy, providing a physically meaningful proxy for wave-driven processes in coastal environments [66,67].

2.4. Statistical Analysis

2.4.1. Statistical Indicators

The performance of the SPM bio-optical algorithms was evaluated using the following statistical metrics: Root Mean Square Deviation (RMSD), Median Absolute Percentage Difference (MAPD Median Ratio), slope, Coefficient of Determination (R²), and Median Ratio (MR). These metrics were calculated by comparing model-derived and in situ concentrations of SPM.

$$RMSD=\sqrt{\sum _{i=1}^N{\displaystyle \frac{{\left[{log}_{10}\left(X_i^{mod}\right)-{log}_{10}\left(X_i^{obs}\right)\right]}^2}{N}}}$$

(8)

$$MAPD=median\left[{\displaystyle \frac{\left|{log}_{10}\left(X_i^{mod}\right)-{log}_{10}\left(X_i^{obs}\right)\right|}{{log}_{10}\left(X_i^{obs}\right)}}\right]\times 100$$

(9)

$$MR=median\left({\displaystyle \frac{X_i^{mod}}{X_i^{obs}}}\right)$$

(10)

where $X_i^{mod}$ and ${ X}_i^{mod}$ are the SPM in situ and model-derived concentrations, respectively.

2.4.2. Time Series Decomposition Method

Monthly SPM composites were generated using daily MODIS OC-SMART data from July 2002 to December 2022. To investigate the temporal variability and long-term changes in water quality in the study area, the resulting monthly time series and the monthly data on wind, currents, river discharge, and precipitation were analyzed using the Census X11 decomposition method and the seasonal Mann–Kendall trend test.

The Census X11 method [35] is widely used in ocean color studies [33,36,68,69] and was employed here to decompose each variable’s time series $X\left(t\right)$ into three additive components: a seasonal signal $S\left(t\right)$, an interannual component $T\left(t\right)$, and an irregular component $I\left(t\right)$, such that:

$$X\left(t\right)=S\left(t\right)+T\left(t\right)+I\left(t\right)$$

(11)

This decomposition was performed on a pixel-by-pixel basis across the study area. This method employs an iterative application of bandpass filters and explicitly allows for interannual variability in the seasonal cycle. This makes it particularly well-suited to describing non-stationary seasonal dynamics in coastal waters [36]. The trend-cycle component from the X11 decomposition has been shown to effectively capture non-linear trajectories in ocean color records [36], providing valuable insights into the multi-annual evolution of water quality indicators in response to environmental and anthropogenic pressures.

Interannual variability in SPM was assessed using normalized fields, obtained by dividing each SPM map by its respective long-term average. This normalization reduces the effect of differences in absolute concentrations between pixels and emphasizes relative temporal fluctuations. This allows interannual variability to be detected independently of spatial concentration gradients.

To complement the decomposition analysis, the non-parametric seasonal Mann–Kendall test [70] was applied to each time series to detect monotonic trends, with statistical significance evaluated at the probability p < 0.05 level. The magnitude of the detected trends was estimated using the seasonal Sen’s slope estimator [71] and expressed as a rate of change (RC) in percent per year.

3. Results

3.1. A Combined Bio-Optical Model for SPM Estimation Based on OWTs

A combined model was developed to estimate SPM concentrations by blending two existing bio-optical algorithms, based on the 5 OWTs classification performed on the in situ data set. This followed the method described in Section 2.2.2. This approach was adopted to account for the optical complexity of coastal environments, where a wide range of water types is typically encountered [28]. The Supplementary Materials (Figure S1) illustrates this by showing the spatial distribution of the most frequently occurring OWT class at each pixel over the study area, highlighting that the five classes adequately capture the optical diversity of the region. The two SPM inversion models selected for blending were chosen based on a comparative analysis of twelve SPM algorithms applied to optically diverse coastal waters [72]. This study, which focused on the application of the GLORIA in situ dataset to the Sentinel-3 Ocean and Land Colour Instrument (OLCI), revealed that the formulations proposed by Novoa et al. [73] and Jiang et al. [74] demonstrated the best overall performance. However, these models rely on spectral bands that are not available in MODIS data, specifically, the 620 nm band used by Jiang et al. [74], and the 856 nm band used by Novoa et al. [73]. To address this limitation, alternative algorithms with reliable performance using the available MODIS bands were selected. Subirade et al. [72] reported that the Siswanto et al. [75] algorithm performs best in clear waters, particularly for SPM concentrations between 1 and 10 mg·L⁻¹. Additionally, Han et al. [76] model, in both their visible-only and NIR-inclusive versions, demonstrated consistent performance across a range of conditions and were thus suitable for use with MODIS OC-SMART bands.

These two algorithms were therefore tuned and adapted to the MODIS OC-SMART spectral bands and to the coastal OWTs. The coefficients for the Siswanto et al. [75] algorithm were recalibrated using in situ data corresponding to OWT classes 1 to 3, representing oligotrophic to mesotrophic waters. For OWT classes 4 and 5, which correspond to turbid waters, we applied the NIR-based approach from Han et al. [76]. This approach performs well in high-SPM conditions due to the strong relationship between NIR reflectance and suspended particle backscattering [77,78].

The adapted version of the Siswanto et al. [75] algorithm (R² = 0.66, N =185) for MODIS OC-SMART (${SPM}_S$) is expressed as follows:

$${\mathit{log}}_{10}{SPM}_S=0.7752+15.994\left(R_{rs}\left(547\right)+R_{rs}\left(667\right)\right)-0.7103\left({\displaystyle \frac{R_{rs}\left(488\right)}{R_{rs }\left(547\right)}}\right)$$

(12)

The NIR-based model from Han et al. [76] has been adapted for use with MODIS OC-SMART (${SPM}_{NIR }$) and uses water-leaving reflectance at 748 nm (R² = 0.79, N = 304). It is given by the following formula:

$${SPM}_{NIR }= {\displaystyle \frac{2181.029\times R_w\left(748\right)}{1- \frac{R_w\left(748\right)}{0.3852}}}$$

(13)

where $R_w\left(748\right)= {\pi R}_{rs}\left(748\right)$.

The final $SPM$ estimate is obtained by combining the two models using the OWTs probabilities as described in Section 2.2.2:

$$SPM= {\displaystyle \frac{W_1{SPM}_S+W_2{SPM}_{NIR}}{W_1+W_2}}$$

(14)

where $W_1=p_{{OWT}_1}+p_{{OWT}_2}+p_{{OWT}_3}$, and $W_2=p_{{OWT}_4}+p_{{OWT}_5}$, each belonging to probabilities of ${OWT}_1$ to ${OWT}_5$ which are computed from Equation (5). The combined model was validated using both in situ and satellite-derived $R_{rs}$(λ) data.

Figure 4 shows the results of validating the combined model [75,76] using in situ (Figure 4a) and OC-SMART satellite-derived reflectance data (Figure 4b). The model shows a strong correlation between the estimated and measured SPM values in the in situ dataset (R² = 0.80, slope = 0.94), with relatively low error metrics (RMSD = 0.28, MAPD = 18.91%). These results confirm the robustness of the combined approach across a wide range of optical water types, particularly in turbid conditions (OWT 4 and 5).

Figure 4. Validation of the new inversion algorithm for estimating SPM from $R_{rs}$ (λ): Siswanto et al. [75] and Han et al. [76] combination based on optical water types (OWT 1–3 and 4–5, respectively). (a) in situ validation, (b) satellite matchup validation (MODIS—OC-SMART).

Satellite-based validation using OC-SMART shows slightly lower performance (R² = 0.63, slope = 0.73), with greater dispersion (MAPD = 50.43%). This degradation is expected due to the additional uncertainties associated with atmospheric correction and satellite-derived reflectance. Nevertheless, the model still provides a reasonable estimate of SPM concentrations in coastal waters, demonstrating its potential for large-scale monitoring.

This new combined model was then applied to 20 years of MODIS OC-SMART data (2002–2022) in order to estimate the dynamics of SPM in Brazilian coastal waters over the last two decades.

3.2. Spatio-Temporal Dynamics of SPM

3.2.1. SPM Seasonal Variation

The seasonal distribution of SPM in the study area is illustrated by monthly average maps derived from MODIS-Aqua data processed with OC-SMART (1 km resolution, 2002–2022). Figure 5 shows the average SPM concentrations in the austral winter (e.g., July) and summer (e.g., December) months. Overall, SPM concentrations are higher during the austral winter, with elevated levels extending farther offshore than in summer. Near the coast, higher concentrations are likely due to enhanced resuspension processes. In both seasons, the highest concentrations are consistently observed near the coast, decreasing towards the open sea. Coefficient of variation (CV) maps (Figure 5c,d) show relatively low temporal variability across most of the shelf domain (CV < 40%), which further indicates persistent seasonal patterns over the 20-year period.

Figure 5. (a,b) Satellite-derived monthly mean surface suspended particulate matter (SPM) concentrations (mg·L⁻¹) from MODIS OC-SMART (2002–2022) illustrating SPM distribution during (a) austral summer (e.g., December) and (b) austral winter (e.g., July). (c,d) Coefficient of variation (%) of SPM concentrations during (c) austral summer (December) and (d) austral winter (July). White arrows represent current circulation during (a) austral summer (December 2019) and (b) austral winter (July 2019).

The discharge of the Paraíba do Sul River shows no significant correlation with SPM dynamics within the APIC at Point A (Figure 5), as assessed using Spearman’s correlation (r), (r = −0.2, p < 0.05, N = 246). Moderately elevated concentrations are observed near the river mouth (≥10 mg·L⁻¹; Figure 5a,b), likely due to localized inputs. However, the overall contribution is minimal, as the river’s flow is heavily regulated to supply urban areas with water, generate hydroelectric power and support agriculture [44,79,80,81]. Consequently, sediment transport from land to sea via the river is substantially reduced.

The X11 seasonal variation in SPM (Figure 6) appears to be positively correlated with the X11 seasonal variation in the meridional (V) component of surface currents and wind within the APIC. Higher SPM concentrations coincide with positive values of the V component, indicating enhanced northward transport during periods of elevated SPM.

Figure 6. (a) Time series of Seasonal variation (X11 outputs) of SPM (black), meridional current (blue), (b) Time series of Seasonal variation (X11 outputs) of SPM and meridional wind (green) extracted at Point A, located offshore (20 m depth) the Açu Port (Figure 5a).

However, the correlation between SPM seasonality and seasonal variation in the meridional wind component (r = 0.43, p < 0.001, N = 246) was weaker than with the meridional component of surface currents (r = 0.65, p < 0.001, N = 246), highlighting the dominant role of current-driven transport relative to direct wind forcing. The seasonal component of wave power also shows a strong positive correlation with the seasonal variation in SPM (r = 0.74, p < 0.001, N = 246; Figure S3), indicating that wave-induced resuspension contributes substantially to the observed seasonal signal.

In contrast, the zonal (U) component exerts little influence on SPM variability. There was no correlation found between SPM seasonality and the zonal current component (r = −0.1, p > 0.1, N = 246), which indicates that east–west advection does not contribute to SPM transport in the APIC. Similarly, the zonal wind component showed only a moderate correlation with SPM (r = 0.37, p < 0.05, N = 246), Correlation maps of seasonal SPM with the seasonal components of currents and winds (see Supplementary Materials, Figure S4) confirm these patterns spatially. The maps highlight a robust and coherent positive correlation between SPM and the meridional current across the APIC, supporting the dominance of northward transport processes.

In addition, the seasonal component of precipitation, extracted using the X11 method exhibits a clear seasonal pattern and shows a negative correlation with the seasonal component of SPM at point A, which is located next to the APIC (r = –0.64, p < 0.05, N = 246) (see Supplementary Materials, Figure S5). This inverse relationship suggests a dilution effect associated with increased rainfall.

3.2.2. SPM Interannual Variability

Figure 7 shows the annual average of monotonic trends in SPM observed from MODIS data gathered across the study area between 2002 and 2022.

Figure 7. Temporal Rate of Change (RC) in SPM between 2002–2022 within the study area, showing only statistically significant changes (p < 0.05).

Figure 7 displays the significant monotonic trends (p < 0.05) observed from the analysis of 20 years of monthly SPM data from MODIS (in $\%·{\mathrm{year}}^{-1}$). The most pronounced and persistent increases were located immediately offshore of the APIC, with hotspots extending along the 20–50 m isobaths to the south and southeast of the port area. In this area, SPM increased by 20–60% between 2002 and 2022. In contrast, the Paraíba do Sul River mouth showed a slight decrease (~10%) in SPM over the 20-year period, with only a few pixels in front of the river exhibiting a restricted but significant downward trend, which is consistent with previous studies [44,82].

Decreasing SPM patterns were observed offshore in oceanic water (bathymetry > 50 m), which is probably related to oceanic processes that are beyond the scope of this study. The significant increase in SPM from 2002 to 2022 observed in front of the Açu Port area (Figure 7) is detailed in Figure 8a. Figure 8b,c show the corresponding time series: Figure 8b represents the raw SPM, while Figure 8c shows the normalized interannual variation derived from the X11 decomposition. Both figures are averaged over the area in Figure 8a where a significant increase was detected (RC > 0, p < 0.05, N = 246).

Figure 8. (a) Map showing the spatial extent of the surface area used for extracting the time series. (b)Time series of the raw SPM (c) and Census X11 interannual term from 2002 to 2022, both (b,c) averaged over the area within the Açu Port where the RC is positive (RC > 0) and statistically significant (p < 0.05, N = 246). The red dot marks the APIC location. The black line represents the normalized average, and the purple shading shows ±1 standard deviation.

To evaluate whether the increasing SPM trend detected near the APIC could be influenced by uncertainties in the satellite retrievals, a Monte Carlo error-propagation analysis was performed (the distribution of simulated trends is shown in Supplementary Materials, Figure S6). Retrieval errors were first quantified using the MAPD value (50.4%) computed from the matchup dataset by comparing in situ and satellite-derived SPM values. Based on this error metric, 1000 perturbed monthly SPM series were generated by adding random noise to the original time series [47,83]. The noise level was defined by a normal distribution with a mean of 0 and an effective standard deviation (${\sigma }_{eff}$).

$${\sigma }_{eff }=MAPD/\sqrt{N_{eff} }$$

(15)

where $N_{eff}$ is the effective number of independent observations contributing to each monthly value [84]. In this case, $N_{eff}$ was set to 30, corresponding approximately to one independent observation per day. For each simulation, the Seasonal Mann–Kendall test and Sen’s slope were recomputed, yielding 1000 estimates of the long-term RC.

Near the APIC hotspot (Figure 8a), SPM increased by 20–60% between 2002 and 2022, while the spatially averaged RC reached 0.82%·year⁻¹ (~+16% over 20 years). The Monte Carlo ensemble produced a mean RC of 0.80%·year⁻¹, which is nearly identical to the original estimate (bias = −0.02%·year⁻¹), with a 95% confidence interval of 0.67–0.94%·year⁻¹. The perturbed trends showed a mean absolute relative error of 7.2%, and all simulated Sen slopes remained positive. These results demonstrate that realistic retrieval uncertainties cannot reverse the trend and only marginally affect the magnitude of the observed long-term increase in SPM near the APIC.

The spatial correlation map between the interannual variation in SPM across the entire study area and the reference time series extracted from the Açu Port region (where statistically significant increases in SPM were observed) (Figure 9a) reveals a coherent spatial pattern over a larger spatial area. High correlation values (r > 0.8, p < 0.001, N = 246) are concentrated near the APIC area, indicating that SPM dynamics in this zone closely follow those observed at the port. Further north, the correlation gradually decreases but remains positive, suggesting that SPM variability in northern coastal waters may be influenced by similar drivers. This spatial continuity implies a possible northward propagation of SPM-related changes, which may be linked to hydrodynamic processes such as coastal currents or wind-driven transport. This could indicate that the influence of APIC-related activities extends beyond the immediate vicinity of the port. Indeed, trend analyses suggest a slight increase in SPM towards this northern area (Figure 9b), although this is not statistically significant.

Figure 9. (a) Correlation map (p < 0.05, N = 246) between the interannual variation in SPM in the entire study area and the referenced time series of interannual variation extracted from the Açu Port region exhibiting significant SPM increase shown in Figure 8c. (b) RC of SPM including both significant and non-significant variations.

3.3. Evolution of the Environmental Conditions

3.3.1. Land Use and Cover

The land cover and use maps for 2002 and 2022 (Figure 10), derived from the MapBiomas platform, illustrate the profound transformation of the coastal landscape following the construction of the APIC, from late 2006/early 2007 to 2014. Before the development of the port in 2002, the area was characterized by natural features such as beach dunes, herbaceous-shrub vegetation, and forest patches, interspersed with small farming zones. The region exhibited minimal urban occupation, with water bodies limited to natural lagoons and river channels (Figure 10a). By contrast, the 2022 map reveals extensive changes in land cover, driven by the expansion of industrial and urban infrastructure associated with the APIC (Figure 10b). This transformation includes the construction of artificial water bodies, storage areas, access roads, and industrial facilities. Concurrently, natural landscapes such as dunes and herbaceous vegetation have partially given way to industrial land use and urban development.

Figure 10. Land use and land cover maps for the years 2002 and 2022 within the Açu Port (MapBiomas). (a) 2002 land cover map, (b) 2022 land cover map.

3.3.2. Combined Impact of Industrialization and Environmental Forcings on the Regional SPM Dynamics

Analysis of SPM interannual variation in relation to APIC’s industrial development reveals three distinct phases: pre-port, construction and operational (Figure 11).

Figure 11. Impact of industrialization (Açu) on the SPM variation. (a) Correlation map between industrialization and SPM interannual variation between 2002 and 2022. (b) Time series of the interannual variation in SPM (black) and the industrialization growth (red).

A clear connection can be observed between port activity and SPM dynamics (Figure 11b), which can be summarized as follows:

• 2002–2006, Pre-Açu Port Phase: industrial activity in São João da Barra was minimal, with around 300 hectares of industrialized land. SPM levels remained relatively stable and low, reflecting limited anthropogenic disturbance.
• 2006–2014, Construction Phase: significant dredging activities began as port infrastructure took shape. The installation of Terminal 1 (T1) began in 2008 and involved around 20 million m³ of dredged material. This was followed by the construction of a 3 km bridge and a breakwater. In 2011, dredging for Terminal 2 (T2) commenced, totaling ~65 million m³ of sediments [85]. Despite a moderate increase in the industrial area (to ~500 hectares by 2012), SPM levels showed marked increase in interannual variability starting in 2007, likely due to sediment resuspension and coastal disturbance. Elevated SPM levels persisted throughout this phase, peaking between 2008 and 2012.
• 2014–2015, Early Operational Phase: industrialization accelerated, with the industrialized area expanding from ~500 hectares in 2012 to over 900 hectares by 2022, representing nearly 40% of the municipality [37]. SPM levels decreased slightly between 2014 and 2015, possibly due to a reduction in dredging intensity and the stabilization of port infrastructure.
• 2016, Intensification of Operations: SPM concentrations increased again, corresponding to the activation of the T-OIL terminal and the onset of intense traffic from Very Large Crude Carriers (VLCCs) [86]. In the same year, the T1 deepening permitting process introduced the Adaptive Dredging Plan, combining dredging operations with spatial zoning to enhance sediment management. The work was scheduled for 15 months [85].
• 2018, Pipeline Ruptures: according to Technical Report 21 282–301 [87], two pipeline ruptures occurred in March 2018, leading to a temporary suspension of operations. A significant drop in SPM levels was observed that year, likely linked to the rupture events. When operations resumed in December 2018, SPM levels remained below those of 2016 levels but began to rise alongside the increase in industrial activity.
• 2021–2022, Maintenance Dredging and Peak SPM: SPM levels rose again, coinciding with large volumes of sediment deposition totaling 676,000 m³ at T1 (88% fluid mud) and 938,300 m³ at T2 (76% fluid mud). This period also saw the implementation of maintenance dredging and sediment management strategies to mitigate siltation, particularly during periods of heavy rainfall, which were identified as key drivers of erosion and sediment transport based on monitoring data collected between January 2021 and October 2022 [27].

Figure 11a shows the spatial correlation between the interannual variation in SPM and the progression of industrialization in the Açu region. The strongest positive correlations are observed in the waters adjacent to the Açu Port, suggesting that industrial activities have significantly impacted sediment dynamics in the immediate vicinity of the port. Notably, the influence predominantly extends towards the north, as evidenced by a clear northward gradient of high correlation values. This pattern suggests that sediment transport processes favor the northward dispersion of suspended material, amplifying the industrial impact in this direction. This is also indicated by the correlation between the SPM X11 time series in the area (showing a significant increase in front of the Açu Port) and the SPM interannual pattern in the northern part of this coastal domain (Figure 9a), as well as by the presence of a slight increase in SPM over the 2002–2022 time period in the corresponding area (Figure 9b).

To explain the regional interannual dynamics of SPM, we compared the Census X11 interannual components of SPM, wind, currents, and waves. Monotonic trend analysis revealed no significant long-term trends in the wind components or wave power within the study area (p > 0.05), while the currents showed a few patchy and contradictory trends that were not aligned with the significant increase in SPM. This indicates that the observed rise in SPM cannot be explained by persistent changes in natural hydrodynamic regimes.

Correlation analysis of the X11 interannual component of SPM and hydrodynamic drivers reveals a distinct asymmetry in the potential impact of currents, winds, and waves (Figure 12). Currents have little effect; neither the meridional (V) nor the zonal (U) components are significantly correlated with SPM at the interannual scale between 2002 and 2022 near the port (point A, Figure 12a,b), while correlations in the northern area are weak, patchy, and inconsistent across space.

Figure 12. Correlation maps between interannual components of SPM, wind and current. (a) correlation map between the interannual component of SPM and meridional current (V). (b) correlation map between the interannual component of SPM and zonal current (U). (c) correlation map between the interannual component of SPM and meridional wind (V). (d) correlation map between the interannual component of SPM and zonal wind (U). Only significant correlations (p < 0.05, N = 246) are shown.

In contrast, winds exhibit stronger interannual correlations with SPM, particularly in shallow waters (<20 m). The meridional (V) wind component is especially influential (Figure 12c), which is consistent with the recurrent northward dispersion of turbid waters (r > 0.5, p < 0.001, N = 246). This indicates that enhanced SPM concentrations are consistently associated with northward coastal transport at the interannual scale. The zonal (U) wind component (Figure 12d) also plays a role, exhibiting moderate positive correlations offshore of the APIC (r > 0.4, p < 0.001, N = 246).

The interannual component of wave power also shows a moderate but significant correlation with SPM variability (r = 0.34, p < 0.05, N = 246, Figure S7).

The role of the meridional components of winds and current is further illustrated by the time series at Point A (Figure 13). The V-current fluctuates around zero, with weak interannual signals; however, peaks in SPM often coincide with periods of northward flow in the raw signal, indicating their seasonal influence. In contrast, the V-wind shows a much stronger correspondence with SPM variability: northerly winds consistently align with elevated SPM levels, confirming their role in modulating and transporting sediments northward. Occasional lags between wind or current (1 to 2 months) changes and SPM peaks suggest sediment remobilization or delayed transport, as observed in other coastal systems [88,89].

Figure 13. (a) Time series of V-current raw (black), interannual variation in V-current (red), and interannual variation in SPM (blue) at point A parallel to the APIC. (b) Time series of V-wind raw (black), interannual variation in V-wind (red), and interannual variation in SPM (blue) at point A parallel to the APIC (bathy = 20 m).

4. Discussion

4.1. A Combined Bio-Optical Model for SPM Estimation Based on OWTs

A new combined bio-optical model has been developed for the application of MODIS in a case study of the Açu Port area. A key advantage of this method is its OWT-based structure [30,31,32], which has been trained using the GLORIA marine coastal dataset. This allows the model to be applied reliably even in regions lacking in situ data, provided that the optical conditions match those of the training OWTs. The performance of the combined model is consistent with values typically reported for SPM retrievals in coastal waters. In situ validation (Figure 4a) falls within the upper range of what has been achieved in previous algorithm evaluations using multispectral sensors (R² = 0.80; MAPD ≈ 19%). Similar levels of agreement have been reported for SPM retrievals in turbid environments, with R² typically ranging from 0.70 to 0.85 (e.g., [73,74,76]). These results indicate that the OWT-based combination performs reliably within its intended optical domain.

The reduced performance of satellite matchups (Figure 4b; R² = 0.63; MAPD ≈ 50%) is consistent with previous studies, which suggest that atmospheric correction and sensor-specific limitations generally result in weaker correlations than those obtained through direct in situ reflectance. Several studies report R² values between 0.50 and 0.70, as well as relative errors often exceeding 40%, for MODIS and OLCI in coastal and estuarine waters [72,76,78]. Such degradation is typical in highly dynamic regions with strong turbidity gradients and complex particulate assemblages. In this context, the performance achieved here remains within the expected range for regional-scale applications. Overall, the results demonstrate that combining two adapted SPM formulations through OWT probabilities improves the robustness of the algorithm while maintaining accuracy that is comparable to, or better than values reported for similar coastal environments. This approach is therefore suitable for analyzing long-term SPM dynamics from MODIS OC-SMART.

In this study, we used the GLORIA dataset [49] because it provides broad optical coverage representative of coastal and shelf waters, consistent with the conditions observed around the Açu Port region (Figure S1). The recently published data BRAZA [90], covering mostly inland waters, might complement the GLORIA data set and further enable the development of SPM inversion model adapted to applications along the land–sea continuum and will represent the objective of future studies.

4.2. Spatio-Temporal Dynamics of SPM

The spatio-temporal patterns of SPM revealed by MODIS-Aqua and the Census X11 decomposition indicate that seasonal hydrodynamics, interannual variability, and human activities jointly control suspended sediment dynamics in the APIC region. Seasonal analyses show systematically higher SPM concentrations during austral winter, with enhanced offshore extension and a clear northward gradient along the coast, whereas interannual analyses highlight a marked long-term increase in SPM immediately offshore of the Açu Port, particularly after the onset of port construction. Together, these observations point to a system where natural forcing defines the short-term variability while anthropogenic pressures progressively reshape the long-term baseline. These changes occur in the context of profound land-use transformation and rapid industrial development along the northern coast of Rio de Janeiro.

The circulation in the study area is influenced by both mesoscale and coastal processes. At the mesoscale, the region is affected by the southward flow of the Brazil Current (BC), which transports warm surface waters along the southeastern coast of Brazil [41,91]. Seasonal reversals may occur during the austral winter (Figure 5), when coastal currents can flow northward [26]. On the inner shelf, circulation is mainly wind-driven; NE winds promote southeastward flow, while southerly winds associated with cold fronts can reverse this pattern [92]. The alternating wave climate generates longshore sediment transport in opposite directions depending on the season, a process long recognized as key in shaping coastal plains in southeastern Brazil [93,94]. These hydrodynamic regimes provide the physical framework within which SPM is mobilized, redistributed, and exported.

4.2.1. Seasonal Variation in SPM and Environmental Forcings

The moderate correlation between the seasonal variation in SPM and seasonal variation in V wind (Figure 6b) compared to the good correlation of SPM with V current (Figure 6a), suggests that wind forcing primarily acts as a trigger, while the effective seasonal redistribution of suspended material is controlled by the dynamics of surface currents [95,96]. Similar delayed responses have been documented, where wind-induced perturbations require time to evolve into established circulation patterns through mesoscale instabilities [97]. Thus, the weaker wind SPM relationship compared to the stronger current SPM relationship is consistent with a temporal lag between wind variability and SPM seasonality, as illustrated in the climatological averages of wind, current, and SPM shown in the Supplementary Materials (Figure S2). This lagged response underscores the importance of cumulative hydrodynamic adjustments rather than instantaneous forcing.

This northward transport coincides with periods of elevated SPM concentrations within the coastal domain, as illustrated in Figure 5b, and likely contributes to the broader dispersion of suspended material observed primarily northward along the coast during the winter months, possibly due to resuspension processes and riverine inputs. Coastal winds predominantly blow from the northeast but tend to shift towards the south, southeast, and southwest in austral autumn and winter [42,43,44]. This further promotes northward coastal circulation through Ekman transport, influencing nearshore current patterns and sediment dynamics. The seasonal alignment between enhanced meridional circulation and elevated SPM thus suggests a strong coupling between hydrodynamic forcing and sediment redistribution [98].

The good correlation of seasonal SPM with seasonal variation in wave power (Figure S3) highlights the role of surface waves in enhancing sediment resuspension [67,98,99], particularly during periods of increased wave energy in austral winter [100]. In contrast to wind, which primarily acts as a trigger for circulation variability, waves directly mobilize bottom sediments in the coastal and inner-shelf domains. This resuspension elevates nearshore SPM concentrations, which can be subsequently redistributed by northward coastal currents, reinforcing the seasonal pattern of higher wintertime SPM levels. This two-step mechanism wave-driven resuspension followed by current-driven transport is consistent with sediment dynamics observed in other energetic coastal systems [95,98,101].

The limited correlation of seasonal SPM with seasonal variation in zonal wind and currents (Figure S4b–d) suggests that it plays a limited role in modulating SPM dynamics compared to the dominant effect of meridional processes. By contrast, the relationship with the meridional wind is more fragmented and restricted to specific sectors, which is consistent with its weaker and more indirect influence. In contrast, correlations with the zonal current are spatially incoherent and close to zero across most of the region, which reinforces the negligible role of east–west advection. The zonal wind displays slightly more structured patterns than the zonal current, but its influence remains weak compared to that of the meridional processes. Overall, the correlation maps (Figure S4) indicate that SPM seasonality in the APIC is primarily governed by northward transport.

The seasonal variability of precipitation is negatively correlated with SPM seasonality within the APIC (Figure S5). This relationship is likely to be driven by a dilution effect: increased rainfall reduces concentrations of particles in surface waters by increasing the input of fresh water and limiting the resuspension of sediment [102]. Such dilution effects have also been described in coastal systems where enhanced rainfall reduces suspended sediment concentrations by increasing freshwater input and stabilizing the water column [82,88,102].

4.2.2. Interplay Between Port Development and Environmental Forcing in Shaping SPM Interannual Variability

The long-term MODIS analysis, based on the seasonal Mann–Kendall test and Sen’s slope estimator, reveals a marked and spatially coherent increase (20–60%) in SPM in the vicinity of the APIC (Figure 7). The concentration of significant positive trends around the APIC indicates that the increase in SPM is limited to the port-influenced coastal area, rather than affecting the wider region. Such a pattern is typical of areas where maritime infrastructure alters hydrodynamics, enhances resuspension, and increases sediment availability [103,104,105]. In contrast, the lack of comparable positive trends near the Paraíba do Sul River suggests that changes in fluvial supply are not the main driver.

The nonlinear interannual modulation of SPM reveals a clear shift in the sediment regime of the APIC region. SPM concentrations were at their lowest before the onset of port construction in late 2006/early 2007. After this period, a gradual increase is observed, accompanied by pronounced interannual fluctuations (Figure 8c). These variations likely reflect the combined influence of seasonal dynamics and external factors such as water circulation, wind patterns and the intensification of port-related activities and operations.

The spatial correlation analysis (Figure 9a) reinforces this interpretation by revealing that the interannual SPM signal observed near the APIC is not entirely confined to the port area. High correlations extend northward along the coast, suggesting that part of the variability detected at the port propagates beyond its immediate influence zone. This pattern is consistent with northward-moving coastal circulation and wind-driven transport, which can redistribute suspended material along the shoreline. Although the northern trends are weaker and not statistically significant (Figure 9b), the coherent spatial structure indicates that APIC-related changes may affect adjacent coastal waters through alongshore sediment transport.

These results point to a port-driven alteration of coastal sediment dynamics, which is further supported by land-use and land-cover changes (Figure 10). The transition from a predominantly natural system to a highly industrialized landscape [25,27] highlights the magnitude of human-induced disturbance and supports the observed trends in coastal water quality parameters, including the increase in SPM near the port area. The spatial expansion of industrial structures, artificial water bodies, and dredged zones is consistent with rising sediment availability and altered local hydrodynamics [106,107].

Overall, the temporal alignment between industrial growth, major dredging activities, and changes in SPM levels suggests a strong anthropogenic influence (Figure 11b). The progressive industrialization of São João da Barra, particularly the intensification of port operations since 2016 appears to have played a major role in shaping SPM dynamics. These patterns emphasize the need for long-term sediment management and environmental monitoring as industrial activities expand.

In addition, the spatial correlation of SPM interannual variation with the industrialization timeline (Figure 11a) further indicates that the influence of the APIC is not strictly local but tends to extend northward. The northward gradient of high correlation values aligns with the alongshore transport pattern identified earlier (Figure 9), suggesting that port-related disturbances are progressively transferred to adjacent coastal waters through sediment pathways oriented in this direction.

Climate-related impacts do not provide a convincing explanation for the observed SPM increase. Scheel et al., 2023 [26] underscore the importance of incorporating local information in climate change assessments for ports, and their analyses indicate that the Açu Port is expected to remain relatively resilient to sea-level rise and wave condition changes before 2070. No significant sea-level–driven impacts are projected before this timeframe, and wave conditions should remain largely stable. The main projected change concerns wind speed, with potential increases of up to 10% after 2070, which may influence sediment transport in the longer term. However, these projections lie outside the temporal scope of the present analysis.

The long-term increase in SPM offshore of the APIC, including its northward and eastward extension, therefore cannot be attributed to interannual fluctuations in currents (Figure 12a,b). As shown previously, the influence of currents is mainly seasonal, with short-term changes in the meridional component occasionally contributing to episodic northward sediment export. The absence of significant long-term trends in the current components confirms their limited capacity to drive persistent SPM increases.

In contrast, winds emerge as an important driver of sediment redistribution, transporting suspended material both alongshore and offshore [108]. Although wind components do not exhibit significant long-term trends, their interannual fluctuations correlate strongly with SPM variability in shallow waters (Figure 12c,d), especially for the meridional (V) component. This pattern is consistent with the recurrent northward dispersion of suspended sediments observed throughout the study area.

Wave power also contributes to modulating SPM concentrations. While it shows no monotonic long-term trend, its year-to-year fluctuations coincide with SPM peaks (Figure S6), reflecting the capacity of waves to resuspend bottom sediments and increase their availability for advection [98,99]. This mechanism is particularly evident near the port area (Point A, Figure 12), where interannual SPM maxima align with episodes of increased wave energy. However, the weaker correlation between waves and SPM, compared to winds, suggests that waves primarily enhance sediment resuspension rather than drive long-term increases.

Given the absence of significant long-term trends in winds, waves, and currents, the persistent rise in SPM offshore of the APIC cannot be explained by natural hydrodynamic forcing alone. It most likely reflects the intensification of industrial and port activities, which increase sediment availability within the system. In this context, winds modulate the direction and efficiency of sediment transport, whereas waves amplify resuspension processes, together sustaining interannual variability and promoting the northward dispersal of suspended sediments [108,109,110].

Similar patterns have been documented in other major port-affected systems. In the Yangtze River estuary, for example, Fan et al. [111] showed that dredging activities can re-suspend large quantities of bottom material, which can then be transported far beyond the dredging zone and increase SPM concentrations in adjacent regions. The authors also report that shipping activities disturbed surface sediments and modified the spatial distribution of SPM, while large engineering structures altered sediment pathways and produced persistent changes in estuarine sediment regimes. These mechanisms closely mirror the increase and northward redistribution of SPM observed near the Açu Port, indicating that port development in dynamic coastal environments tends to produce long-lasting, spatially coherent modifications to sediment availability and transport.

Beyond these drivers, sea surface height (SSH) has shown an increasing trend in the study area over the past two decades (8 to 9 cm in 20 years, corresponding to approximately 4 mm year⁻¹; Figure S8). This rate is consistent with the global mean sea level rise derived from satellite altimetry, which has accelerated to about 4 mm year⁻¹ in recent decades [112,113]. Along the Brazilian coast, including our study region, sea level rise occurs at variable rates depending on local oceanographic and climatic conditions [114,115,116,117]. Rising SSH may influence shoreline stability and intensify erosion processes, thereby increasing sediment availability for resuspension and transport [118,119,120]. While our dataset does not allow us to fully assess this mechanism, it is plausible that long-term SSH changes could exacerbate the impacts of industrial activities on regional SPM dynamics. Confirming this hypothesis will require longer observational records and targeted analyses explicitly linking SSH variability to coastal erosion and sediment fluxes.

4.3. Limitations and Perspectives for Multi-Scale Monitoring

This study has clearly demonstrated the potential of long-lasting ocean color time series (here MODIS 1 km resolution data) for characterizing SPM variability in a complex and industrialized coastal environment although limitations for describing the most nearshore domain due to reduced spatial resolution. The 20-year MODIS record (2002–2022) captured the evolution of SPM both before the installation of the Açu Port and throughout its subsequent construction and operational phases, allowing the identification of clear seasonal patterns, interannual fluctuations, and a significant long-term increase coinciding with the intensification of industrial activities. Through the combination of MODIS monthly composites with the Census X-11 decomposition, the seasonal Mann–Kendall test, and Sen’s slope estimator, the sensor proved capable of resolving coherent and persistent changes in sediment dynamics at regional scale. While its moderate spatial resolution does not allow the identification of fine-scale dredging plumes or small disposal features, the results show that MODIS provides robust information on broad-scale coastal processes and long-term environmental evolution. The reliability of these long-term patterns is further supported by the use of OC-SMART for atmospheric correction. Its multilayer neural-network architecture and enhanced cloud-screening capabilities [34] ensure stable, spatially consistent ${\mathrm{R}}_{\mathrm{r}\mathrm{s}}$(λ) retrievals in this turbid, optically complex region. In addition, OC-SMART has recently been validated for MODIS-Aqua coastal applications [47], providing additional confidence in the consistency of the 20-year SPM time series.

This study illustrates the impact of the Açu Port installation on surface SPM dynamics for the first time using a 20-year ocean-color time series. The available information on port activities consists mainly of qualitative descriptions of construction phases, dredging episodes, and operational milestones, with no quantitative records of dredged volumes or sediment fluxes associated with pipelines. This limits the ability to attribute precise magnitudes to the observed changes. In addition, no operational hydrosedimentary model exists for the region, preventing a full three-dimensional assessment of sediment transport pathways. Despite these constraints, the long-term MODIS analysis provides an initial regional diagnosis that establishes the baseline necessary for future quantitative modeling and process-based studies.

Future work could benefit from complementing this approach with higher-resolution sensors such as the Sentinel-2 Multispectral Instrument (MSI) or the Landsat-8/9 Operational Land Imager (OLI), in order to better resolve nearshore structures and localized sediment patterns. A multi-scale framework combining the temporal continuity of MODIS (the only long-lasting mono-sensor time series able to depict the pre and post port phases) with the spatial detail of high-resolution missions (able to describe specific events and the nearshore domain) would further enhance the monitoring of rapidly evolving coastal systems.

5. Conclusions

This study presents a comprehensive satellite-based approach to quantifying long-term changes in coastal sediment dynamics, focusing on the impact of industrial port infrastructure. A new SPM retrieval model was developed by combining two optimized bio-optical algorithms based on OWTs to ensure accurate performance across a wide range of coastal optical conditions for MODIS applications. Validation using both in situ and satellite data confirmed the reliability of this approach, particularly in turbid environments, where the accuracy of conventional models is often reduced.

When applied to 20 years (2002–2022) of MODIS-Aqua imagery, the model revealed significant spatial and temporal variations in SPM concentrations along the northern coast of Rio de Janeiro state. The most significant and statistically robust increases occurred near the APIC, with trends reaching 60% in two decades. These increases coincided with major phases of industrial development, including large-scale dredging and port expansion, indicating a significant anthropogenic influence. Natural processes also modulate SPM dynamics: currents exert a strong seasonal influence by driving the plume northward, winds show significant correlations with SPM at the interannual scale, and waves enhance sediment resuspension in coastal and inner-shelf domains which amplify short- to interannual-scale variability. However, since no long-term increase was detected in either wind intensity or wave power, their role remains limited to modulation rather than driving the persistent rise in SPM. Together, winds and waves continue to redistribute suspended material and sustain the observed northward and eastward sediment dispersion, but industrialization remains the dominant driver of the long-term increase in SPM near the APIC.

The amount of sediment that needs to be dredged in the APIC is closely linked to SPM concentrations, which are subject to tidal, meteorological, and climatological variability. In addition to these natural drivers, anthropogenic activities such as dredging, shipping, and sediment disposal can modify the distribution of SPM by enhancing the processes of dispersal, resuspension, and buffering [109,121,122,123]. Elevated SPM levels in adjacent coastal waters have significant implications for water quality and the functioning of ecosystems. Sediment particles carry trace metals and hydrophobic organic contaminants [124] and also contribute to nutrient inputs [125]. While SPM can stimulate primary production, excessive enrichment increases the risk of eutrophication, oxygen depletion, and fish mortality [126,127]. Furthermore, high turbidity reduces light penetration, thereby altering phytoplankton bloom dynamics and affecting higher trophic levels [68]. These results demonstrate that long-term increases in SPM offshore of the APIC are primarily linked to a port-driven supply of sediment, while natural circulation patterns regulate its redistribution and transport. This interplay between anthropogenic pressures and hydrodynamic modulation emphasizes the importance of both sediment management and ecological monitoring in preserving coastal ecosystems amid expanding industrial activity.

Satellite ocean color remote sensing has become an indispensable tool for monitoring SPM dynamics and assessing water quality in complex and impacted coastal zones. Long-term datasets from optical sensors such as MODIS provide extensive spatial coverage and historical depth, enabling the consistent detection of environmental trends over time and across space. The methodological framework applied here, which integrates atmospheric correction, OWT-based algorithm fusion, and statistical time-series analysis, offers a robust, scalable, and transferable strategy for identifying human-induced changes in coastal systems. AS well as supporting environmental assessments and constraining hydro-sedimentary models, these tools enhance our ability to distinguish anthropogenic signals from natural variability. Although this approach has been demonstrated for the Brazilian coast, and particularly the heavily industrialized Açu Port region, it can be used as a valuable template for similar applications in other coastal areas undergoing rapid industrialization and facing sediment management challenges.

Nonetheless, several limitations must be acknowledged. The progressive degradation of the MODIS-Aqua sensor since 2022 has affected the continuity and reliability of the time series. While multi-sensor datasets such as GlobColour and OC-CCI offer extended coverage, their lower reliability in optically complex nearshore waters constrains their utility. Furthermore, validation efforts are hindered by the limited availability and spatial representativeness of in situ measurements, particularly during key periods of industrial change. Future work should prioritize the integration of higher-resolution sensors such as Sentinel2-MSI or data from the recently launched hyperspectral NASA mission Plankton, Aerosol, Cloud, Ocean Ecosystem (PACE), which promise improved retrieval accuracy in coastal zones.