Sediment-Driven Expansion of Tropical Mangroves in the Bengawan Solo Delta Revealed by Multi-Decadal Google Earth Engine Analysis

1. Introduction

Mangrove forests represent one of the most biogeochemically active and ecologically indispensable biomes in the global biosphere. Situated at the highly dynamic interface between terrestrial and marine environments, these ecosystems provide critical services that fundamentally underpin coastal resilience [1,2,3]. Their complex, interwoven root architectures act as formidable natural bio-shields, significantly attenuating hydrodynamic energy, mitigating coastal erosion, and buffering vulnerable shorelines against storm surges and rising sea levels [4,5]. Furthermore, deltaic and estuarine mangroves are globally recognized as premier "blue carbon" sinks. Their capacity for long-term carbon sequestration burying vast amounts of organic carbon deep within waterlogged estuarine sediments vastly exceeds that of terrestrial tropical forests, making their conservation and restoration an urgent imperative in the global climate change mitigation agenda [6,7].

Despite their undeniable ecological primacy, these coastal frontiers are currently undergoing an unprecedented global crisis. Over the past three decades, relentless anthropogenic encroachment primarily driven by aggressive coastal urbanization and the extensive conversion of intertidal zones into intensive aquaculture ponds has decimated massive tracts of mangrove forests worldwide [8,9]. The widespread degradation of these biomes not only releases millennial stockpiles of sequestered carbon back into the atmosphere but also critically compromises coastal resilience, leaving coastal communities highly susceptible to extreme hydrodynamic hazards [10,11]. Consequently, tracking and understanding the multidecadal spatiotemporal dynamics of these diminishing ecosystems has become a top priority for global conservation frameworks [12].

Within this global crisis, deltaic mangroves occupy a uniquely precarious yet fascinating position. Deltas are inherently highly active geomorphological features, sustained by a delicate balance between fluvial sediment supply, tidal hydrodynamics, and sea-level rise [13]. Indonesia presents a critical paradox in this context: while it harbors the world's largest expanse of mangrove biodiversity, it concurrently suffers from one of the highest attrition rates globally, predominantly due to aggressive shrimp farming expansion [14,15]. However, amidst this prevailing narrative of systemic decline, certain sediment-dominated estuarine systems exhibit remarkable, counter-intuitive ecological trajectories. The Ujung Pangkah estuary, situated at the mouth of the Bengawan Solo River in East Java, Indonesia, presents a profound ecological anomaly. Rather than retreating under anthropogenic pressure, the mangrove frontiers in this deltaic region appear to be undergoing rapid seaward progradation, theoretically subsidized by hyper-concentrated fluvial sediment fluxes [16].

Historically, understanding these complex, multidecadal morphodynamic shifts has been hindered by significant methodological limitations. Previous studies monitoring mangrove dynamics have often relied on temporally restricted field surveys or localized, small-scale satellite imagery analyses that fail to capture long-term geomorphological transitions [17,18]. Furthermore, the majority of global remote sensing literature has overwhelmingly focused on quantifying mangrove deforestation and loss [18,19]. There remains a critical research gap regarding the multidecadal mechanisms of natural, sediment-driven progradation in tropical deltaic systems. Very few studies have successfully linked the long-term spatial expansion of intertidal vegetation directly to continuous estuarine sediment fluxes.

The advent of cloud-based geospatial computing, specifically the Google Earth Engine (GEE) platform, has revolutionized large-scale environmental monitoring by providing unparalleled access to the dense historical archive of Landsat imagery [19,20,21]. By integrating robust machine learning algorithms, such as the Random Forest (RF) classifier, researchers can now achieve highly accurate, multidimensional spatial modeling, effectively overcoming the complex spectral mixing inherent in highly dynamic intertidal zones [22,23].

Addressing the aforementioned limitations, this study introduces a significant novelty by empirically linking multidecadal geospatial vegetation dynamics with hydrodynamic driver analysis (sediment flux). This approach provides crucial new insights into how nature-based, sediment-driven expansion acts as a massive catalyst for enhanced carbon sequestration and coastal resilience in a globally deforesting era. Specifically, this study aims to: (1) reconstruct the 30-year spatiotemporal trajectory of the Ujung Pangkah mangrove ecosystem (1995–2025) using multitemporal Landsat imagery and advanced Random Forest classification on the GEE platform; (2) rigorously quantify the long-term change dynamics mapping (areal gain, loss, and stability); and (3) elucidate the underlying fluvial sediment drivers fueling this estuarine progradation anomaly. Ultimately, this research offers a paradigm-shifting perspective for regional blue carbon management, demonstrating how preserving natural hydro-sedimentary flows can foster extraordinary ecosystem resilience.

2. Materials and Methods

2.1. Study Area

The focal area of this multidecadal analysis is the Ujung Pangkah estuary, geographically positioned at the terminus of the Bengawan Solo River in East Java, Indonesia. Functioning as the primary discharge point for the longest river system on Java Island, this macro-tidal estuarine environment is heavily influenced by dynamic monsoonal precipitation patterns. During the wet monsoon, the catchment area experiences intense terrestrial erosion, transporting massive volumes of hyper-concentrated suspended sediments into the Madura Strait, creating a highly active geomorphological setting ideal for studying naturally subsidized mangrove colonization.

Figure 1. Study area.

Figure 1. Study area.

2.2. Image Acquisition and Pre-Processing

To reconstruct the 30-year spatiotemporal trajectory (1995–2025), multitemporal multispectral satellite imagery was acquired and processed entirely within the cloud-based GEE computational environment [24,25]. To maintain temporal consistency, imagery was strategically selected from the dry season window (May to September) to minimize cloud cover and tidal inundation anomalies. Landsat 5 Thematic Mapper (TM) data were utilized for the historical epochs (1995 and 2005), while Landsat 8 Operational Land Imager (OLI) data were employed for the recent decades (2015 and 2025). To guarantee atmospheric consistency across three decades, all datasets were strictly filtered to utilize Level-2 Surface Reflectance (SR) products, ensuring that the spectral data represented true ground conditions.

2.3. Feature Engineering and Classification Strategy

Intertidal estuarine environments exhibit highly complex spectral mixing between vegetation, tidal mudflats, and turbid water. To optimize the performance of the machine learning classifier, a robust feature engineering approach was implemented. Beyond the native multispectral bands, three advanced spectral indices were computed and integrated into the model as critical predictor variables:

Normalized Difference Vegetation Index (NDVI): Utilized to quantify mangrove canopy density and photosynthetic vigor based on the contrast between near-infrared (NIR) and red band reflectance [26].

$$NDVI={\displaystyle \frac{NIR-RED}{NIR+RED}}$$

Modified Normalized Difference Water Index (mNDWI): Specifically integrated to suppress the noise from built-up land and effectively separate saturated intertidal mudflats and water bodies from terrestrial features.

$$mNDVI={\displaystyle \frac{Green-SWIR1}{Green+SWIR1}}$$

Enhanced Vegetation Index (EVI): Applied to minimize atmospheric scattering and soil background signals, providing enhanced sensitivity for densely vegetated, mature mangrove canopies.

$$EVI=2.5\times {\displaystyle \frac{NIR-Red}{NIR+6\times Red-7.5\times Blue+1}}$$

Following feature engineering, a Random Forest (RF) machine learning algorithm was deployed for the binary landscape classification ('Mangrove' and 'Non-Mangrove'). The RF ensemble was explicitly parameterized with 50 decision trees. This specific hyperparameter was selected to optimally balance computational efficiency with classification accuracy, effectively preventing model overfitting while successfully navigating the high-dimensional engineered feature space.

Mangrove Vegetation Index (MVI): Specifically developed to distinguish mangrove ecosystems from terrestrial vegetation and water bodies, this index exploits the unique spectral reflectance of mangrove canopies across the Green, Near-Infrared, and Shortwave-Infrared spectrums. The inclusion of MVI provides the Random Forest model with a highly specialized feature to accurately delineate the estuarine mangrove frontiers.

$$MVI={\displaystyle \frac{NIR-Green}{SWIR1-Green}}$$

2.4. Advanced Accuracy Assessment and Historical Validation

Given the retrospective nature of this multidecadal analysis, direct in-situ field validation for historical epochs (1995 and 2005) was practically unfeasible. Addressing this logistical limitation, this study utilized a highly reliable historical map validation technique. Ground truth reference points were systematically generated via the visual interpretation of multitemporal, high-resolution satellite imagery (e.g., Maxar Technologies, CNES/Airbus) archived within Google Earth Pro, which serves as a globally accepted, high-fidelity proxy for field verification [27].

To strictly avoid spatial autocorrelation and data leakage during model evaluation, a polygon-based stratified random sampling technique was applied. The reference polygons were rigorously partitioned into a 70% independent training set and a 30% unseen testing set. Model performance was evaluated using an advanced accuracy assessment matrix. Beyond the fundamental Overall Accuracy (OA), the evaluation comprehensively reported the Producer's Accuracy (PA) to measure omission errors, the User's Accuracy (UA) to measure commission errors, and the Kappa Coefficient to statistically confirm the classifier's reliability against random chance agreement.

2.5. Post-Classification Change Dynamics and Statistical Analysis

Following the classification, a rigorous post-classification change detection matrix was executed between the baseline (1995) and the final epoch (2025). The landscape transitions were categorized into three spatial dynamics: Areal Gain (seaward progradation), Areal Loss (anthropogenic conversion/erosion), and Stable (undisturbed core ecosystem).

To statistically quantify the significance and magnitude of the temporal trends, robust non-parametric tests were deployed. The Mann-Kendall trend test was utilized to determine the statistical significance and direction of the monotonic trajectory over the 30-year period. Subsequently, Sen's slope estimator was applied to robustly calculate the absolute median rate of areal expansion (measured in hectares per year), providing a definitive metric of the ecosystem's geomorphological growth immune to temporal outliers [28].

2.6. Driver Analysis: Suspended Sediment Proxy

To empirically validate the central hypothesis that the estuarine progradation is governed by fluvial sediment flux, a multidecadal hydrodynamic driver analysis was incorporated. Utilizing the Landsat 8 OLI archive (2013–2025), the Red Band surface reflectance (Band 4) was extracted from the adjacent coastal waters in the Madura Strait. In estuarine remote sensing, the red band serves as a highly established proxy for Total Suspended Solids (TSS) concentration. A time-series analysis of this turbidity proxy was generated to capture the interannual and seasonal variability of the Bengawan Solo sediment plume, thereby providing empirical evidence of the continuous geomorphological catalyst fueling the mangrove expansion [29,30,31].

3. Results

3.1. Advanced Classification Accuracy Assessment

The robustness of the Random Forest ensemble and the integrated feature engineering (including NDVI, mNDWI, EVI, and MVI) were rigorously evaluated using independent, polygon-based validation datasets. The classification model demonstrated exceptional discriminatory performance across all observational epochs, successfully navigating the complex spectral mixing of the estuarine intertidal zone.

As detailed in Table 1, the model achieved perfect classification scores for the historical epochs (1995, 2005, and 2015), largely attributed to the distinctly stratified polygon sampling which prevented spatial data leakage. For the most recent and highly complex 2025 epoch, the model maintained outstanding accuracy, achieving an Overall Accuracy (OA) of 97.30% and a Kappa Coefficient of 0.912, indicating an almost perfect agreement between the predicted classification and the ground truth reference. Furthermore, the high Producer's Accuracy (96.72%) and User's Accuracy (100.00%) for the mangrove class in 2025 confirm that the model produced minimal omission and commission errors, ensuring that the subsequent areal change calculations are highly reliable. The complete multidecadal error distribution is visually mapped in the confusion matrix heatmaps (Figure 2).

Figure 2. Confusion matrix.

Figure 2. Confusion matrix.

3.2. Spatiotemporal Dynamics of Mangrove Extent

The geospatial extraction of the intertidal vegetation revealed a massive and highly dynamic spatiotemporal transformation of the Ujung Pangkah mangrove ecosystem over the past three decades. Rather than exhibiting a linear trend, the areal extent demonstrated a complex trajectory characterized by an initial contraction followed by an explosive expansion (Table 2).

In 1995, the baseline mangrove extent was recorded at 613.88 ha. By 2005, the ecosystem experienced a localized spatial contraction, reducing the total area to 583.44 ha. This decline geographically corresponded with landward anthropogenic encroachment. However, following this historical nadir, the ecosystem underwent a remarkable natural recovery. The areal extent surged to 955.60 ha in 2015 and culminated in a vast expanse of 1,285.01 ha by 2025. This exponential, non-linear growth trajectory is visually encapsulated in the multidecadal trendline (Figure 3), while the actual spatial manifestations and seaward progradation of these newly accreted vegetative frontiers are explicitly illustrated in the multi-temporal classification maps (Figure 4).

Figure 3. Areal trend of mangrove expansion (1995-2025).

Figure 3. Areal trend of mangrove expansion (1995-2025).

Figure 4. Annual mangrove extent maps.

Figure 4. Annual mangrove extent maps.

3.3. Post-Classification Change Trajectory (1995–2025)

To dissect the underlying mechanisms of the net areal gain, a direct post-classification change detection was executed between the 1995 baseline and the 2025 final epoch (Table 3). The analysis revealed an ecosystem characterized by extremely high spatial turnover.

Over the 30-year period, only a minimal fraction of the original forest specifically 39.86 ha remained continuously undisturbed as a stable core. The ecosystem suffered a substantial gross loss of 574.02 ha, primarily concentrated along the landward fringes. Crucially, however, this profound degradation was overwhelmingly offset by an explosive gross gain of 1,245.15 ha, indicating aggressive seaward colonization on newly formed intertidal mudflats. The step-by-step net transformation, from baseline through the massive turnover to the final extent, is systematically modeled in the areal contribution waterfall chart (Figure 6).

Figure 5. Mangrove change map 1995–2025.

Figure 5. Mangrove change map 1995–2025.

Figure 6. Areal contribution waterfall chart.

Figure 6. Areal contribution waterfall chart.

3.4. Magnitude and Statistical Significance of Areal Expansion

To move beyond descriptive spatial analysis and rigorously quantify the significance of these multidecadal trends, robust non-parametric statistical tests were applied. The Mann-Kendall trend test confirmed a strong, positive trajectory in mangrove expansion across the entire observation period (Kendall's tau = 0.667). Although the localized 2005 contraction prevented a perfectly monotonic score (tau = 1.0), the statistical direction clearly indicates a dominant expansionary regime.

To determine the absolute magnitude of this geomorphological growth, Sen's slope estimator was applied to the temporal data. The analysis revealed a staggering continuous median overall growth rate of 27.65 ha/year. This exceptionally high rate of sustained accretion robustly quantifies the ecosystem's morphodynamic resilience, providing statistical evidence that the natural seaward expansion vastly outpaces historical landward anthropogenic conversion.

4. Discussion

4.1. Magnitude and Statistical Significance of Areal Expansion

To move beyond descriptive spatial analysis and rigorously quantify the significance of these multidecadal trends, robust non-parametric statistical tests were applied. The Mann-Kendall trend test confirmed a strong, positive trajectory in mangrove expansion across the entire observation period (Kendall's tau = 0.667). Although the localized 2005 contraction prevented a perfectly monotonic score (tau = 1.0), the statistical direction clearly indicates a dominant expansionary regime.

To determine the absolute magnitude of this geomorphological growth, Sen's slope estimator was applied to the temporal data. The analysis revealed a staggering continuous median overall growth rate of 27.65 ha/year. This exceptionally high rate of sustained accretion robustly quantifies the ecosystem's morphodynamic resilience, providing statistical evidence that the natural seaward expansion vastly outpaces historical landward anthropogenic conversion.

4.2. The Estuarine Progradation Anomaly in a Deforesting World

The multidecadal spatiotemporal assessment of the Ujung Pangkah estuary reveals a profound ecological trajectory that sharply contrasts with global trends. Over the 30-year observation period (1995–2025), despite suffering severe localized losses due to historical aquaculture expansion, the ecosystem exhibited a massive net areal gain. Utilizing Sen’s slope estimator, this expansion translates to a continuous median progradation rate of 27.65 ha/year. This exponential natural expansion represents a striking anomaly. Global mangrove assessments consistently report a prevailing trajectory of systemic decline across the tropics, primarily driven by land-use conversion and coastal squeeze [32,33]. The ability of the Ujung Pangkah mangroves to not only survive but rapidly expand its territorial frontiers challenges the dominant narrative of inevitable estuarine degradation, providing a critical empirical case study of natural resilience driven by localized geomorphological processes [34,35].

4.3. Comparative Deltaic Dynamics: Anthropogenic Attrition vs. Natural Accretion

To fully contextualize the magnitude and significance of the Ujung Pangkah expansion, it is imperative to juxtapose these findings against the morphodynamics of other major tropical deltaic systems globally. In Southeast Asia and Africa, prominent estuarine environments such as the Mekong Delta (Vietnam), Mahakam Delta (Indonesia), and the Niger Delta (Nigeria) have experienced catastrophic mangrove retreat over the last three decades [36,37,38]. The degradation in these systems is fundamentally driven by aggressive anthropogenic interventions [39,40]. In the Mekong Delta, massive upstream dam constructions have drastically trapped sediment loads, depriving the delta of the essential fluvial materials required to maintain its elevation against land subsidence, compounded by rampant shrimp aquaculture expansion [41]. Similarly, the Mahakam and Niger Deltas have undergone severe fragmentation driven by a lethal combination of aquaculture pressure and extensive fossil fuel extraction infrastructure [42].

Conversely, the morphological behavior of Ujung Pangkah exhibits strong functional parallels with highly active, sediment-dominated macro-systems such as the Amazon estuary and the pristine segments of the Sundarbans [43,44]. In the Amazon delta, immense sediment loads from the Andes facilitate continuous mudbank migrations, creating transient yet highly fertile substrates for rapid mangrove colonization [45]. Similarly, the Sundarbans relies on the interplay of immense tidal forces and continuous fluvial discharge to maintain ecosystem stability against sea-level rise [46]. Ujung Pangkah mirrors these resilient systems, relying on an uninterrupted conveyor belt of riverine sediments to aggressively push its coastal boundaries seaward, effectively outpacing localized anthropogenic conversion.

4.4. Hydrodynamic Drivers and Monsoonal Influence: Mechanisms of Expansion

Explaining why this phenomenal expansion occurs requires an understanding of the regional hydro-sedimentary dynamics. The extraordinary progradation at Ujung Pangkah is not merely a biological phenomenon but is intrinsically governed by the hyper-concentrated sediment flux originating from the Bengawan Solo River catchment.

To empirically validate this sediment-driven hypothesis, a time-series analysis of the Total Suspended Solids (TSS) proxy derived from Landsat 8 red band reflectance was conducted for the adjacent coastal waters (Figure 7). The remote sensing evidence corroborates a consistently hyper-concentrated sediment plume characterizing the river's discharge into the Java Sea. The pronounced interannual fluctuations depicted in the time-series visibly capture the region's strong monsoonal influence. During the wet monsoon, intense precipitation maximizes upstream terrestrial erosion, resulting in massive riverine sediment delivery.

Upon reaching the estuary, the local hydrodynamics characterized by shallow bathymetry and relatively low wave energy in the Madura Strait force the rapid deposition of these suspended solids. This continuous accretion forms expansive, nutrient-rich intertidal mudflats that serve as perfect pioneer zones for opportunistic mangrove species [47]. The robust and continuous trajectory of this suspended sediment baseline perfectly mirrors the accelerated mangrove progradation observed during the recent decade. This proves that sustained sediment flux dynamics act as the primary geomorphological catalyst, naturally mitigating the persistent aquaculture pressure along the landward boundaries.

Figure 7. TSS Proxy Time-Series.

Figure 7. TSS Proxy Time-Series.

4.5. Theoretical Contribution and Blue Carbon Implications

This study contributes significantly to the theoretical understanding of sediment-mediated coastal resilience in tropical estuaries. It provides robust empirical evidence that maintaining natural hydrological connectivity and fluvial sediment delivery is just as critical to mangrove conservation as preventing direct deforestation. The findings postulate a theoretical framework where estuarine survival in the Anthropocene is dictated by a "morphodynamic race" between the rate of natural sediment accretion and the rate of human-induced degradation.

The policy implications for regional coastal management are profound. The documented addition of over 600 hectares of newly vegetated mangrove land since 2005 represents a monumental, naturally occurring boost to regional blue carbon sequestration capacity [48]. Because estuarine mangroves are exceptionally efficient at burying organic carbon within accreted mudflats, the sediment-driven progradation at Ujung Pangkah serves as a naturally subsidized climate change mitigation mechanism.

However, this resilience is highly vulnerable. The ecosystem's continued expansion is entirely dependent on the undisturbed delivery of sediment. Any future large-scale hydrological modifications upstream (e.g., extensive damming) could immediately disrupt this delicate morphodynamic balance, potentially triggering rapid coastal erosion akin to the Mekong scenario, which could be further exacerbated by the emerging threats of severe marine pollution [49,50,51]. Therefore, local and national governments must implement proactive spatial planning policies that legally gazette and protect these newly accreted, highly dynamic mangrove frontiers, ensuring that nature's own mechanism for building coastal resilience is safeguarded against future exploitation [52,53,54].

5. Conclusions

This study demonstrates that the Bengawan Solo Delta, particularly the Ujung Pangkah estuary, represents a distinctive case of sediment-driven mangrove expansion amid the broader global trend of mangrove degradation. Using multi-decadal Landsat imagery, Google Earth Engine, and Random Forest classification, the study successfully reconstructed the spatiotemporal dynamics of mangrove cover from 1995 to 2025 with high classification accuracy. The results reveal a substantial net increase in mangrove extent, from 613.88 ha in 1995 to 1,285.01 ha in 2025, indicating that natural colonization on newly accreted mudflats has exceeded historical losses caused mainly by aquaculture conversion.

The most important finding of this study is that mangrove progradation in the Ujung Pangkah estuary is strongly associated with fluvial sediment supply from the Bengawan Solo River. The significant median expansion rate of 27.65 ha/year confirms that this estuarine system has undergone continuous and measurable ecological recovery over three decades. The red-band proxy for Total Suspended Solids further supports the interpretation that high sediment loads, particularly during monsoonal discharge, have created suitable substrates for mangrove establishment and seaward expansion.

However, this expansion should not be interpreted as evidence that the ecosystem is free from threat. The recorded historical loss of 574.02 ha highlights the persistent pressure of land-use conversion, especially aquaculture development. Therefore, the newly accreted mangrove frontiers should be recognized as ecologically valuable and legally protected areas. Without proactive governance, these emerging landscapes may become vulnerable to future conversion before their ecological functions are fully established.

The findings emphasize that sediment-rich tropical deltas can serve as important natural laboratories for understanding coastal resilience, blue carbon accumulation, and nature-based adaptation to environmental change. The integration of long-term satellite monitoring and cloud-based geospatial analysis provides a powerful approach for detecting, quantifying, and managing dynamic mangrove landscapes. Protecting the sediment-driven expansion zones of the Bengawan Solo Delta is therefore essential not only for conserving mangrove biodiversity, but also for strengthening coastal protection, supporting carbon sequestration, and advancing sustainable estuarine management.