{"paper_id":"2c6153d8-6e2a-46f7-ae32-561070e73d90","body_text":"Multi-season Sentinel-2 reveals conservation-relevant transition-zone dynamics in a Phragmites communis–Suaeda japonica coastal wetland mosaic | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Multi-season Sentinel-2 reveals conservation-relevant transition-zone dynamics in a Phragmites communis–Suaeda japonica coastal wetland mosaic Gapseong Jekal, Yong Hwan Kim, Seung Hyeon Lee, Ji Weon Yun, Dae Yeol Kim, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8731780/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Coastal salt-marsh mosaics reorganize rapidly under hydro-climatic variability, yet many remote-sensing studies rely on single-season analyses, limiting conservation-relevant understanding of boundary transitions. This study developed a multi-season Sentinel-2 framework to monitor interannual dynamics in a Phragmites communis–Suaeda japonica–tidal-flat mosaic in Suncheon Bay, Korea. We analyzed 37 Sentinel-2A images acquired in spring and autumn from 2019, 2022, 2023, and 2024. NDVI, MNDWI, and SSVI were used to compare Decision Tree, Gradient Boosting, and Random Forest classifiers, and multi-year vegetation change was assessed through area, persistence, and inter-class transitions. Random Forest showed the best performance, with a mean overall accuracy of 85.7% and a kappa coefficient of 0.79. NDVI was consistently informative across seasons, whereas MNDWI contributed more strongly in spring and SSVI in autumn. The largest reorganization occurred during 2019–2022, when 9.82 ha (21.2%) underwent vegetation-type transitions. Suaeda japonica declined from 8.91 ha to 4.01 ha and recovered only partially by 2024, while most changes were concentrated along vegetation boundaries rather than stable cores. Multi-season remote sensing can move beyond technical classification by identifying vulnerable halophytic transition zones and priority areas for adaptive monitoring and management. The framework is therefore useful for tracking weakening mosaic heterogeneity in coastal wetlands of high conservation value. Coastal wetlands Phragmites communis Suaeda japonica Sentinel-2 Random Forest classification Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Introduction Coastal wetlands provide critical ecosystem functions, including biodiversity conservation, habitat provision, water purification, and shoreline stabilization(Huang et al. 2025 ). As transitional ecotones linking terrestrial and marine systems, they are regarded as globally important nature-based solutions for climate adaptation and coastal protection(Yi et al. 2024 ). However, climate change, sea-level rise, large-scale reclamation, aquaculture development, and the invasion of alien species such as Spartina alterniflora have led to rapid reductions in coastal wetland area and degradation of their ecological functions(Sun et al. 2021 ; Bai et al. 2025 ; Huang et al. 2025 ). These threats are recognized not only at the regional level but also internationally(Ke et al. 2024 ). Both the Ramsar Convention and the United Nations Sustainable Development Goals (SDGs) emphasize the necessity of quantitative wetland monitoring(Peng et al. 2023 ). In particular, understanding the spatial distribution and interannual dynamics of wetland vegetation provides critical insights into ecological processes and resilience, enabling adaptive and science-based management decisions(Yi et al. 2024 ). Phragmites communis ( P. communis ) and Suaeda japonica ( S. japonica ) are critical indicator species for monitoring ecological dynamics in East Asian coastal wetlands(Zhang et al. 2021 ). They cover approximately 75.9% of vegetation area in Korean coastal wetlands while in regions such as Yancheng, the Yellow River Delta, and the Liaohe Delta in China, their closely related species Phragmites australis and Suaeda salsa are dominant(Lee 2019 ; Sun et al. 2021 ; Yi et al. 2024 ). P. communis exhibits high biomass, long growth periods, deep rooting systems, and strong photosynthetic capacity, serving as a marker of wetland expansion and structural stability(Chaudhary et al. 2023 ). Conversely, S. japonica is an annual halophyte highly sensitive to climatic and hydrological variability, establishing in saline clay soils and turning purple in autumn due to the accumulation of betacyanin pigments as a physiological stress response(Lee et al. 2014 ; Chung 2018 ). This ecological contrast makes paired monitoring of the two species a direct means to track interannual vegetation turnover in coastal wetland mosaics(Gitay et al. 2011 ). Prior studies report relatively persistent expansion of Phragmites stands, whereas Suaeda distributions can fluctuate more strongly with external conditions, contributing to marked variability in wetland composition (Yi et al. 2024 ; Huang et al. 2025 ). Accordingly, simultaneous monitoring of these two species provides a practical basis for tracking vegetation distribution and turnover in tidal wetlands(Qiu et al. 2024 ). Long-term remote-sensing analyses in major Yellow Sea deltas show pronounced reorganization of Phragmites–Suaeda mosaics, where the areal extents of Phragmites and Suaeda marshes alternate between decline and recovery over decadal scales, and Suaeda typically exhibits larger interannual fluctuations than Phragmites(Chang et al. 2025 ). P. communis and S. japonica , the dominant vegetation in coastal wetlands, exhibit distinct seasonal color transitions: P. communis maintains green during the growing season before turning yellow in senescence, whereas S. japonica accumulates betacyanin in autumn, shifting from green to purple(Chung 2018 ; Huang et al. 2025 ). However, previous studies often relied primarily on the Normalized Difference Vegetation Index (NDVI), without sufficiently accounting for these species-specific phenological and spectral characteristics(Huang et al. 2025 ). As a result, spectral mixing among Suaeda salsa(S. salsa) , green vegetation, and bare ground frequently reduced classification accuracy(Ke et al. 2024 ). To address this issue, several indices specifically designed for S. salsa —such as the Suaeda salsa Vegetation Index (SSVI)—were introduced and achieved partial success(Huang et al. 2025 ). Since S. salsa and S. japonica are congeneric halophytes that share similar phenological patterns and pigment-related spectral traits, previous studies on S. salsa can reasonably be used as a methodological reference for research on S. japonica (Lee et al. 2014 ; Wang et al. 2022 ; Huang et al. 2025 ). Nevertheless, even these studies largely relied on single-date or short-term imagery, limiting their ability to capture the strong spatial heterogeneity and dynamic shifts that characterize coastal wetland mosaics (Sun et al. 2021 ). These gaps highlight the need for multi-temporal, seasonally optimized remote sensing approaches that can detect both interannual variability and spatially complex vegetation dynamics(Clemente et al. 2023 ). Beyond improving classification accuracy, coastal wetland monitoring must support conservation planning and management by identifying where ecological character is being maintained and where it is becoming vulnerable(Hansen et al., 2021 ). In mosaic wetlands such as Suncheon Bay, changes in vegetation boundaries are especially important because they indicate whether habitat heterogeneity is being retained or progressively simplified(Larkin, 2018 ). From a conservation perspective, the critical question is therefore not only how much vegetation area changes, but also where transitions occur, which vegetation type is retreating, and which boundary sectors require priority monitoring or field inspection(Yang et al., 2022 ). A multi-season remote-sensing framework can contribute to this need by detecting spatially explicit transition zones and by providing an operational basis for adaptive monitoring in Ramsar coastal wetlands(Lopatin et al., 2026 ). This study aims to monitor the multi-temporal changes of P. communis and S. japonica , two dominant species in East Asian coastal wetlands, by leveraging vegetation indices derived from species-specific phenological stages (March–April for the growing season, October–November for the senescent stage) across multiple years (2019, 2022, 2023, 2024). As P. communis represents stability and S. japonica reflects vulnerability, quantifying their spatiotemporal changes and linking them to environmental variables provides critical insights into the resilience and vulnerability of coastal wetland ecosystems(Peter Sheng et al. 2022 ; Song et al. 2022 ). The specific objectives of this study are as follows: (1) To develop and evaluate a multi-season Sentinel-2 classification framework for monitoring Phragmites communis and Suaeda japonica in a coastal wetland mosaic. (2) To quantify multi-year (2019–2024) vegetation dynamics through area change, persistence, and inter-class transitions, with emphasis on boundary transition zones. (3) To examine temporal associations between vegetation dynamics and hydro-meteorological conditions as a basis for future field-based monitoring. Materials and Methods 2.1 Study area Suncheon Bay is a representative coastal wetland located in Suncheon, Jeollanam-do, on the southern coast of the Republic of Korea. It is characterized by extensive mudflats formed at the confluence of the Dongcheon and Isacheon rivers (Fig. 1 )(Hong et al. 2015 ). The total area of the wetland is approximately 27 km², of which about 5.4 km² of intertidal wetlands are dominated by P. communis and S. japonica communities(Chaudhary et al. 2023 ). In 2006, Suncheon Bay was designated as the first Ramsar wetland in Korea, in addition to being designated as a national Wetland Protected Area, thereby underscoring its international recognition as a region of high conservation value(Byaruhanga and Kigoolo 2005 ). The P. communis community, which accounted for about 0.96 km² (9.6% of the surveyed area) in 2011, is the most extensive vegetation type in Suncheon Bay(Kong et al. 2014 ). By contrast, the S. japonica community covers a smaller area of approximately 0.166 km² (1.6%). S. japonica primarily occurs in low-lying salt marshes with moderate seawater exchange and silt deposition(Chung et al. 2021 ). However, as an annual halophyte, it responds sensitively to environmental changes, resulting in high variability in its areal extent(Lee et al. 2014 ). conservation(Jiao et al. 2020 ; Chaudhary et al. 2023 ). 2.2 Materials In this study, Sentinel-2A satellite imagery was acquired using Google Earth Engine (GEE) (Table 1). The target years were 2019, 2022, 2023, and 2024, with images collected for March, April, October, and November of each year. During preprocessing, any images containing cloud, shadow, or snow within the study boundary were visually inspected and excluded from the analysis.(Liu et al. 2024 ). Considering that Suncheon Bay exhibits a tidal range exceeding 3 m, only images with a mean value of the Modified Normalized Difference Water Index (MNDWI) less than 0 within the study boundary were selected, as this indicates conditions where vegetation cover is dominant over water bodies(Sun et al. 2021 ). In total, 37 Sentinel-2 images were compiled, and high-resolution reference datasets were established to enable temporal validation of vegetation distribution in Suncheon Bay. Multispectral UAV imagery with a spatial resolution of 6 cm per pixel was acquired on November 15, 2024, targeting P. communis and S. japonica communities. Additional UAV imagery from September 2019, October 2022, and October 2023, together with Google Earth data and high-resolution drone imagery from 2024, were integrated to construct annual reference maps for accuracy validation (Table 1). Table 1. Remote sensing datasets and reference materials Data source Specification resolution Usage Observation Dates Sentinel-2A Bands B2 (Blue), B3 (Green), B4 (Red), B8 (NIR), B11 (SWIR) 10 m Classification training and monitoring 2019 : Mar-01, Mar-16, Apr-15, Apr-18, Oct-20, Nov-06, Nov-09 2022 : Mar-15, Apr-04, Apr-24, Oct-14, Oct-16, Oct-19, Oct-31, Nov-03, Nov-05, Nov-13, Nov-18 2023 : Mar-03, Mar-13, Apr-02, Apr-27, Oct-11, Oct-21, Nov-08, Nov-13, Nov-20, Nov-25 2024 : Mar-02, Apr-01, Apr-13, Apr-18, Oct-08, Oct-10, Nov-07, Nov-12, Nov-22 Multispectral UAV imagery Red, Green, Blue, Red edge, NIR 6 cm/pixel Accuracy validation Nov-15, 2024 Reference data UAV imagery, Google Earth - Accuracy validation Sep-2019, Oct-2022, Oct-2023 2.3 Training Area and Vegetation Index Construction The training areas were established based on prior research and field survey results, targeting regions where vegetation was consistently maintained across the years 2019, 2022, 2023, and 2024 (Fig. 2 ). Training areas were selected for P. communis , S. japonica , and tidal flats, and these identical training areas were consistently applied to Sentinel-2 imagery from each of the four years. NDVI, MNDWI, and SSVI were derived from Sentinel-2 images acquired in March, April, October, and November of 2019, 2022, 2023, and 2024, respectively. Values within each training polygon at each time point were extracted (Table 2). Outliers were removed using a threshold of 1.5 times the interquartile range (IQR), and mean values of NDVI, MNDWI, and SSVI were calculated for each vegetation type and season(Zhang et al. 2021 ; Qiu et al. 2024 ). An integrated training dataset was constructed to comprehensively capture seasonal and interannual spectral variability, incorporating all samples from the four years(Fig. 3 , Table 3). Table 2. Vegetation indices (NDVI, MNDWI, SSVI) used for wetland vegetation classification Vegetation Index Formula Reference NDVI (NIR - RED) / (NIR + RED) (Rouse Jr, Haas et al. 1973) MNDWI (GREEN - SWIR) / (GREEN + SWIR) (Xu 2006 ) SSVI NDVI × ((BLUE - GREEN) / (GREEN - RED)) 2 (Ying et al. 2019 ) 2.4 Classification and Validation Wetland and coastal vegetation mapping poses significant challenges due to spectral heterogeneity, tidal inundation variability, and the coexistence of mixed land-cover types, requiring classification methods that are both robust and interpretable(Mahdavi et al. 2018 ). Tree-based algorithms are particularly suitable for such environments because they can model nonlinear relationships and maintain stability in the presence of noisy or correlated input variables(Berhane et al. 2018 ). Among these, Decision Tree (DT), Random Forest (RF), and Gradient Boosting (GB) have consistently demonstrated strong performance in vegetation classification tasks(Berhane et al. 2018 ). Comparing these classifiers allows for an integrated evaluation of the trade-offs between interpretability (DT), robustness to data variability (RF), and predictive capability (GB), providing a balanced perspective on model transparency and accuracy(Jamali et al. 2021 ). Therefore, comparing DT, RF, and GB offers a methodologically sound and ecologically interpretable approach for monitoring long-term vegetation dynamics in complex coastal wetlands(Jamali et al. 2021 ). All classifiers were trained on the same dataset and applied separately to four study years (2019, 2022, 2023, and 2024), resulting in a total of 12 classification maps (4 years × 3 classifiers). In addition, RF feature importance was derived to quantitatively assess the relative contributions of NDVI, MNDWI, and SSVI to seasonal and interannual classification (Table 2). Accuracy assessment was conducted for each year using field surveys, UAV imagery, and satellite data(Peng et al. 2023 ). Specifically, field survey data and UAV imagery acquired in November 2024, UAV imagery from September 2019, October 2022, and October 2023, as well as Google Earth imagery, were used as reference data. For validation, 600 random points (200 per class) were generated, from which confusion matrices were constructed(Liu et al. 2024 ). From these, Overall Accuracy (OA), Kappa coefficient, Producer’s Accuracy (PA), and User’s Accuracy (UA) were calculated for each year(Huang et al. 2025 ). 2.5 Spatiotemporal Change Analysis and Exploratory Environmental Pattern Similarity Based on a comparison of the RF, DT, and GB algorithms, the classifier with the highest accuracy was selected to quantify the spatiotemporal changes in P. communis and S. japonica communities from 2019, 2022, 2023, 2024. Subsequently, the interannual variations in the areas of P. communis and S. japonica were analyzed in relation to climatic and environmental variables. For this purpose, data were collected on precipitation, temperature (mean, maximum, minimum), humidity, tidal level (maximum, minimum, mean), wind speed (mean, maximum), sunshine duration, number of monsoon days, and number of heatwave days(Fu et al. 2022 ). All variables were normalized to a range of 0–1 to enable direct comparison across years(Table 4). The analysis was conducted by calculating RMSE between normalized vegetation area changes (2019, 2022, 2023, 2024) and normalized environmental variables. Lower RMSE values indicated higher similarity in trends, and variables with RMSE ≤ 0.25 were considered as key candidates. Given the limited number of study years (n = 4), this analysis was designed as an exploratory approach, focusing on identifying potential trend similarities rather than verifying causal relationships. Table 4. Environmental Variables for Vegetation Change Analysis Environmental Variable Unit Source Range Heatwave days days Korea Meteorological Administration (KMA, Suncheon City) 6–10 Sunshine duration hr KMA, Suncheon City 2093.8–2254.8 Mean humidity % KMA, Suncheon City 70.0–78.3 Total precipitation mm KMA, Suncheon City 987.0–2139.5 Mean temperature ℃ KMA, Suncheon City 13.1–13.8 Maximum temperature ℃ KMA, Suncheon City 19.3–19.8 Minimum temperature ℃ KMA, Suncheon City 7.7–8.9 Monsoon days days KMA, Suncheon City 16.4–22.2 Maximum wind speed m/s KMA, Suncheon City 9.1–11.7 Mean wind speed m/s KHOA (Gohung Balpo Station, DT0026) 1.7–2.0 Maximum marine wind speed m/s KHOA (Gohung Balpo Station, DT0026) 9.1–11.7 Maximum tide level m KHOA (Gohung Balpo Station, DT0026) 4.18–4.50 Minimum tide level m KHOA (Gohung Balpo Station, DT0026) -0.64 – -0.24 Mean tide level m KHOA (Gohung Balpo Station, DT0026) 1.929 – 1.964 Results 3.1 Performance and Seasonal Sensitivity of Machine Learning Classifiers for Multi-Year Wetland Vegetation Mapping Using multi-season vegetation indices (NDVI, MNDWI, and SSVI), Random Forest (RF) consistently outperformed Gradient Boosting (GB) and Decision Tree (DT) (Fig. 5 ). RF achieved the highest mean overall accuracy (OA = 85.7%) and Kappa (κ = 0.79), compared with GB (OA = 79.1%, κ = 0.68) and DT (OA = 69.2%, κ = 0.53), indicating that RF provides robust performance for multi-year coastal wetland vegetation mapping. (Fig. 4 , Fig. 5 ). RF variable importance further revealed phenology-driven seasonal shifts in the diagnostic value of indices(Tian et al. 2016 ) (Fig. 6 ). NDVI remained the most influential predictor across all seasons (0.46 in March; 0.37 in April; 0.38 in October; 0.36 in November), supporting its role as a stable baseline indicator of vegetation greenness(Pettorelli et al. 2005 ). In contrast, MNDWI was strongly spring-weighted (0.39–0.41 in March–April) but declined in autumn (0.27–0.30), suggesting that early-season wetness/hydrological signals contribute disproportionately to class separability in tidal wetlands(Narron et al. 2022 ). This seasonal contrast is plausibly driven by the combination of sparse early-season canopy conditions and strong tidal moisture/background effects in spring, whereas denser late-season vegetation reduces the relative separability provided by wetness signals(Surasinghe et al. 2025 ). Conversely, SSVI showed an autumn-weighted pattern, contributing minimally in spring (0.15–0.22) but increasing sharply in October–November (0.34–0.35), consistent with the seasonally expressed pigment dynamics of S. japonica (Ke et al. 2024 ). Together, these patterns demonstrate that while NDVI provides consistent discrimination, season-specific indices (MNDWI in spring; SSVI in autumn) add critical, phenology-relevant separability, supporting seasonally tailored index combinations for improved classification(Zeng et al. 2022 ). 3.2 Dynamics in the Multi-Seasonal Distribution of Phragmites communis and Suaeda japonica Across the years (2019–2024), Suncheon Bay showed a directional reorganization of vegetation–bare ground mosaics, characterized by a long-term expansion of Phragmites communis and a sharp contraction with only partial recovery of Suaeda japonica , accompanied by marked fluctuations in tidal-flat extent (Fig. 7; Table 5). S. japonica declined from 8.91 ha (2019) to 4.01 ha (2022; −55.0%), with modest recovery in 2023–2024 (4.55–4.74 ha) but remaining 46.8% below 2019. In contrast, P. communis increased to 19.77 ha in 2024 (+ 15.3% relative to 2019) after a temporary decline in 2022, while tidal flats expanded in 2022 (+ 29.3% relative to 2019) and then contracted by 2024. Vegetation type transitions in Suncheon Bay were consistently concentrated along vegetation boundaries (interfaces among Phragmites communis , Suaeda japonica , and tidal flats), rather than within vegetation cores (Figs. 8 – 9 ). In particular, core areas dominated by P. communis remained largely stable across the study period. The largest reorganization occurred in 2019→2022, when 9.82 ha (21.2%) underwent vegetation type transitions, dominated by loss of S. japonica to tidal flats (5.05 ha) and transitions from P. communis to tidal flats (4.25 ha), indicating pronounced boundary restructuring associated with S. japonica decline. Stability increased thereafter (85.1% unchanged in 2022→2023), with limited boundary transitions persisting (e.g., S. japonica →tidal flat 2.46 ha; tidal flat→ P. 2.28 ha). In 2023→2024, 81.1% of the area remained unchanged, and residual transitions were again confined to the same boundary zones(Lee et al. 2014 ). Overall, vegetation type transitions were boundary-driven, with the most substantial change occurring during 2019–2022 and linked to S. japonica decline(Peng et al. 2022 ). 3.3 Environmental Associations with Species-Specific Vegetation Shifts in a Coastal Wetland Ecosystem All environmental variables were normalized to a 0–1 scale for interannual comparability (Fig. 10 ). The area of P. communis showed strong trend similarity with humidity and temperature conditions, with RMSE values below 0.25 for mean humidity (0.10), minimum temperature (0.15), mean tide level (0.18), mean temperature (0.23), and total precipitation (0.23). These results indicate that the interannual trajectory of P. communis was temporally similar to variation in moisture- and inundation-related environmental conditions. This pattern is broadly consistent with previous studies showing that reed-dominated wetland vegetation responds sensitively to hydrological and salinity-related conditions under favorable climatic regimes (Qiu et al. 2024 ). In contrast, S. japonica showed relatively low RMSE values for total precipitation (0.20) and mean humidity (0.23), suggesting that its interannual decline and partial recovery were temporally associated with hydro-meteorological fluctuation. Previous studies have reported that reduced rainfall can increase soil salinity and suppress halophytic vegetation cover, whereas increased precipitation and humidity may alleviate salinity stress and improve moisture availability for germination and growth (Zhang et al. 2021 ; Ke et al. 2024 ). In this context, the observed dynamics of S. japonica may reflect sensitivity to shifts in hydro-saline conditions in the coastal wetland environment. However, because this analysis was based on four annual observations and did not include direct field measurements of salinity, sediment deposition, or micro-topography, these results should be interpreted as exploratory environmental associations rather than verified causal drivers. Discussion 4.1. Classification Performance and Ecological Implications of Spectral Indices This study demonstrates that combining multi-temporal Sentinel-2 observations with seasonally optimized vegetation indices substantially improves both classification accuracy and ecological interpretability in coastal wetlands. The Random Forest (RF) model achieved the highest overall accuracy (85.7%) and Kappa coefficient (0.785), confirming its robustness for discriminating vegetation types under the hydrologically dynamic conditions of Suncheon Bay. Seasonal index analysis revealed clear seasonal contrasts in index sensitivity associated with vegetation growth stages and hydrological regimes(Dronova and Taddeo 2022 ). In spring, NDVI and MNDWI were dominant, reflecting the combined effects of early vegetative growth and strong surface moisture contrasts. During this period, sparse canopy development and frequent tidal inundation amplify the spectral separability associated with water content and wet soil backgrounds, making MNDWI particularly effective at distinguishing halophytic vegetation from tidal flats(Dronova and Taddeo 2022 ). Similar spring-dominant roles of wetness-sensitive indices have been reported in salt-marsh systems with strong tidal influence, where hydrological conditions exert primary control over early-season spectral responses(O’Connell et al. 2017; Narron et al. 2022 ). In contrast, autumn classification was primarily driven by NDVI and SSVI, corresponding to peak biomass conditions and species-specific pigment dynamics. While NDVI captured overall canopy density and biomass accumulation(Mutanga et al. 2012 ), SSVI provided complementary information by emphasizing red spectral responses associated with betacyanin accumulation in Suaeda species during late-season physiological stress and senescence(Ying et al. 2019 ; Zhang et al. 2024 ); this autumn-weighted importance of pigment-sensitive indices is consistent with previous findings for Suaeda spp. in saline wetlands(Ying et al. 2019 ). However, NDVI-only approaches are limited in salt-marsh environments because greenness-based signals may miss pigment-driven physiological changes in halophytes, whereas SSVI isolates red-band signals linked to betacyanin accumulation and improves autumn species-level discrimination(Marchesini et al. 2016 ; Lu et al. 2020 ). 4.2. Species-Specific Ecological Implications: Structural Stability of Phragmites communis and Environmental Vulnerability of Suaeda japonica The long-term expansion of P. communis and decline of S. japonica indicate a directional reorganization of the Suncheon Bay salt-marsh mosaic rather than a simple vegetation gain–loss pattern. From 2019 to 2024, P. communis increased by 15.3% (from 17.15 to 19.77 ha), whereas S. japonica remained 46.8% below its 2019 extent by 2024, despite partial recovery after 2022. This contrast is broadly consistent with regional observations that Phragmites -dominated vegetation often shows greater persistence, whereas halophytic Suaeda communities fluctuate more strongly under changing coastal wetland conditions (Huang et al. 2024 ). The increase in P. communis may reflect its perennial life cycle, clonal expansion capacity, and structural persistence, all of which can contribute to sediment trapping, reduced flow velocity, and organic matter accumulation (Moore et al. 2012 ). In addition, the relatively low RMSE values between P. communis area and mean humidity (0.10), minimum temperature (0.15), mean tide level (0.18), mean temperature (0.23), and total precipitation (0.23) suggest temporal similarity with moisture- and inundation-related environmental variation, consistent with previous studies on reed-dominated wetland vegetation (Chambers et al. 1999 ; Qiu et al. 2024 ). However, from a conservation perspective, reed expansion should not be interpreted uncritically as structural change with mixed ecological implications. Although greater Phragmites cover may enhance structural stabilization and shoreline buffering, it may also reduce habitat heterogeneity if it progressively replaces more environmentally sensitive halophytic vegetation. Moreover, previous studies suggest that reed-dominated wetlands can involve biogeochemical trade-offs, including increased soil carbon accumulation and potentially elevated methane-related processes (Lee and Nepf 2024 ; Ury et al. 2024 ). By contrast, the marked reduction of S. japonica suggests greater sensitivity to short-term hydro-meteorological variability and transition-zone instability. S. japonica declined from 8.91 ha in 2019 to 4.01 ha in 2022 (− 55.0%), and recovered only partially to 4.74 ha by 2024, indicating that its distribution remained substantially contracted over the study period. Because this annual halophyte typically occupies lower and more environmentally exposed salt-marsh positions, changes in precipitation, humidity, inundation regime, sedimentation, and salinity balance may disproportionately affect its germination and establishment (Flowers and Colmer 2008 ; Sun et al. 2010 ). This interpretation is supported by relatively low RMSE values for total precipitation (0.20) and mean humidity (0.23), which indicate that S. japonica area exhibited temporal similarity with hydro-meteorological fluctuation rather than stable persistence (O’Connell et al. 2017). Previous studies likewise show that reduced rainfall can increase soil salinity and limit halophytic vegetation cover, whereas wetter conditions may alleviate salinity stress and improve establishment conditions (Zhang et al. 2021 ; Ke et al. 2024 ). In this context, the ecological significance of the observed change lies not only in the decline of a single species, but in the potential weakening of mosaic heterogeneity through contraction of vulnerable halophytic boundary habitats. Thus, the coupled expansion of P. communis and retreat of S. japonica may be interpreted as a shift toward a more homogeneous marsh structure, with important implications for habitat diversity and conservation-oriented wetland management (Steinigeweg et al. 2023 ). 4.3. Management Implications and Future Directions for Coastal Wetland Conservation Interannual shifts in plant occupancy are a common feature of coastal wetlands, where small changes in elevation, hydroperiod, and sediment dynamics can reorganize vegetation–tidal flat mosaics over relatively short timescales (Morris et al. 2002 ). In this context, the contrasting trajectories observed in Suncheon Bay—namely, the expansion of Phragmites communis and contraction of Suaeda japonica —suggest that conservation monitoring should not focus solely on dominant vegetation extent, but should also track dynamic boundary zones where ecological reorganization is most likely to occur (Friedrichs and Perry 2001 ). Because many of the observed transitions were concentrated along vegetation boundaries rather than within stable cores, these transition areas may provide more sensitive signals of early habitat change than interior marsh patches. From a conservation planning perspective, the expansion of P. communis may reflect increasing structural persistence in some sectors, but it should not be interpreted uncritically as a uniformly positive ecological outcome (Kirwan and Murray 2007 ; Cui et al. 2019 ). Management attention should instead consider where reed expansion occurs, particularly whether it encroaches into halophytic transition habitats and coincides with the reduction of S. japonica patches (Cui et al. 2019 ; Fagherazzi et al. 2020 ). Conversely, the decline of S. japonica highlights the vulnerability of low-lying halophytic zones, which may serve as priority monitoring areas and potential early-warning habitats under hydrological or geomorphic stress (Runion et al. 2025 ). Accordingly, the practical value of this study lies in supporting the identification of vulnerable transition sectors, tracking habitat homogenization, and guiding where field-based investigation should be prioritized. Future research should integrate multi-season remote sensing with targeted field measurements, including salinity, micro-topography, sediment dynamics, and hydrological connectivity, in order to better identify threshold conditions associated with rapid vegetation turnover (Wang et al. 2020 ; Wang et al. 2021 ). Such an approach would improve the ecological interpretation of remotely sensed transitions and strengthen the basis for adaptive conservation and management in Suncheon Bay and other East Asian coastal wetlands. Conclusion This study developed a multi-year (2019–2024), multi-season Sentinel-2 framework to monitor the reorganization of a Phragmites communis – Suaeda japonica –tidal-flat mosaic in the Suncheon Bay salt marsh. Among the tested classifiers, Random Forest showed the most reliable performance (overall accuracy = 85.7%, κ = 0.79). Seasonal variable contributions further showed that NDVI provided a consistent baseline for vegetation discrimination, while MNDWI was more informative in spring and SSVI was more informative in autumn, reflecting seasonal differences in hydrological and phenological sensitivity (Li et al. 2015 ; Ying et al. 2019 ; Duan et al. 2025 ). The main ecological finding was that wetland change was expressed less through simple net area shifts than through repeated vegetation transitions concentrated along boundary zones. The largest reorganization occurred during 2019–2022, when 9.82 ha (21.2%) underwent class transitions and S. japonica declined sharply from 8.91 ha to 4.01 ha (− 55.0%). Although overall stability increased in subsequent periods, residual turnover continued in the same transition sectors, indicating that vegetation boundaries remained the most dynamic parts of the mosaic. These results suggest that conservation and management in Suncheon Bay should move beyond tracking class extent alone and instead prioritize boundary transition zones, where habitat heterogeneity is most likely to weaken or be reconfigured. In this sense, simultaneous monitoring of P. communis and S. japonica provides a practical basis for identifying conservation-relevant priority areas and supporting adaptive management in coastal salt marshes (Liu et al. 2026 ). This study has some limitations, including the use of four target years and 10 m Sentinel-2 imagery, which may not fully resolve longer-term variability or fine-scale boundary dynamics. Moreover, the environmental comparison was exploratory and should be interpreted as indicating associations rather than causal mechanisms. Declarations Funding Declaration This work was supported by a grant from the Korea Environment Industry & Technology Institute (KEITI) through Creation Restoration and Management Technology of Carbon Accumulated Abandoned Paddy Wetland Project, funded by the Korea Ministry of Environment (MOE) (2022003630004). This work is financially supported by Korea Ministry of Land, Infrastructure and Transport(MOLIT) as 「Innovative Talent Education Program for Smart City」 Author Contributions JKGS Conceptualization; study design; data curation; formal analysis; visualization; interpretation of results; writing—original draft; writing, editing. KYH Field data acquisition; UAV-based multispectral data collection; contribution to study conceptualization. LSH Support in QGIS-based spatial analysis; remote sensing data interpretation; technical guidance. YJW Validation of classification outputs; assistance with reference dataset construction and accuracy assessment. KDY Ecological interpretation of Suaeda japonica dynamics; guidance on halophyte physiological responses. SYK (Corresponding Author) Supervision; project administration; methodological guidance; critical revision of the manuscript; oversight of research direction; responsibility for correspondence and manuscript submission. Ethics and Consent to Participate declarations: not applicable. References Bai Q et al (2025) Vegetation dynamics induced by climate change and human activities: Implications for coastal wetland restoration. J Environ Manage 384:125594 Berhane TM et al (2018) Decision-tree, rule-based, and random forest classification of high-resolution multispectral imagery for wetland mapping and inventory. Remote Sens 10(4):580 Byaruhanga A, Kigoolo S (2005) Information Sheet on Ramsar Wetlands (RIS). Lutembe Bay Wetland System Ramsar Information Sheet on Ramsar Wetlands Chambers RM et al (1999) Expansion of Phragmites australis into tidal wetlands of North America. Aquat Bot 64(3–4):261–273 Chang M et al (2025) Time series satellite remote sensing reveals annual distribution, composition, and trajectory of tidal wetlands in the Yellow River Delta. Estuarine. Coastal Shelf Sci 319:109264 Chaudhary DR et al (2023) Soil microbial community composition along chronosequence of the introduction of Phragmites australis at Suncheon Bay. Estuarine. Coastal Shelf Sci 290:108405 Chung J et al (2021) Spatial distribution of halophytes and environment factors in salt marshes along the eastern Yellow Sea. J Ecol Environ 45(1):28 Chung SH (2018) Features and functions of purple pigment compound in halophytic plant Suaeda japonica: Antioxidant/anticancer activities and osmolyte function in halotolerance. Korean J Plant Resour 31(4):342–354 Clemente KJE et al (2023) The vulnerability and resilience of seagrass ecosystems to marine heatwaves in New Zealand: a remote sensing analysis of seascape metrics using PlanetScope imagery. Remote Sens Ecol Conserv 9(6):803–819 Cui L et al (2019) Phragmites australis meets Suaeda salsa on the red beach: Effects of an ecosystem engineer on salt-marsh litter decomposition. Sci Total Environ 693:133477 Dronova I, Taddeo S (2022) Remote sensing of phenology: Towards the comprehensive indicators of plant community dynamics from species to regional scales. J Ecol 110(7):1460–1484 Duan R et al (2025) A novel ensemble model for multi-temporal forest vegetation classification: integrating spectral-temporal features and topographic constraints. Big Earth Data 9(4):938–964 Fagherazzi S et al (2020) Salt marsh dynamics in a period of accelerated sea level rise. Journal of Geophysical Research: Earth Surface 125(8): e2019JF005200 Flowers TJ, Colmer TD (2008) Salinity tolerance in halophytes. New phytologist: 945–963 Friedrichs CT, Perry JE (2001) Tidal salt marsh morphodynamics: a synthesis. J Coastal Res : 7–37 Fu B et al (2022) Spatio-temporal monitoring of marsh vegetation phenology and its response to hydro-meteorological factors using CCDC algorithm with optical and SAR images: In case of Honghe National Nature Reserve. China Sci Total Environ 843:156990 Gitay H et al (2011) A framework for assessing the vulnerability of wetlands to climate change, Secretariat of the Ramsar Convention Hong SH et al (2015) The 2009-based detailed distribution pattern and area of Phragmites communis-dominant and Suaeda japonica-dominant communities on the Suncheon-bay and Beolgyo estuarine wetlands. J Wetlands Res 17(1):26–37 Huang W et al (2025) A study on the classification of coastal wetland vegetation based on the Suaeda salsa index and its phenological characteristics. Ecol Ind 170:113021 Huang Y et al (2024) Habitat quality evaluation and pattern simulation of coastal salt marsh wetlands. Sci Total Environ 945:174003 Ishtiaq KS, Abdul-Aziz OI (2020) Ecological parameter reductions, environmental regimes, and characteristic process diagram of carbon dioxide fluxes in coastal salt marshes. Sci Rep 10(1):15732 Jamali A et al (2021) A Comparison of Tree-Based Algorithms for Complex Wetland Classification Using the Google Earth Engine. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XLVI-4/W5-2021: 313–319 Jiao L et al (2020) Fine-scale distribution patterns of phragmites australis populations across an environmental gradient in the salt marsh wetland of dunhuang. China Sustain 12(4):1671 Ke Y et al (2024) Suaeda salsa spectral index for Suaeda salsa mapping and fractional cover estimation in intertidal wetlands. ISPRS J Photogrammetry Remote Sens 207:104–121 Kirwan ML, Murray AB (2007) A coupled geomorphic and ecological model of tidal marsh evolution. Proceedings of the National Academy of Sciences 104(15): 6118–6122 Kong B-w et al (2014) Distribution of Halophytes at Coastal Wetland in Suncheon Bay. J Korean Soc Environ Technol 15(2):130–139 Lee EI, Nepf H (2024) Marsh restoration in front of seawalls is an economically justified nature-based solution for coastal protection. Commun Earth Environ 5(1):605 Lee M (2019) Distribution Patterns of Halophytes in the Coastal Area in Korea. The sea: the journal of the Korean society of oceanography 24(1) Lee Y-K, Estuarine et al (2014) Coastal Shelf Sci 151: 347–354 Li L et al (2015) Evaluation of MODIS spectral indices for monitoring hydrological dynamics of a small. seasonally-flooded wetland South Spain Wetlands 35(5):851–864 Liu B et al (2026) Dynamic changes in the sea-land convergence zone: Long-term coastal wetland dynamics, landscape pattern evolution, and human-activity impacts in Zhejiang. Ecol Ind 182:114583 Liu X et al (2024) Monitoring Wetland Changes and Analyzing the Spartina Alterniflora Invasion in the Yellow River Delta Over the Past 30 Years Based on Google Earth Engine. IEEE Geosci Remote Sens Lett 21:1–5 Lu X et al (2020) Monitoring suaeda salsa spectral response to salt conditions in coastal wetlands: A case study in dafeng elk national nature reserve. China Remote Sens 12(17):2700 Mahdavi S et al (2018) Remote sensing for wetland classification: A comprehensive review. GIScience remote Sens 55(5):623–658 Marchesini VA et al (2016) Spectral detection of stress-related pigments in salt-lake succulent halophytic shrubs. Int J Appl Earth Obs Geoinf 52:457–463 Moore GE et al (2012) Belowground biomass of Phragmites australis in coastal marshes. Northeastern Naturalist 19(4):611–626 Morris JT et al (2002) Responses of coastal wetlands to rising sea level. Ecology 83(10):2869–2877 Mutanga O et al (2012) High density biomass estimation for wetland vegetation using WorldView-2 imagery and random forest regression algorithm. Int J Appl Earth Obs Geoinf 18:399–406 Narron CR et al (2022) Flooding in Landsat across tidal systems (FLATS): An index for intermittent tidal filtering and frequency detection in salt marsh environments. Ecol Ind 141:109045 O'Connell JL et al (2017) The Tidal Marsh Inundation Index (TMII): An inundation filter to flag flooded pixels and improve MODIS tidal marsh vegetation time-series analysis. Remote Sens Environ 201:34–46 Peng H et al (2022) Monitoring spatial and temporal dynamics of wetland vegetation and their response to hydrological conditions in a large seasonal lake with time series Landsat data. Ecol Ind 142:109283 Peng K et al (2023) Continental-scale wetland mapping: A novel algorithm for detailed wetland types classification based on time series Sentinel-1/2 images. Ecol Ind 148:110113 Peter Sheng Y et al (2022) Coastal marshes provide valuable protection for coastal communities from storm-induced wave, flood, and structural loss in a changing climate. Sci Rep 12(1):3051 Pettorelli N et al (2005) Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol Evol 20(9):503–510 Qiu M et al (2024) Spatio-temporal changes and hydrological forces of wetland landscape pattern in the Yellow River Delta during 1986–2022. Landscape Ecol 39(3):51 Rouse JW Jr et al (1973) Monitoring the vernal advancement and retrogradation (green wave effect) of natural vegetation Runion KD et al (2025) Early warning signs of salt marsh drowning indicated by widespread vulnerability from declining belowground plant biomass. Proceedings of the National Academy of Sciences 122(26): e2425501122 Song Z et al (2022) Assessing the ecosystem health of coastal wetland vegetation (Suaeda salsa) using the pressure state response model, a case of the Liao River estuary in China. Int J Environ Res Public Health 19(1):546 Steinigeweg CS et al (2023) Conquering new frontiers: The effect of vegetation establishment and environmental interactions on the expansion of tidal marsh systems. Estuaries Coasts 46(6):1515–1535 Sun C et al (2021) Plant species classification in salt marshes using phenological parameters derived from Sentinel-2 pixel-differential time-series. Remote Sens Environ 256:112320 Sun Z et al (2010) Effects of sediment burial disturbance on seedling survival and growth of Suaeda salsa in the tidal wetland of the Yellow River estuary. Plant Soil 337(1):457–468 Surasinghe TD et al (2025) Leveraging Phenology to Assess Seasonal Variations of Plant Communities for Mapping Dynamic Ecosystems. Remote Sens 17(10):1778 Tian S et al (2016) Random forest classification of wetland landcovers from multi-sensor data in the arid region of Xinjiang. China Remote Sens 8(11):954 Ury EA et al (2024) Managing the Global Wetland Methane-Climate Feedback: A Review of Potential Options. Glob Change Biol 30(11):e17585 Wang C et al (2020) Identifying hydro-geomorphological conditions for state shifts from bare tidal flats to vegetated tidal marshes. Remote Sens 12(14):2316 Wang Q et al (2021) How hydrological connectivity regulates the plant recovery process in salt marshes. J Appl Ecol 58(6):1314–1324 Wang X et al (2022) Genus Suaeda: Advances in phytology, chemistry, pharmacology and clinical application (1895–2021). Pharmacol Res 179:106203 Wetson AM et al (2008) Do conditions during dormancy influence germination of Suaeda maritima? Ann Botany 101(9):1319–1327 Xu H (2006) Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. Int J Remote Sens 27(14):3025–3033 Yi W et al (2024) An enhanced monitoring method for spatio-temporal dynamics of salt marsh vegetation using google earth engine. Estuarine. Coastal Shelf Sci 298:108658 Ying L et al (2019) Development of a Vegetation Index for Suaeda glauca Using GF-1 WFV Imagery. J Wuhan Univ – Inform Sci Ed 44(12):1823–1831 Zeng J et al (2022) A phenology-based vegetation index classification (PVC) algorithm for coastal salt marshes using Landsat 8 images. Int J Appl Earth Obs Geoinf 110:102776 Zhang S et al (2024) Monitoring of chlorophyll content in local saltwort species Suaeda salsa under water and salt stress based on the PROSAIL-D model in coastal wetland. Remote Sens Environ 306:114117 Zhang Y et al (2021) Dynamic response of phragmites australis and Suaeda salsa to climate change in the Liaohe Delta Wetland. J Meteorological Res 35(1):157–171 Hansen BD, Szabo JK, Fuller RA, Clemens RS, Rogers DI, Milton DA (2021) Insights from long-term shorebird monitoring for tracking change in ecological character of Australasian Ramsar sites. Biol Conserv 260:109189 Larkin DJ (2018) Wetland heterogeneity. The Wetland Book: I: Structure and Function, Management, and Methods. Springer Netherlands, pp 177–182 Lopatin J, Araya-López R, Dronova I (2026) Remotely sensed phenology reveals environmental and management controls on coastal wetland plant communities. Ecol Inf, 103610 Yang X, Zhu Z, Qiu S, Kroeger KD, Zhu Z, Covington S (2022) Detection and characterization of coastal tidal wetland change in the northeastern US using Landsat time series. Remote Sens Environ 276:113047 Table 3 and 5 Table 3 and 5 are available in the Supplementary Files section. Additional Declarations No competing interests reported. Supplementary Files Table35.docx GraphicalAbstract.jpg Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {\"props\":{\"pageProps\":{\"initialData\":{\"identity\":\"rs-8731780\",\"acceptedTermsAndConditions\":true,\"allowDirectSubmit\":true,\"archivedVersions\":[],\"articleType\":\"Research Article\",\"associatedPublications\":[],\"authors\":[{\"id\":609069944,\"identity\":\"c5e32a69-a913-4fc6-a03c-cceb90aa0ea1\",\"order_by\":0,\"name\":\"Gapseong Jekal\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Seoul National University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Gapseong\",\"middleName\":\"\",\"lastName\":\"Jekal\",\"suffix\":\"\"},{\"id\":609069945,\"identity\":\"67cd512a-fc9f-4259-8030-d44104b16919\",\"order_by\":1,\"name\":\"Yong Hwan Kim\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Seoul National University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Yong\",\"middleName\":\"Hwan\",\"lastName\":\"Kim\",\"suffix\":\"\"},{\"id\":609069946,\"identity\":\"6708ca9c-e774-4f50-a17a-a949bd8671d6\",\"order_by\":2,\"name\":\"Seung Hyeon Lee\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Seoul National University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Seung\",\"middleName\":\"Hyeon\",\"lastName\":\"Lee\",\"suffix\":\"\"},{\"id\":609069948,\"identity\":\"09c72dc0-b884-49cf-90bb-5f552fa4c20f\",\"order_by\":3,\"name\":\"Ji Weon Yun\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Seoul National University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Ji\",\"middleName\":\"Weon\",\"lastName\":\"Yun\",\"suffix\":\"\"},{\"id\":609069952,\"identity\":\"8441c2be-eecd-4196-872e-103acc072d85\",\"order_by\":4,\"name\":\"Dae Yeol Kim\",\"email\":\"\",\"orcid\":\"\",\"institution\":\"Seoul National University\",\"correspondingAuthor\":false,\"prefix\":\"\",\"firstName\":\"Dae\",\"middleName\":\"Yeol\",\"lastName\":\"Kim\",\"suffix\":\"\"},{\"id\":609069954,\"identity\":\"9abe65e0-c32c-4f0c-856e-ca3fc07f8b90\",\"order_by\":5,\"name\":\"YoungKeun Song\",\"email\":\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA20lEQVRIiWNgGAWjYFAD9gYog4doLTwHSNYikUCkFvn23sOveSoO58lHPn/2mIfBTp6B5+wDvFoMzpxLs+Y5c7jY8HaOuTEPQ7JhA2+7AX4tEjlmxrxthxM3zs5hk+ZhYE5g4Gcj4LAZMC0zjz8DaqknrIXhRo7xY5CW+RIMZkAthxMYeNvw6zA4c8aMcc6Z9MQNPDlmknMMjhu28Rwj4LD2HuMPbyqsE+e3H38m8aaiWp6fJ42AwxgY2KRAMWFwAGwpkEtQAwMD88cfIOsaiFA6CkbBKBgFIxMAAAU2Pc7dwrQxAAAAAElFTkSuQmCC\",\"orcid\":\"\",\"institution\":\"Seoul National University\",\"correspondingAuthor\":true,\"prefix\":\"\",\"firstName\":\"YoungKeun\",\"middleName\":\"\",\"lastName\":\"Song\",\"suffix\":\"\"}],\"badges\":[],\"createdAt\":\"2026-01-29 12:54:40\",\"currentVersionCode\":1,\"declarations\":\"\",\"doi\":\"10.21203/rs.3.rs-8731780/v1\",\"doiUrl\":\"https://doi.org/10.21203/rs.3.rs-8731780/v1\",\"draftVersion\":[],\"editorialEvents\":[],\"editorialNote\":\"\",\"failedWorkflow\":false,\"files\":[{\"id\":105037948,\"identity\":\"558dce30-e162-49d4-a2f6-48d7f49cf771\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:41:00\",\"extension\":\"png\",\"order_by\":1,\"title\":\"Figure 1\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":4045535,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003e(A) Research boundary and vegetation training regions within the Suncheon Bay Wetland; (B) Location of Suncheon Bay in Jeollanam-do, Republic of Korea; (C) High-resolution drone image acquired in November 2024 showing \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e and \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure1.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/265b83476106e05b9678eb39.png\"},{\"id\":105037950,\"identity\":\"01dd2e43-eacc-4ae8-9f6d-332555917ece\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:41:00\",\"extension\":\"png\",\"order_by\":2,\"title\":\"Figure 2\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":58194,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eWorkflow of Vegetation Dynamics Monitoring in Suncheon Bay\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure2.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/8694773722988bce43791e47.png\"},{\"id\":105037917,\"identity\":\"4c3dfa50-433c-4f04-9794-ed282b553812\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:59\",\"extension\":\"png\",\"order_by\":3,\"title\":\"Figure 3\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1125375,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eMulti-Year Seasonal Trends of Vegetation Indices (NDVI, MNDWI, SSVI) for \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e, \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e, and Tidal Flats (March, April, October, November)\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure3.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/df456680ae4fd1eefdf32bc8.png\"},{\"id\":105037987,\"identity\":\"334afc05-5f7d-4fd5-9b11-859b334c7ebd\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:41:14\",\"extension\":\"png\",\"order_by\":4,\"title\":\"Figure 4\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":4901504,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eMulti-year classification results of wetland vegetation (2019, 2022, 2023, 2024) using RF, DT, and GB models\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure4.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/90a4c2051dfeae6a4117dea3.png\"},{\"id\":105037985,\"identity\":\"17789318-96d7-4d59-ae91-45c11fda8361\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:41:14\",\"extension\":\"png\",\"order_by\":5,\"title\":\"Figure 5\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":72922,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eClassification performance comparison: mean overall accuracy and Kappa across classifiers\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure5.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/cbfca4888da27d2cfe24f128.png\"},{\"id\":105037913,\"identity\":\"c4eca17e-fd96-41ba-81b4-3acc4bfaf2e7\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:58\",\"extension\":\"png\",\"order_by\":6,\"title\":\"Figure 6\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":170700,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eVariable Importance in Random Forest for Multi-Temporal Vegetation Classification\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure6.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/ca83b24b371bb7831a9c9b20.png\"},{\"id\":105037916,\"identity\":\"dbbc25c2-6719-48f7-a60c-c3d600e09212\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:58\",\"extension\":\"png\",\"order_by\":7,\"title\":\"Figure 7\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":890799,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eDistribution Changes of \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e and \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e (2019, 2022, 2023, 2024)\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure7.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/ce79c5da85bf31bace5664b2.png\"},{\"id\":105037792,\"identity\":\"0624c3c1-b1a3-49ed-9cf2-c9e53d17a128\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:34\",\"extension\":\"png\",\"order_by\":8,\"title\":\"Figure 8\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1255669,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eSpatio-temporal transitions of vegetation communities (2019–2024)\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure8.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/7b3e492ccc892f5253054448.png\"},{\"id\":105037822,\"identity\":\"d42364a2-ec58-4c15-95e3-16ed5e1606d1\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:42\",\"extension\":\"png\",\"order_by\":9,\"title\":\"Figure 9\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":1646531,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eSankey diagram of spatio-temporal transitions in \\u003cem\\u003ePhragmites communis(PC)\\u003c/em\\u003e, \\u003cem\\u003eSuaeda japonica(SJ)\\u003c/em\\u003e, and Tidal Flat(TF) (2019–2024)\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure9.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/87258f37d0d3fd71b4b71f0a.png\"},{\"id\":105038460,\"identity\":\"ae036af0-fda7-4657-babb-2203c4d41fa3\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:43:45\",\"extension\":\"png\",\"order_by\":10,\"title\":\"Figure 10\",\"display\":\"\",\"copyAsset\":false,\"role\":\"figure\",\"size\":794570,\"visible\":true,\"origin\":\"\",\"legend\":\"\\u003cp\\u003eNormalized trends of \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e and \\u003cem\\u003eSuaeda Japonica\\u003c/em\\u003e area and correlated environmental variables (RMSE ≤ 0.25).\\u003c/p\\u003e\",\"description\":\"\",\"filename\":\"Figure10.png\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/acd61bda418e4fae1fac1167.png\"},{\"id\":105040120,\"identity\":\"4b725162-02bc-45ef-b42c-42d47fcc5691\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:48:17\",\"extension\":\"pdf\",\"order_by\":0,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"manuscript-pdf\",\"size\":16723159,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"manuscript.pdf\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/01a2d99d-6888-4bbb-9628-7bf440967054.pdf\"},{\"id\":105037769,\"identity\":\"61d727fc-4ef3-4c11-97c1-4f72ad2d62e0\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:26\",\"extension\":\"docx\",\"order_by\":1,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":17665,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"Table35.docx\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/9488b92459a35da8e67eb5c9.docx\"},{\"id\":105037788,\"identity\":\"180f8919-98c4-4376-a15c-c8c7270d870a\",\"added_by\":\"auto\",\"created_at\":\"2026-03-20 07:40:34\",\"extension\":\"jpg\",\"order_by\":2,\"title\":\"\",\"display\":\"\",\"copyAsset\":false,\"role\":\"supplement\",\"size\":4387153,\"visible\":true,\"origin\":\"\",\"legend\":\"\",\"description\":\"\",\"filename\":\"GraphicalAbstract.jpg\",\"url\":\"https://assets-eu.researchsquare.com/files/rs-8731780/v1/75ae11eb76692feb0926efa6.jpg\"}],\"financialInterests\":\"No competing interests reported.\",\"formattedTitle\":\"Multi-season Sentinel-2 reveals conservation-relevant transition-zone dynamics in a Phragmites communis–Suaeda japonica coastal wetland mosaic\",\"fulltext\":[{\"header\":\"Introduction\",\"content\":\"\\u003cp\\u003eCoastal wetlands provide critical ecosystem functions, including biodiversity conservation, habitat provision, water purification, and shoreline stabilization(Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). As transitional ecotones linking terrestrial and marine systems, they are regarded as globally important nature-based solutions for climate adaptation and coastal protection(Yi et al. \\u003cspan citationid=\\\"CR60\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). However, climate change, sea-level rise, large-scale reclamation, aquaculture development, and the invasion of alien species such as \\u003cem\\u003eSpartina alterniflora\\u003c/em\\u003e have led to rapid reductions in coastal wetland area and degradation of their ecological functions(Sun et al. \\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e; Bai et al. \\u003cspan citationid=\\\"CR1\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e; Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). These threats are recognized not only at the regional level but also internationally(Ke et al. \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Both the Ramsar Convention and the United Nations Sustainable Development Goals (SDGs) emphasize the necessity of quantitative wetland monitoring(Peng et al. \\u003cspan citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e). In particular, understanding the spatial distribution and interannual dynamics of wetland vegetation provides critical insights into ecological processes and resilience, enabling adaptive and science-based management decisions(Yi et al. \\u003cspan citationid=\\\"CR60\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003e \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e (\\u003cem\\u003eP. communis\\u003c/em\\u003e) and \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e (\\u003cem\\u003eS. japonica\\u003c/em\\u003e) are critical indicator species for monitoring ecological dynamics in East Asian coastal wetlands(Zhang et al. \\u003cspan citationid=\\\"CR64\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). They cover approximately 75.9% of vegetation area in Korean coastal wetlands while in regions such as Yancheng, the Yellow River Delta, and the Liaohe Delta in China, their closely related species \\u003cem\\u003ePhragmites australis\\u003c/em\\u003e and \\u003cem\\u003eSuaeda salsa\\u003c/em\\u003e are dominant(Lee \\u003cspan citationid=\\\"CR28\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e; Sun et al. \\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e; Yi et al. \\u003cspan citationid=\\\"CR60\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). \\u003cem\\u003eP. communis\\u003c/em\\u003e exhibits high biomass, long growth periods, deep rooting systems, and strong photosynthetic capacity, serving as a marker of wetland expansion and structural stability(Chaudhary et al. \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e). Conversely, \\u003cem\\u003eS. japonica\\u003c/em\\u003e is an annual halophyte highly sensitive to climatic and hydrological variability, establishing in saline clay soils and turning purple in autumn due to the accumulation of betacyanin pigments as a physiological stress response(Lee et al. \\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e2014\\u003c/span\\u003e; Chung \\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e2018\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eThis ecological contrast makes paired monitoring of the two species a direct means to track interannual vegetation turnover in coastal wetland mosaics(Gitay et al. \\u003cspan citationid=\\\"CR17\\\" class=\\\"CitationRef\\\"\\u003e2011\\u003c/span\\u003e). Prior studies report relatively persistent expansion of \\u003cem\\u003ePhragmites\\u003c/em\\u003e stands, whereas \\u003cem\\u003eSuaeda\\u003c/em\\u003e distributions can fluctuate more strongly with external conditions, contributing to marked variability in wetland composition (Yi et al. \\u003cspan citationid=\\\"CR60\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e; Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). Accordingly, simultaneous monitoring of these two species provides a practical basis for tracking vegetation distribution and turnover in tidal wetlands(Qiu et al. \\u003cspan citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Long-term remote-sensing analyses in major Yellow Sea deltas show pronounced reorganization of \\u003cem\\u003ePhragmites\\u0026ndash;Suaeda\\u003c/em\\u003e mosaics, where the areal extents of \\u003cem\\u003ePhragmites\\u003c/em\\u003e and Suaeda marshes alternate between decline and recovery over decadal scales, and \\u003cem\\u003eSuaeda\\u003c/em\\u003e typically exhibits larger interannual fluctuations than Phragmites(Chang et al. \\u003cspan citationid=\\\"CR5\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003e \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e, the dominant vegetation in coastal wetlands, exhibit distinct seasonal color transitions: \\u003cem\\u003eP. communis\\u003c/em\\u003e maintains green during the growing season before turning yellow in senescence, whereas \\u003cem\\u003eS. japonica\\u003c/em\\u003e accumulates betacyanin in autumn, shifting from green to purple(Chung \\u003cspan citationid=\\\"CR8\\\" class=\\\"CitationRef\\\"\\u003e2018\\u003c/span\\u003e; Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). However, previous studies often relied primarily on the Normalized Difference Vegetation Index (NDVI), without sufficiently accounting for these species-specific phenological and spectral characteristics(Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). As a result, spectral mixing among \\u003cem\\u003eSuaeda salsa(S. salsa)\\u003c/em\\u003e, green vegetation, and bare ground frequently reduced classification accuracy(Ke et al. \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eTo address this issue, several indices specifically designed for \\u003cem\\u003eS. salsa\\u003c/em\\u003e\\u0026mdash;such as the \\u003cem\\u003eSuaeda salsa\\u003c/em\\u003e Vegetation Index (SSVI)\\u0026mdash;were introduced and achieved partial success(Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). Since \\u003cem\\u003eS. salsa\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e are congeneric halophytes that share similar phenological patterns and pigment-related spectral traits, previous studies on \\u003cem\\u003eS. salsa\\u003c/em\\u003e can reasonably be used as a methodological reference for research on \\u003cem\\u003eS. japonica\\u003c/em\\u003e(Lee et al. \\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e2014\\u003c/span\\u003e; Wang et al. \\u003cspan citationid=\\\"CR57\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e; Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). Nevertheless, even these studies largely relied on single-date or short-term imagery, limiting their ability to capture the strong spatial heterogeneity and dynamic shifts that characterize coastal wetland mosaics (Sun et al. \\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). These gaps highlight the need for multi-temporal, seasonally optimized remote sensing approaches that can detect both interannual variability and spatially complex vegetation dynamics(Clemente et al. \\u003cspan citationid=\\\"CR9\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eBeyond improving classification accuracy, coastal wetland monitoring must support conservation planning and management by identifying where ecological character is being maintained and where it is becoming vulnerable(Hansen et al., \\u003cspan citationid=\\\"CR65\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). In mosaic wetlands such as Suncheon Bay, changes in vegetation boundaries are especially important because they indicate whether habitat heterogeneity is being retained or progressively simplified(Larkin, \\u003cspan citationid=\\\"CR66\\\" class=\\\"CitationRef\\\"\\u003e2018\\u003c/span\\u003e). From a conservation perspective, the critical question is therefore not only how much vegetation area changes, but also where transitions occur, which vegetation type is retreating, and which boundary sectors require priority monitoring or field inspection(Yang et al., \\u003cspan citationid=\\\"CR68\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). A multi-season remote-sensing framework can contribute to this need by detecting spatially explicit transition zones and by providing an operational basis for adaptive monitoring in Ramsar coastal wetlands(Lopatin et al., \\u003cspan citationid=\\\"CR67\\\" class=\\\"CitationRef\\\"\\u003e2026\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eThis study aims to monitor the multi-temporal changes of \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e, two dominant species in East Asian coastal wetlands, by leveraging vegetation indices derived from species-specific phenological stages (March\\u0026ndash;April for the growing season, October\\u0026ndash;November for the senescent stage) across multiple years (2019, 2022, 2023, 2024). As \\u003cem\\u003eP. communis\\u003c/em\\u003e represents stability and \\u003cem\\u003eS. japonica\\u003c/em\\u003e reflects vulnerability, quantifying their spatiotemporal changes and linking them to environmental variables provides critical insights into the resilience and vulnerability of coastal wetland ecosystems(Peter Sheng et al. \\u003cspan citationid=\\\"CR43\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e; Song et al. \\u003cspan citationid=\\\"CR48\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eThe specific objectives of this study are as follows:\\u003c/p\\u003e \\u003cp\\u003e(1) To develop and evaluate a multi-season Sentinel-2 classification framework for monitoring Phragmites communis and Suaeda japonica in a coastal wetland mosaic.\\u003c/p\\u003e \\u003cp\\u003e(2) To quantify multi-year (2019\\u0026ndash;2024) vegetation dynamics through area change, persistence, and inter-class transitions, with emphasis on boundary transition zones.\\u003c/p\\u003e \\u003cp\\u003e(3) To examine temporal associations between vegetation dynamics and hydro-meteorological conditions as a basis for future field-based monitoring.\\u003c/p\\u003e\"},{\"header\":\"Materials and Methods\",\"content\":\"\\u003cdiv id=\\\"Sec3\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.1 Study area\\u003c/h2\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eSuncheon Bay is a representative coastal wetland located in Suncheon, Jeollanam-do, on the southern coast of the Republic of Korea. It is characterized by extensive mudflats formed at the confluence of the Dongcheon and Isacheon rivers (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig1\\\" class=\\\"InternalRef\\\"\\u003e1\\u003c/span\\u003e)(Hong et al. \\u003cspan citationid=\\\"CR18\\\" class=\\\"CitationRef\\\"\\u003e2015\\u003c/span\\u003e). The total area of the wetland is approximately 27 km\\u0026sup2;, of which about 5.4 km\\u0026sup2; of intertidal wetlands are dominated by \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e communities(Chaudhary et al. \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e). In 2006, Suncheon Bay was designated as the first Ramsar wetland in Korea, in addition to being designated as a national Wetland Protected Area, thereby underscoring its international recognition as a region of high conservation value(Byaruhanga and Kigoolo \\u003cspan citationid=\\\"CR3\\\" class=\\\"CitationRef\\\"\\u003e2005\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eThe \\u003cem\\u003eP. communis\\u003c/em\\u003e community, which accounted for about 0.96 km\\u0026sup2; (9.6% of the surveyed area) in 2011, is the most extensive vegetation type in Suncheon Bay(Kong et al. \\u003cspan citationid=\\\"CR26\\\" class=\\\"CitationRef\\\"\\u003e2014\\u003c/span\\u003e). By contrast, the \\u003cem\\u003eS. japonica\\u003c/em\\u003e community covers a smaller area of approximately 0.166 km\\u0026sup2; (1.6%). \\u003cem\\u003eS. japonica\\u003c/em\\u003e primarily occurs in low-lying salt marshes with moderate seawater exchange and silt deposition(Chung et al. \\u003cspan citationid=\\\"CR7\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). However, as an annual halophyte, it responds sensitively to environmental changes, resulting in high variability in its areal extent(Lee et al. \\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e2014\\u003c/span\\u003e). conservation(Jiao et al. \\u003cspan citationid=\\\"CR23\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e; Chaudhary et al. \\u003cspan citationid=\\\"CR6\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec4\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.2 Materials\\u003c/h2\\u003e \\u003cp\\u003eIn this study, Sentinel-2A satellite imagery was acquired using Google Earth Engine (GEE) (Table\\u0026nbsp;1). The target years were 2019, 2022, 2023, and 2024, with images collected for March, April, October, and November of each year. During preprocessing, any images containing cloud, shadow, or snow within the study boundary were visually inspected and excluded from the analysis.(Liu et al. \\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Considering that Suncheon Bay exhibits a tidal range exceeding 3 m, only images with a mean value of the Modified Normalized Difference Water Index (MNDWI) less than 0 within the study boundary were selected, as this indicates conditions where vegetation cover is dominant over water bodies(Sun et al. \\u003cspan citationid=\\\"CR50\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). In total, 37 Sentinel-2 images were compiled, and high-resolution reference datasets were established to enable temporal validation of vegetation distribution in Suncheon Bay. Multispectral UAV imagery with a spatial resolution of 6 cm per pixel was acquired on November 15, 2024, targeting \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e communities. Additional UAV imagery from September 2019, October 2022, and October 2023, together with Google Earth data and high-resolution drone imagery from 2024, were integrated to construct annual reference maps for accuracy validation (Table\\u0026nbsp;1).\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"No\\\" id=\\\"Taba\\\" border=\\\"1\\\"\\u003e \\u003ccolgroup cols=\\\"1\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTable\\u0026nbsp;1. Remote sensing datasets and reference materials\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"No\\\" id=\\\"Tabb\\\" border=\\\"1\\\"\\u003e \\u003ccolgroup cols=\\\"5\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c5\\\" colnum=\\\"5\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eData source\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eSpecification\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eresolution\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eUsage\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eObservation Dates\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eSentinel-2A\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eBands B2 (Blue), B3 (Green), B4 (Red), B8 (NIR), B11 (SWIR)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e10 m\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eClassification training and monitoring\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003e2019\\u003c/b\\u003e: Mar-01, Mar-16, Apr-15, Apr-18, Oct-20, Nov-06, Nov-09\\u003c/p\\u003e \\u003cp\\u003e\\u003cb\\u003e2022\\u003c/b\\u003e: Mar-15, Apr-04, Apr-24, Oct-14, Oct-16, Oct-19, Oct-31, Nov-03, Nov-05, Nov-13, Nov-18\\u003c/p\\u003e \\u003cp\\u003e\\u003cb\\u003e2023\\u003c/b\\u003e: Mar-03, Mar-13, Apr-02, Apr-27, Oct-11, Oct-21, Nov-08, Nov-13, Nov-20, Nov-25\\u003c/p\\u003e \\u003cp\\u003e\\u003cb\\u003e2024\\u003c/b\\u003e: Mar-02, Apr-01, Apr-13, Apr-18, Oct-08, Oct-10, Nov-07, Nov-12, Nov-22\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eMultispectral UAV imagery\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eRed, Green, Blue, Red edge, NIR\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e6 cm/pixel\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eAccuracy validation\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eNov-15, 2024\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003e\\u003cb\\u003eReference data\\u003c/b\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eUAV imagery, Google Earth\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e-\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eAccuracy validation\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c5\\\"\\u003e \\u003cp\\u003eSep-2019, Oct-2022, Oct-2023\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec5\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.3 Training Area and Vegetation Index Construction\\u003c/h2\\u003e \\u003cp\\u003eThe training areas were established based on prior research and field survey results, targeting regions where vegetation was consistently maintained across the years 2019, 2022, 2023, and 2024 (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig2\\\" class=\\\"InternalRef\\\"\\u003e2\\u003c/span\\u003e). Training areas were selected for \\u003cem\\u003eP. communis\\u003c/em\\u003e, \\u003cem\\u003eS. japonica\\u003c/em\\u003e, and tidal flats, and these identical training areas were consistently applied to Sentinel-2 imagery from each of the four years. NDVI, MNDWI, and SSVI were derived from Sentinel-2 images acquired in March, April, October, and November of 2019, 2022, 2023, and 2024, respectively. Values within each training polygon at each time point were extracted (Table\\u0026nbsp;2). Outliers were removed using a threshold of 1.5 times the interquartile range (IQR), and mean values of NDVI, MNDWI, and SSVI were calculated for each vegetation type and season(Zhang et al. \\u003cspan citationid=\\\"CR64\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e; Qiu et al. \\u003cspan citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). An integrated training dataset was constructed to comprehensively capture seasonal and interannual spectral variability, incorporating all samples from the four years(Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig3\\\" class=\\\"InternalRef\\\"\\u003e3\\u003c/span\\u003e, Table\\u0026nbsp;3).\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"No\\\" id=\\\"Tabc\\\" border=\\\"1\\\"\\u003e \\u003ccolgroup cols=\\\"1\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTable\\u0026nbsp;2. Vegetation indices (NDVI, MNDWI, SSVI) used for wetland vegetation classification\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"No\\\" id=\\\"Tabd\\\" border=\\\"1\\\"\\u003e \\u003ccolgroup cols=\\\"3\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eVegetation Index\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eFormula\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eReference\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eNDVI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e(NIR - RED)\\u003c/p\\u003e \\u003cp\\u003e/ (NIR\\u0026thinsp;+\\u0026thinsp;RED)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e(Rouse Jr, Haas et al. 1973)\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMNDWI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e(GREEN - SWIR)\\u003c/p\\u003e \\u003cp\\u003e/ (GREEN\\u0026thinsp;+\\u0026thinsp;SWIR)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e(Xu \\u003cspan citationid=\\\"CR59\\\" class=\\\"CitationRef\\\"\\u003e2006\\u003c/span\\u003e)\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eSSVI\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eNDVI \\u0026times; ((BLUE - GREEN)\\u003c/p\\u003e \\u003cp\\u003e/ (GREEN - RED))\\u003csup\\u003e2\\u003c/sup\\u003e\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003e(Ying et al. \\u003cspan citationid=\\\"CR61\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e)\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec6\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.4 Classification and Validation\\u003c/h2\\u003e \\u003cp\\u003eWetland and coastal vegetation mapping poses significant challenges due to spectral heterogeneity, tidal inundation variability, and the coexistence of mixed land-cover types, requiring classification methods that are both robust and interpretable(Mahdavi et al. \\u003cspan citationid=\\\"CR34\\\" class=\\\"CitationRef\\\"\\u003e2018\\u003c/span\\u003e). Tree-based algorithms are particularly suitable for such environments because they can model nonlinear relationships and maintain stability in the presence of noisy or correlated input variables(Berhane et al. \\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2018\\u003c/span\\u003e). Among these, Decision Tree (DT), Random Forest (RF), and Gradient Boosting (GB) have consistently demonstrated strong performance in vegetation classification tasks(Berhane et al. \\u003cspan citationid=\\\"CR2\\\" class=\\\"CitationRef\\\"\\u003e2018\\u003c/span\\u003e). Comparing these classifiers allows for an integrated evaluation of the trade-offs between interpretability (DT), robustness to data variability (RF), and predictive capability (GB), providing a balanced perspective on model transparency and accuracy(Jamali et al. \\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). Therefore, comparing DT, RF, and GB offers a methodologically sound and ecologically interpretable approach for monitoring long-term vegetation dynamics in complex coastal wetlands(Jamali et al. \\u003cspan citationid=\\\"CR22\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eAll classifiers were trained on the same dataset and applied separately to four study years (2019, 2022, 2023, and 2024), resulting in a total of 12 classification maps (4 years \\u0026times; 3 classifiers). In addition, RF feature importance was derived to quantitatively assess the relative contributions of NDVI, MNDWI, and SSVI to seasonal and interannual classification (Table\\u0026nbsp;2). Accuracy assessment was conducted for each year using field surveys, UAV imagery, and satellite data(Peng et al. \\u003cspan citationid=\\\"CR42\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e). Specifically, field survey data and UAV imagery acquired in November 2024, UAV imagery from September 2019, October 2022, and October 2023, as well as Google Earth imagery, were used as reference data. For validation, 600 random points (200 per class) were generated, from which confusion matrices were constructed(Liu et al. \\u003cspan citationid=\\\"CR32\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). From these, Overall Accuracy (OA), Kappa coefficient, Producer\\u0026rsquo;s Accuracy (PA), and User\\u0026rsquo;s Accuracy (UA) were calculated for each year(Huang et al. \\u003cspan citationid=\\\"CR19\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec7\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e2.5 Spatiotemporal Change Analysis and Exploratory Environmental Pattern Similarity\\u003c/h2\\u003e \\u003cp\\u003eBased on a comparison of the RF, DT, and GB algorithms, the classifier with the highest accuracy was selected to quantify the spatiotemporal changes in \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e communities from 2019, 2022, 2023, 2024. Subsequently, the interannual variations in the areas of \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e were analyzed in relation to climatic and environmental variables. For this purpose, data were collected on precipitation, temperature (mean, maximum, minimum), humidity, tidal level (maximum, minimum, mean), wind speed (mean, maximum), sunshine duration, number of monsoon days, and number of heatwave days(Fu et al. \\u003cspan citationid=\\\"CR16\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). All variables were normalized to a range of 0\\u0026ndash;1 to enable direct comparison across years(Table\\u0026nbsp;4).\\u003c/p\\u003e \\u003cp\\u003eThe analysis was conducted by calculating RMSE between normalized vegetation area changes (2019, 2022, 2023, 2024) and normalized environmental variables. Lower RMSE values indicated higher similarity in trends, and variables with RMSE\\u0026thinsp;\\u0026le;\\u0026thinsp;0.25 were considered as key candidates. Given the limited number of study years (n\\u0026thinsp;=\\u0026thinsp;4), this analysis was designed as an exploratory approach, focusing on identifying potential trend similarities rather than verifying causal relationships.\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"No\\\" id=\\\"Tabe\\\" border=\\\"1\\\"\\u003e \\u003ccolgroup cols=\\\"1\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTable\\u0026nbsp;4. Environmental Variables for Vegetation Change Analysis\\u003c/p\\u003e \\u003cp\\u003e \\u003cdiv class=\\\"gridtable\\\"\\u003e\\u003ctable float=\\\"No\\\" id=\\\"Tabf\\\" border=\\\"1\\\"\\u003e \\u003ccolgroup cols=\\\"4\\\"\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c1\\\" colnum=\\\"1\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c2\\\" colnum=\\\"2\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c3\\\" colnum=\\\"3\\\"\\u003e\\u003c/div\\u003e \\u003cdiv align=\\\"left\\\" class=\\\"colspec\\\" colname=\\\"c4\\\" colnum=\\\"4\\\"\\u003e\\u003c/div\\u003e \\u003cthead\\u003e \\u003ctr\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eEnvironmental Variable\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003eUnit\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eSource\\u003c/p\\u003e \\u003c/th\\u003e \\u003cth align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003eRange\\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003ctbody\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eHeatwave days\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003edays\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKorea Meteorological Administration\\u003c/p\\u003e \\u003cp\\u003e(KMA, Suncheon City)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e6\\u0026ndash;10\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eSunshine duration\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003ehr\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e2093.8\\u0026ndash;2254.8\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMean humidity\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e%\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e70.0\\u0026ndash;78.3\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eTotal precipitation\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003emm\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e987.0\\u0026ndash;2139.5\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMean temperature\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e℃\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e13.1\\u0026ndash;13.8\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximum temperature\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e℃\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e19.3\\u0026ndash;19.8\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMinimum temperature\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003e℃\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e7.7\\u0026ndash;8.9\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMonsoon days\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003edays\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e16.4\\u0026ndash;22.2\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximum wind speed\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003em/s\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKMA, Suncheon City\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e9.1\\u0026ndash;11.7\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMean wind speed\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003em/s\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKHOA (Gohung Balpo Station, DT0026)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.7\\u0026ndash;2.0\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximum marine wind speed\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003em/s\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKHOA (Gohung Balpo Station, DT0026)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e9.1\\u0026ndash;11.7\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMaximum tide level\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003em\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKHOA (Gohung Balpo Station, DT0026)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e4.18\\u0026ndash;4.50\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMinimum tide level\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003em\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKHOA (Gohung Balpo Station, DT0026)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e-0.64 \\u0026ndash; -0.24\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003ctr\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c1\\\"\\u003e \\u003cp\\u003eMean tide level\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c2\\\"\\u003e \\u003cp\\u003em\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c3\\\"\\u003e \\u003cp\\u003eKHOA (Gohung Balpo Station, DT0026)\\u003c/p\\u003e \\u003c/td\\u003e \\u003ctd align=\\\"left\\\" colname=\\\"c4\\\"\\u003e \\u003cp\\u003e1.929 \\u0026ndash; 1.964\\u003c/p\\u003e \\u003c/td\\u003e \\u003c/tr\\u003e \\u003c/tbody\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/th\\u003e \\u003c/tr\\u003e \\u003c/thead\\u003e \\u003c/colgroup\\u003e \\u003c/table\\u003e\\u003c/div\\u003e \\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"Results\",\"content\":\"\\u003cdiv id=\\\"Sec9\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.1 Performance and Seasonal Sensitivity of Machine Learning Classifiers for Multi-Year Wetland Vegetation Mapping\\u003c/h2\\u003e \\u003cp\\u003eUsing multi-season vegetation indices (NDVI, MNDWI, and SSVI), Random Forest (RF) consistently outperformed Gradient Boosting (GB) and Decision Tree (DT) (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e). RF achieved the highest mean overall accuracy (OA\\u0026thinsp;=\\u0026thinsp;85.7%) and Kappa (κ\\u0026thinsp;=\\u0026thinsp;0.79), compared with GB (OA\\u0026thinsp;=\\u0026thinsp;79.1%, κ\\u0026thinsp;=\\u0026thinsp;0.68) and DT (OA\\u0026thinsp;=\\u0026thinsp;69.2%, κ\\u0026thinsp;=\\u0026thinsp;0.53), indicating that RF provides robust performance for multi-year coastal wetland vegetation mapping. (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig4\\\" class=\\\"InternalRef\\\"\\u003e4\\u003c/span\\u003e, Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig5\\\" class=\\\"InternalRef\\\"\\u003e5\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eRF variable importance further revealed phenology-driven seasonal shifts in the diagnostic value of indices(Tian et al. \\u003cspan citationid=\\\"CR53\\\" class=\\\"CitationRef\\\"\\u003e2016\\u003c/span\\u003e) (Fig.\\u0026nbsp;\\u003cspan refid=\\\"Fig6\\\" class=\\\"InternalRef\\\"\\u003e6\\u003c/span\\u003e). NDVI remained the most influential predictor across all seasons (0.46 in March; 0.37 in April; 0.38 in October; 0.36 in November), supporting its role as a stable baseline indicator of vegetation greenness(Pettorelli et al. \\u003cspan citationid=\\\"CR44\\\" class=\\\"CitationRef\\\"\\u003e2005\\u003c/span\\u003e). In contrast, MNDWI was strongly spring-weighted (0.39\\u0026ndash;0.41 in March\\u0026ndash;April) but declined in autumn (0.27\\u0026ndash;0.30), suggesting that early-season wetness/hydrological signals contribute disproportionately to class separability in tidal wetlands(Narron et al. \\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). This seasonal contrast is plausibly driven by the combination of sparse early-season canopy conditions and strong tidal moisture/background effects in spring, whereas denser late-season vegetation reduces the relative separability provided by wetness signals(Surasinghe et al. \\u003cspan citationid=\\\"CR52\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003e \\u003c/p\\u003e \\u003cp\\u003eConversely, SSVI showed an autumn-weighted pattern, contributing minimally in spring (0.15\\u0026ndash;0.22) but increasing sharply in October\\u0026ndash;November (0.34\\u0026ndash;0.35), consistent with the seasonally expressed pigment dynamics of \\u003cem\\u003eS. japonica\\u003c/em\\u003e (Ke et al. \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). Together, these patterns demonstrate that while NDVI provides consistent discrimination, season-specific indices (MNDWI in spring; SSVI in autumn) add critical, phenology-relevant separability, supporting seasonally tailored index combinations for improved classification(Zeng et al. \\u003cspan citationid=\\\"CR62\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec10\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e3.2 Dynamics in the Multi-Seasonal Distribution of Phragmites communis and Suaeda japonica\\u003c/h2\\u003e \\u003cp\\u003eAcross the years (2019\\u0026ndash;2024), Suncheon Bay showed a directional reorganization of vegetation\\u0026ndash;bare ground mosaics, characterized by a long-term expansion of \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e and a sharp contraction with only partial recovery of \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e, accompanied by marked fluctuations in tidal-flat extent (Fig.\\u0026nbsp;7; Table\\u0026nbsp;5). \\u003cem\\u003eS. japonica\\u003c/em\\u003e declined from 8.91 ha (2019) to 4.01 ha (2022; \\u0026minus;55.0%), with modest recovery in 2023\\u0026ndash;2024 (4.55\\u0026ndash;4.74 ha) but remaining 46.8% below 2019. In contrast, \\u003cem\\u003eP. communis\\u003c/em\\u003e increased to 19.77 ha in 2024 (+\\u0026thinsp;15.3% relative to 2019) after a temporary decline in 2022, while tidal flats expanded in 2022 (+\\u0026thinsp;29.3% relative to 2019) and then contracted by 2024.\\u003c/p\\u003e \\u003cp\\u003eVegetation type transitions in Suncheon Bay were consistently concentrated along vegetation boundaries (interfaces among \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e, \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e, and tidal flats), rather than within vegetation cores (Figs. \\u003cspan refid=\\\"Fig7\\\" class=\\\"InternalRef\\\"\\u003e8\\u003c/span\\u003e\\u0026ndash;\\u003cspan refid=\\\"Fig8\\\" class=\\\"InternalRef\\\"\\u003e9\\u003c/span\\u003e). In particular, core areas dominated by \\u003cem\\u003eP. communis\\u003c/em\\u003e remained largely stable across the study period. The largest reorganization occurred in 2019\\u0026rarr;2022, when 9.82 ha (21.2%) underwent vegetation type transitions, dominated by loss of \\u003cem\\u003eS. japonica\\u003c/em\\u003e to tidal flats (5.05 ha) and transitions from \\u003cem\\u003eP. communis\\u003c/em\\u003e to tidal flats (4.25 ha), indicating pronounced boundary restructuring associated with \\u003cem\\u003eS. japonica\\u003c/em\\u003e decline. Stability increased thereafter (85.1% unchanged in 2022\\u0026rarr;2023), with limited boundary transitions persisting (e.g., \\u003cem\\u003eS. japonica\\u003c/em\\u003e\\u0026rarr;tidal flat 2.46 ha; tidal flat\\u0026rarr;\\u003cem\\u003eP. 2.28\\u003c/em\\u003e ha). In 2023\\u0026rarr;2024, 81.1% of the area remained unchanged, and residual transitions were again confined to the same boundary zones(Lee et al. \\u003cspan citationid=\\\"CR29\\\" class=\\\"CitationRef\\\"\\u003e2014\\u003c/span\\u003e). Overall, vegetation type transitions were boundary-driven, with the most substantial change occurring during 2019\\u0026ndash;2022 and linked to \\u003cem\\u003eS. japonica\\u003c/em\\u003e decline(Peng et al. \\u003cspan citationid=\\\"CR41\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e).\\u003c/p\\u003e\\n\\u003cdiv id=\\\"Sec11\\\" class=\\\"Section2\\\"\\u003e\\n \\u003ch2\\u003e3.3 Environmental Associations with Species-Specific Vegetation Shifts in a Coastal Wetland Ecosystem\\u003c/h2\\u003e\\n \\u003cp\\u003eAll environmental variables were normalized to a 0\\u0026ndash;1 scale for interannual comparability (Fig. \\u003cspan refid=\\\"Fig9\\\" class=\\\"InternalRef\\\"\\u003e10\\u003c/span\\u003e). The area of \\u003cem\\u003eP. communis\\u003c/em\\u003e showed strong trend similarity with humidity and temperature conditions, with RMSE values below 0.25 for mean humidity (0.10), minimum temperature (0.15), mean tide level (0.18), mean temperature (0.23), and total precipitation (0.23). These results indicate that the interannual trajectory of \\u003cem\\u003eP. communis\\u003c/em\\u003e was temporally similar to variation in moisture- and inundation-related environmental conditions. This pattern is broadly consistent with previous studies showing that reed-dominated wetland vegetation responds sensitively to hydrological and salinity-related conditions under favorable climatic regimes (Qiu et al. \\u003cspan citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e).\\u003c/p\\u003e\\n \\u003cp\\u003eIn contrast, \\u003cem\\u003eS. japonica\\u003c/em\\u003e showed relatively low RMSE values for total precipitation (0.20) and mean humidity (0.23), suggesting that its interannual decline and partial recovery were temporally associated with hydro-meteorological fluctuation. Previous studies have reported that reduced rainfall can increase soil salinity and suppress halophytic vegetation cover, whereas increased precipitation and humidity may alleviate salinity stress and improve moisture availability for germination and growth (Zhang et al. \\u003cspan citationid=\\\"CR64\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e; Ke et al. \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). In this context, the observed dynamics of \\u003cem\\u003eS. japonica\\u003c/em\\u003e may reflect sensitivity to shifts in hydro-saline conditions in the coastal wetland environment. However, because this analysis was based on four annual observations and did not include direct field measurements of salinity, sediment deposition, or micro-topography, these results should be interpreted as exploratory environmental associations rather than verified causal drivers.\\u003c/p\\u003e\\n\\u003c/div\\u003e\"},{\"header\":\"Discussion\",\"content\":\"\\u003cdiv id=\\\"Sec13\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.1. Classification Performance and Ecological Implications of Spectral Indices\\u003c/h2\\u003e \\u003cp\\u003eThis study demonstrates that combining multi-temporal Sentinel-2 observations with seasonally optimized vegetation indices substantially improves both classification accuracy and ecological interpretability in coastal wetlands. The Random Forest (RF) model achieved the highest overall accuracy (85.7%) and Kappa coefficient (0.785), confirming its robustness for discriminating vegetation types under the hydrologically dynamic conditions of Suncheon Bay. Seasonal index analysis revealed clear seasonal contrasts in index sensitivity associated with vegetation growth stages and hydrological regimes(Dronova and Taddeo \\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). In spring, NDVI and MNDWI were dominant, reflecting the combined effects of early vegetative growth and strong surface moisture contrasts. During this period, sparse canopy development and frequent tidal inundation amplify the spectral separability associated with water content and wet soil backgrounds, making MNDWI particularly effective at distinguishing halophytic vegetation from tidal flats(Dronova and Taddeo \\u003cspan citationid=\\\"CR11\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). Similar spring-dominant roles of wetness-sensitive indices have been reported in salt-marsh systems with strong tidal influence, where hydrological conditions exert primary control over early-season spectral responses(O\\u0026rsquo;Connell et al. 2017; Narron et al. \\u003cspan citationid=\\\"CR39\\\" class=\\\"CitationRef\\\"\\u003e2022\\u003c/span\\u003e). In contrast, autumn classification was primarily driven by NDVI and SSVI, corresponding to peak biomass conditions and species-specific pigment dynamics.\\u003c/p\\u003e \\u003cp\\u003eWhile NDVI captured overall canopy density and biomass accumulation(Mutanga et al. \\u003cspan citationid=\\\"CR38\\\" class=\\\"CitationRef\\\"\\u003e2012\\u003c/span\\u003e), SSVI provided complementary information by emphasizing red spectral responses associated with betacyanin accumulation in Suaeda species during late-season physiological stress and senescence(Ying et al. \\u003cspan citationid=\\\"CR61\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e; Zhang et al. \\u003cspan citationid=\\\"CR63\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e); this autumn-weighted importance of pigment-sensitive indices is consistent with previous findings for \\u003cem\\u003eSuaeda spp.\\u003c/em\\u003e in saline wetlands(Ying et al. \\u003cspan citationid=\\\"CR61\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e). However, NDVI-only approaches are limited in salt-marsh environments because greenness-based signals may miss pigment-driven physiological changes in halophytes, whereas SSVI isolates red-band signals linked to betacyanin accumulation and improves autumn species-level discrimination(Marchesini et al. \\u003cspan citationid=\\\"CR35\\\" class=\\\"CitationRef\\\"\\u003e2016\\u003c/span\\u003e; Lu et al. \\u003cspan citationid=\\\"CR33\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec14\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.2. Species-Specific Ecological Implications: Structural Stability of \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e and Environmental Vulnerability of \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e\\u003c/h2\\u003e \\u003cp\\u003eThe long-term expansion of \\u003cem\\u003eP. communis\\u003c/em\\u003e and decline of \\u003cem\\u003eS. japonica\\u003c/em\\u003e indicate a directional reorganization of the Suncheon Bay salt-marsh mosaic rather than a simple vegetation gain\\u0026ndash;loss pattern. From 2019 to 2024, \\u003cem\\u003eP. communis\\u003c/em\\u003e increased by 15.3% (from 17.15 to 19.77 ha), whereas \\u003cem\\u003eS. japonica\\u003c/em\\u003e remained 46.8% below its 2019 extent by 2024, despite partial recovery after 2022. This contrast is broadly consistent with regional observations that \\u003cem\\u003ePhragmites\\u003c/em\\u003e-dominated vegetation often shows greater persistence, whereas halophytic \\u003cem\\u003eSuaeda\\u003c/em\\u003e communities fluctuate more strongly under changing coastal wetland conditions (Huang et al. \\u003cspan citationid=\\\"CR20\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). The increase in \\u003cem\\u003eP. communis\\u003c/em\\u003e may reflect its perennial life cycle, clonal expansion capacity, and structural persistence, all of which can contribute to sediment trapping, reduced flow velocity, and organic matter accumulation (Moore et al. \\u003cspan citationid=\\\"CR36\\\" class=\\\"CitationRef\\\"\\u003e2012\\u003c/span\\u003e). In addition, the relatively low RMSE values between \\u003cem\\u003eP. communis\\u003c/em\\u003e area and mean humidity (0.10), minimum temperature (0.15), mean tide level (0.18), mean temperature (0.23), and total precipitation (0.23) suggest temporal similarity with moisture- and inundation-related environmental variation, consistent with previous studies on reed-dominated wetland vegetation (Chambers et al. \\u003cspan citationid=\\\"CR4\\\" class=\\\"CitationRef\\\"\\u003e1999\\u003c/span\\u003e; Qiu et al. \\u003cspan citationid=\\\"CR45\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). However, from a conservation perspective, reed expansion should not be interpreted uncritically as structural change with mixed ecological implications. Although greater \\u003cem\\u003ePhragmites\\u003c/em\\u003e cover may enhance structural stabilization and shoreline buffering, it may also reduce habitat heterogeneity if it progressively replaces more environmentally sensitive halophytic vegetation. Moreover, previous studies suggest that reed-dominated wetlands can involve biogeochemical trade-offs, including increased soil carbon accumulation and potentially elevated methane-related processes (Lee and Nepf \\u003cspan citationid=\\\"CR27\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e; Ury et al. \\u003cspan citationid=\\\"CR54\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eBy contrast, the marked reduction of \\u003cem\\u003eS. japonica\\u003c/em\\u003e suggests greater sensitivity to short-term hydro-meteorological variability and transition-zone instability. \\u003cem\\u003eS. japonica\\u003c/em\\u003e declined from 8.91 ha in 2019 to 4.01 ha in 2022 (\\u0026minus;\\u0026thinsp;55.0%), and recovered only partially to 4.74 ha by 2024, indicating that its distribution remained substantially contracted over the study period. Because this annual halophyte typically occupies lower and more environmentally exposed salt-marsh positions, changes in precipitation, humidity, inundation regime, sedimentation, and salinity balance may disproportionately affect its germination and establishment (Flowers and Colmer \\u003cspan citationid=\\\"CR14\\\" class=\\\"CitationRef\\\"\\u003e2008\\u003c/span\\u003e; Sun et al. \\u003cspan citationid=\\\"CR51\\\" class=\\\"CitationRef\\\"\\u003e2010\\u003c/span\\u003e). This interpretation is supported by relatively low RMSE values for total precipitation (0.20) and mean humidity (0.23), which indicate that \\u003cem\\u003eS. japonica\\u003c/em\\u003e area exhibited temporal similarity with hydro-meteorological fluctuation rather than stable persistence (O\\u0026rsquo;Connell et al. 2017). Previous studies likewise show that reduced rainfall can increase soil salinity and limit halophytic vegetation cover, whereas wetter conditions may alleviate salinity stress and improve establishment conditions (Zhang et al. \\u003cspan citationid=\\\"CR64\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e; Ke et al. \\u003cspan citationid=\\\"CR24\\\" class=\\\"CitationRef\\\"\\u003e2024\\u003c/span\\u003e). In this context, the ecological significance of the observed change lies not only in the decline of a single species, but in the potential weakening of mosaic heterogeneity through contraction of vulnerable halophytic boundary habitats. Thus, the coupled expansion of \\u003cem\\u003eP. communis\\u003c/em\\u003e and retreat of \\u003cem\\u003eS. japonica\\u003c/em\\u003e may be interpreted as a shift toward a more homogeneous marsh structure, with important implications for habitat diversity and conservation-oriented wetland management (Steinigeweg et al. \\u003cspan citationid=\\\"CR49\\\" class=\\\"CitationRef\\\"\\u003e2023\\u003c/span\\u003e).\\u003c/p\\u003e \\u003c/div\\u003e \\u003cdiv id=\\\"Sec15\\\" class=\\\"Section2\\\"\\u003e \\u003ch2\\u003e4.3. Management Implications and Future Directions for Coastal Wetland Conservation\\u003c/h2\\u003e \\u003cp\\u003eInterannual shifts in plant occupancy are a common feature of coastal wetlands, where small changes in elevation, hydroperiod, and sediment dynamics can reorganize vegetation\\u0026ndash;tidal flat mosaics over relatively short timescales (Morris et al. \\u003cspan citationid=\\\"CR37\\\" class=\\\"CitationRef\\\"\\u003e2002\\u003c/span\\u003e). In this context, the contrasting trajectories observed in Suncheon Bay\\u0026mdash;namely, the expansion of \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e and contraction of \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e\\u0026mdash;suggest that conservation monitoring should not focus solely on dominant vegetation extent, but should also track dynamic boundary zones where ecological reorganization is most likely to occur (Friedrichs and Perry \\u003cspan citationid=\\\"CR15\\\" class=\\\"CitationRef\\\"\\u003e2001\\u003c/span\\u003e). Because many of the observed transitions were concentrated along vegetation boundaries rather than within stable cores, these transition areas may provide more sensitive signals of early habitat change than interior marsh patches.\\u003c/p\\u003e \\u003cp\\u003eFrom a conservation planning perspective, the expansion of \\u003cem\\u003eP. communis\\u003c/em\\u003e may reflect increasing structural persistence in some sectors, but it should not be interpreted uncritically as a uniformly positive ecological outcome (Kirwan and Murray \\u003cspan citationid=\\\"CR25\\\" class=\\\"CitationRef\\\"\\u003e2007\\u003c/span\\u003e; Cui et al. \\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e). Management attention should instead consider where reed expansion occurs, particularly whether it encroaches into halophytic transition habitats and coincides with the reduction of \\u003cem\\u003eS. japonica\\u003c/em\\u003e patches (Cui et al. \\u003cspan citationid=\\\"CR10\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e; Fagherazzi et al. \\u003cspan citationid=\\\"CR13\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e). Conversely, the decline of \\u003cem\\u003eS. japonica\\u003c/em\\u003e highlights the vulnerability of low-lying halophytic zones, which may serve as priority monitoring areas and potential early-warning habitats under hydrological or geomorphic stress (Runion et al. \\u003cspan citationid=\\\"CR47\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e). Accordingly, the practical value of this study lies in supporting the identification of vulnerable transition sectors, tracking habitat homogenization, and guiding where field-based investigation should be prioritized.\\u003c/p\\u003e \\u003cp\\u003eFuture research should integrate multi-season remote sensing with targeted field measurements, including salinity, micro-topography, sediment dynamics, and hydrological connectivity, in order to better identify threshold conditions associated with rapid vegetation turnover (Wang et al. \\u003cspan citationid=\\\"CR55\\\" class=\\\"CitationRef\\\"\\u003e2020\\u003c/span\\u003e; Wang et al. \\u003cspan citationid=\\\"CR56\\\" class=\\\"CitationRef\\\"\\u003e2021\\u003c/span\\u003e). Such an approach would improve the ecological interpretation of remotely sensed transitions and strengthen the basis for adaptive conservation and management in Suncheon Bay and other East Asian coastal wetlands.\\u003c/p\\u003e \\u003c/div\\u003e\"},{\"header\":\"Conclusion\",\"content\":\"\\u003cp\\u003eThis study developed a multi-year (2019\\u0026ndash;2024), multi-season Sentinel-2 framework to monitor the reorganization of a \\u003cem\\u003ePhragmites communis\\u003c/em\\u003e\\u0026ndash;\\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e\\u0026ndash;tidal-flat mosaic in the Suncheon Bay salt marsh. Among the tested classifiers, Random Forest showed the most reliable performance (overall accuracy\\u0026thinsp;=\\u0026thinsp;85.7%, κ\\u0026thinsp;=\\u0026thinsp;0.79). Seasonal variable contributions further showed that NDVI provided a consistent baseline for vegetation discrimination, while MNDWI was more informative in spring and SSVI was more informative in autumn, reflecting seasonal differences in hydrological and phenological sensitivity (Li et al. \\u003cspan citationid=\\\"CR30\\\" class=\\\"CitationRef\\\"\\u003e2015\\u003c/span\\u003e; Ying et al. \\u003cspan citationid=\\\"CR61\\\" class=\\\"CitationRef\\\"\\u003e2019\\u003c/span\\u003e; Duan et al. \\u003cspan citationid=\\\"CR12\\\" class=\\\"CitationRef\\\"\\u003e2025\\u003c/span\\u003e).\\u003c/p\\u003e \\u003cp\\u003eThe main ecological finding was that wetland change was expressed less through simple net area shifts than through repeated vegetation transitions concentrated along boundary zones. The largest reorganization occurred during 2019\\u0026ndash;2022, when 9.82 ha (21.2%) underwent class transitions and \\u003cem\\u003eS. japonica\\u003c/em\\u003e declined sharply from 8.91 ha to 4.01 ha (\\u0026minus;\\u0026thinsp;55.0%). Although overall stability increased in subsequent periods, residual turnover continued in the same transition sectors, indicating that vegetation boundaries remained the most dynamic parts of the mosaic. These results suggest that conservation and management in Suncheon Bay should move beyond tracking class extent alone and instead prioritize boundary transition zones, where habitat heterogeneity is most likely to weaken or be reconfigured. In this sense, simultaneous monitoring of \\u003cem\\u003eP. communis\\u003c/em\\u003e and \\u003cem\\u003eS. japonica\\u003c/em\\u003e provides a practical basis for identifying conservation-relevant priority areas and supporting adaptive management in coastal salt marshes (Liu et al. \\u003cspan citationid=\\\"CR31\\\" class=\\\"CitationRef\\\"\\u003e2026\\u003c/span\\u003e). This study has some limitations, including the use of four target years and 10 m Sentinel-2 imagery, which may not fully resolve longer-term variability or fine-scale boundary dynamics. Moreover, the environmental comparison was exploratory and should be interpreted as indicating associations rather than causal mechanisms.\\u003c/p\\u003e\"},{\"header\":\"Declarations\",\"content\":\"\\u003cp\\u003e\\u003cstrong\\u003eFunding Declaration\\u003c/strong\\u003e\\u003c/p\\u003e\\n\\u003cp\\u003eThis work was supported by a grant from the Korea Environment Industry \\u0026amp; Technology Institute (KEITI) through Creation Restoration and Management Technology of Carbon Accumulated Abandoned Paddy Wetland Project, funded by the Korea Ministry of Environment (MOE) (2022003630004).\\u0026nbsp;This work is financially supported by Korea Ministry of Land, Infrastructure and Transport(MOLIT) as\\u0026nbsp;「Innovative Talent Education Program for\\u0026nbsp;Smart\\u0026nbsp;City」\\u003c/p\\u003e\\n\\u003cp\\u003eAuthor Contributions\\u003c/p\\u003e\\n\\u003cp\\u003eJKGS\\u003cbr\\u003e\\u0026nbsp;Conceptualization; study design; data curation; formal analysis; visualization; interpretation of results; writing\\u0026mdash;original draft; writing, editing.\\u003c/p\\u003e\\n\\u003cp\\u003eKYH\\u003cbr\\u003e\\u0026nbsp;Field data acquisition; UAV-based multispectral data collection; contribution to study conceptualization.\\u003c/p\\u003e\\n\\u003cp\\u003eLSH\\u003cbr\\u003e\\u0026nbsp;Support in QGIS-based spatial analysis; remote sensing data interpretation; technical guidance.\\u003c/p\\u003e\\n\\u003cp\\u003eYJW\\u003cbr\\u003e\\u0026nbsp;Validation of classification outputs; assistance with reference dataset construction and accuracy assessment.\\u003c/p\\u003e\\n\\u003cp\\u003eKDY\\u003cbr\\u003eEcological interpretation of \\u003cem\\u003eSuaeda japonica\\u003c/em\\u003e dynamics; guidance on halophyte physiological responses.\\u003c/p\\u003e\\n\\u003cp\\u003eSYK (Corresponding Author)\\u003cbr\\u003e\\u0026nbsp;Supervision; project administration; methodological guidance; critical revision of the manuscript; oversight of research direction; responsibility for correspondence and manuscript submission.\\u003c/p\\u003e\\n\\u003cp\\u003eEthics and Consent to Participate declarations: not applicable.\\u003c/p\\u003e\"},{\"header\":\"References\",\"content\":\"\\u003col\\u003e\\u003cli\\u003e\\u003cspan\\u003eBai Q et al (2025) Vegetation dynamics induced by climate change and human activities: Implications for coastal wetland restoration. J Environ Manage 384:125594\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eBerhane TM et al (2018) Decision-tree, rule-based, and random forest classification of high-resolution multispectral imagery for wetland mapping and inventory. Remote Sens 10(4):580\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eByaruhanga A, Kigoolo S (2005) Information Sheet on Ramsar Wetlands (RIS). Lutembe Bay Wetland System Ramsar Information Sheet on Ramsar Wetlands\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChambers RM et al (1999) Expansion of Phragmites australis into tidal wetlands of North America. Aquat Bot 64(3\\u0026ndash;4):261\\u0026ndash;273\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChang M et al (2025) Time series satellite remote sensing reveals annual distribution, composition, and trajectory of tidal wetlands in the Yellow River Delta. Estuarine. Coastal Shelf Sci 319:109264\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChaudhary DR et al (2023) Soil microbial community composition along chronosequence of the introduction of Phragmites australis at Suncheon Bay. Estuarine. Coastal Shelf Sci 290:108405\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChung J et al (2021) Spatial distribution of halophytes and environment factors in salt marshes along the eastern Yellow Sea. J Ecol Environ 45(1):28\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eChung SH (2018) Features and functions of purple pigment compound in halophytic plant Suaeda japonica: Antioxidant/anticancer activities and osmolyte function in halotolerance. Korean J Plant Resour 31(4):342\\u0026ndash;354\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eClemente KJE et al (2023) The vulnerability and resilience of seagrass ecosystems to marine heatwaves in New Zealand: a remote sensing analysis of seascape metrics using PlanetScope imagery. Remote Sens Ecol Conserv 9(6):803\\u0026ndash;819\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eCui L et al (2019) Phragmites australis meets Suaeda salsa on the red beach: Effects of an ecosystem engineer on salt-marsh litter decomposition. Sci Total Environ 693:133477\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eDronova I, Taddeo S (2022) Remote sensing of phenology: Towards the comprehensive indicators of plant community dynamics from species to regional scales. J Ecol 110(7):1460\\u0026ndash;1484\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eDuan R et al (2025) A novel ensemble model for multi-temporal forest vegetation classification: integrating spectral-temporal features and topographic constraints. Big Earth Data 9(4):938\\u0026ndash;964\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFagherazzi S et al (2020) Salt marsh dynamics in a period of accelerated sea level rise. Journal of Geophysical Research: Earth Surface 125(8): e2019JF005200\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFlowers TJ, Colmer TD (2008) Salinity tolerance in halophytes. New phytologist: 945\\u0026ndash;963\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFriedrichs CT, Perry JE (2001) Tidal salt marsh morphodynamics: a synthesis. J Coastal Res : 7\\u0026ndash;37\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eFu B et al (2022) Spatio-temporal monitoring of marsh vegetation phenology and its response to hydro-meteorological factors using CCDC algorithm with optical and SAR images: In case of Honghe National Nature Reserve. China Sci Total Environ 843:156990\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eGitay H et al (2011) A framework for assessing the vulnerability of wetlands to climate change, Secretariat of the Ramsar Convention\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHong SH et al (2015) The 2009-based detailed distribution pattern and area of Phragmites communis-dominant and Suaeda japonica-dominant communities on the Suncheon-bay and Beolgyo estuarine wetlands. J Wetlands Res 17(1):26\\u0026ndash;37\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHuang W et al (2025) A study on the classification of coastal wetland vegetation based on the Suaeda salsa index and its phenological characteristics. Ecol Ind 170:113021\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHuang Y et al (2024) Habitat quality evaluation and pattern simulation of coastal salt marsh wetlands. Sci Total Environ 945:174003\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eIshtiaq KS, Abdul-Aziz OI (2020) Ecological parameter reductions, environmental regimes, and characteristic process diagram of carbon dioxide fluxes in coastal salt marshes. Sci Rep 10(1):15732\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eJamali A et al (2021) A Comparison of Tree-Based Algorithms for Complex Wetland Classification Using the Google Earth Engine. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences XLVI-4/W5-2021: 313\\u0026ndash;319\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eJiao L et al (2020) Fine-scale distribution patterns of phragmites australis populations across an environmental gradient in the salt marsh wetland of dunhuang. China Sustain 12(4):1671\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKe Y et al (2024) Suaeda salsa spectral index for Suaeda salsa mapping and fractional cover estimation in intertidal wetlands. ISPRS J Photogrammetry Remote Sens 207:104\\u0026ndash;121\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKirwan ML, Murray AB (2007) A coupled geomorphic and ecological model of tidal marsh evolution. Proceedings of the National Academy of Sciences 104(15): 6118\\u0026ndash;6122\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eKong B-w et al (2014) Distribution of Halophytes at Coastal Wetland in Suncheon Bay. J Korean Soc Environ Technol 15(2):130\\u0026ndash;139\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLee EI, Nepf H (2024) Marsh restoration in front of seawalls is an economically justified nature-based solution for coastal protection. Commun Earth Environ 5(1):605\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLee M (2019) Distribution Patterns of Halophytes in the Coastal Area in Korea. The sea: the journal of the Korean society of oceanography 24(1)\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLee Y-K, Estuarine et al (2014) Coastal Shelf Sci 151: 347\\u0026ndash;354\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLi L et al (2015) Evaluation of MODIS spectral indices for monitoring hydrological dynamics of a small. seasonally-flooded wetland South Spain Wetlands 35(5):851\\u0026ndash;864\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLiu B et al (2026) Dynamic changes in the sea-land convergence zone: Long-term coastal wetland dynamics, landscape pattern evolution, and human-activity impacts in Zhejiang. Ecol Ind 182:114583\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLiu X et al (2024) Monitoring Wetland Changes and Analyzing the Spartina Alterniflora Invasion in the Yellow River Delta Over the Past 30 Years Based on Google Earth Engine. IEEE Geosci Remote Sens Lett 21:1\\u0026ndash;5\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLu X et al (2020) Monitoring suaeda salsa spectral response to salt conditions in coastal wetlands: A case study in dafeng elk national nature reserve. China Remote Sens 12(17):2700\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMahdavi S et al (2018) Remote sensing for wetland classification: A comprehensive review. GIScience remote Sens 55(5):623\\u0026ndash;658\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMarchesini VA et al (2016) Spectral detection of stress-related pigments in salt-lake succulent halophytic shrubs. Int J Appl Earth Obs Geoinf 52:457\\u0026ndash;463\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMoore GE et al (2012) Belowground biomass of Phragmites australis in coastal marshes. Northeastern Naturalist 19(4):611\\u0026ndash;626\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMorris JT et al (2002) Responses of coastal wetlands to rising sea level. Ecology 83(10):2869\\u0026ndash;2877\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eMutanga O et al (2012) High density biomass estimation for wetland vegetation using WorldView-2 imagery and random forest regression algorithm. Int J Appl Earth Obs Geoinf 18:399\\u0026ndash;406\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eNarron CR et al (2022) Flooding in Landsat across tidal systems (FLATS): An index for intermittent tidal filtering and frequency detection in salt marsh environments. Ecol Ind 141:109045\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eO'Connell JL et al (2017) The Tidal Marsh Inundation Index (TMII): An inundation filter to flag flooded pixels and improve MODIS tidal marsh vegetation time-series analysis. Remote Sens Environ 201:34\\u0026ndash;46\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003ePeng H et al (2022) Monitoring spatial and temporal dynamics of wetland vegetation and their response to hydrological conditions in a large seasonal lake with time series Landsat data. Ecol Ind 142:109283\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003ePeng K et al (2023) Continental-scale wetland mapping: A novel algorithm for detailed wetland types classification based on time series Sentinel-1/2 images. Ecol Ind 148:110113\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003ePeter Sheng Y et al (2022) Coastal marshes provide valuable protection for coastal communities from storm-induced wave, flood, and structural loss in a changing climate. Sci Rep 12(1):3051\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003ePettorelli N et al (2005) Using the satellite-derived NDVI to assess ecological responses to environmental change. Trends Ecol Evol 20(9):503\\u0026ndash;510\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eQiu M et al (2024) Spatio-temporal changes and hydrological forces of wetland landscape pattern in the Yellow River Delta during 1986\\u0026ndash;2022. Landscape Ecol 39(3):51\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eRouse JW Jr et al (1973) Monitoring the vernal advancement and retrogradation (green wave effect) of natural vegetation\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eRunion KD et al (2025) Early warning signs of salt marsh drowning indicated by widespread vulnerability from declining belowground plant biomass. Proceedings of the National Academy of Sciences 122(26): e2425501122\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSong Z et al (2022) Assessing the ecosystem health of coastal wetland vegetation (Suaeda salsa) using the pressure state response model, a case of the Liao River estuary in China. Int J Environ Res Public Health 19(1):546\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSteinigeweg CS et al (2023) Conquering new frontiers: The effect of vegetation establishment and environmental interactions on the expansion of tidal marsh systems. Estuaries Coasts 46(6):1515\\u0026ndash;1535\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSun C et al (2021) Plant species classification in salt marshes using phenological parameters derived from Sentinel-2 pixel-differential time-series. Remote Sens Environ 256:112320\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSun Z et al (2010) Effects of sediment burial disturbance on seedling survival and growth of Suaeda salsa in the tidal wetland of the Yellow River estuary. Plant Soil 337(1):457\\u0026ndash;468\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eSurasinghe TD et al (2025) Leveraging Phenology to Assess Seasonal Variations of Plant Communities for Mapping Dynamic Ecosystems. Remote Sens 17(10):1778\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eTian S et al (2016) Random forest classification of wetland landcovers from multi-sensor data in the arid region of Xinjiang. China Remote Sens 8(11):954\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eUry EA et al (2024) Managing the Global Wetland Methane-Climate Feedback: A Review of Potential Options. Glob Change Biol 30(11):e17585\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWang C et al (2020) Identifying hydro-geomorphological conditions for state shifts from bare tidal flats to vegetated tidal marshes. Remote Sens 12(14):2316\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWang Q et al (2021) How hydrological connectivity regulates the plant recovery process in salt marshes. J Appl Ecol 58(6):1314\\u0026ndash;1324\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWang X et al (2022) Genus Suaeda: Advances in phytology, chemistry, pharmacology and clinical application (1895\\u0026ndash;2021). Pharmacol Res 179:106203\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eWetson AM et al (2008) Do conditions during dormancy influence germination of Suaeda maritima? Ann Botany 101(9):1319\\u0026ndash;1327\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eXu H (2006) Modification of normalised difference water index (NDWI) to enhance open water features in remotely sensed imagery. Int J Remote Sens 27(14):3025\\u0026ndash;3033\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eYi W et al (2024) An enhanced monitoring method for spatio-temporal dynamics of salt marsh vegetation using google earth engine. Estuarine. Coastal Shelf Sci 298:108658\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eYing L et al (2019) Development of a Vegetation Index for Suaeda glauca Using GF-1 WFV Imagery. J Wuhan Univ \\u0026ndash; Inform Sci Ed 44(12):1823\\u0026ndash;1831\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eZeng J et al (2022) A phenology-based vegetation index classification (PVC) algorithm for coastal salt marshes using Landsat 8 images. Int J Appl Earth Obs Geoinf 110:102776\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eZhang S et al (2024) Monitoring of chlorophyll content in local saltwort species Suaeda salsa under water and salt stress based on the PROSAIL-D model in coastal wetland. Remote Sens Environ 306:114117\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eZhang Y et al (2021) Dynamic response of phragmites australis and Suaeda salsa to climate change in the Liaohe Delta Wetland. J Meteorological Res 35(1):157\\u0026ndash;171\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eHansen BD, Szabo JK, Fuller RA, Clemens RS, Rogers DI, Milton DA (2021) Insights from long-term shorebird monitoring for tracking change in ecological character of Australasian Ramsar sites. Biol Conserv 260:109189\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLarkin DJ (2018) Wetland heterogeneity. The Wetland Book: I: Structure and Function, Management, and Methods. Springer Netherlands, pp 177\\u0026ndash;182\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eLopatin J, Araya-L\\u0026oacute;pez R, Dronova I (2026) Remotely sensed phenology reveals environmental and management controls on coastal wetland plant communities. Ecol Inf, 103610\\u003c/span\\u003e\\u003c/li\\u003e \\u003cli\\u003e\\u003cspan\\u003eYang X, Zhu Z, Qiu S, Kroeger KD, Zhu Z, Covington S (2022) Detection and characterization of coastal tidal wetland change in the northeastern US using Landsat time series. Remote Sens Environ 276:113047\\u003c/span\\u003e\\u003c/li\\u003e\\u003c/ol\\u003e\"},{\"header\":\"Table 3 and 5\",\"content\":\"\\u003cp\\u003eTable 3 and 5 are available in the Supplementary Files section.\\u003c/p\\u003e\"}],\"fulltextSource\":\"\",\"fullText\":\"\",\"funders\":[],\"hasAdminPriorityOnWorkflow\":false,\"hasManuscriptDocX\":true,\"hasOptedInToPreprint\":true,\"hasPassedJournalQc\":\"\",\"hasAnyPriority\":false,\"hideJournal\":true,\"highlight\":\"\",\"institution\":\"\",\"isAcceptedByJournal\":false,\"isAuthorSuppliedPdf\":false,\"isDeskRejected\":\"\",\"isHiddenFromSearch\":false,\"isInQc\":false,\"isInWorkflow\":false,\"isPdf\":false,\"isPdfUpToDate\":true,\"isWithdrawnOrRetracted\":false,\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true},\"keywords\":\"Coastal wetlands, Phragmites communis, Suaeda japonica, Sentinel-2, Random Forest classification\",\"lastPublishedDoi\":\"10.21203/rs.3.rs-8731780/v1\",\"lastPublishedDoiUrl\":\"https://doi.org/10.21203/rs.3.rs-8731780/v1\",\"license\":{\"name\":\"CC BY 4.0\",\"url\":\"https://creativecommons.org/licenses/by/4.0/\"},\"manuscriptAbstract\":\"\\u003cp\\u003eCoastal salt-marsh mosaics reorganize rapidly under hydro-climatic variability, yet many remote-sensing studies rely on single-season analyses, limiting conservation-relevant understanding of boundary transitions. This study developed a multi-season Sentinel-2 framework to monitor interannual dynamics in a Phragmites communis\\u0026ndash;Suaeda japonica\\u0026ndash;tidal-flat mosaic in Suncheon Bay, Korea. We analyzed 37 Sentinel-2A images acquired in spring and autumn from 2019, 2022, 2023, and 2024. NDVI, MNDWI, and SSVI were used to compare Decision Tree, Gradient Boosting, and Random Forest classifiers, and multi-year vegetation change was assessed through area, persistence, and inter-class transitions. Random Forest showed the best performance, with a mean overall accuracy of 85.7% and a kappa coefficient of 0.79. NDVI was consistently informative across seasons, whereas MNDWI contributed more strongly in spring and SSVI in autumn. The largest reorganization occurred during 2019\\u0026ndash;2022, when 9.82 ha (21.2%) underwent vegetation-type transitions. Suaeda japonica declined from 8.91 ha to 4.01 ha and recovered only partially by 2024, while most changes were concentrated along vegetation boundaries rather than stable cores. Multi-season remote sensing can move beyond technical classification by identifying vulnerable halophytic transition zones and priority areas for adaptive monitoring and management. The framework is therefore useful for tracking weakening mosaic heterogeneity in coastal wetlands of high conservation value.\\u003c/p\\u003e\",\"manuscriptTitle\":\"Multi-season Sentinel-2 reveals conservation-relevant transition-zone dynamics in a Phragmites communis–Suaeda japonica coastal wetland mosaic\",\"msid\":\"\",\"msnumber\":\"\",\"nonDraftVersions\":[{\"code\":1,\"date\":\"2026-03-20 07:05:45\",\"doi\":\"10.21203/rs.3.rs-8731780/v1\",\"editorialEvents\":[{\"type\":\"communityComments\",\"content\":0}],\"status\":\"published\",\"journal\":{\"display\":true,\"email\":\"info@researchsquare.com\",\"identity\":\"researchsquare\",\"isNatureJournal\":false,\"hasQc\":true,\"allowDirectSubmit\":true,\"externalIdentity\":\"\",\"sideBox\":\"\",\"snPcode\":\"\",\"submissionUrl\":\"/submission\",\"title\":\"Research Square\",\"twitterHandle\":\"researchsquare\",\"acdcEnabled\":true,\"dfaEnabled\":false,\"editorialSystem\":\"\",\"reportingPortfolio\":\"\",\"inReviewEnabled\":false,\"inReviewRevisionsEnabled\":true}}],\"origin\":\"\",\"ownerIdentity\":\"2b23cc58-d9ba-42c3-a6f1-44dfe2cf217f\",\"owner\":[],\"postedDate\":\"March 20th, 2026\",\"published\":true,\"recentEditorialEvents\":[],\"rejectedJournal\":[],\"revision\":\"\",\"amendment\":\"\",\"status\":\"posted\",\"subjectAreas\":[],\"tags\":[],\"updatedAt\":\"2026-03-20T07:05:45+00:00\",\"versionOfRecord\":[],\"versionCreatedAt\":\"2026-03-20 07:05:45\",\"video\":\"\",\"vorDoi\":\"\",\"vorDoiUrl\":\"\",\"workflowStages\":[]},\"version\":\"v1\",\"identity\":\"rs-8731780\",\"journalConfig\":\"researchsquare\"},\"__N_SSP\":true},\"page\":\"/article/[identity]/[[...version]]\",\"query\":{\"redirect\":\"/article/rs-8731780\",\"identity\":\"rs-8731780\",\"version\":[\"v1\"]},\"buildId\":\"XKTyCvWXoU3ODBz1xrDgd\",\"isFallback\":false,\"isExperimentalCompile\":false,\"dynamicIds\":[84888],\"gssp\":true,\"scriptLoader\":[]}","source_license":"CC-BY-4.0","license_restricted":false}