Using Machine Learning and Hyperspectral Satellite to Monitor Suspended Sediment Concentration under Dredging Engineering

Using Machine Learning and Hyperspectral Satellite to Monitor Suspended Sediment Concentration under Dredging Engineering | 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 Using Machine Learning and Hyperspectral Satellite to Monitor Suspended Sediment Concentration under Dredging Engineering Pengfei Wang, Xu Chang, Bo Chen, Jianhu Fan, Qingjun Hao This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8033000/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 To address the threat of significantly elevated suspended sediment concentration (SSC) and its diffusion to coastal ecosystems caused by intense seabed sediment disturbance during large cross-sea transportation infrastructure construction. Using GF-5B hyperspectral satellite images and synchronous in-situ data, we developed SSC remote sensing inversion models for dredging-disturbed environments, including Random Forest (RF), Support Vector Machine (SVM), Linear Regression (LR) and empirical models. By comparing these models, we found that the RF model outperformed other models, with the RF model performing best (test set R 2 = 0.773, RMSE = 0.062 kg/m 3 ). By using the RF model to invert SSC from satellite data, we found that there was a high SSC in the dredging area (0.2-0.4684 kg/m 3 ) and a low level in surrounding areas (< 0.15 kg/m 3 ), with obvious spatial differences. This confirms that GF-5B hyperspectral data combined with machine learning enables high-precision SSC monitoring in dredging areas, providing a scientific basis for marine engineering ecological risk early warning and environmental protection optimization. satellite remote sensing marine construction random forest environmental monitoring Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Highlights Remote sensing inversion models for SSC in dredging-disturbed environments has been developed. RF model outperformed other models High-precision SSC monitoring in dredging areas has been achieved. The SSC in dredging areas significantly changes with the construction environment. 1 Introduction Large-scale transportation infrastructure induces intense seabed sediment disturbance, leading to orders-of-magnitude increases in SSC. The resulting suspended sediments disperse via hydrodynamic processes, exerting substantial impacts on coastal ecosystems through multiple physiological and ecological mechanisms. First, suspended sediments significantly attenuate aquatic light fields, thereby reducing phytoplankton photosynthetic; Second, as a physical stressor, elevated SSC can induce physiological stress in sensitive habitats (Bilotta et al., 2008; Betbeder et al., 2015 ). Given that SSC dynamics represent critical environmental parameters for monitoring sediment transport processes and the ecological impacts of marine engineering projects, the inversion of SSC has become essential for scientifically quantifying disturbance intensity, implementing early-warning systems for ecological risks, and optimizing environmentally sustainable construction protocols (Bento et al., 2025 ) In remote sensing-based water quality retrieval, predominant methodologies include empirical algorithms, semi-empirical algorithms, physics-based models and machine learning approaches. Empirical algorithms develop statistical regression models that relate in-situ measurements to spectral data. While these models achieve acceptable accuracy within the spatiotemporal bounds of their training data, their generalization capacity is limited for large-scale or cross-regional applications (Politi et al., 2015 ; Binding et al., 2005 ; Li et al., 2016 ). Semi-empirical algorithms by integrating spectral response mechanisms with in-situ data, however, their performance critically depends on sensor-specific spectral resolution and band configurations, increasing the complexity of data acquisition and implementation challenges (Chang et al., 2015 ; Yang et al., 2022 )Physics-based analytical approaches, grounded in aquatic radiative transfer theory, often face computational constraints in optically complex waters due to difficulties in quantifying inherent optical properties (Yang et al., 2011 ). In contrast, machine learning models offer distinct advantages for SSC retrieval, particularly in their superior handling of complex spectral-SSC relationships and large-scale spatial patterns. In recent years, machine learning-driven frameworks integrating multi-source satellite imagery have been widely applied. For instance, Kazemzadeh et al. ( 2013 ) integrated MODIS imagery with in-situ measurements (2003–2011) to develop SSC retrieval models for analyzing spatiotemporal patterns in the Bahmanshir River Basin. Iran. Qiu et al. (2013) estimated SSC in the Yellow River Estuary using an optical model established via a method specifying optimal band difference ratios, based on five sets of in-situ aerial survey data. Maciel et al. ( 2021 ) achieved remote sensing inversion of inland water transparency in Brazil on the Google Earth Engine platform, demonstrating that ML significantly outperformed semi-analytical algorithms and confirming the feasibility of large-scale monitoring on cloud-based platforms (Kazemzadeh et al., 2013 ; Qiu et al., 2013; Maciel et al., 2021 ). This study focuses on marine dredging areas, which locate at the Xiamen Third East Passage serving as an important highway corridor connecting Xiamen Island, Xiang’an District, and Xiang’an New Airport, adjacent to the core zone of the Chinese White Dolphin (Sousa chinensis) National Nature Reserve and the buffer zone of the lancelet (Branchiostoma) nature reserve. Using GF-5B hyperspectral imagery and concurrent in-situ measurements, we develop novel machine learning frameworks for SSC retrieval in such ecologically sensitive environments (Francke et al., 2008 ; McFeeters et al., 1996; Zhang et al., 2023 ). Three machine learning-models are established: RF, SVM, LR. Through systematic comparison with representative empirical algorithms, a comprehensive accuracy assessment is conducted to evaluate model performance, aiming to enable high-precision monitoring of dredging-induced SSC dynamics in these ecologically critical zones. By quantifying the magnitude of SSC impacts across the construction area, this research provides evidence-based decision support for formulating targeted environmental protocols and optimizing construction strategies, which is crucial for protecting the habitats of the Branchiostoma and Sousa chinensis. (Fig. 1 ) (Kuhn et al., 2019 ; Juan et al., 2025 ; Tu et al., 2025; Li et al., 2025 ; Sun et al., 2022 ; Guo et al., 2023 ). 2 Materials and Methods 2.1 Inversion principle Machine learning methods for retrieving SSC from satellite data establish nonlinear statistical relationships between multispectral or hyperspectral remote-sensing reflectance and concurrent in-situ measurements. The trained model facilitates large-scale SSC retrieval by applying this function to independent satellite imagery. Figure 2 illustrates the inversion mechanism of suspended sediment concentration in the ocean studied in this study. (Singh et al., 2020 ). 2.2 Suspended-sediment samples On 27 May 2025 (10:00–11:00 local time), in-situ turbidity measurements were collected at 0.5m depth in the dredging area using a portable turbidimeter, with simultaneous georeferencing via handheld GPS. Thirty-four monitoring points were established, yielding 24 valid measurements following quality control and outlier removal (Fig. 3 ). Samples from this layer underwent laboratory analysis. Accurately weighed sediment aliquots were mixed with pure water to prepare serial suspended sediment dilutions. Turbidity was measured for each dilution using the same field-deployed turbidimeter model, establishing a quantitative SSC-turbidity calibration relationship (Fig. 4 ). The fitted equation (Eq. ( 1 )) subsequently converted all 24 field turbidity measurements into corresponding SSC values. 1 2.3 Satellite data This study utilized remote sensing imagery acquired by the Advanced Hyperspectral Imager (AHSI) aboard the GF5B satellite. Developed by the China Aerospace Science and Technology Corporation, GF5B was successfully launched into orbit from the Taiyuan Satellite Launch Center on 7 September 2021. As a core payload of the satellite, the AHSI sensor provides 30m spatial resolution. The imagery employed in this research was captured on 27 May 2025 at 10:48:52 UTC + 8, exhibiting 0% cloud coverage over the study area, thereby meeting all data quality requirements for subsequent analysis. Preprocessing of raw satellite data was performed in ENVI 5.6, comprising two core procedures: radiometric calibration and atmospheric correction using the FLAASH model (Pan et al., 2022 ; Zhang et al., 2010 ). Initially, radiometric calibration converted the sensor-recorded digital number (DN) values into surface apparent radiance. Subsequently, the FLAASH atmospheric correction module was applied to precisely remove atmospheric path radiance and mitigate absorption and scattering effects. This processing chain ultimately generated surface reflectance data. (Fig. 5 ). 3 Results and Discussions 3.1 Correlation analysis Before establishing the SSC inversion model, this study employs Pearson correlation (Eq. ( 2 )) analysis to rapidly screen out spectral bands sensitive to sediment. Meanwhile, it eliminates wavelengths with low contribution rates to optimize the model structure, thereby enhancing the accuracy and efficiency of the inversion model. 2 The results of the correlation analysis are presented in Fig. 6 . In this study, the top four bands(Table 1 )with the highest correlation coefficients were selected as modeling features to mitigate the challenges of data sparsity in high-dimensional spaces and the risk of overfitting when the number of features approaches or exceeds the sample size. This feature selection strategy effectively reduces dimensionality while preserving critical spectral information, thereby enhancing the stability and generalization capacity of the predictive model. Table 1 The first four bands with the strongest correlation Wavelength(nm) Correlation 669.25 0.8038 703.46 0.8001 673.53 0.7963 652.15 0.7936 3.2 Machine learning model Building upon the top 4 bands-identified via correlation analysis in the preceding section as those most responsive to SSC-this study employs the reflectance data of these bands as core feature parameters to construct a satellite remote sensing-based inversion model for suspended sediment concentration. Three machine learning algorithms are utilized: RF, SVM and LR. The Fig. 7 presents the analytical results of the random forest model for satellite remote sensing inversion of suspended sediment concentration: The model was configured with parameters of n_estimators = 500 and max_depth = 5. For the training set, the coefficient of determination (R 2 ) was 0.896 with a root mean square error (RMSE) of 0.039 kg/m 3 . For the test set, R 2 was 0.773, RMSE was 0.062 kg/m 3 . The Fig. 8 presents the results of satellite remote sensing inversion analysis of suspended sediment concentration based on the SVM model: The model parameters were set as follows: Support Vector Regression with kernel='rbf', C = 100, gamma = 0.1, and epsilon = 0.1. For the training set, the coefficient of determination was 0.609, the root mean square error was 0.076 kg/m 3 . For the test set, R 2 was 0.800, RMSE was 0.058 kg/m 3 . The Fig. 9 presents the results of satellite remote sensing inversion analysis of suspended sediment concentration based on the LR model. The training set exhibited an R 2 of 0.622, an RMSE of 0.074 kg/m 3 . The test set achieved an R 2 of 0.597, an RMSE of 0.075 kg/m 3 . 3.3 Empirical model The remote sensing inversion model for suspended sediment concentration developed in this study comprehensively analyzes reflectance data from four characteristic bands and establishes multiple mathematical models to explore the quantitative relationship between spectral reflectance and suspended sediment concentration. The results indicate that the cubic polynomial model based on the average reflectance of the four bands performs the best, with a coefficient of determination of 0.6588. All models demonstrate a clear nonlinear relationship between suspended sediment concentration and reflectance, and cubic polynomial models generally outperform linear models. The results are shown in Table 2 . Table 2 Model performance comparison table Rank Model Type Optimal Configuration R 2 1 Band Combination Four-band average cubic polynomial 0.6588 2 Single Band R669.25 band cubic polynomial 0.6563 3 Band Combination R703.46×R652.15 cubic polynomial 0.6529 4 Band Combination R703.46 + R652.15 cubic polynomial 0.6525 5 Single Band R703.46 band cubic polynomial 0.6455 3.4 Model application Based on the above comparison, the RF model performed the best. In this study, the RF algorithm was used to invert SSC on satellite images. This algorithm achieved the quantitative inversion of SSC and could obtain SSC concentration prediction results with a 30-meter spatial resolution (Wang et al., 2023 ). Figure 10 shows that the SSC on April 16th, May 27th, and July 17th, 2025 exhibited significant dynamic variations due to dredging. The hydrodynamic conditions were stable on April 16th, when no dredging operations were conducted. The SSC was below 0.15 kg/m 3 and was evenly distributed, with no high-concentration zones. Multiple dredgers operated on May 27th, significantly disturbing the sediment. The SSC increased sharply due to propeller agitation and sediment excavation. A significant transformation occurred on July 17th due to changes in the construction scheme. The SSC distribution changed from concentrated to dispersed. The monitoring data showed that the SSC stabilized in the range of 0.3–0.5 kg/m 3 and decreased in the surrounding area. Despite the significant increase in the SSC in the construction area, the anti-pollution curtains provided an effective physical barrier, significantly reducing the diffusion of sediments to other areas. Tidal and ocean currents resulted in the horizontal migration of some of the suspended sediments, and sediment settling occurred over short distances. Therefore, the variation in the SSC was confined to the construction area and its immediate vicinity, without causing large-scale diffuse pollution. At the same time, The SSC in the sea areas near the two reserves remained at the natural background level, indicating a minimal impact of construction activities on the sensitive ecological reserves. What is more, this study optimized the monitoring method to reduce retrieval errors in the SSC, resulting from the difficulty of accurately distinguishing water and land information in nearshore areas because of mixed spectral signatures. Therefore, an SSC prediction method was developed based on the RF model. In practical applications, multispectral reflectance data from a single sampling point or a batch import of reflectance datasets of a region is required to output the SSC values for the target area rapidly. This approach is not affected by interference from complex nearshore environments, significantly reducing errors caused by boundary effects. 3.5 Study limitations This study has several limitations that should be clearly noted. First, the small sample size may limit the model's ability to capture spatial heterogeneity; single-temporal data cannot reflect the dynamic changes of SSC during the dredging period. Second, hydrodynamic factors such as tides and ocean currents were not incorporated, which may affect the model's generalization ability. Nevertheless, the research results still have important practical value: they provide a scientific basis for optimizing construction plans and implementing ecological protection measures. This study verifies the effectiveness of "GF-5B hyperspectral data combined with machine learning" in SSC inversion in dredging areas, and provides a promotable technical paradigm for coastal engineering environmental monitoring. 4 Conclusion Based on the results presented, the following conclusions can be drawn: First, Among the four models (RF, SVM, LR, empirical models), the RF model performs best with a test set R² of 0.773 and RMSE of 0.062 kg/m³, reliably capturing the spectral-SSC nonlinear relationship. Second, the study area’s SSC shows obvious spatiotemporal differences: high SSC (0.2-0.4684 kg/m³) in the dredging area and low SSC (< 0.15 kg/m³) in surrounding areas; SSC rises sharply during dredging and is low and uniform without operations. Third, Anti-pollution curtains effectively block sediment diffusion. SSC changes are limited to the construction area and its vicinity, keeping SSC in the nearby Chinese White Dolphin and lancelet reserve waters at the natural background level. Declarations All authors have read, understood, and have complied as applicable with the statement on “Ethical responsibilities of Authors” as found in the Instructions for Authors. All authors declare that there are no potential conflicts of interest. This study does not involve human participants or animals. All researchers involved in this study have been fully informed of the research content and submission matters, and have given their consent. FUNDING The work described in this paper was fully supported by Scientific Research Funds of Huaqiao University (Project No. 20BS202) and Research Funds by China Railway Southeast Investment Co., Ltd. (Project No. 2025HH314). Author Contribution Pengfei Wang compiled the manuscript and provided the main ideas and methods. Xu Chang completed the proofreading of the paper and provided financial support. Jianhu Fan, Bo Chen and Qingjun Hao provided data support and financial support. ACKNOWLEDGMENTS We would like to acknowledge China Railway Southeast Investment Co., Ltd. for making data publicly available. We would like to thank anonymous reviewers, associate editor and editor for providing useful comments that significantly improved this study. References Bento, A. M., Fazeres-Ferradosa, T., Rosa-Santos, P., & Taveira-Pinto, F. (2025). Risk analysis in ocean and maritime engineering. Ocean Engineering , 120579. https://doi.org/10.1016/j.oceaneng.2025.120579 Betbeder, J., Rapinel, S., Corgne, S., Pottier, E., & Hubert-Moy, L. (2015). TerraSAR-X dual-pol time-series for mapping of wetland vegetation. ISPRS Journal of Photogrammetry and Remote Sensing , 107, 90–98. https://doi.org/10.1016/j.isprsjprs.2015.05.001 Bilotta, G. S., & Brazier, R. E. (2008). Understanding the influence of suspended solids on water quality and aquatic biota. Water research , 42(12), 2849–2861. https://doi.org/10.1016/j.watres.2008.03.018 Binding, C. E., Bowers, D. G., & Mitchelson-Jacob, E. G. (2005). Estimating suspended sediment concentrations from ocean colour measurements in moderately turbid waters; the impact of variable particle scattering properties. Remote sensing of Environment , 94(3), 373–383. https://doi.org/10.1016/j.rse.2004.11.002 Chang, N. B., Imen, S., & Vannah, B. (2015). Remote sensing for monitoring surface water quality status and ecosystem state in relation to the nutrient cycle: a 40-year perspective. Critical Reviews in Environmental Science and Technology , 45(2), 101–166. https://doi.org/10.1080/10643389.2013.829981 Francke, T., López-Tarazón, J. A., & Schröder, B. (2008). Estimation of suspended sediment concentration and yield using linear models, random forests and quantile regression forests. Hydrological Processes : An International Journal, 22(25), 4892–4904. https://doi.org/10.1002/hyp.7110 Guo, Y., Gui, D., Liu, W., Xie, Q., & Wu, Y. (2023). Hormonal biomarkers provide insights into the reproductive biology and pollutants–associated health hazards of endangered dolphins. Chemosphere , 337, 139328. https://doi.org/10.1016/j.chemosphere.2023.139328 Juan, N. P., Valdecantos, V. N., & Troch, P. (2025). Advancing artificial intelligence in ocean and maritime engineering: Trends, progress, and future directions. Ocean Engineering , 339, 122077. https://doi.org/10.1016/j.oceaneng.2025.122077 Kazemzadeh, M. B., Ayyoubzadeh, S. A., & Moridnezhad, A. (2013). Remote sensing of temporal and spatial variations of suspended sediment concentration in Bahmanshir Estuary, Iran. Indian Journal of Science and Technology , 6(8), 5036–5045. Online ISSN: 0974–5645 Kuhn, C., de Matos Valerio, A., Ward, N., Loken, L., Sawakuchi, H. O., Kampel, M., … Butman, D. (2019). Performance of Landsat-8 and Sentinel-2 surface reflectance products for river remote sensing retrievals of chlorophyll-a and turbidity. Remote Sensing of Environment , 224, 104–118. https://doi.org/10.1016/j.rse.2019.01.023 Li, X., Zhu, J., Yuan, R., Qiu, C., & Wu, H. (2016). Sediment trapping in the Changjiang Estuary: Observations in the North Passage over a spring-neap tidal cycle. Estuarine, Coastal and Shelf Science , 177, 8–19. https://doi.org/10.1016/j.ecss.2016.05.004 Li, Z., Yan, Y., Zhao, Z., Xu, Y., & Hu, Y. (2025). Numerical study on hydrodynamic effects of intermittent or sinusoidal coordination of pectoral fins to achieve spontaneous nose-up pitching behavior in dolphins. Ocean Engineering , 337, 121854. https://doi.org/10.1016/j.oceaneng.2025.121854 Maciel, D. A., Barbosa, C. C. F., de Moraes Novo, E. M. L., Júnior, R. F., & Begliomini, F. N. (2021). Water clarity in Brazilian water assessed using Sentinel-2 and machine learning methods. ISPRS Journal of Photogrammetry and Remote Sensing , 182, 134–152. https://doi.org/10.1016/j.isprsjprs.2021.10.009 McFeeters, S. K. (1996). The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. International journal of remote sensing , 17(7), 1425–1432. https://doi.org/10.1080/01431169608948714 Pan, F., Wang, B., Zhang, Y., Zhang, X., Cai, Y., Liu, H., … Wang, X. (2022). Bioavailability,(im) mobilization kinetics, and spatiotemporal patterns of arsenic and cadmium in surficial sediments of a river–estuary–coast system. Journal of Hydrology , 612, 128140. https://doi.org/10.1016/j.jhydrol.2022.128140 Politi, E., Cutler, M. E. J., & Rowan, J. S. (2015). Evaluating the spatial transferability and temporal repeatability of remote-sensing-based lake water quality retrieval algorithms at the European scale: a meta-analysis approach. International Journal of Remote Sensing , 36 (11), 2995–3023. https://doi.org/10.1080/01431161.2015.1054962 Qiu, Z. (2013). A simple optical model to estimate suspended particulate matter in Yellow River Estuary. Optics Express , 21(23), 27891–27904. https://doi.org/10.1364/OE.21.027891 Singh, A. T., Sharma, P., Sharma, C., Laluraj, C. M., Patel, L., Pratap, B., … Thamban, M. (2020). Water discharge and suspended sediment dynamics in the Chandra River, Western Himalaya. Journal of Earth System Science , 129(1), 206. https://doi.org/10.1007/s12040-020-01455-4 Sun, X., Guo, L., Luo, D., Yu, R. Q., Yu, X., Liang, Y., … Wu, Y. (2022). Long-term increase in mortality of Indo-Pacific humpback dolphins (Sousa chinensis) in the Pearl River Estuary following anthropic activities: Evidence from the stranded dolphin mortality analysis from 2003 to 2017. Environmental Pollution , 307 , 119526. https://doi.org/10.1016/j.envpol.2022.119526 Tu, Y., Yao, Q., Zhou, Y., & Zhu, Z. (2025). Prediction of maximum dynamic shear modulus of undisturbed marine soils in the eastern coast of China based on machine learning methods. Ocean Engineering , 321, 120382. https://doi.org/10.1016/j.oceaneng.2025.120382 Wang, H., Zhang, C., Zhou, B., Xue, S., Jia, P., & Zhu, X. (2023). Prediction of triaxial mechanical properties of rocks based on mesoscopic finite element numerical simulation and multi-objective machine learning. Journal of King Saud University-Science , 35(7), 102846. https://doi.org/10.1016/j.jksus.2023.102846 Yang, G., Wang, X. H., Zhong, Y., & Oliver, T. S. (2022). Modelling study on the sediment dynamics and the formation of the flood-tide delta near Cullendulla Beach in the Batemans Bay, Australia. Marine Geology , 452, 106910. https://doi.org/10.1016/j.margeo.2022.106910 Yang, W., Matsushita, B., Chen, J., & Fukushima, T. (2011). Estimating constituent concentrations in case II waters from MERIS satellite data by semi-analytical model optimizing and look-up tables. Remote sensing of environment , 115(5), 1247–1259. https://doi.org/10.1016/j.rse.2011.01.007 Zhang, M., Tang, J., Dong, Q., Song, Q., & Ding, J. (2010). Retrieval of total suspended matter concentration in the Yellow and East China Seas from MODIS imagery. Remote Sensing of Environment , 114(2), 392–403. https://doi.org/10.1016/j.rse.2009.09.016 Zhang, X., Qiao, W., Lu, Y., Huang, J., & Xiao, Y. (2023). Quantitative analysis of the influence of the Xiaolangdi reservoir on water and sediment in the middle and lower reaches of the Yellow River. International Journal of Environmental Research and Public Health , 20(5), 4351. https://doi.org/10.3390/ijerph20054351 Additional Declarations No competing interests reported. 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-8033000","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":544800814,"identity":"4894fb81-f3f8-404b-bc3f-2515e9672fa6","order_by":0,"name":"Pengfei Wang","email":"","orcid":"","institution":"School of Civil Engineering, Huaqiao University","correspondingAuthor":false,"prefix":"","firstName":"Pengfei","middleName":"","lastName":"Wang","suffix":""},{"id":544800815,"identity":"f8eb6c13-8f90-47d5-ad9a-e8c21d16dd5e","order_by":1,"name":"Xu Chang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA60lEQVRIiWNgGAWjYDACCQglxyZ/+BiEeYBILcZ8EmxppGlJnCfBY0acFv7Zzc8efm07bMwm3fPtwds2Bjm+GwmMnwvwWXLnmLmxzJnDcmwyZ7cbzm1jMJa8kcAsPQOPFgOJBDNpiQqgLQy526R52xgSN9xIYGPmwasl/Zu0hMHhxDaGnGcgLfVEaMkxk/xQAdQikcMG0pJgQEiLxI2cMmmGM+nGbDzHzCTnnJMwnHnmYbM0Pi38M9K3Sf5ss5aTb29+JvGmzEae73jywc/4tIAAwhk84GhibCCgAajkB0LLKBgFo2AUjAJMAACuLUcqabD4uAAAAABJRU5ErkJggg==","orcid":"","institution":"School of Civil Engineering, Huaqiao University","correspondingAuthor":true,"prefix":"","firstName":"Xu","middleName":"","lastName":"Chang","suffix":""},{"id":544800816,"identity":"f7ec2b90-7b57-4cc0-8e7d-ba07a6426870","order_by":2,"name":"Bo Chen","email":"","orcid":"","institution":"China Railway Southeast Investment Co., Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Bo","middleName":"","lastName":"Chen","suffix":""},{"id":544800817,"identity":"de8b5333-f186-45a9-a548-1e648e55dd7d","order_by":3,"name":"Jianhu Fan","email":"","orcid":"","institution":"China Railway Southeast Investment Co., Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Jianhu","middleName":"","lastName":"Fan","suffix":""},{"id":544800818,"identity":"ebb66cfe-e074-4740-aba3-f6f5ffbbc635","order_by":4,"name":"Qingjun Hao","email":"","orcid":"","institution":"China Railway Southeast Investment Co., Ltd.","correspondingAuthor":false,"prefix":"","firstName":"Qingjun","middleName":"","lastName":"Hao","suffix":""}],"badges":[],"createdAt":"2025-11-05 01:23:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8033000/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8033000/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":97894669,"identity":"f8b1cfb5-2860-4369-abd3-82bf4be376d9","added_by":"auto","created_at":"2025-12-10 15:32:52","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":347404,"visible":true,"origin":"","legend":"\u003cp\u003eThe rough diagram of relative positions\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/d7496e8b7bdf4cc4608031a8.png"},{"id":97720469,"identity":"24dcf02c-5b27-402a-9eb7-ca1522bd7f3f","added_by":"auto","created_at":"2025-12-08 15:35:30","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":72711,"visible":true,"origin":"","legend":"\u003cp\u003eThe inversion mechanism of suspended sediment concentration in the ocean studied in this study\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/43dadc40919facd05559cfcc.png"},{"id":97720470,"identity":"4b7da974-2e33-442c-8e3b-ca8a7696fe2f","added_by":"auto","created_at":"2025-12-08 15:35:30","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":331718,"visible":true,"origin":"","legend":"\u003cp\u003eWorkflow of field data collection: (a) Dredging operational environment; (b) In-situ measurement of seawater turbidity; (c) Spatial distribution of measurement points; (d) Sediment sampling at dredging depths.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/2dea7629db8af91738bcb430.png"},{"id":97720472,"identity":"b7e46622-ab1d-4443-a6de-75451f38d99c","added_by":"auto","created_at":"2025-12-08 15:35:30","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":101333,"visible":true,"origin":"","legend":"\u003cp\u003eTurbidity and suspended sediment concentration were fitted to the curve\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/d073f4e38241ded348feffb2.png"},{"id":97895661,"identity":"9f531275-4454-43a0-8c67-142002969af9","added_by":"auto","created_at":"2025-12-10 15:34:38","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":613420,"visible":true,"origin":"","legend":"\u003cp\u003eSpectral reflectance\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/f7800e765cda1f346dc81358.png"},{"id":97895544,"identity":"fcef8b40-032b-4e88-b350-dc67e51f2b23","added_by":"auto","created_at":"2025-12-10 15:34:25","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":105036,"visible":true,"origin":"","legend":"\u003cp\u003ePearson correlation coefficient between band and reflectance\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/67530b8575a5bb59b1743766.png"},{"id":97893837,"identity":"04746688-521e-4061-ac90-b2693cb5702d","added_by":"auto","created_at":"2025-12-10 15:31:21","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":83181,"visible":true,"origin":"","legend":"\u003cp\u003eRF inversion model\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/787891c8e81e213073b3b99a.png"},{"id":97720478,"identity":"65a5e03e-f6a9-4bf0-b979-d3c399dc89b7","added_by":"auto","created_at":"2025-12-08 15:35:30","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":86378,"visible":true,"origin":"","legend":"\u003cp\u003eSVM inversion model\u003c/p\u003e","description":"","filename":"8.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/bfd9bbea072be8db6204ce1a.png"},{"id":97895900,"identity":"735c0627-7bae-4c99-a1cd-7bbb554b2b82","added_by":"auto","created_at":"2025-12-10 15:35:20","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":83260,"visible":true,"origin":"","legend":"\u003cp\u003eLR inversion model\u003c/p\u003e","description":"","filename":"9.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/1c4c6e3f836d721789435aab.png"},{"id":97720474,"identity":"02ee1a34-946d-4c32-9144-7ee2e103820c","added_by":"auto","created_at":"2025-12-08 15:35:30","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":336241,"visible":true,"origin":"","legend":"\u003cp\u003eInversion results of suspended sediment concentration in the construction area\u003c/p\u003e","description":"","filename":"10.png","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/e0c37e1e398f0e4a516e09a8.png"},{"id":100154082,"identity":"1c555b9d-eb9a-43d3-8233-13a2ca6544dd","added_by":"auto","created_at":"2026-01-13 13:54:32","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2460728,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8033000/v1/2617039a-f607-42ee-810e-03d928f961b6.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Using Machine Learning and Hyperspectral Satellite to Monitor Suspended Sediment Concentration under Dredging Engineering","fulltext":[{"header":"Highlights","content":"\u003cul\u003e\n\u003cli\u003eRemote sensing inversion models for SSC in dredging-disturbed environments has been developed.\u003c/li\u003e\n\u003cli\u003eRF model outperformed other models\u003c/li\u003e\n\u003cli\u003eHigh-precision SSC monitoring in dredging areas has been achieved.\u003c/li\u003e\n\u003cli\u003eThe SSC in dredging areas significantly changes with the construction environment.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"1 Introduction","content":"\u003cp\u003eLarge-scale transportation infrastructure induces intense seabed sediment disturbance, leading to orders-of-magnitude increases in SSC. The resulting suspended sediments disperse via hydrodynamic processes, exerting substantial impacts on coastal ecosystems through multiple physiological and ecological mechanisms. First, suspended sediments significantly attenuate aquatic light fields, thereby reducing phytoplankton photosynthetic; Second, as a physical stressor, elevated SSC can induce physiological stress in sensitive habitats (Bilotta et al., 2008; Betbeder et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). Given that SSC dynamics represent critical environmental parameters for monitoring sediment transport processes and the ecological impacts of marine engineering projects, the inversion of SSC has become essential for scientifically quantifying disturbance intensity, implementing early-warning systems for ecological risks, and optimizing environmentally sustainable construction protocols (Bento et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2025\u003c/span\u003e)\u003c/p\u003e\u003cp\u003eIn remote sensing-based water quality retrieval, predominant methodologies include empirical algorithms, semi-empirical algorithms, physics-based models and machine learning approaches. Empirical algorithms develop statistical regression models that relate in-situ measurements to spectral data. While these models achieve acceptable accuracy within the spatiotemporal bounds of their training data, their generalization capacity is limited for large-scale or cross-regional applications (Politi et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Binding et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2005\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Semi-empirical algorithms by integrating spectral response mechanisms with in-situ data, however, their performance critically depends on sensor-specific spectral resolution and band configurations, increasing the complexity of data acquisition and implementation challenges (Chang et al., \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Yang et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2022\u003c/span\u003e)Physics-based analytical approaches, grounded in aquatic radiative transfer theory, often face computational constraints in optically complex waters due to difficulties in quantifying inherent optical properties (Yang et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2011\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn contrast, machine learning models offer distinct advantages for SSC retrieval, particularly in their superior handling of complex spectral-SSC relationships and large-scale spatial patterns. In recent years, machine learning-driven frameworks integrating multi-source satellite imagery have been widely applied. For instance, Kazemzadeh et al. (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2013\u003c/span\u003e) integrated MODIS imagery with in-situ measurements (2003\u0026ndash;2011) to develop SSC retrieval models for analyzing spatiotemporal patterns in the Bahmanshir River Basin. Iran. Qiu et al. (2013) estimated SSC in the Yellow River Estuary using an optical model established via a method specifying optimal band difference ratios, based on five sets of in-situ aerial survey data. Maciel et al. (\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e) achieved remote sensing inversion of inland water transparency in Brazil on the Google Earth Engine platform, demonstrating that ML significantly outperformed semi-analytical algorithms and confirming the feasibility of large-scale monitoring on cloud-based platforms (Kazemzadeh et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2013\u003c/span\u003e; Qiu et al., 2013; Maciel et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eThis study focuses on marine dredging areas, which locate at the Xiamen Third East Passage serving as an important highway corridor connecting Xiamen Island, Xiang\u0026rsquo;an District, and Xiang\u0026rsquo;an New Airport, adjacent to the core zone of the Chinese White Dolphin (Sousa chinensis) National Nature Reserve and the buffer zone of the lancelet (Branchiostoma) nature reserve. Using GF-5B hyperspectral imagery and concurrent in-situ measurements, we develop novel machine learning frameworks for SSC retrieval in such ecologically sensitive environments (Francke et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2008\u003c/span\u003e; McFeeters et al., 1996; Zhang et al., \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Three machine learning-models are established: RF, SVM, LR. Through systematic comparison with representative empirical algorithms, a comprehensive accuracy assessment is conducted to evaluate model performance, aiming to enable high-precision monitoring of dredging-induced SSC dynamics in these ecologically critical zones. By quantifying the magnitude of SSC impacts across the construction area, this research provides evidence-based decision support for formulating targeted environmental protocols and optimizing construction strategies, which is crucial for protecting the habitats of the Branchiostoma and Sousa chinensis. (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) (Kuhn et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Juan et al., \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Tu et al., 2025; Li et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Sun et al., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Guo et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e"},{"header":"2 Materials and Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Inversion principle\u003c/h2\u003e\u003cp\u003eMachine learning methods for retrieving SSC from satellite data establish nonlinear statistical relationships between multispectral or hyperspectral remote-sensing reflectance and concurrent in-situ measurements. The trained model facilitates large-scale SSC retrieval by applying this function to independent satellite imagery. Figure\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e illustrates the inversion mechanism of suspended sediment concentration in the ocean studied in this study. (Singh et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Suspended-sediment samples\u003c/h2\u003e\u003cp\u003eOn 27 May 2025 (10:00\u0026ndash;11:00 local time), in-situ turbidity measurements were collected at 0.5m depth in the dredging area using a portable turbidimeter, with simultaneous georeferencing via handheld GPS. Thirty-four monitoring points were established, yielding 24 valid measurements following quality control and outlier removal (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eSamples from this layer underwent laboratory analysis. Accurately weighed sediment aliquots were mixed with pure water to prepare serial suspended sediment dilutions. Turbidity was measured for each dilution using the same field-deployed turbidimeter model, establishing a quantitative SSC-turbidity calibration relationship (Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe fitted equation (Eq.\u0026nbsp;(\u003cspan refid=\"Equ1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)) subsequently converted all 24 field turbidity measurements into corresponding SSC values.\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Satellite data\u003c/h2\u003e\u003cp\u003eThis study utilized remote sensing imagery acquired by the Advanced Hyperspectral Imager (AHSI) aboard the GF5B satellite. Developed by the China Aerospace Science and Technology Corporation, GF5B was successfully launched into orbit from the Taiyuan Satellite Launch Center on 7 September 2021. As a core payload of the satellite, the AHSI sensor provides 30m spatial resolution. The imagery employed in this research was captured on 27 May 2025 at 10:48:52 UTC\u0026thinsp;+\u0026thinsp;8, exhibiting 0% cloud coverage over the study area, thereby meeting all data quality requirements for subsequent analysis.\u003c/p\u003e\u003cp\u003ePreprocessing of raw satellite data was performed in ENVI 5.6, comprising two core procedures: radiometric calibration and atmospheric correction using the FLAASH model (Pan et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Zhang et al., \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Initially, radiometric calibration converted the sensor-recorded digital number (DN) values into surface apparent radiance. Subsequently, the FLAASH atmospheric correction module was applied to precisely remove atmospheric path radiance and mitigate absorption and scattering effects. This processing chain ultimately generated surface reflectance data. (Fig.\u0026nbsp;\u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e5\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Results and Discussions","content":"\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Correlation analysis\u003c/h2\u003e\u003cp\u003eBefore establishing the SSC inversion model, this study employs Pearson correlation (Eq.\u0026nbsp;(\u003cspan refid=\"Equ2\" class=\"InternalRef\"\u003e2\u003c/span\u003e)) analysis to rapidly screen out spectral bands sensitive to sediment. Meanwhile, it eliminates wavelengths with low contribution rates to optimize the model structure, thereby enhancing the accuracy and efficiency of the inversion model.\u003cdiv id=\"Equ2\" class=\"Equation\"\u003e\u003cdiv class=\"EquationNumber\"\u003e2\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e\u003cp\u003eThe results of the correlation analysis are presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig6\" class=\"InternalRef\"\u003e6\u003c/span\u003e. In this study, the top four bands(Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e)with the highest correlation coefficients were selected as modeling features to mitigate the challenges of data sparsity in high-dimensional spaces and the risk of overfitting when the number of features approaches or exceeds the sample size. This feature selection strategy effectively reduces dimensionality while preserving critical spectral information, thereby enhancing the stability and generalization capacity of the predictive model.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eThe first four bands with the strongest correlation\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"2\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eWavelength(nm)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCorrelation\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e669.25\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.8038\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e703.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.8001\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e673.53\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.7963\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e652.15\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.7936\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/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Machine learning model\u003c/h2\u003e\u003cp\u003eBuilding upon the top 4 bands-identified via correlation analysis in the preceding section as those most responsive to SSC-this study employs the reflectance data of these bands as core feature parameters to construct a satellite remote sensing-based inversion model for suspended sediment concentration. Three machine learning algorithms are utilized: RF, SVM and LR.\u003c/p\u003e\u003cp\u003eThe Fig.\u0026nbsp;\u003cspan refid=\"Fig7\" class=\"InternalRef\"\u003e7\u003c/span\u003e presents the analytical results of the random forest model for satellite remote sensing inversion of suspended sediment concentration: The model was configured with parameters of n_estimators\u0026thinsp;=\u0026thinsp;500 and max_depth\u0026thinsp;=\u0026thinsp;5. For the training set, the coefficient of determination (R\u003csup\u003e2\u003c/sup\u003e) was 0.896 with a root mean square error (RMSE) of 0.039 kg/m\u003csup\u003e3\u003c/sup\u003e. For the test set, R\u003csup\u003e2\u003c/sup\u003e was 0.773, RMSE was 0.062 kg/m\u003csup\u003e3\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe Fig.\u0026nbsp;\u003cspan refid=\"Fig8\" class=\"InternalRef\"\u003e8\u003c/span\u003e presents the results of satellite remote sensing inversion analysis of suspended sediment concentration based on the SVM model: The model parameters were set as follows: Support Vector Regression with kernel='rbf', C\u0026thinsp;=\u0026thinsp;100, gamma\u0026thinsp;=\u0026thinsp;0.1, and epsilon\u0026thinsp;=\u0026thinsp;0.1. For the training set, the coefficient of determination was 0.609, the root mean square error was 0.076 kg/m\u003csup\u003e3\u003c/sup\u003e. For the test set, R\u003csup\u003e2\u003c/sup\u003e was 0.800, RMSE was 0.058 kg/m\u003csup\u003e3\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe Fig.\u0026nbsp;\u003cspan refid=\"Fig9\" class=\"InternalRef\"\u003e9\u003c/span\u003e presents the results of satellite remote sensing inversion analysis of suspended sediment concentration based on the LR model. The training set exhibited an R\u003csup\u003e2\u003c/sup\u003e of 0.622, an RMSE of 0.074 kg/m\u003csup\u003e3\u003c/sup\u003e. The test set achieved an R\u003csup\u003e2\u003c/sup\u003e of 0.597, an RMSE of 0.075 kg/m\u003csup\u003e3\u003c/sup\u003e.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Empirical model\u003c/h2\u003e\u003cp\u003eThe remote sensing inversion model for suspended sediment concentration developed in this study comprehensively analyzes reflectance data from four characteristic bands and establishes multiple mathematical models to explore the quantitative relationship between spectral reflectance and suspended sediment concentration. The results indicate that the cubic polynomial model based on the average reflectance of the four bands performs the best, with a coefficient of determination of 0.6588. All models demonstrate a clear nonlinear relationship between suspended sediment concentration and reflectance, and cubic polynomial models generally outperform linear models.\u003c/p\u003e\u003cp\u003eThe results are shown in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eModel performance comparison table\u003c/p\u003e\u003c/div\u003e\u003c/caption\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=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eRank\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eModel Type\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eOptimal Configuration\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eR\u003csup\u003e2\u003c/sup\u003e\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBand Combination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eFour-band average cubic polynomial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.6588\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSingle Band\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eR669.25 band cubic polynomial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.6563\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBand Combination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eR703.46\u0026times;R652.15 cubic polynomial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.6529\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eBand Combination\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eR703.46\u0026thinsp;+\u0026thinsp;R652.15 cubic polynomial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.6525\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eSingle Band\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c3\"\u003e\u003cp\u003eR703.46 band cubic polynomial\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.6455\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/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.4 Model application\u003c/h2\u003e\u003cp\u003eBased on the above comparison, the RF model performed the best. In this study, the RF algorithm was used to invert SSC on satellite images. This algorithm achieved the quantitative inversion of SSC and could obtain SSC concentration prediction results with a 30-meter spatial resolution (Wang et al., \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFigure\u0026nbsp;\u003cspan refid=\"Fig10\" class=\"InternalRef\"\u003e10\u003c/span\u003e shows that the SSC on April 16th, May 27th, and July 17th, 2025 exhibited significant dynamic variations due to dredging. The hydrodynamic conditions were stable on April 16th, when no dredging operations were conducted. The SSC was below 0.15 kg/m\u003csup\u003e3\u003c/sup\u003e and was evenly distributed, with no high-concentration zones. Multiple dredgers operated on May 27th, significantly disturbing the sediment. The SSC increased sharply due to propeller agitation and sediment excavation. A significant transformation occurred on July 17th due to changes in the construction scheme. The SSC distribution changed from concentrated to dispersed. The monitoring data showed that the SSC stabilized in the range of 0.3\u0026ndash;0.5 kg/m\u003csup\u003e3\u003c/sup\u003e and decreased in the surrounding area.\u003c/p\u003e\u003cp\u003eDespite the significant increase in the SSC in the construction area, the anti-pollution curtains provided an effective physical barrier, significantly reducing the diffusion of sediments to other areas. Tidal and ocean currents resulted in the horizontal migration of some of the suspended sediments, and sediment settling occurred over short distances. Therefore, the variation in the SSC was confined to the construction area and its immediate vicinity, without causing large-scale diffuse pollution. At the same time, The SSC in the sea areas near the two reserves remained at the natural background level, indicating a minimal impact of construction activities on the sensitive ecological reserves.\u003c/p\u003e\u003cp\u003eWhat is more, this study optimized the monitoring method to reduce retrieval errors in the SSC, resulting from the difficulty of accurately distinguishing water and land information in nearshore areas because of mixed spectral signatures. Therefore, an SSC prediction method was developed based on the RF model. In practical applications, multispectral reflectance data from a single sampling point or a batch import of reflectance datasets of a region is required to output the SSC values for the target area rapidly. This approach is not affected by interference from complex nearshore environments, significantly reducing errors caused by boundary effects.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.5 Study limitations\u003c/h2\u003e\u003cp\u003eThis study has several limitations that should be clearly noted. First, the small sample size may limit the model's ability to capture spatial heterogeneity; single-temporal data cannot reflect the dynamic changes of SSC during the dredging period. Second, hydrodynamic factors such as tides and ocean currents were not incorporated, which may affect the model's generalization ability. Nevertheless, the research results still have important practical value: they provide a scientific basis for optimizing construction plans and implementing ecological protection measures. This study verifies the effectiveness of \"GF-5B hyperspectral data combined with machine learning\" in SSC inversion in dredging areas, and provides a promotable technical paradigm for coastal engineering environmental monitoring.\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Conclusion","content":"\u003cp\u003eBased on the results presented, the following conclusions can be drawn:\u003c/p\u003e\u003cp\u003eFirst, Among the four models (RF, SVM, LR, empirical models), the RF model performs best with a test set R\u0026sup2; of 0.773 and RMSE of 0.062 kg/m\u0026sup3;, reliably capturing the spectral-SSC nonlinear relationship.\u003c/p\u003e\u003cp\u003eSecond, the study area\u0026rsquo;s SSC shows obvious spatiotemporal differences: high SSC (0.2-0.4684 kg/m\u0026sup3;) in the dredging area and low SSC (\u0026lt;\u0026thinsp;0.15 kg/m\u0026sup3;) in surrounding areas; SSC rises sharply during dredging and is low and uniform without operations.\u003c/p\u003e\u003cp\u003eThird, Anti-pollution curtains effectively block sediment diffusion. SSC changes are limited to the construction area and its vicinity, keeping SSC in the nearby Chinese White Dolphin and lancelet reserve waters at the natural background level.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eAll authors have read, understood, and have complied as applicable with the statement on \u0026ldquo;Ethical responsibilities of Authors\u0026rdquo; as found in the Instructions for Authors. All authors declare that there are no potential conflicts of interest. This study does not involve human participants or animals. All researchers involved in this study have been fully informed of the research content and submission matters, and have given their consent.\u003c/p\u003e\u003ch2\u003eFUNDING\u003c/h2\u003e\u003cp\u003eThe work described in this paper was fully supported by Scientific Research Funds of Huaqiao University (Project No. 20BS202) and Research Funds by China Railway Southeast Investment Co., Ltd. (Project No. 2025HH314).\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003ePengfei Wang compiled the manuscript and provided the main ideas and methods. Xu Chang completed the proofreading of the paper and provided financial support. Jianhu Fan, Bo Chen and Qingjun Hao provided data support and financial support.\u003c/p\u003e\u003ch2\u003eACKNOWLEDGMENTS\u003c/h2\u003e\u003cp\u003eWe would like to acknowledge China Railway Southeast Investment Co., Ltd. for making data publicly available. We would like to thank anonymous reviewers, associate editor and editor for providing useful comments that significantly improved this study.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eBento, A. M., Fazeres-Ferradosa, T., Rosa-Santos, P., \u0026amp; Taveira-Pinto, F. (2025). Risk analysis in ocean and maritime engineering. \u003cem\u003eOcean Engineering\u003c/em\u003e, 120579. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.oceaneng.2025.120579\u003c/span\u003e\u003cspan address=\"10.1016/j.oceaneng.2025.120579\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBetbeder, J., Rapinel, S., Corgne, S., Pottier, E., \u0026amp; Hubert-Moy, L. (2015). TerraSAR-X dual-pol time-series for mapping of wetland vegetation. \u003cem\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/em\u003e, 107, 90\u0026ndash;98. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.isprsjprs.2015.05.001\u003c/span\u003e\u003cspan address=\"10.1016/j.isprsjprs.2015.05.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBilotta, G. S., \u0026amp; Brazier, R. E. (2008). Understanding the influence of suspended solids on water quality and aquatic biota. \u003cem\u003eWater research\u003c/em\u003e, 42(12), 2849\u0026ndash;2861. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.watres.2008.03.018\u003c/span\u003e\u003cspan address=\"10.1016/j.watres.2008.03.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBinding, C. E., Bowers, D. G., \u0026amp; Mitchelson-Jacob, E. G. (2005). Estimating suspended sediment concentrations from ocean colour measurements in moderately turbid waters; the impact of variable particle scattering properties. \u003cem\u003eRemote sensing of Environment\u003c/em\u003e, 94(3), 373\u0026ndash;383. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2004.11.002\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2004.11.002\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChang, N. B., Imen, S., \u0026amp; Vannah, B. (2015). Remote sensing for monitoring surface water quality status and ecosystem state in relation to the nutrient cycle: a 40-year perspective. \u003cem\u003eCritical Reviews in Environmental Science and Technology\u003c/em\u003e, 45(2), 101\u0026ndash;166. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10643389.2013.829981\u003c/span\u003e\u003cspan address=\"10.1080/10643389.2013.829981\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFrancke, T., L\u0026oacute;pez-Taraz\u0026oacute;n, J. A., \u0026amp; Schr\u0026ouml;der, B. (2008). Estimation of suspended sediment concentration and yield using linear models, random forests and quantile regression forests. \u003cem\u003eHydrological Processes\u003c/em\u003e: An International Journal, 22(25), 4892\u0026ndash;4904. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/hyp.7110\u003c/span\u003e\u003cspan address=\"10.1002/hyp.7110\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGuo, Y., Gui, D., Liu, W., Xie, Q., \u0026amp; Wu, Y. (2023). Hormonal biomarkers provide insights into the reproductive biology and pollutants\u0026ndash;associated health hazards of endangered dolphins. \u003cem\u003eChemosphere\u003c/em\u003e, 337, 139328. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chemosphere.2023.139328\u003c/span\u003e\u003cspan address=\"10.1016/j.chemosphere.2023.139328\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJuan, N. P., Valdecantos, V. N., \u0026amp; Troch, P. (2025). Advancing artificial intelligence in ocean and maritime engineering: Trends, progress, and future directions. \u003cem\u003eOcean Engineering\u003c/em\u003e, 339, 122077. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.oceaneng.2025.122077\u003c/span\u003e\u003cspan address=\"10.1016/j.oceaneng.2025.122077\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKazemzadeh, M. B., Ayyoubzadeh, S. A., \u0026amp; Moridnezhad, A. (2013). Remote sensing of temporal and spatial variations of suspended sediment concentration in Bahmanshir Estuary, Iran. \u003cem\u003eIndian Journal of Science and Technology\u003c/em\u003e, 6(8), 5036\u0026ndash;5045. Online ISSN: 0974\u0026ndash;5645\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKuhn, C., de Matos Valerio, A., Ward, N., Loken, L., Sawakuchi, H. O., Kampel, M., \u0026hellip; Butman, D. (2019). Performance of Landsat-8 and Sentinel-2 surface reflectance products for river remote sensing retrievals of chlorophyll-a and turbidity. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, 224, 104\u0026ndash;118. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2019.01.023\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2019.01.023\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, X., Zhu, J., Yuan, R., Qiu, C., \u0026amp; Wu, H. (2016). Sediment trapping in the Changjiang Estuary: Observations in the North Passage over a spring-neap tidal cycle. \u003cem\u003eEstuarine, Coastal and Shelf Science\u003c/em\u003e, 177, 8\u0026ndash;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ecss.2016.05.004\u003c/span\u003e\u003cspan address=\"10.1016/j.ecss.2016.05.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLi, Z., Yan, Y., Zhao, Z., Xu, Y., \u0026amp; Hu, Y. (2025). Numerical study on hydrodynamic effects of intermittent or sinusoidal coordination of pectoral fins to achieve spontaneous nose-up pitching behavior in dolphins. \u003cem\u003eOcean Engineering\u003c/em\u003e, 337, 121854. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.oceaneng.2025.121854\u003c/span\u003e\u003cspan address=\"10.1016/j.oceaneng.2025.121854\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaciel, D. A., Barbosa, C. C. F., de Moraes Novo, E. M. L., J\u0026uacute;nior, R. F., \u0026amp; Begliomini, F. N. (2021). Water clarity in Brazilian water assessed using Sentinel-2 and machine learning methods. \u003cem\u003eISPRS Journal of Photogrammetry and Remote Sensing\u003c/em\u003e, 182, 134\u0026ndash;152. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.isprsjprs.2021.10.009\u003c/span\u003e\u003cspan address=\"10.1016/j.isprsjprs.2021.10.009\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcFeeters, S. K. (1996). The use of the Normalized Difference Water Index (NDWI) in the delineation of open water features. \u003cem\u003eInternational journal of remote sensing\u003c/em\u003e, 17(7), 1425\u0026ndash;1432. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01431169608948714\u003c/span\u003e\u003cspan address=\"10.1080/01431169608948714\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePan, F., Wang, B., Zhang, Y., Zhang, X., Cai, Y., Liu, H., \u0026hellip; Wang, X. (2022). Bioavailability,(im) mobilization kinetics, and spatiotemporal patterns of arsenic and cadmium in surficial sediments of a river\u0026ndash;estuary\u0026ndash;coast system. \u003cem\u003eJournal of Hydrology\u003c/em\u003e, 612, 128140. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jhydrol.2022.128140\u003c/span\u003e\u003cspan address=\"10.1016/j.jhydrol.2022.128140\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePoliti, E., Cutler, M. E. J., \u0026amp; Rowan, J. S. (2015). Evaluating the spatial transferability and temporal repeatability of remote-sensing-based lake water quality retrieval algorithms at the European scale: a meta-analysis approach. \u003cem\u003eInternational Journal of Remote Sensing\u003c/em\u003e, \u003cem\u003e36\u003c/em\u003e(11), 2995\u0026ndash;3023. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/01431161.2015.1054962\u003c/span\u003e\u003cspan address=\"10.1080/01431161.2015.1054962\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eQiu, Z. (2013). A simple optical model to estimate suspended particulate matter in Yellow River Estuary. \u003cem\u003eOptics Express\u003c/em\u003e, 21(23), 27891\u0026ndash;27904. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1364/OE.21.027891\u003c/span\u003e\u003cspan address=\"10.1364/OE.21.027891\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSingh, A. T., Sharma, P., Sharma, C., Laluraj, C. M., Patel, L., Pratap, B., \u0026hellip; Thamban, M. (2020). Water discharge and suspended sediment dynamics in the Chandra River, Western Himalaya. \u003cem\u003eJournal of Earth System Science\u003c/em\u003e, 129(1), 206. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s12040-020-01455-4\u003c/span\u003e\u003cspan address=\"10.1007/s12040-020-01455-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSun, X., Guo, L., Luo, D., Yu, R. Q., Yu, X., Liang, Y., \u0026hellip; Wu, Y. (2022). Long-term increase in mortality of Indo-Pacific humpback dolphins (Sousa chinensis) in the Pearl River Estuary following anthropic activities: Evidence from the stranded dolphin mortality analysis from 2003 to 2017. \u003cem\u003eEnvironmental Pollution\u003c/em\u003e, \u003cem\u003e307\u003c/em\u003e, 119526. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.envpol.2022.119526\u003c/span\u003e\u003cspan address=\"https://doi.org/10.1016/j.envpol.2022.119526\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e Tu, Y., Yao, Q., Zhou, Y., \u0026amp; Zhu, Z. (2025). Prediction of maximum dynamic shear modulus of undisturbed marine soils in the eastern coast of China based on machine learning methods. \u003cem\u003eOcean Engineering\u003c/em\u003e, 321, 120382. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.oceaneng.2025.120382\u003c/span\u003e\u003cspan address=\"10.1016/j.oceaneng.2025.120382\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWang, H., Zhang, C., Zhou, B., Xue, S., Jia, P., \u0026amp; Zhu, X. (2023). Prediction of triaxial mechanical properties of rocks based on mesoscopic finite element numerical simulation and multi-objective machine learning. \u003cem\u003eJournal of King Saud University-Science\u003c/em\u003e, 35(7), 102846. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jksus.2023.102846\u003c/span\u003e\u003cspan address=\"10.1016/j.jksus.2023.102846\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, G., Wang, X. H., Zhong, Y., \u0026amp; Oliver, T. S. (2022). Modelling study on the sediment dynamics and the formation of the flood-tide delta near Cullendulla Beach in the Batemans Bay, Australia. \u003cem\u003eMarine Geology\u003c/em\u003e, 452, 106910. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.margeo.2022.106910\u003c/span\u003e\u003cspan address=\"10.1016/j.margeo.2022.106910\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYang, W., Matsushita, B., Chen, J., \u0026amp; Fukushima, T. (2011). Estimating constituent concentrations in case II waters from MERIS satellite data by semi-analytical model optimizing and look-up tables. \u003cem\u003eRemote sensing of environment\u003c/em\u003e, 115(5), 1247\u0026ndash;1259. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2011.01.007\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2011.01.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, M., Tang, J., Dong, Q., Song, Q., \u0026amp; Ding, J. (2010). Retrieval of total suspended matter concentration in the Yellow and East China Seas from MODIS imagery. \u003cem\u003eRemote Sensing of Environment\u003c/em\u003e, 114(2), 392\u0026ndash;403. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.rse.2009.09.016\u003c/span\u003e\u003cspan address=\"10.1016/j.rse.2009.09.016\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang, X., Qiao, W., Lu, Y., Huang, J., \u0026amp; Xiao, Y. (2023). Quantitative analysis of the influence of the Xiaolangdi reservoir on water and sediment in the middle and lower reaches of the Yellow River. \u003cem\u003eInternational Journal of Environmental Research and Public Health\u003c/em\u003e, 20(5), 4351. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/ijerph20054351\u003c/span\u003e\u003cspan address=\"10.3390/ijerph20054351\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\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":"[email protected]","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":"satellite remote sensing, marine construction, random forest, environmental monitoring","lastPublishedDoi":"10.21203/rs.3.rs-8033000/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8033000/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eTo address the threat of significantly elevated suspended sediment concentration (SSC) and its diffusion to coastal ecosystems caused by intense seabed sediment disturbance during large cross-sea transportation infrastructure construction. Using GF-5B hyperspectral satellite images and synchronous in-situ data, we developed SSC remote sensing inversion models for dredging-disturbed environments, including Random Forest (RF), Support Vector Machine (SVM), Linear Regression (LR) and empirical models. By comparing these models, we found that the RF model outperformed other models, with the RF model performing best (test set R\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.773, RMSE\u0026thinsp;=\u0026thinsp;0.062 kg/m\u003csup\u003e3\u003c/sup\u003e). By using the RF model to invert SSC from satellite data, we found that there was a high SSC in the dredging area (0.2-0.4684 kg/m\u003csup\u003e3\u003c/sup\u003e) and a low level in surrounding areas (\u0026lt;\u0026thinsp;0.15 kg/m\u003csup\u003e3\u003c/sup\u003e), with obvious spatial differences. This confirms that GF-5B hyperspectral data combined with machine learning enables high-precision SSC monitoring in dredging areas, providing a scientific basis for marine engineering ecological risk early warning and environmental protection optimization.\u003c/p\u003e","manuscriptTitle":"Using Machine Learning and Hyperspectral Satellite to Monitor Suspended Sediment Concentration under Dredging Engineering","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-08 15:35:25","doi":"10.21203/rs.3.rs-8033000/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","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":"01b67f75-9515-411f-a397-0fa271951afa","owner":[],"postedDate":"December 8th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-13T13:53:38+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-08 15:35:25","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8033000","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8033000","identity":"rs-8033000","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}