Scholarlink

<p>Suspended particulate matter (SPM) is a critical indicator of river water quality, influencing light availability, thermal regimes, and ecosystem functioning, yet its seasonal and interannual dynamics remain poorly quantified in many large river systems. The present study addresses this gap by conducting a seasonally resolved, multi-year (2019–2023) assessment of SPM concentration and surface water temperature (LST<sub>SW</sub>) in an understudied stretch of the Ganga River in Sahibganj, India, using Sentinel-2 and Landsat-8 OLI/TIRS imagery. The objectives were to (i) characterize seasonal and interannual variability in SPM under contrasting hydrological seasons, (ii) systematically compare the performance of six machine learning regression models for seasonal SPM prediction, and (iii) evaluate the contribution of spectral indices and ancillary variables to model accuracy. Results reveal pronounced seasonal variability, with reduced SPM concentrations in post-monsoon periods and lower variability in pre-monsoon seasons. Model evaluation across seasons demonstrates that ensemble tree-based approaches, especially random forest (RF) and bagged regression random forest (BGR<sub>RF</sub>), consistently outperform linear and boosting-based models, highlighting their robustness for satellite-based SPM prediction in riverine environments characterized by strong seasonal contrasts. While the analysis is constrained by the absence of in situ validation and relies on satellite-derived SPM proxies, the findings underscore the value of integrating multisensor remote sensing with ensemble machine learning for near-real-time and seasonal monitoring of sediment dynamics. Overall, this study presents a transferable and scalable framework for seasonally consistent SPM assessment, supporting improved river water quality monitoring and adaptive management in data-limited regions.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Potentials of machine learning models in estimating the seasonal variability of suspended particulate matters in river ecosystems

  • 期刊论文

  • 发表日期：

    2026年3月5日

查看全文
  • Akash Roy,
  • Kirti Avishek

摘要

Suspended particulate matter (SPM) is a critical indicator of river water quality, influencing light availability, thermal regimes, and ecosystem functioning, yet its seasonal and interannual dynamics remain poorly quantified in many large river systems. The present study addresses this gap by conducting a seasonally resolved, multi-year (2019–2023) assessment of SPM concentration and surface water temperature (LSTSW) in an understudied stretch of the Ganga River in Sahibganj, India, using Sentinel-2 and Landsat-8 OLI/TIRS imagery. The objectives were to (i) characterize seasonal and interannual variability in SPM under contrasting hydrological seasons, (ii) systematically compare the performance of six machine learning regression models for seasonal SPM prediction, and (iii) evaluate the contribution of spectral indices and ancillary variables to model accuracy. Results reveal pronounced seasonal variability, with reduced SPM concentrations in post-monsoon periods and lower variability in pre-monsoon seasons. Model evaluation across seasons demonstrates that ensemble tree-based approaches, especially random forest (RF) and bagged regression random forest (BGRRF), consistently outperform linear and boosting-based models, highlighting their robustness for satellite-based SPM prediction in riverine environments characterized by strong seasonal contrasts. While the analysis is constrained by the absence of in situ validation and relies on satellite-derived SPM proxies, the findings underscore the value of integrating multisensor remote sensing with ensemble machine learning for near-real-time and seasonal monitoring of sediment dynamics. Overall, this study presents a transferable and scalable framework for seasonally consistent SPM assessment, supporting improved river water quality monitoring and adaptive management in data-limited regions.