Scholarlink

<p>Wetlands, being critical biodiversity hotspots, are facing severe threats from anthropogenic activities globally. This study examined the water quality of Deepor Beel, a Ramsar site in Assam, India, through an integrated approach combining Water Quality Index (WQI), statistical correlation and ensemble machine learning modelling. Water samples were collected during pre- and post-monsoon seasons from ten sites to analyze 20 physicochemical and heavy metal parameters using standard methods. The WQI was computed using the weighted arithmetic method and correlation analysis was performed to assess relationships among parameters. Random Forest (RF) and Bagging Random Forest (B-RF) algorithms were employed to predict WQI, with B-RF showing superior performance (<i>R</i> = 0.916, RMSE = 0.0315) during the post-monsoon season. The results revealed significant spatial and seasonal variations in water quality, with sites 5, 6, 8, 9 and 10 near dumping grounds and urban inflows exhibiting poor water quality due to elevated levels of biochemical oxygen demand (BOD), chemical oxygen demand (COD), total suspended solids (TSS), dissolved oxygen (DO), copper, arsenic. Thus, these areas require priority management to restore the wetland’s ecological integrity. The proposed framework demonstrates the potential of integrating conventional WQI assessment with ensemble ML models for predictive water quality monitoring and management in Ramsar wetlands globally.</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

Assessing and Predicting Wetland Water Quality Using Water Quality Index and Ensemble Machine Learning Models: A Case Study of Deepor Beel (Ramsar Site), Assam, India

  • 期刊论文

  • 发表日期：

    2026年1月7日

查看全文
  • Tamal Kanti Saha,
  • Haroon Sajjad,
  • Yatendra Sharma,
  • Manjit Das,
  • Rayees Ali

摘要

Wetlands, being critical biodiversity hotspots, are facing severe threats from anthropogenic activities globally. This study examined the water quality of Deepor Beel, a Ramsar site in Assam, India, through an integrated approach combining Water Quality Index (WQI), statistical correlation and ensemble machine learning modelling. Water samples were collected during pre- and post-monsoon seasons from ten sites to analyze 20 physicochemical and heavy metal parameters using standard methods. The WQI was computed using the weighted arithmetic method and correlation analysis was performed to assess relationships among parameters. Random Forest (RF) and Bagging Random Forest (B-RF) algorithms were employed to predict WQI, with B-RF showing superior performance (R = 0.916, RMSE = 0.0315) during the post-monsoon season. The results revealed significant spatial and seasonal variations in water quality, with sites 5, 6, 8, 9 and 10 near dumping grounds and urban inflows exhibiting poor water quality due to elevated levels of biochemical oxygen demand (BOD), chemical oxygen demand (COD), total suspended solids (TSS), dissolved oxygen (DO), copper, arsenic. Thus, these areas require priority management to restore the wetland’s ecological integrity. The proposed framework demonstrates the potential of integrating conventional WQI assessment with ensemble ML models for predictive water quality monitoring and management in Ramsar wetlands globally.