Scholarlink

<p>This study investigates the efficacy of combining spectral unmixing techniques with Convolutional Neural Networks (CNN) to assess water quality parameters, specifically chlorophyll-a and turbidity, in Chembarapakkam, Puzhal, and Poondi lakes across different seasons. Traditional in situ measurements, while accurate, often fall short in capturing the spatial and temporal variability of these parameters across large water bodies. To address this limitation, spectral unmixing was employed first to analyze satellite imagery, providing a foundational understanding of the spatial distribution and seasonal variations of chlorophyll-a and turbidity. Building on the insights from spectral unmixing, a CNN model was developed to refine these predictions and enhance accuracy. The CNN model demonstrated high precision and reliability, with R² values ranging from 0.9595 to 0.9898 for chlorophyll-a and 0.9781 to 0.9932 for turbidity and consistently low RMSE values below one across all seasons. Integrating spectral unmixing data into the CNN framework significantly improved model accuracy, particularly in handling complex environmental dynamics. Notably, post-monsoon heavy rainfall led to increased runoff, elevating turbidity levels and reducing chlorophyll-a concentrations. This combined approach offers a comprehensive method for monitoring water quality, proving highly effective in supporting the management of freshwater ecosystems.</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

Lake Water Quality Assessment Using a Combined Approach of Multispectral Unmixing and CNN Technique

  • 期刊论文

  • 发表日期：

    2026年4月18日

查看全文
  • Raghul Mageswaran,
  • Porchelvan Ponnusamy

摘要

This study investigates the efficacy of combining spectral unmixing techniques with Convolutional Neural Networks (CNN) to assess water quality parameters, specifically chlorophyll-a and turbidity, in Chembarapakkam, Puzhal, and Poondi lakes across different seasons. Traditional in situ measurements, while accurate, often fall short in capturing the spatial and temporal variability of these parameters across large water bodies. To address this limitation, spectral unmixing was employed first to analyze satellite imagery, providing a foundational understanding of the spatial distribution and seasonal variations of chlorophyll-a and turbidity. Building on the insights from spectral unmixing, a CNN model was developed to refine these predictions and enhance accuracy. The CNN model demonstrated high precision and reliability, with R² values ranging from 0.9595 to 0.9898 for chlorophyll-a and 0.9781 to 0.9932 for turbidity and consistently low RMSE values below one across all seasons. Integrating spectral unmixing data into the CNN framework significantly improved model accuracy, particularly in handling complex environmental dynamics. Notably, post-monsoon heavy rainfall led to increased runoff, elevating turbidity levels and reducing chlorophyll-a concentrations. This combined approach offers a comprehensive method for monitoring water quality, proving highly effective in supporting the management of freshwater ecosystems.