Scholarlink

<p>Chlorophyll-a, as an indicator of phytoplankton biomass, holds fundamental significance in qualitative and ecological studies of aquatic environments. Considering the unique advantages of remote sensing technology, this study employed MODIS-Aqua sensor images to estimate chlorophyll-a concentrations along the southern coast of the Caspian Sea. A key initial finding was the inapplicability of standard NASA Level-2 products (both Chl-a and Rrs), which failed to yield any valid data for our study area due to prohibitive quality flags, highlighting the limitations of standard atmospheric correction algorithms in this optically complex region. Therefore, an end-to-end approach using artificial neural networks (ANNs) was developed to model the relationship directly from Level-1B calibrated radiance data. Various structures of multilayer feedforward neural networks were analyzed, with a focus on the impact of the number of training samples on optimizing network performance. A final dataset of 176 matchups, derived from in-situ field observations, was utilized for training and evaluating the network’s performance. The optimal architecture was identified as an optimized two-hidden-layer neural network. This model successfully established a statistically significant predictive relationship where standard methods failed, achieving an accuracy with an RMSE = 0.2245&#xa0;mg/m³ and an R² = 0.087. The analysis confirmed that while the relationship between raw radiance and Chl-a is complex, the ANN approach provides a viable and necessary strategy for monitoring these challenging coastal waters.</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

MODIS-Derived Chlorophyll-a in the Caspian Sea Using Artificial Neural Networks

  • 期刊论文

  • 发表日期：

    2026年1月2日

查看全文
  • Hajar Sadat Alizadeh Moghaddam,
  • Siamak Boudaghpour,
  • Ali Akbar Abkar,
  • Mehdi Gholamalifard

摘要

Chlorophyll-a, as an indicator of phytoplankton biomass, holds fundamental significance in qualitative and ecological studies of aquatic environments. Considering the unique advantages of remote sensing technology, this study employed MODIS-Aqua sensor images to estimate chlorophyll-a concentrations along the southern coast of the Caspian Sea. A key initial finding was the inapplicability of standard NASA Level-2 products (both Chl-a and Rrs), which failed to yield any valid data for our study area due to prohibitive quality flags, highlighting the limitations of standard atmospheric correction algorithms in this optically complex region. Therefore, an end-to-end approach using artificial neural networks (ANNs) was developed to model the relationship directly from Level-1B calibrated radiance data. Various structures of multilayer feedforward neural networks were analyzed, with a focus on the impact of the number of training samples on optimizing network performance. A final dataset of 176 matchups, derived from in-situ field observations, was utilized for training and evaluating the network’s performance. The optimal architecture was identified as an optimized two-hidden-layer neural network. This model successfully established a statistically significant predictive relationship where standard methods failed, achieving an accuracy with an RMSE = 0.2245 mg/m³ and an R² = 0.087. The analysis confirmed that while the relationship between raw radiance and Chl-a is complex, the ANN approach provides a viable and necessary strategy for monitoring these challenging coastal waters.