Related Experiment Video
Updated: Jun 19, 2026
Investigating the Relationship between Sea Surface Chlorophyll and Major Features of the South China Sea with Satellite Information
Published on: June 13, 2020
Satellite algorithms for retrieving dissolved organic carbon concentrations in Chinese lakes
Dong Liu1, Evangelos Spyrakos2, Andrew Tyler2
1Key Laboratory of Lake and Watershed Science for Water Security, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China; School of Biological and Environmental Science, University of Stirling, Stirling FK9 4LA, United Kingdom.
This study developed a hybrid machine learning algorithm to accurately estimate dissolved organic carbon (DOC) in lakes using satellite data. The new method improves upon traditional algorithms for large-scale water quality monitoring.
Area of Science:
- Environmental Science
- Remote Sensing
- Water Quality Monitoring
Background:
- Dissolved organic carbon (DOC) impacts lake water quality and carbon cycling.
- Existing regional algorithms struggle with large-scale DOC retrieval.
- Satellite-based monitoring requires accurate and scalable methods for DOC estimation.
Purpose of the Study:
- To investigate feasible satellite algorithms for retrieving DOC concentrations from OLCI/Sentinel-3 imagery across diverse Chinese lakes.
- To compare the performance of traditional regression methods with a novel hybrid machine learning approach.
- To map the spatial distribution of DOC concentrations in large Chinese lakes.
Main Methods:
- Bio-optical measurements from 55 lakes were analyzed.
- Relationships between DOC, colored dissolved organic matter absorption, and spectral reflectance were investigated for freshwater and saline lakes.
- Traditional linear regression and a hybrid machine learning algorithm were developed and validated.
- Satellite monitoring of 370 large lakes was conducted.
Main Results:
- Bio-optical characteristics and DOC-retrieval relationships differed significantly between freshwater and saline lakes.
- Traditional methods yielded high mean absolute percent differences (55.68-66.44%) for DOC estimation.
- The hybrid machine learning algorithm achieved a substantially lower MAPD of 18.16%.
- Spatial analysis revealed higher DOC concentrations in northwest China and lower concentrations in the southeast.
Conclusions:
- A hybrid machine learning approach offers a more reliable method for large-scale satellite-based DOC monitoring in lakes.
- Distinguishing between freshwater and saline lake characteristics is crucial for accurate DOC retrieval.
- Satellite monitoring provides valuable insights into the spatial patterns of lake DOC concentrations.
