[An artificial neural network model for lake color inversion using TM imagery].

Area of Science:

• Environmental Science
• Remote Sensing
• Water Quality Monitoring

Background:

• Lake water quality assessment is crucial for environmental management.
• Traditional monitoring methods can be labor-intensive and spatially limited.
• Satellite remote sensing offers a synoptic and efficient approach for water quality studies.

Purpose of the Study:

• To develop and validate an artificial neural network model for inversing lake water quality parameters from TM imagery.
• To assess the eutrophic status of lakes using remote sensing data.
• To investigate the accuracy and applicability of the developed model.

Main Methods:

• Construction of a Backpropagation (BP) neural network model.
• Utilizing Landsat TM imagery data for water quality parameter retrieval.
• Conducting satellite synchronous monitoring experiments for model calibration and validation.

Main Results:

• Successfully inversed concentrations of suspended solids (SS), chemical oxygen demand (CODMn), dissolved oxygen (DO), total nitrogen (T-N), total phosphorus (T-P), and chlorophyll-a (chlo-a).
• Achieved good accuracy with a relative error controllable below 25% for the inversed parameters.
• Analyzed sources of simulation error and proposed methods for model improvement.

Conclusions:

• The developed BP neural network model is effective for estimating lake water quality parameters from TM imagery.
• The model shows successful application in the investigation, analysis, and estimation of lake water quality based on small-scale experiments.
• This approach provides a valuable tool for routine lake water quality monitoring and eutrophication assessment.