Quantification of Diffusive Methane Emissions from a Large Eutrophic Lake with Satellite Imagery

Area of Science:

• Environmental Science
• Remote Sensing
• Climate Science

Background:

• Lakes are significant sources of methane (CH4), a potent greenhouse gas.
• Quantifying lake CH4 emissions is challenging due to high spatial and temporal variability.
• Satellite remote sensing offers a promising approach for large-scale, high-resolution emission monitoring.

Purpose of the Study:

• To develop and validate a satellite-based method for predicting diffusive methane emissions from lakes.
• To assess the performance of various machine learning models in estimating CH4 fluxes.
• To identify key environmental drivers influencing lake methane emissions.

Main Methods:

• Utilized Aqua/MODIS satellite imagery (2003-2020) and in situ data (2011-2017) from Lake Taihu.
• Compared eight machine learning models for predicting diffusive CH4 emissions.
• Employed the random forest (RF) model, validated with satellite-derived variables (chlorophyll a, water temperature, light attenuation, PAR).

Main Results:

• The random forest model demonstrated the highest accuracy (R² = 0.65, MRE = 21%) in predicting CH4 emissions.
• Satellite-derived variables (chlorophyll a, water temperature, diffuse attenuation coefficient, PAR) were key predictors.
• Reconstructed historical CH4 emissions (2003-2020) showed a long-term increase linked to climate warming and algal blooms.

Conclusions:

• Satellite remote sensing, particularly with the RF model, can effectively map lake CH4 emissions.
• Environmental factors like water temperature and algal presence mechanistically drive CH4 emissions.
• This approach provides crucial spatiotemporal data for understanding aquatic greenhouse gas budgets and climate change impacts.