Inversion diffuse attenuation coefficient of photosynthetically active radiation based on deep learning

Area of Science:

• Oceanography
• Remote Sensing
• Biogeochemistry

Background:

• Accurate estimation of K_d(PAR) is crucial for understanding aquatic physical, chemical, and biological processes.
• Traditional methods for estimating K_d(PAR) have limitations in stability and accuracy.

Purpose of the Study:

• To develop and evaluate a deep learning model (DLKPAR) for improved remote estimation of K_d(PAR).

Main Methods:

• A deep learning model (DLKPAR) was developed for K_d(PAR) estimation.
• The model's performance was evaluated using in situ NOMAD and Argo data, compared against MODIS data.

Main Results:

• DLKPAR demonstrated superior accuracy and stability compared to traditional empirical and semi-analytical algorithms.
• DLKPAR achieved lower RMSD (0.028 m^-1) and MARD (0.14), with a higher R² (0.94) than existing methods.
• The model showed improved K_d(PAR) inversion accuracy when matched with NOMAD, Argo, and MODIS data.

Conclusions:

• DLKPAR provides a reliable method for estimating K_d(PAR) from ocean color remote sensing.
• The model facilitates accurate assessment of visible light attenuation in the upper ocean, aiding biogeochemical research.