A Novel Reconstruction Method for Missing Chlorophyll-A Concentration

Cloud cover often results in incomplete satellite-derived chlorophyll-a (Chl-a) data, posing challenges for practical applications. This study introduces a novel approach, Data INterpolation Masking Autoencoders (DINMAE), to reconstruct missing Chl-a data. The performance of DINMAE was evaluated using Himawari-9 Chl-a product data and compared with existing methods. The algorithm demonstrates its efficacy with an R
²
of 0.998 and an MSE of 0.001, significantly outperforming other advanced algorithms by 4 to 9 times in terms of MSE. Detailed qualitative and quantitative analyses confirm that DINMAE proficiently reconstructs missing data and accurately captures minute variations in Chl-a levels. The findings suggest that DINMAE is an exceptionally effective algorithm for data reconstruction, particularly suited for oceanic data products.