Gap filling in long-term AOD time series using Iterative Singular Spectrum Analysis at a tropical AERONET site

Aerosol optical depth (AOD) is widely used in studies related to atmospheric monitoring, climate variability, and air quality, with the Aerosol RObotic NETwork (AERONET) providing high-quality records. However, AERONET AOD time series frequently contain observational gaps associated with cloud cover, especially in tropical regions characterized by strong hydrological seasonality. In this study, the performance of Iterative Singular Spectrum Analysis (ISSA) was evaluated for gap-filling an incomplete AOD time series from the AERONET Cuiabá-Miranda station (AOD-AERONET) in the Central-West region of Brazil. Daily AOD-AERONET data between 2005 and 2024 were used, with approximately $$47%$$ 47 % of the daily observations identified as gaps and showing a predominantly seasonal distribution. The application of ISSA to the gap-filling of the AOD series involved the evaluation of different initialization and cross-validation strategies using model performance metrics, as well as different reconstruction parameters. The results showed that the climatological initialization, based on the Day of the Year (DOY) mean, outperformed the global-mean initialization, producing smaller errors and greater temporal consistency of the reconstruction, and that the cross-validation design plays an important role in the selection of optimal parameters.