The Impact of Choice of Solar Spectral Irradiance Model on Atmospheric Correction of Landsat 8 OLI Satellite Data

Many atmospheric correction schemes of radiancebased optical satellite data require the selection of normalized solar spectral irradiance models at the top of atmosphere (TOA). However, there is no scientific consensus in literature as to which available model is most suitable. This article examines five commonly used models applied to Landsat 8 Operational Land Imager (OLI) TOA radiance and reflectance products to assess the accuracy and stability between models used to derive surface reflectance products. It is assumed that the calibration of the United States Geological Survey (USGS) Landsat 8 OLI TOA reflectance and radiance products are accurate to currently claimed levels. The results show that the retrieved surface reflectance can exhibit significant variations when different solar irradiance models are used, especially in the OLI coastal blue band at 443 nm. From the five solar irradiance models, the Kurucz 2005 model showed the least bias compared with OLI TOA reflectance product and least variance in surface reflectance. Furthermore, improvement was obtained by adjusting the total solar irradiance (TSI) normalization, and additional validation was provided using observed in situ water leaving reflectance data. The results from this article are particularly relevant to aquatic applications and to satellite sensors that provide TOA radiance such as previous Landsat and other current and historical missions.

Automated summary

This study examined the impact of five commonly used solar irradiance models on the correction of Landsat 8 OLI satellite data. The results showed significant variations in surface reflectance when different solar irradiance models were used. The Kurucz 2005 model was found to have the least bias and variance compared to OLI TOA reflectance product.