Evaluation of surface reflectance retrieval over diverse surface types using SREM algorithm in varied aerosol conditions for coarse to medium resolution data from multiple spaceborne sensors

ABSTRACT

Surface reflectance (SR) is one of the most important components for deriving different biophysical parameters from remotely sensed data. For accurate retrieval of SR, it is essential to accurately estimate the degree of atmospheric alteration to the target signals reaching at-sensor. Simplified and robust SR estimation method (SREM) is one such approximate estimation method that could demonstrate the retrieval of SR from top-of-atmosphere (ToA) signals, without using any precalculated LUTs or aerosol information. The present study attempts to assess the accuracy of the SREM algorithm over varied Indian land surfaces using data from Landsat 8, Sentinel 2, AWiFS and LISS III. The results indicate that while SREM performs quite well for Sentinel 2 over most land surface types, the performance is good for Landsat 8 as well except over water and in coastal aerosol band. SREM retrievals also show high correlation for AWiFS and LISS III but the estimated values are much lower than level 2 SR due to smaller ToA reflectance values. Furthermore, while aerosol load does impact the performance adversely, it is least impactful for Sentinel 2 and LISS III datasets and most over Landsat 8, particularly over water. Additionally, while AWiFS and LISS III SR retrievals are underestimated at all spectral bands, retrieval is slightly underestimated from green to SWIR band in case of Sentinel 2 and SWIR band in case of Landsat 8. Overall, the findings demonstrate that while SREM will need to be modified to adapt to LISS III and AWiFS retrievals, it can, nevertheless, be readily used as a simplified method for quick and approximate estimation of SR over the Indian region using data from Landsat 8 and Sentinel 2.

KEYWORDS:

• SREM
• surface reflectance estimation
• Landsat 8
• Sentinel 2
• LISS III
• AWiFS

Acknowledgements

The authors are thankful to NRSC and IIRS for providing access to LISS III and AWiFS datasets. We are also thankful to USGS, ESA Copernicus, and MODIS teams for the datasets used in the study.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data Availability statement

LISS III and AWiFS can be ordered from National Remote Sensing Centre (NRSC), ISRO’s Bhoonidhi portal (https://bhoonidhi.nrsc.gov.in/bhoonidhi/index.html). Landsat 8 can be freely downloaded from USGS Earthexplorer (https://earthexplorer.usgs.gov/). Sentinel 2 can be freely downloaded from ESA Copernicus Open Access Hub (https://scihub.copernicus.eu/dhus/#/home). MODIS data is freely available at https://ladsweb.modaps.eosdis.nasa.gov/).