Improving atmospheric correction for KOMPSAT-3A by optimizing the 6SV look-up table

We present techniques for enhancing the look-up table (LUT) method, using the Second Simulation of a Satellite Signal in the Solar Spectrum Vector (6 SV) model for atmospheric correction (AC) and accurate surface reflectance (SR) computation across the multispectral (MS) KOMPSAT-3A channels. We impr...

Full description

Saved in:
Bibliographic Details
Published in:Remote sensing letters 2024-05, Vol.15 (5), p.514-525
Main Authors: Jung, Daeseong, Seong, Noh-Hun, Choi, Sungwon, Sim, Suyoung, Woo, Jongho, Kim, Nayeon, Park, Sungwoo, Han, Kyung-Soo
Format: Article
Language:English
Subjects:
6SV
Artificial neural networks
Atmospheric correction
Atmospheric effects
Channels
Computation
Interpolation
KOMPSAT-3A
look-up table
Lookup tables
Neural networks
Optical analysis
Optical thickness
radiative transfer
Reflectance
Remote sensing
Wavelengths
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:We present techniques for enhancing the look-up table (LUT) method, using the Second Simulation of a Satellite Signal in the Solar Spectrum Vector (6 SV) model for atmospheric correction (AC) and accurate surface reflectance (SR) computation across the multispectral (MS) KOMPSAT-3A channels. We improved the LUT-based SR accuracy using three interpolation and prediction methods for AC coefficients: minimum curvature surface (MCS), six-dimensional linear interpolation (6D), and a deep neural network (DNN). When assessed based on the solar zenith angle (SZA) and aerosol optical depth (AOD), MCS had limitations interpolating atmospheric effects, particularly at shorter wavelengths. The DNN method had high predictive ability for atmospheric effect variability with an increased SZA, but struggled to predict minor changes at low SZAs. In comparison, the 6D method, operating in real-time and considering all AC input variables, consistently retrieved high-quality SR across all MS channels. Our findings offer insights into each method's strengths and limitations, guiding future remote sensing research and applications.
ISSN:2150-704X
2150-7058
DOI:10.1080/2150704X.2024.2343130