Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI

In the process of radiometric calibration, the corrections for bidirectional reflectance distribution functions (BRDFs) and spectral band adjustment factors (SBAFs) are crucial. Time-series MODIS images are commonly used to construct BRDFs by using the Ross–Li model in current research. However, the...

Full description

Saved in:
Bibliographic Details
Main Authors: Xiang Zhou, Yidan Chen, Yong Xie, Jie Han, Wen Shao
Format: Article
Language:English
Published: MDPI AG 2024-11-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/16/22/4337
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1846152519376437248
author Xiang Zhou
Yidan Chen
Yong Xie
Jie Han
Wen Shao
author_facet Xiang Zhou
Yidan Chen
Yong Xie
Jie Han
Wen Shao
author_sort Xiang Zhou
collection DOAJ
description In the process of radiometric calibration, the corrections for bidirectional reflectance distribution functions (BRDFs) and spectral band adjustment factors (SBAFs) are crucial. Time-series MODIS images are commonly used to construct BRDFs by using the Ross–Li model in current research. However, the Ross–Li BRDF model is based on the linear relationship between the kernel models and is unable to take into account the nonlinear relationship between them. Furthermore, when using SBAF to account for spectral difference, a radiative transfer model is often used, but it requires many parameters to be set, which may introduce more errors and reduce the calibration accuracy. To address these issues, the random forest algorithm and a spectral interpolation convolution method using the Sentinel-2/multispectral instrument (MSI) are proposed in this study, in which the HuanJing-2A (HJ-2A)/charge-coupled device (CCD3) sensor is taken as an example, and the Dunhuang radiometric calibration site (DRCS) is used as a radiometric delivery platform. Firstly, a BRDF model by using the random forest algorithm of the DRCS is constructed using time-series MODIS images, which corrects the viewing geometry difference. Secondly, the BRDF correction coefficients, MSI reflectance, and relative spectral responses (RSRs) of CCD3 are used to correct the spectral differences. Finally, with the validation results, the maximum relative error between the calibration results of the proposed method and the official calibration coefficients (OCCs) published by the China Centre for Resources Satellite Data and Application (CRESDA) is 3.38%. When tested using the Baotou sandy site, the proposed method is better than the OCCs of the average relative errors calculated for all the bands except for the near-infrared (NIR) band, which has a larger error. Additionally, the effects of the light-matching method and the radiative transfer method, different approaches to constructing the BRDF model, using SBAF to account for spectral differences, different BRDF sources, as well as the imprecise viewing geometrical parameters, spectral interpolation method, and geometric positioning error, on the calibration results are analyzed. Results indicate that the cross-calibration coefficients obtained using the random forest algorithm and the proposed spectral interpolation method are more applicable to the CCD3; thus, they also account for the nonlinear relationships between the kernel models and reduce the error due to the radiative transfer model. The total uncertainty of the proposed method in all bands is less than 5.16%.
format Article
id doaj-art-c44bc43637da409fa5c455b776e1b7d5
institution Kabale University
issn 2072-4292
language English
publishDate 2024-11-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-c44bc43637da409fa5c455b776e1b7d52024-11-26T18:20:28ZengMDPI AGRemote Sensing2072-42922024-11-011622433710.3390/rs16224337Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSIXiang Zhou0Yidan Chen1Yong Xie2Jie Han3Wen Shao4College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, ChinaSchool of Geographical Sciences, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Geographical Sciences, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaSchool of Geography and Geomatics, Xuchang University, Xuchang 461000, ChinaSchool of Geographical Sciences, Nanjing University of Information Science and Technology, Nanjing 210044, ChinaIn the process of radiometric calibration, the corrections for bidirectional reflectance distribution functions (BRDFs) and spectral band adjustment factors (SBAFs) are crucial. Time-series MODIS images are commonly used to construct BRDFs by using the Ross–Li model in current research. However, the Ross–Li BRDF model is based on the linear relationship between the kernel models and is unable to take into account the nonlinear relationship between them. Furthermore, when using SBAF to account for spectral difference, a radiative transfer model is often used, but it requires many parameters to be set, which may introduce more errors and reduce the calibration accuracy. To address these issues, the random forest algorithm and a spectral interpolation convolution method using the Sentinel-2/multispectral instrument (MSI) are proposed in this study, in which the HuanJing-2A (HJ-2A)/charge-coupled device (CCD3) sensor is taken as an example, and the Dunhuang radiometric calibration site (DRCS) is used as a radiometric delivery platform. Firstly, a BRDF model by using the random forest algorithm of the DRCS is constructed using time-series MODIS images, which corrects the viewing geometry difference. Secondly, the BRDF correction coefficients, MSI reflectance, and relative spectral responses (RSRs) of CCD3 are used to correct the spectral differences. Finally, with the validation results, the maximum relative error between the calibration results of the proposed method and the official calibration coefficients (OCCs) published by the China Centre for Resources Satellite Data and Application (CRESDA) is 3.38%. When tested using the Baotou sandy site, the proposed method is better than the OCCs of the average relative errors calculated for all the bands except for the near-infrared (NIR) band, which has a larger error. Additionally, the effects of the light-matching method and the radiative transfer method, different approaches to constructing the BRDF model, using SBAF to account for spectral differences, different BRDF sources, as well as the imprecise viewing geometrical parameters, spectral interpolation method, and geometric positioning error, on the calibration results are analyzed. Results indicate that the cross-calibration coefficients obtained using the random forest algorithm and the proposed spectral interpolation method are more applicable to the CCD3; thus, they also account for the nonlinear relationships between the kernel models and reduce the error due to the radiative transfer model. The total uncertainty of the proposed method in all bands is less than 5.16%.https://www.mdpi.com/2072-4292/16/22/4337HuanJing-2A (HJ-2A)Sentinel-2cross-calibrationbidirectional reflectance distribution function (BRDF)random forest algorithmspectral difference
spellingShingle Xiang Zhou
Yidan Chen
Yong Xie
Jie Han
Wen Shao
Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI
Remote Sensing
HuanJing-2A (HJ-2A)
Sentinel-2
cross-calibration
bidirectional reflectance distribution function (BRDF)
random forest algorithm
spectral difference
title Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI
title_full Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI
title_fullStr Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI
title_full_unstemmed Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI
title_short Radiometric Cross-Calibration of HJ-2A/CCD3 Using the Random Forest Algorithm and a Spectral Interpolation Convolution Method with Sentinel-2/MSI
title_sort radiometric cross calibration of hj 2a ccd3 using the random forest algorithm and a spectral interpolation convolution method with sentinel 2 msi
topic HuanJing-2A (HJ-2A)
Sentinel-2
cross-calibration
bidirectional reflectance distribution function (BRDF)
random forest algorithm
spectral difference
url https://www.mdpi.com/2072-4292/16/22/4337
work_keys_str_mv AT xiangzhou radiometriccrosscalibrationofhj2accd3usingtherandomforestalgorithmandaspectralinterpolationconvolutionmethodwithsentinel2msi
AT yidanchen radiometriccrosscalibrationofhj2accd3usingtherandomforestalgorithmandaspectralinterpolationconvolutionmethodwithsentinel2msi
AT yongxie radiometriccrosscalibrationofhj2accd3usingtherandomforestalgorithmandaspectralinterpolationconvolutionmethodwithsentinel2msi
AT jiehan radiometriccrosscalibrationofhj2accd3usingtherandomforestalgorithmandaspectralinterpolationconvolutionmethodwithsentinel2msi
AT wenshao radiometriccrosscalibrationofhj2accd3usingtherandomforestalgorithmandaspectralinterpolationconvolutionmethodwithsentinel2msi