Correction for the Sun-Angle Effect on the NDVI Based on Path Length

• Published in: IEEE transactions on geoscience and remote sensing Vol. 61; pp. 1 - 17
• Main Authors: Liu, Xinli, Chen, Yaoyao, Mu, Xihan, Yan, Guangjian, Xie, Donghui, Ma, Xuanlong, Yan, Kai, Song, Wanjuan, Liu, Zhigang
• Format: Journal Article
• Language: English
• Published: New York IEEE 2023; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Bidirectional reflectance; Bidirectional reflectance distribution function (BRDF); Coefficient of variation; Data models; Distribution functions; Error reduction; Geometry; kernel-driven model; MODIS; normalized difference vegetation index (NDVI); Normalized difference vegetative index; path length (PL); Reflectance; Reflectivity; Remote monitoring; Remote sensing; Root-mean-square errors; Soil; sun-angle effect; Swath width; Trigonometric functions; Variation; Vegetation; Vegetation mapping
• ISSN: 0196-2892; 1558-0644
• DOI: 10.1109/TGRS.2023.3322780

Changes in the sun zenith angle (SZA) alter the normalized difference vegetation index (NDVI) and introduce uncertainties into the estimation of vegetation biochemical and biophysical parameters. For the NDVI obtained from narrow swath width sensors, there is not a unified and easy-to-use approach to correct the sun-angle effect. In this study, the cosine correction model (CCM) was proposed to reduce the sun-angle effect on NDVI based on the path length (PL) of light calculated from the SZA without the need for multiangle observations. The PL was found to be closely correlated to the simple ratio vegetation index (SR) and can mitigate the impact on the NDVI caused by SZA variations. The CCM performed well when correcting the sun-angle effect on NDVI for different types of data. After correction for the simulated data (e.g., the reference SZA of 10°), the coefficient of variation (CV) of the NDVI concerning SZA variations from 10° to 60° was reduced by 5.42%, and the root-mean-square error (RMSE) was reduced by 0.049. For the field-measured data, the CV of the NDVI under various SZAs was reduced by up to 5.55% after correction, and the maximum difference between the uncorrected and corrected NDVI was 0.099. The RMSE of corrected nadir NDVI from MODIS satellite data was reduced by 34.2% on average. The CCM, as an easily implemented method, can attenuate the sun-angle effect on NDVI without relying on the bidirectional reflectance distribution function (BRDF) products and hence has the potential to improve the accuracy of remote sensing monitoring of vegetation dynamics.