Normalized Difference Vegetation Index Maps of Pure Pixels over China's mainland for Estimation of Fractional Vegetation Cover

Abstract. Fractional Vegetation Cover (FVC) is an important vegetation structure factor for applications in agriculture, forestry, ecology, etc. Due to its simplicity, the normalized difference vegetation index (NDVI)-based mixture model is widely used to estimate FVC from remotely sensed data. However, the accuracy and efficiency of FVC estimation require the precise calculation of two key parameters: the NDVI of fully covered vegetation and bare soil. Despite their importance, these two endmember NDVI values have not yet been produced as large-scale maps. Traditional statistical methods for obtaining endmember NDVI from satellite datasets highly rely on the assumption that a certain amount of pure pixels of vegetation and soil must be present, which is often invalid for many areas. This study generated 30 m resolution maps of endmember NDVI across China using the MultiVI algorithm incorporating multi-angle remote sensing data. The quality and accuracy of the endmember NDVI maps were evaluated using various validation data, including statistically obtained pure NDVI, soil spectra from a soil library, and field-measured FVC. The NDVI values for bare soil derived from the MultiVI algorithm were consistent with those obtained from the soil spectral library. Additionally, the FVC estimated using the MultiVI-derived endmember NDVI and the VI-based mixture model exhibited reasonable accuracy when compared to the field measurements. The root mean square deviation (RMSD) values for MultiVI FVC were below 0.13 in the Heihe, Hebei, and Three Gorges Reservoir regions of China. Furthermore, the MultiVI FVC outperformed those calculated using the statistical methods. The endmember NDVI maps provide a convenient and reliable source of key parameters for the accurate and rapid estimation of FVC at large scales. The 30 m pure NDVI maps are free access at https://zenodo.org/records/14060222 (Zhao et al., 2024).