A novel index for directly indicating fractional vegetation cover based on spectral differences between vegetation and soil

The green fractional vegetation cover (FVC) is an essential parameter used to characterize the spatial pattern of vegetation coverage. Remote sensing provides the most efficient way to estimate FVC at regional and global scales. However, existing FVC-estimation approaches based on remote sensing fail to achieve high accuracy, broad applicability, and ease of use simultaneously, thus limiting their practical implementation. Based on the unique spectral shapes of green vegetation and soil within the visible to near-infrared spectrum (400–1000 nm), we proposed the vegetation coverage index (VCI), which is a novel index for directly indicating the FVC. VCI utilizes the spectral reflectance from the blue, green, red, and near-infrared bands to quantify the vegetation signal and soil signal as 1 and 0, respectively, and then establishes a quantitative relationship with FVC through the linear spectral mixing model. The performance of VCI in FVC estimation was first tested using simulated datasets generated by the radiative transfer model LESS under varying factors, including the vegetation structure, leaf area index, soil background, and solar zenith angle. It was then validated at 15 in-situ test sites in China, using UAV-derived reference FVC and Sentinel-2 surface reflectance data. These sites covered 10 vegetation and land cover types, 4 phenological phases, and 10 soil types. Additionally, VCI was compared against existing FVC products across another 40 in-situ comparative sites in China, using Landsat-8/9, Sentinel-3, and MODOCGA data at spatial resolutions of 30 m, 300 m, and 1000 m, respectively. Simulation results demonstrated that VCI performed comparably or slightly better than the dimidiate pixel model (DPM), reducing the root mean square error (RMSE) by 0.21 % to 14.42 %. Validation at 15 test sites showed that during the green-up to peak phase, when pixels are primarily composed of green vegetation and soil, VCI and DPM exhibited similar average accuracy (VCI: RMSE = 0.13; DPM: RMSE = 0.12). In contrast, during the peak to dormancy phase, with the presence of non-photosynthetic vegetation, VCI (RMSE = 0.11) clearly outperformed DPM (RMSE = 0.21), achieving a 46.8 % reduction in RMSE. At the 40 comparative sites, VCI yielded RMSE comparable to the MultiVI FVC product and outperformed the GEOV3 FVC and GLASS FVC products, with RMSE reductions of 20.00 % and 30.77 %, respectively. VCI provides a simple and efficient approach for FVC estimation through basic spectral band calculations. Moreover, VCI demonstrates broad applicability across widely used remote sensing sensors, including the Sentinel-2 Multispectral Instrument, Sentinel-3 Ocean and Land Colour Instrument, Landsat-8 Operational Land Imager, and Moderate-Resolution Imaging Spectroradiometer, showing strong potential for FVC monitoring across various spatial and temporal scales.