A Daily, Multi-scale (10m/2.5m) Multispectral and Vegetation Index Dataset of Chenba'erhu Banner, Inner Mongolia (2024)

Fine-scale grassland monitoring is crucial for ecological conservation and sustainable animal husbandry, but has long been constrained by the spatio-temporal resolution trade-off of remote sensing data. This research produced a daily multi-scale remote sensing dataset covering the 2024 growing season (March to October) in Chenba'herhu Banner, Inner Mongolia. The dataset includes a 10 m product covering the entire banner and a 2.5 m enhanced product for the Bayankuren Town, both containing multispectral imagery and vegetation indices (NDVI, EVI). Based on MODIS MCD43A4 and Sentinel-2 L2A products, this research reconstructed a continuous MODIS time series using the Whittaker smoother algorithm. Subsequently, a daily seamless 10 m dataset was generated through a "predict-and-update" chain spatio-temporal fusion strategy. Finally, using Bayan Kuren Town as a demonstration area, the 10 m imagery was enhanced to 2.5 m resolution using the SEN2SR super-resolution model. The quality assessment indicates high accuracy: the cross-validation for MODIS data reconstruction yielded an average coefficient of determination (R²) greater than 0.983. The indirect validation of the super-resolution model showed an average Structural Similarity Index (SSIM) above 0.94 and a Spectral Angle Mapper (SAM) below 0.07 rad. This dataset can support for research on fine-scale phenology monitoring, and ecosystem model optimization.