Development of a hyperspectral spatio-temporal surface albedo dataset for Earth

Surface albedo is a crucial component of accurate radiative transfer simulations of Earth's system, playing a key role in calculating the planet's energy budget. The MODIS Surface Reflectance dataset (MCD43C3, Version 6.1) provides detailed albedo maps across seven spectral bands, enabling the monitoring of daily and yearly changes in planetary surface albedo. However, a comprehensive set of albedo maps covering the entire wavelength range is essential for simulating radiance spectra and accurately retrieving atmospheric and cloud properties in Earth's remote sensing. Braghiere et al. (2023) highlighted the impact of simplistic assumptions on albedo maps in Earth System Models, estimating a 3.55 W m-2 divergence in radiative forcing when using hyperspectral albedo maps instead of the commonly employed two broadband albedo value approach. They find that omitting the hyperspectral nature of Earth’s surface causes deviation in many climatological patterns, such as precipitation and surface temperature, over regional scales. We average the MODIS datasets over a 10-years period for different times of the year, obtaining a MODIS climatological dataset. Thanks to both high spatial and temporal resolution, we study albedo seasonal and spatial variability in the seven MODIS bands, obtaining estimates of the surface reflectivity as a function of space and time. This MODIS climatological average is the starting point to generate hyperspectral albedo maps using a Principal Component Analysis (PCA) regression algorithm. Combining different datasets of hyperspectral reflectance laboratory measurements for various dry soils, vegetation surfaces, and mixtures of both, we reconstruct the albedo maps in the entire wavelength range from 400 to 2500 nm. We obtain hyperspectral albedo maps with a spatial resolution of 0.05° in latitude and longitude, a spectral resolution of 10 nm, and a temporal resolution of 8 days. The hyperspectral albedo maps are validated against SEVIRI and TROPOMI land surface products. Using the spectral dimension of our albedo maps, we select different land surface types such as forests, deserts, cities and icy surfaces, and we integrate their spectral profiles over entire regions. In this way, it is possible to reconstruct regional spectral patterns which are the combination of typical vegetation and surface spectral features, like the Vegetation Red Edge. In addition, we study the seasonal variability of every region averaging spatially integrated spectra over three months period. From these seasonal spectra, we clearly see the impact of snow cover over different regions, the difference between wet and dry seasons over boreal forests and the formation of lakes over Greenland during the boreal summer. This hyperspectral albedo dataset will lead to more refined calculations of Earth's energy budget, its seasonal variability, and could be used to improve climate simulations.