A Physics-Based Framework for PRISMA-Like Hyperspectral Data Simulation in Urban Environment

Physically based synthetic hyperspectral datasets are required to support the development and testing of atmospheric and aerosol-related methodologies, particularly in situations where the availability of cloud-free hyperspectral observations is limited. In this study, a physically consistent framework is presented for the generation of PRISMA-like hyperspectral spectra by integrating atmospheric composition fields from the Copernicus Atmosphere Monitoring Service, spectrally resolved aerosol optical properties derived with FlexAOD, broadband surface reflectance prototypes obtained from the Harmonized Landsat--Sentinel dataset, and full radiative-transfer simulations performed with libRadtran under realistic PRISMA acquisition geometries.The framework is evaluated using six PRISMA acquisitions over the Rome metropolitan area, encompassing a range of surface types, atmospheric conditions and viewing geometries. A two-stage spectral consistency assessment is adopted to characterise the relationship between simulated and observed spectra. In the first stage, a spectral distance metric based on PRISMA Level~2C surface reflectance is used to derive scene-independent radiometric partitions of the observations. In the second stage, empirical constraints based on NDVI, NDBI and VgNIR-BI, derived from PRISMA Level~1 radiance data, are employed to ensure the physical consistency of the simulated spectra across vegetation, soil and urban surface classes.Results indicate a satisfactory radiometric agreement between simulated and real PRISMA spectra, with the majority of the synthetic curves falling within the observed mean $\pm3\sigma$ spectral envelopes for all major land-cover classes. The spectral distance metric exhibits stable, scene-independent behaviour, while the empirical spectral indices remain consistent across all analysed acquisitions. The proposed framework supports the integration of PRISMA observations with physically based synthetic spectra, enabling the construction of large, internally consistent hyperspectral datasets suitable for future studies e.g. focused on aerosol optical properties, aerosol speciation and hyperspectral sensitivity analyses. The proposed framework is readily extendable to similar hyperspectral sensors, including EnMAP and the upcoming Copernicus CHIME mission.