Simulating at-sensor hyperspectral satellite data for inland water algal blooms

Hyperspectral data have been overshadowed by multispectral data for studying algal blooms for decades. However, newer hyperspectral missions, including the recent Plankton, Aerosol, Cloud, ocean Ecosystem (PACE) Ocean Color Instrument (OCI), are opening the doors to accessible hyperspectral data, at spatial and temporal resolutions comparable to ocean color and multispectral missions. Simulation studies can help to understand the potential of these hyperspectral sensors prior to launch and without extensive field data collection. This study introduces a toolkit, HySIMU (HYperspectral SIMUlator), capable of simulating at-sensor data for hyperspectral sensors from simulated (or real) ground truth images. Six ground truth models are generated by populating distribution models that range from simulated to semi-realistic patterns of algal bloom targets, with various sets of spectral records. These ground truth models are run through HySIMU to simulate at-sensor PACE OCI and PRISMA radiance images using a radiative transfer model. The utility of these simulated images is demonstrated through the estimation of chlorophyll-a concentration from the simulated HySIMU models using the red-Near Infrared 2-band ratio and fluorescence line height (FLH) algorithms. Overall R² and RMSE, excluding poorly performing outliers, range from ~0.4-0.9 and 2.4-41.8 μg/L, respectively. While simulated PRISMA models can resolve fine-scale features better than simulated PACE OCI models, their performance metrics are not consistently better. This may be evidence of the impacts of spectral mixing and averaging at the scale of the spatial resolution of the two sensors.