Hyperspectral observation from Landsat: an optimised spectral enhancement method for water colour remote sensing

Benefiting from its spatio-temporal coverage and high spatial resolution, Landsat series imagery has great potential in freshwater environment monitoring. However, the broad spectral bands strongly inhibit the accurate retrieval of water parameters with complex optical properties, such as the chlorophyll-a concentration (Cchla). Therefore, an optical-water-type-assisted, bio-optical-based hyperspectral reconstruction (BBHR) method (OWT-BBHR) was proposed to improve the performance of the BBHR in the near-infrared (NIR) spectral band. Three critical datasets, namely, GLORIA, China Lakes, and In situ, were used to evaluate the proposed method. It was found that the proposed OWT-BBHR method effectively suppressed the reconstruction errors to approximately 710 nm, which is highly effective for Cchla estimation in freshwater. The reconstruction prominently strengthens the correlation between the band-ratio indices and Cchla, as validated by the increase in r2 values of the Cchla model from 0.14, 0.69 and 0.17 to 0.43, 0.75 and 0.50, respectively. When applied to match-up samples from the In situ and China Lakes datasets, the R value increased from 0.02 and -0.69 to 0.55 and 0.67, respectively. This finding indicates that the reconstruction effectively controlled the spatial scale effect of the Landsat imagery. The source code for this work is publicly available at: https://github.com/gylhenau-hash/OWT-BBHR.git.