A practical and efficient model for benthic habitat parameters retrieval in optically shallow waters from four-band multispectral imagery

Abstract

Water depth (H), bottom reflectance ($R_b\left(\lambda \right)$), chlorophyll-a concentration (Chl) and seawater transparency (SDD) are key parameters in assessing shallow benthic habitats. However, due to limited in-situ data, restricted hyperspectral imagery or dynamic water quality, it is challenging to develop a generalized remote sensing model for decoupling water column and benthic signals across diverse regions and time periods. In this study, a Modified Log rotation Ratio and Semi-analytical (MLR-S) model is proposed for diffuse attenuation coefficient $K_d\left(\lambda \right)$ retrieval by adjusting inherent optical properties (P, G, X), from widely used four-band multispectral images without relying on truth data. A new Binary Quadratic Polynomial Relationship (BQPR) between P and the logarithmic values of blue-green bands was developed upon a sparse set of self-inferred points (SIPs). Then, large-scale and optimal $K_d\left(\lambda \right)$ was determined. Finally, using the monthly mean water column reflectance dataset ${\mathrm{MMCR}}_w\left(\lambda \right)$ constructed from 23-year MODIS data, $R_b\left(\lambda \right)$, Chl and SDD can be efficiently derived. Synchronous validation at Dazhou Island confirmed the effectiveness of the MLR-S model, with mean relative errors of 23.77%, 15.26% and 9.91% for blue, green and red $R_b$, respectively, and 36.69% for Chl. Further validation of $R_b\left(\lambda \right)$ in Heron Reef, Aitutaki Island and Buck Island demonstrated the model’s robustness across diverse habitats. Compared to the ICESat-2 Along Track Benthic Reflectance (ATBR) model, the MLR-S model yielded more reasonable and reliable $R_b\left(\lambda \right)$ in deeper waters. Additionally, it outperformed the traditional semi-analytical model, achieving greater accuracy while being over 100 times more computationally efficient. This model offers promising potential for large-scale, high-resolution monitoring of benthic habitats using four-band multispectral data, supporting improved understanding of reef health and informed decision-making.