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

• A new binary quadratic polynomial relationship for K d λ retrieval is created. • An efficient model for shallow benthic habitat parameters is built. • The model only requires four-band multispectral imagery without ground truth data. • The model is based on radiative transfer process in optically shallow waters. • Four cases show robustness and adaptability for bottom and water column properties. Water depth ( H ), bottom reflectance ( R b λ ), 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 λ 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 λ was determined. Finally, using the monthly mean water column reflectance dataset MMCR w λ constructed from 23-year MODIS data, R b λ , 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 λ 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 λ 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.