Researcher Bangyi Tao from our institute, along with his collaborators, published a research paper titled "Bathymetry Retrieval Algorithm Based on Hyperspectral Features of Pure Water Absorption from 570 to 600 nm" in the prestigious remote sensing journal IEEE Transactions on Geoscience and Remote Sensing. The first author of the paper is Dr. Zhongqiang Wu, a jointly supervised PhD student from our institute and Nanjing University (currently a lecturer at Hainan Normal University), with Bangyi Tao serving as the corresponding author. Collaborators include Researcher Zhihua Mao and Senior Engineer Haiqing Huang from our institute. Our institute is the primary affiliation for this paper.
Airborne hyperspectral remote sensing is a promising method for obtaining high spatial resolution depth measurements in shallow coastal and island reef waters. The widely applied semi-analytical model proposed by Professor Zhongping Lee (Lee et al., 1998; 1999), known as the HOPE model, currently serves as the predominant hyperspectral remote sensing model for water depth retrieval. Existing algorithms for hyperspectral water depth inversion mostly build upon the HOPE model and typically utilize broad spectral ranges (400-800 nm) of water column hyperspectral reflectance. However, recent advancements in sensor spectral resolution have not significantly improved the accuracy of traditional hyperspectral water depth inversion algorithms. This limitation is primarily attributed to the assumptions regarding the accuracy of water optical properties and bottom reflectance spectra shapes within these algorithms.
Therefore, current efforts in improving hyperspectral algorithms focus on identifying more reliable hyperspectral features to minimize the impact of inherent variability in water optical properties and seafloor reflectance on water depth retrieval.
Figure 1(a) shows the absorption spectra of various components; (b) illustrates the contribution of pure water to total absorption.
This study proposes a novel water depth retrieval algorithm, named HOPE-PW, based exclusively on hyperspectral reflectance within the 570-600 nm spectral range with a spectral resolution of 3.5 nm. This algorithm innovatively utilizes high spectral resolution pure water absorption characteristics within the narrow spectral region of 570-600 nm. In many optically shallow coastal areas, particularly coral reef environments, pure water absorption dominates the total absorption within this spectral range. Apart from variations in water depth, the spectral behavior of reflectance at 570-600 nm is primarily governed by the sharp increase in pure water absorption coefficient with wavelength, allowing simplification of other optical properties such as phytoplankton/CDOM absorption, particle backscattering, and seafloor reflectance using flat spectral shapes assumptions. Therefore, the HOPE-PW algorithm proposed in this paper only requires solving for four unknown parameters based on these characteristics.
Validation of the algorithm was conducted using multiple hyperspectral datasets, including data from domestically produced AMMIS sensors and airborne LiDAR bathymetry data, demonstrating its effectiveness. Comparison with the latest HOPE-BRUCE algorithm using publicly available PRISM hyperspectral data from 15 different locations including Palau, Guam, and the Great Barrier Reef confirmed the algorithm's applicability. Additionally, due to the reduced number of spectral calculations, the algorithm's computational efficiency was improved threefold.
Table 1 Comparison of HOPE-PW and HOPE-BRUCE algorithms and parameterization schemes.
Figure 2. Inversion results of local water depths at Beidao Island. (a) Water depth map derived from LIDAR, (b) water depth map inverted using HOPE-BRUCE, and (c) water depth map inverted using HOPE-PW at Beidao Island. Scatter plots comparing airborne LIDAR water depths with inversion results from (d) HOPE-BRUCE and (e) HOPE-PW. Maximum water depth inversion based on domestic AMMIS hyperspectral data from Shanghai Institute of Technical Physics exceeds 30 meters.
Figure 3. Comparison between HOPE-PW and HOPE-BRUCE inversion results based on PRISM hyperspectral data in global regions (Palau, Great Barrier Reef, Hawaii, Florida, etc.). Both algorithms demonstrate universality.
