Abstract:

Mangroves, as an important component of the coastal ecosystem, the determination of soil organic carbon (SOC) content within them is of great significance for evaluating the carbon storage capacity of the coastal ecosystem. At present, research on mangrove soil is relatively scarce both at home and abroad. To address the issues of unclear spectral characteristics of mangrove soil and the difficulty in exploring SOC sensitive spectral subdomains, this study innovatively proposed the continuous wavelet spectral similarity angle (CSS) analysis method to systematically analyze the spectral response mechanism of mangrove soils. It also developed the soil sensitive spectral subdomain capture (CT-2DCOS) method to achieve accurate extraction of sensitive spectral subdomains for soil organic carbon (SOC) across multiple scenarios. Using in-situ full-band hyperspectral data (350~2500 nm) as the data source, this study combined the aforementioned methods to investigate the spectral reflection mechanisms of mangrove soils under three scenarios (different depths, tree species, and habitats), and further developed an adaptive ensemble learning (AEL) model to complete high-precision inversion of SOC content under these three scenarios. On this basis, it quantified the degree of influence of the three scenarios on SOC content via factor analysis and revealed the intrinsic correlations between mangrove SOC content and depth, tree species, and habitat by integrating significance tests. The results showed that: ① The soil spectra in the 400~800 nm band interval exhibited significant correlations with mangrove SOC content, among which the linear correlation between the spectra around 600 nm and SOC content was more prominent.②The sensitive spectral subdomains of soils at different depths were mainly concentrated in the 350~800 nm band, those of soils under different tree species were dominated by the 600~900 nm band, while those of soils in different habitats were distributed in two band intervals (350~900 nm and 1500~2200 nm).③The AEL model effectively achieved high-precision inversion of SOC content. Among the 42 inversion schemes, the coefficient of determination (R²) ranged from 0.46 to 0.98 within the 0~60 cm soil depth range, the 0~10 cm soil layer showed the optimal SOC inversion effect (R²=0.96), and this layer also had the highest SOC content, accounting for 25.05%; among the 5 mangrove tree species, the soil under Bruguiera sexangula forests exhibited the best SOC inversion effect (R²=0.97) and the highest SOC content, accounting for 27.16%; among the 3 habitats, the near-natural restoration area had the highest SOC inversion accuracy, with SOC content accounting for 45.24%. This study systematically clarifies the spectral response mechanism of mangrove soil under different scenarios, accurately captures its SOC diagnostic spectral bands, and achieves high-precision inversion of SOC content. Among them, the mining of the diagnostic spectral bands of mangrove soil can precisely match the bands of satellite images, providing scientific support for the hyperspectral remote sensing estimation of blue carbon in the coastal zone under large-scale and multi-scenario conditions.

Key words: mangroves, soil organic carbon, measured hyperspectral data, spectral characteristic analysis, depth and species, quantitative inversion