Abstract
Remote sensing is increasingly applied in monitoring biodiversity and ecosystem health, however, its utility for detecting heritable aspects of biodiversity is limited by the challenge of distinguishing genetically adapted traits from plastic phenotypic responses. The gap between genetic information and spectral traits is from the indirect, environment-mediated nature of reflectance signals. In this study, we acquired hyperspectral reflectance data (350-2500 nm) combined with specific-locus amplified fragment sequencing (SLAF-seq) across wild bermudagrass (Cynodon dactylon) populations on Hainan Island and experimental common garden to investigate genetic constraints on spectral variation. We assessed relationships between genetic information and spectral signatures through Partial Mantel Tests and Partial Least Squares Discriminant Analysis (PLS-DA). Leaf spectral-genetic correlations (r = 0.4-0.7; P < 0.05) were detected significantly in the SWIR region between 1900 and 2400 nm, underscoring the role of ecological heterogeneity in shaping adaptive spectral traits. Environmental covariates including annual mean temperature, total annual precipitation, elevation, mean solar radiation, mean water vapor content, presented significant liner correlation (r > 0.4, p < 0.05) with leaf spectral-genetic similarity, which revealed that we disentangled SWIR bands as heritable spectral signals from plastic phenotypic responses. PLS-DA achieved a higher accuracy in classifying populations using natural leaf spectra (Accuracy = 86.76 %; Kappa = 0.86) than using canopy data (Accuracy = 41.74 %; Kappa = 0.37), which is influenced by canopy structure. Our findings highlight that integrating spectral-genetic similarity with environmental drivers can identify inheritable spectral signals, providing a pathway for monitoring the adaptive evolution of biodiversity using remote sensing under rapid global change.