Revised empirical models for estimating coral pigment concentrations from optical reflectance spectra

Spectral reflectance provides a potentially non-invasive approach to estimating coral pigments. We developed revised models using partial least-squares regression trained on paired reflectance and high-performance liquid chromatography pigment data from five coral species. Robust cross-validation yielded useful models for chlorophyll <i>a</i>, peridinin, diadinoxanthin, and <i>β</i>-carotene, though chlorophyll <i>c</i>₂ and diatoxanthin performed poorly, likely from chromatographic or degradation issues. Target factor analysis confirmed all pigments were spectrally present. Applying the models to >4000 spectra from 130 corals across 16 genera produced biologically reasonable estimates consistent with Symbiodiniaceae pigment suites. Although prediction errors limit high-precision physiology, these models enable scalable, non-destructive monitoring of coral pigment dynamics.