RESUMO
Vegetated coastal ecosystems (VCE), encompassing tidal marshes, mangroves, and seagrass, serve as significant 'blue' carbon (C) sinks. Improving our understanding of VCE soils and their spatial and temporal dynamics is essential for conservation efforts. Conventional methods to characterise the dynamics and provenance of VCE soils and measure their total organic carbon (TOC) and inorganic carbon (TIC) contents are cumbersome and expensive. We recorded the MIR spectra and measured the TOC and TIC content of 323 subsamples across consistent depths from 106 soil core samples. Using the spectra of each VCE, we determined their mineral and organic composition by depth. We then used a regression tree algorithm, cubist, to model TOC and TIC contents. We rigorously validated the models to test their performance with a 10-fold cross-validation, bootstrapping, and a separate random test dataset. Our analysis revealed distinct mineralogical and organic MIR signatures in VCE soils correlated with their position within the seascape. The spectra showed decreased clay minerals and increased quartz and carbonate with distance from freshwater inputs. The mineralogy of tidal marsh and mangrove soils differed with depth, showing larger absorptions due to carbonate and quartz and weakening clay minerals and organics absorptions. The mineralogy of the seagrass soils remained the same with depth. The cubist models to estimate TOC and TIC content were accurate (Lin's concordance correlation, ρc≥ 0.92 and 0.93 respectively) and interpretable, confirming our understanding of C in these systems. These findings shed light on the provenance of the soils and help quantify the flux and accumulation of TOC and TIC, which is crucial for informing VCE conservation. Moreover, our results show that MIR spectroscopy could help scale the measurements cost-effectively, for example, in carbon crediting schemes and to improve inventories. The approach could advance blue C science and contribute to their conservation and protection.
