Mapping salt marsh soil properties using imaging spectroscopy

Tidal salt marshes sequester and store blue carbon at both short and long time scales. Marsh soils shape and maintain the ecosystem by supporting complex biogeochemical reactions, deposition of sediment, and accumulation of organic matter. In this study, we examined the potential of imaging spectroscopy techniques to indirectly quantify and map tidal marsh soil properties at a National Estuarine Research Reserve in Georgia, USA. A framework was developed to combine modern digital image processing techniques for marsh soil mapping, including object-based image analysis (OBIA), machine learning modeling, and ensemble analysis. We also evaluated the efficacy of airborne hyperspectral sensors in estimating marsh soil properties compared to spaceborne multispectral sensors, WorldView-2 and QuickBird. The pros and cons of object-based modeling and mapping were assessed and compared with traditional pixel-based mapping methods. The results showed that the designed framework was effective in quantifying and mapping three marsh soil properties using the composite reflectance from salt marsh environment: soil salinity, soil water content, and soil organic matter content. Multispectral sensors were successful in quantifying soil salinity and soil water content but failed to model soil organic matter. The study also demonstrated the value of minimum noise fraction transformation and ensemble analysis techniques for marsh soil mapping. The results suggest that imaging spectroscopy based modeling is a promising tool to quantify and map marsh soil properties at a local scale, and is a potential alternative to traditional soil data acquisition to support carbon cycle research and the conservation and restoration of tidal marshes.