Forecasting tidal marsh elevation and habitat change through fusion of Earth observations and a process model

Abstract

AbstractReducing uncertainty in data inputs at relevant spatial scales can improve tidal marsh forecasting models, and their usefulness in coastal climate change adaptation decisions. The Marsh Equilibrium Model (MEM), a one‐dimensional mechanistic elevation model, incorporates feedbacks of organic and inorganic inputs to project elevations under sea‐level rise scenarios. We tested the feasibility of deriving two key MEM inputs—average annual suspended sediment concentration (SSC) and aboveground peak biomass—from remote sensing data in order to apply MEM across a broader geographic region. We analyzed the precision and representativeness (spatial distribution) of these remote sensing inputs to improve understanding of our study region, a brackish tidal marsh in San Francisco Bay, and to test the applicable spatial extent for coastal modeling. We compared biomass and SSC models derived from Landsat 8, DigitalGlobe WorldView‐2, and hyperspectral airborne imagery. Landsat 8‐derived inputs were evaluated in a MEM sensitivity analysis. Biomass models were comparable although peak biomass from Landsat 8 best matched field‐measured values. The Portable Remote Imaging Spectrometer SSC model was most accurate, although a Landsat 8 time series provided annual average SSC estimates. Landsat 8‐measured peak biomass values were randomly distributed, and annual average SSC (30 mg/L) was well represented in the main channels (IQR: 29–32 mg/L), illustrating the suitability of these inputs across the model domain. Trend response surface analysis identified significant diversion between field and remote sensing‐based model runs at 60 yr due to model sensitivity at the marsh edge (80–140 cm NAVD88), although at 100 yr, elevation forecasts differed less than 10 cm across 97% of the marsh surface (150–200 cm NAVD88). Results demonstrate the utility of Landsat 8 for landscape‐scale tidal marsh elevation projections due to its comparable performance with the other sensors, temporal frequency, and cost. Integration of remote sensing data with MEM should advance regional projections of marsh vegetation change by better parameterizing MEM inputs spatially. Improving information for coastal modeling will support planning for ecosystem services, including habitat, carbon storage, and flood protection.