Using broad-band NDVI to track tidal submergence patterns in ecosystem-scale NEE measurements in a New England salt marsh

Salt marshes are dynamic systems located at the interface of land and ocean. Their long term stability depends on their ability to accrete sediment and to accumulate organic matter to keep their relative position to the sea level. Both processes are dependent on relative elevation and vegetation cover as well as flooding frequency and intensity. Thus, analyzing the response of salt marsh vegetation to tidal influences is important to understand if or how they can adapt to changes in sea level. To monitor CO2 exchange between marsh and atmosphere, we established a temporary (April-November 2012, 2013) and a year-round site (since December 2012) in a salt marsh at the Plum Island Ecosystems LTER. The sites regularly experience two high and low tides per day, but much of the marsh platform is submerged only during the monthly spring tides. The marsh surface consists of high marsh vegetation (Spartina patens, mix with Spartina alterniflora) intersected by creeks and ponds. Vegetation phenology is monitored as broad-band normalised difference vegetation index (NDVI), calculated from noon measurements of reflectance of photosynthetically active and short-wave radiation with a 10 minute resolution. This resolution allows us to detect sudden changes in the surface characteristics, as e.g. tidal submergence. During the day, NDVI values decrease to zero or negative values as long as the canopy is submerged. We use a fit between water table and NDVI to extrapolate this for the night time flooding, otherwise we fill the time series with 5day running means of ‘low tide' NDVI estimates. The seasonal CO₂ dynamics are driven by the Spartina phenology. However, during periods of spring tides, we observe a decrease in fluxes with each tide during the day and during the night. We model gross primary production (GPP) with a hyperbolic photosynthesis-light equation at the leaf level (represented by NDVI), using the Beer's light extinction equation and integrating down through the canopy. Ecosystem respiration (Reco) is modeled as a function of a basal respiration rate, a Lloyd-Taylor temperature response and NDVI. How much the processes of photosynthesis and respiration are affected by submergence is uncertain, e.g. underwater photosynthesis can be facilitated by leaf gas films and decomposition products can be exported as DIC in the water column. Thus, we fit the model once with NDVI values including submergence and once without submergence to account for the two cases. Flooding of the marsh platform occurred in 6.7% of the time from December 2012 to December 2013. During the growing season, the effect of the tides nearly cancel each other out (May-September 2012: -248 gC m^-2 or -242gC m^-2).(GPP and Reco however, show larger deviations: -818gC m^-2 or -836gC m^-2 and 570gC m^-2 or 594gC m^-2. An uncertainty analysis will show the significance of the difference. While we observe the tidal submergence only during spring tides at the PIE LTER, other marshes are submerged more frequently. We discuss possibilities to use our approach in these systems based on NDVI and water table measurements we have done in a low marsh area (April-November 2013).