A hybrid approach to estimate chromophoric dissolved organic matter in turbid estuaries from satellite measurements: A case study for Tampa Bay

Remote sensing of chromophoric dissolved organic matter (CDOM) from satellite measurements for estuaries has been problematic due to optical complexity of estuarine waters and uncertainties in satellite-derived remote sensing reflectance (Rrs, sr(-1)). Here we demonstrate a hybrid approach to combine empirical and semi-analytical algorithms to derive CDOM absorption coefficient at 443 nm (a(g)(443), m(-1)) in a turbid estuary (Tampa Bay) from MODIS Aqua (MODISA) and SeaWiFS measurements. The approach first used a validated empirical algorithm and a modified quasi-analytical algorithm (QAA) to derive chlorophyll-a concentration (Chla, mg m(-3)) and particulate backscattering coefficient at 443 nm (b(bp)(443), m(-1)), respectively, from which phytoplankton pigment and non-algal particulate absorption coefficient at 443 nm (a(ph)(443) and a(d)(443), m(-1)) were derived with pre-determined bio-optical relationships. Then, the modified QAA was used to estimate the total absorption coefficient at 443 nm (a(t)(443), m(-1)). Finally, a(g)(443) was estimated as (a(t)(443) - a(ph)(443) - a(d)(443) - a(w)(443)) where a(w)(443) is the absorption coefficient of pure water (a constant). Using data collected from 71 field stations and 33 near-concurrent satellite-field matchup data pairs covering a large dynamic range (0.3 - 8 m(-1)), the approach showed ~23% RMS uncertainties in retrieving a(g)(443) when in situ Rrs data (N = 71) were used. The same approach applied to satellite Rrs yielded much higher uncertainties of a(g)(443) (~85%) due to large errors in the satellite-retrieved Rrs(443). When the Rrs(443) was derived from the satellite-retrieved Rrs(550) and then used in the hybrid approach, uncertainties in the retrieved a(g)(443) reduced to ~30% (N = 33). Application of the approach to MODISA and SeaWiFS data led to a 15-year time series of monthly mean a(g)(443) distributions in Tampa Bay between 1998 and 2012. This time series showed significant seasonal and annual variations regulated mainly by river discharge. Testing of the approach over another turbid estuary (Chesapeake Bay, the largest estuary in the U.S.) demonstrated the potential (~25% uncertainties for a limited a(g)(443) range) of using this approach to establish long-term environmental data records (EDRs) of CDOM distributions in other estuaries with similar optical complexity.