Mapping Macrophytes and Algae Scum by Integrating Optical and SAR Satellite Data

The identification of spatial and temporal dynamics of phytoplankton and macrophytes is crucial for deepening the knowledge of lake primary productivity and shifts in trophic status of inland water bodies. Earth Observation (EO) can provide sensitive information on both groups of primary producers, but their possible coexistence within the same site is often not taken into account by satellite-based analyses. Indeed, macrophyte and phytoplankton coexistence is not rare event, especially in shallow eutrophic lakes subject to cyanobacteria blooms, and common methods based on optical VNIR spectral response features for estimating water constituents often fail in distinguishing dense surface accumulation of cyanobacteria forming at peak of bloom events with floating and emergent macrophyte cover. Few authors have tackled this issue in scientific literature, the most effective approach to our knowledge being the one proposed by Oyama et al. (2015; Remote Sensing of Environment, 157: 35-47), based on the exploitation of SWIR range reflectance. On this topic, Bresciani et al. (2014; Remote Sensing of Environment, 146: 124-135) have demonstrated the potential of combined optical and SAR data in delivering accurate information on algae blooms and scum events in Curonian lagoon (Lithuania). In this work, we take advantage of new generation EO sensors (i.e. ESA-Copernicus Sentinel-1 and -2) for investigating the capabilities of optical (broadband multi-spectral) and SAR (C-band) data integration in providing an effective method for distinguishing cyanobacteria scum and floating macrophytes in Lake Taihu (Jiangsu, China). Matchup pairs of Sentinel-2 and Sentinel-1 data acquired with less than 5 day difference have been pre-processed to derive surface reflectance and backscattering coefficient (sigma0), respectively. Statistics of spectral reflectance and Water Adjusted Vegetation Index (WAVI; Villa et al., 2014; Int. J. Appl. Earth. Obs. Geoinf., 30: 113-127) derived from Sentinel-2 data, as well as sigma0 in VV and VH polarization combinations derived from Sentinel-1 data, have been calculated and used to assess the separability of cyanobacteria scum and floating macrophyte pixels response. Finally, a rule-based framework has been designed, parametrized and applied to Sentinel-1 and -2 data to produce maps of algae scum and macrophytes on Lake Taihu in different times of the primary producers cycle, spanning from April to October.