Aquatic vegetation indices assessment through radiative transfer modeling and linear mixture simulation

Although spectral vegetation indices (VIs) have been widely used for remote sensing of vegetation in general, such indices have been traditionally targeted at terrestrial, more than aquatic, vegetation. This study introduces two new VIs specifically targeted at aquatic vegetation: NDAVI and WAVI and assesses their performance in capturing information about aquatic vegetation features by comparison with pre-existing indices: NDVI, SAVI and EVI. The assessment methodology is based on: (i) theoretical radiative transfer modeling of vegetation canopy-backgrounds coupling, and (ii) spectral linear mixture simulation based on real-case endmembers. Two study areas, Lake Garda and Lakes of Mantua, in Northern Italy, and a multisensor dataset have been exploited for our study. Our results demonstrate the advantages of the new indices. In particular, NDAVI and WAVI sensitivity scores to LAI and LIDF parameters were generally higher than pre-existing indices' ones. Radiative transfer modeling and real-case based linear mixture simulation showed a general positive, non-linear correlation of vegetation indices with increasing LAI and vegetation fractional cover (FC), more marked for NDVI and NDAVI. Moreover, NDAVI and WAVI show enhanced capabilities in separating terrestrial from aquatic vegetation response, compared to pre-existing indices, especially of NDVI. The new indices provide good performance in distinguishing aquatic from terrestrial vegetation: NDAVI over low density vegetation (LAI < 0.7-1.0, FC < 40-50%), and WAVI over medium-high density vegetation (LAI > 1.0, FC > 50%). Specific vegetation indices can therefore improve remote sensing applications for aquatic vegetation monitoring.