Modeling of the Douro River Plume Size, Obtained Through Image Segmentation of MERIS Data

Spatial and temporal variation of river plumes can be studied by remote sensing. The main objectives of this letter were to model the Douro River Plume (DRP) size based on image segmentation techniques and to relate it to different parameters such as water volume, last available plume, tide height, and wind speed. Twenty-one MEdium Resolution Imaging Spectrometer (MERIS) scenes of the study area were considered, covering 20 months from 2003 to 2005. Two different segmentation techniques were applied (watershed and region based) to the MERIS scenes in order to estimate the DRP size. Two models were developed in order to relate the DRP size with the water volume, last available plume, tide height, and wind speed. In the first model, the DRP size was modeled using the discharged water volume as input. In the second model, the DRP size was retrieved as a function of the water volume, last available plume, tide height, and wind speed. The last available plume-a factor of crucial importance-was weighted using a forgetting factor, function of the time difference between two consecutive MERIS scenes. A significant correlation was found between DRP dimension and water volume (r = 0.66) during wet season (i.e., excluding dry summer period), because the plume derived from MERIS imagery in the region under study represents river Douro plume only when the river flow exceeds a certain threshold. Despite some particular points, the second model was able to model the plume size obtained through image segmentation, with a mean percentage variation of 34.8% for region-based segmentation method. Segmentation of MERIS data have been shown to be a valid method for modeling the DRP size. Furthermore, DRP size was found to be proportional to water volume, excluding the summer period.