Automated estimation of river bathymetry using change detection based on Landsat imagery and river morphological models

Abstract:

Free global data sets on river bathymetry at global scale are not yet available. While one of the mostly used free elevation datasets, SRTM, provides data on location and elevation of rivers, its quality usually is very limited. This happens mainly because water mask was derived from older satellite imagery, such as Landsat 5, and also because the radar instruments perform bad near water, especially with the presence of vegetation in riparian zone. Additional corrections are required before it can be used for applications such as higher resolution surface water flow simulations. On the other hand, medium resolution satellite imagery from Landsat mission can be used to estimate water mask changes during the last 40 years. Water mask from Landsat imagery can be derived on per-image basis, in some cases, resulting in up to one thousand water masks. For rivers where significant water mask changes can be observed, this information can be used to improve quality of existing digital elevation models in the range between minimum and maximum observed water levels. Furthermore, we can use this information to further estimate river bathymetry using morphological models.

We will evaluate how Landsat imagery can be used to estimate river bathymetry and will point to cases of significant inconsistencies between SRTM and Landsat-based water masks. We will also explore other challenges on a way to automated estimation of river bathymetry using fusion of numerical morphological models and remote sensing data. Some of them include automatic generation of model mesh, estimation of river morphodynamic properties and issues related to spectral method used to analyse optical satellite imagery.

<< Previous Abstract | Next Abstract >>