Day 4

Detailed paper information

Back to list

Paper title Forecasting of River Runoff and Sediment Deposition with Methods of EO4AI for Fairway Modification Prediction
  1. Mariana Damova Mozaika Speaker
  2. Stanko Stankov Mozaika
Form of presentation Poster
  • A7. Hydrology and Water Cycle
    • A7.01 Inland Water Storage and Runoff: Modeling, In Situ Data and Remote Sensing
Abstract text The changes in the runoff and in the alluvial outflow lead to changes in the slope, the depth, meandering, the width of the river bottom and the vegetation. The bed load and the suspended load can change the morphology of the river bed as a result of high runoff. This has a direct impact on the determination of the fairway in navigable rivers. That is why it is of great importance for assisting the maintenance of the navigable rivers to provide with instruments to predict the modifications in the river morphology that will potentially impact the fairway. Achieving this has also effect for understanding of the freshwater cycle, for developing our knowledge of the Earth. To address this problem it is necessary to forecast the sediment deposition amounts and the river runoff and to determine how they will change the river morphology. Predicting sediment deposition potential depends on a variety of meteorological and environmental factors like turbidity, surface reflectance, precipitations, snow cover, soil moisture, vegetation index. Satellite data offer rich variety of datasets, supplying this information. We adopt deep learning to address some specifics of Earth observation data, such as their inconsistency, and generate missing data in the time-series with generative adversarial networks - GANs. And consequently we apply the rendered consistent earth observation data along with in-situ measurements on other deep learning architectures (convolutional neural networks CNNS and LSTMs) to actually generate forecasts for river runoff, water level and sediment deposition by using historic satellite data of the meteorological features listed above, and in-situ measurements for water level, runoff and turbidity. Thus, we employ earth observation data for developing AI based solutions that translates as EO4AI. Further, we report on a series of prediction models and experiments carried out on data from the downstream of the Danube and from Arda that show precision of forecasts with minimal deviation with respect to real measurements. To leverage the applicability of the forecasts on the river morphology in integrated models, we calibrate hydrodynamic models using Telemac, and we demonstrate how the fusion of a complex EO4AI method and geometry mapping produces a solution for a real user need of being aware of upcoming changes in the river fairway of the downstream of the Danube. The satellite data are provided by ADAM via the NoR service of ESA. ADAM provides data access to satellite datasets from different satellites with semantic relevance for the construction of sediment transport and deposition forecast model as discussed above. Finally, we demonstrate a visualization of the forecasted fairway on a GIS component using ESRI ArcGIS server.

This work has been carried out within ESA Contract No 4000133836/21/NL/SC