Remote sensing of water and land elevation for river hydraulic modelling

Climate change is intensifying the frequency of extreme climatic events altering rainfall-runoff patterns and consequently increasing floods and droughts in river systems. To cope with the effects of such climatic events, contingency plans based on accurate hydraulic models are necessary. On top of that, in-situ gauging stations monitoring Water Surface Elevation (WSE) and discharge are decreasing worldwide, and the available in-situ measurements do not provide the necessary spatial coverage. Alternatively, satellite remote sensing and Unmanned Aerial Systems (UAS) hydrometry are playing a more important role in monitoring river systems. Satellite altimetry missions can provide great value in large, remote and data scarce areas to inform hydraulic models, offering a large spatial and temporal coverage. UAS hydrometry offers higher accuracy and spatial resolution that can be fundamental to monitor small rivers and streams. This PhD project explores new high-resolution satellite altimetry datasets and UAS hydrometry technologies to monitor and create models for river systems of different characteristics, from regional to local scales.

One of the main findings of the research was the derivation of ICESat-2 ATL03 product river cross-sections. In the low flow season, there is a large portion of the river cross-section that can be observed, and only a small-submerged portion
needs to be interpolated. River cross-sections are a fundamental input for 1D hydraulic models. ICESat-2 cross-sections were exploited for a regional scale hydraulic model in the Upstream Yellow River, for a 300 km river reach, between two in-situ gauging stations at Jimai and Mentang. The channel width in this river reach varies between 80 to 160 meters. Previous altimetry missions could not deliver accurate measurements in rivers of this size. Moreover, informing the model with observed cross-section facilitates a low parametrization. The model was calibrated against ICESat-2 inland WSE product ATL13. The calibrated model estimates WSE and discharge at any point of the river reach with a WSE error of 0.41 m, which is below errors of highly parametrized models, informed with lower resolution satellite altimetry missions in wider rivers.

UAS hydrometry can measure different hydraulic variables using a variety of payloads, delivering measurements at a cm level accuracy and resolution, and overcoming limitations of traditional in-situ surveys. In this study, we use a LiDAR payload to measure terrain elevation, a radar altimeter to measure WSE and a sonar payload to measure bathymetry. The measurements are used to produce river cross-section in the Ryå stream in Denmark, and to derive a 1D hydraulic model over a 8 km river reach with varying channel width between 10 to 15 meters. This work demonstrated the important role of UAS hydrometry in monitoring small rivers and stream to create local scale models. Additionally, the incorporation of bathymetry measurements gives WSE estimations at a cm level accuracy (<3 cm), and allows to parametrize distributed roughness coefficients to simulate vegetation changes.

Finally, this PhD project exploits the capacity of ICESat -2 cross-sections to improve 2D hydraulic models over braided rivers and floodplain areas. ICESat-2 derived cross-section provide land elevation measurements, mapping the micro-
topography of floodplain, which was observed to be correlated with Sentinel-2 Multi Spectral Imagery (MSI). The ICESat-2 cross-sections are used as target data to train an Artificial Neural Network (ANN) with Sentinel -2 MSI, to correct FabDEM elevation and enhance its resolution to 10-meters in floodplain areas. The corrected Digital Elevation Model (DEM) is used as input of a 2D hydraulic model that estimates Surface Water Extent (SWE) given discharge values. The simulated SWE maps are compared with observed SWE from observed satellite imagery, where the corrected DEM, refer as ANN DEM, clearly reduces the bias in SWE and improves performance metrics conversably compared to simulate SWE using FabDEM.

The results of this PhD project highlights the importance of accurate land elevation and WSE for river hydraulic modelling, for different study cases. For regional studies in medium sized rivers, high-resolution satellite altimetry can deliver meaningful datasets to inform hydraulic models. For local scale studies in small rivers and stream, UAS hydrometry can produce better results. Additionally, the morphology of the river is important for the choice of hydraulic model, where 2D hydraulic models can better represent flow dynamics in braided rivers and produce SWE maps, while 1D hydraulic models are more suitable for confined rivers with well-defined channels.

For future research, hydraulic models can improve by integrating observed bathymetry from UAS hydrometry, and land elevation and WSE from satellite observations with large temporal and spatial coverage. Additionally, incorporating new high-resolution satellite missions, like SWOT, can provide denser WSE and discharge time-series. Moreover, SWOT data and the extended spatial coverage of ICESat-2 expected at the end of mission, will deliver larger reference datasets to potentially improve DEMs in data-scarce and remote regions.