SAR data processing for the detection and monitoring of braided gravelbed rivers morphodynamics

Braided rivers represent one of the most complex forms of natural streams. Characterized by intense bed-load transport and highly dynamic channels, they carry significant naturalistic value and support a multiplicity of ecosystem services. Anthropogenic stressors and environmental changes put under stress hydro-morphological dynamics, biological processes, and ecosystem functioning and services of these fragile environments, necessitating integrated management and conservation strategies to preserve their biodiversity and ecological integrity. From a regulatory perspective, the two European Directives 2007/60/EC (the Floods Directive) and 2000/60/EC (the Water Framework Directive) identify and promote win--win measures that both reduce hydraulic risk and enhance the quality of water bodies. Some examples of win--win measures are river naturalization projects that not only restore river ecosystems to their natural state, enhancing biodiversity and ecosystem services but also provide flood protection, improve water quality, and offer recreational opportunities for local communities. This thesis contributes to the development of scientific knowledge in the previously mentioned areas, facilitating the know-how transfer of expertise from academia to the public institution. Building on these premises, this thesis aims to provide additional insights into the morphodynamics of braided rivers, offering new perspectives on the evolution of morphological indices during flood events and contributing valuable knowledge on how these complex systems respond to external stressors. The PhD thesis has been structured along three parts. The primary goal was to develop an innovative unsupervised algorithm for extracting the spatial and temporal evolution of braided river morphology. This computational framework is tailored for Sentinel--1 Synthetic Aperture Radar (SAR) data, overcoming the limitations imposed by weather conditions and day--night cicles. Moreover, it can be effortlessly adapted to additional SAR imagery databases. In cases where the water class covers only a minimal area of the entire scene, the histogram primarily represents the dry soil class. The framework faces this challenge employing a Self-Adaptive Thresholding Approach (SATA) to achieve a distinct bimodal distribution, enabling the accurate computation of threshold values for the 'dry soil' and 'water' classes. The tool, developed within the Python--API of Google Earth Engine (GEE), allowed us to assess the intra--event inundation dynamics, the estimation of the relationship between hydrometric level and wet area extension, and the assessment of bank erosion phenomena. The second chapter focuses on analyzing how morphological indices, such as the Total Braiding Intensity (TBI) index defined as the number of active channels, the Maximum Channel distance (MCD) defined as the distance between the most external channels, and the Cross-Sectional Cumulative Wetted Area (WA) defined as the sum of the wet area of all chanels in a cross section, correlate with discharge variations during flood events. To achieve this objective, the framework designed for Sentinel--1 images was adapted for use with high--definition imagery from the Italian COSMO--SkyMed satellite constellation. Leveraging the superior ground resolution of 3x3 meters provided by the Italian COSMO--SkyMed satellite constellation, we successfully segmented narrow secondary branches that remained undetected with Sentinel--1's 5x20 meter resolution. Thus obtained, the temporal evolution of the braiding system, enables us to evaluate the temporal evolution and the relationship between the TBI, MCD, and WA indices with increasing discharge values. The last part of the PhD thesis, deals with the assessment of the river bed grain size. The initial concept behind this PhD work was to analyze the potential of Synthetic Aperture Radar (SAR) data in assessing not only river morphology but also the pattern of patches with different grain size. While the initial two parts of the work addressed this, the final section's analysis of SAR data, unfortunately, did not provide significant results. Nevertheless, the subjects of surface roughness and the creation of spatially distributed grain size maps continue to hold significant scientific value in the fields of hydraulic and eco--hydraulic modeling and a key information for river management and renaturation projects. The principal role of this factor led us to slightly shift the research focus towards a detailed investigation of these elements, utilizing orthophotos, digital imagery, and corresponding analytical methods to model patterns of river roughness and grain size. A map illustrating the spatial pattern of grain size at the river reach scale was produced through regression analysis. This analysis correlated the texture properties derived from orthophoto tiles with the d50, d84, d90, and d95 grain size characteristics obtained from digital images, thereby providing considerable support for the implementation of detailed hydraulic models.