Image processing to monitoring morphological changes in river systems: an application to River Paglia (River Tiber basin, central Italy)

Remote sensing is an important tool for risk mitigation and management of natural disasters (Van Westen, 2000). Multispectral images are used in various fields for rapid feature identification over large areas; furthermore, the short time interval in image capturing allows for the temporal analysis and monitoring of feature variations (Sonka et al., 2014). Image processing and classification techniques can also provide interesting applications in fluvial geomorphology, for outlining morpho-sedimentary features (bars, channels, banks and floodplain) at various temporal stages, in order to monitor the evolution of river systems. Monitoring is essential for understanding the dynamics of river systems, particularly in Italian streams that experience rapid channel changes and human disturbance leading to an increase in the risk of flooding in anthropic areas. We show the results of a study aimed at evaluating the planimetric changes which affected the riverbed of the River Paglia (one of the main tributaries of the River Tiber, in central Italy), as a consequence of the flooding event of November 2012. It shows critical characteristics from the morpho-sedimentary dynamic point of view. The riverbed is in a state of sediment-limited non-equilibrium: it is characterized by an intensely active vertical erosion, which led the thalweg to cut through unconsolidated alluvial sediments up and over the bedrock, formed by overconsolidated marine clays. Over time, but especially in the last 50 years, the morphology of the river has changed drastically: the main channel has entrenched and its riverbed has shrunk considerably, leaving most of the bars, which were non-vegetated and active in the past, but now are inactive and covered with dense vegetation of tall trees (abandoned riverbed). Thus, the river has gradually become a “single-channel with low-sinuosity” channel type. An automatic and/or semi-automatic procedure was developed, in order to study the riverbed changes. The procedure starts from the classical photogrammetric techniques, based on multispectral classification, and goes on with post processing operations of pixel aggregation and shadow treatment. The classification also uses the elevation information provided by a Digital Surface Model produced by photogrammetry. The procedure allows for both the identification and classification of the fluvial features in a post flooding condition. Application of the procedure over time permits the evolution of the fluvial dynamics to be monitored in an accurate and inexpensive way, particularly for flood event conditions which lead to major changes in the dynamics of riverbeds.