Evaluating climate change and coastal erosion risks on the Venice coastline: a Machine Learning approach supporting multi-risk scenario analysis

Climate change and its consequences on coastal erosion, flooding and water quality are becoming a major concern for a significant percentage of littorals in the world. This issue is particularly challenging for gentle-sloping sandy coasts which are vulnerable to slow and continuous changes related to rising sea-levels and to extreme storm surge and wave events.

Here we present a multidisciplinary research combining satellite image with machine learning and GIS spatial analysis tools to analyze coastal erosion risk in the Venice shoreline over the period 2015-2019. Firstly, an advanced image preprocessing was performed on satellite images (e.g. co-registration, colors normalization) to prepare the input dataset. Secondly, different supervised and unsupervised machine learning classification methods were tested to accurately define shoreline position by recognizing land-sea areas in each image and the Digital Shoreline Analysis System (DSAS) tool in ArcGis was applied to evaluate the net shoreline movement overtime. Finally, a GIS-based Bayesian Network (BN) approach was developed, to evaluate the probability and uncertainty of coastal erosion risks, and the cascading effects on water quality variation, against multiple ‘what-if’ scenarios related to extreme sea levels and wave conditions under climate change for the period 2040-2050.

According to the spatial resolution of the available data for the case study of Venice (Veneto Region-Italy), the proposed BN-model was trained and validated by considering atmospheric, oceanographic and water quality parameters over the 2015-2019 timeframe, allowing to capture local-scale coastal progression and related driving forces.

Results showed general shoreline stability in the considered reference timeframe. However, the high presence of anthropogenic structures (e.g. jetties, breakwaters) induces the formation of well-delimited hotspots of erosion/accretion. Future trends from the BN-based scenario analysis, according to RCP8.5 scenario within the 2040-2050 period, showed that, even if in minor extent, water quality parameters (i.e. suspended matter, diffuse attenuation) will increase. On the other hand, shoreline evolution trend will face a decreasing probability of the stable class, which in turn will increase instability.

Despite constraints posed by the spatial resolution of the available data for the investigated case, the outcomes of the performed assessment represent valuable information to support adaptive policy pathways in the context of Integrated Coastal Zone Management and Disaster Risk Reduction in the Venice coastal area.