Analysis and assessment of vegetation’s contribution to the prevention of risks associated with the triggering and development of rainfall induced shallow landslides (soil slips)

In recent decades, the frequency and intensity of extreme rainfall events have increased significantly, a phenomenon closely linked to global climate change. These events, exacerbated by rising global temperatures and increased atmospheric moisture capacity, play a critical role in triggering shallow landslides in susceptible areas. Shallow landslides, often involving the surface soil layer, can evolve into debris flows, posing substantial risks to infrastructure, ecosystems, and human lives. The increasing vulnerability of previously stable regions necessitates reevaluating hazard assessments and mitigation strategies to address these evolving risks. Vegetation emerges as a crucial factor in mitigating landslide susceptibility by influencing soil’s mechanical and hydraulic properties, such as cohesion, water retention, and surface runoff. Eco-sustainable solutions involving vegetation offer a promising alternative to conventional engineering approaches, which are often environmentally intrusive and economically demanding. Understanding the complex interplay among rainfall, soil, and vegetation is essential for developing effective land management strategies and climate change adaptation measures. This study will be structured into several sections, with the intent to: (i) investigate historical landslide events and their possible causes; (ii) analyse the state of the art on shallow landslide susceptibility and root reinforcement; (iii) assess the role of vegetation in slope stability across species; (iv) develop predictive models to identify high-susceptibility areas (LEP-AI and PIP-AI) and simulate landslide flow paths (GPE); (v) design eco-friendly stabilization measures and create a fast geo-spatial platform for landslide management (XSLIP); (vi) study a real case application of the designed models on a wide area of EmiliaRomagna region; Specifically, this research integrates physically-based models and advanced machine learning techniques to enhance the understanding and prediction of rainfall-induced shallow landslides, coupled with vegetation. Specifically, the Shallow Landslide Instability Predictor (SLIP) model is employed for its balance between computational efficiency and accuracy, while machine learning approaches, including neural networks and ensemble methods, address the dynamic and non-linear nature of landslide processes. Two machine learning model prototypes are developed: a rapid macro-scale model for filtering susceptible areas based on rainfall events (LEP-AI) and a detailed micro-scale model for pixel-by-pixel slope stability assessment (PIP-AI). The first model, LEP-AI, shows fluctuating results at just a sufficient level, but substantial improvements are needed. The second model, PIP-AI, demonstrates high and consistent performance, which can be further refined when coupled with a physically-based model such as SLIP. Both models exhibit excellent analysis timing, with LEP-AI taking 0.06 seconds for the analysis of 40 municipalities, and PIP-AI taking 1.2 seconds for 4 municipalities with a 15x15 resolution. By bridging classical soil mechanics and data-driven machine learning methodologies, this research establishes a robust framework for landslide prediction and mitigation, advancing the field’s capacity to address climate-driven geohazards effectively.