Railway infrastructure is a critical component of sustainable transportation systems. However, the linear and rigid nature of railway networks makes them highly vulnerable to climate hazards. Moreover, the growing frequency and intensity of climate hazards driven by climate change increase the risks of climate-related failure modes (CRFMs). Despite this growing concern, understanding of CRFMs, their associated risks, and effective adaptation strategies remains limited. This thesis aims to bridge these gaps by conducting extensive analyses of Swedish railway maintenance dataset, supported by a comprehensive literature review. 

To achieve the thesis objectives, a systematic literature review was first conducted to identify climate hazards and associated CRFMs, irrespective of their intensity, likelihood, or geographical variation. The review provides a comprehensive inventory of climate hazards and CRFMs, offers a structured classification framework, and highlights hazard–failure relationships as well as cascading and compound effects of failures. Furthermore, the proposed adaptation measures were reviewed and categorized into physical, operational, ecological, institutional, and behavioural strategies. Key implementation challenges and potential risks of maladaptation were also discussed. In this context, quantitative risk assessment and the use of predictive models were identified as key institutional and operational strategies, respectively, for effective climate adaptation planning.

Building on these insights, the empirical core of this thesis focuses on assessing the risk of CRFMs and developing predictive tools for the Swedish railway infrastructure. To this end, CRFMs were identified through text mining of historical corrective maintenance data covering all five maintenance regions in Sweden. A probabilistic model was developed to quantify the risk of identified CRFMs, considering both maintenance-related and operational consequences. Risk distributions were subsequently estimated using Monte Carlo simulation. The results enable the prioritization of CRFMs at the national scale and a comparative assessment of regional vulnerability across Sweden. 

Finally, a machine learning framework was developed to distinguish between climate-related and non-climatic failure modes by integrating the failure records with meteorological data. The performance of several machine learning algorithms was evaluated across different time horizons prior to failure events. In addition, the analysis identified key climatic variables that significantly influence model performance and failure classification.

Overall, this thesis provides a fundamental basis and practical evidence to support planning for climate-resilient railway infrastructure. Although the empirical analyses focus on the Swedish railway network, the proposed approaches are transferable and can be applied to other railway systems.