Attributing and Scaling Climate Change Impacts on Floods Through Causal Chains

Understanding whether and why observed flood hazards are changing remains a central challenge for hydrology. While event-based attribution has advanced rapidly, robust attribution of changes in observed flood series remains difficult due to the integration of exogenous climatic forcing with endogenous catchment change (e.g. urbanisation, landuse change) and data quality challenges (especially for extremes). Here we develop and apply a causal-chain framework to detect, attribute, and scale changes in annual maximum floods using observational data. Taking the Shannon catchment in Ireland as an exemplar, we i) identify causal chains to reconstruct annual maximum instantaneous discharge from flood-relevant precipitation indices, ii) separate climate-driven and residual components of observed change, iii) evaluate the emergence of a climate change signal in causal chains by regressing (multiple linear regression) local precipitation indices onto global mean surface temperature (GMST), and iv) employ these results to scale local changes in flood magnitude to observed and future changes in GMST. Results show that increases in flood magnitude across the catchment are predominantly climate-driven, with multi-day precipitation totals representing antecedent conditions, particularly annual 30-day maxima and the number of very wet days in winter, emerging as the dominant causal pathways. These precipitation indices exhibit detectable warming-related signals that have emerged from variability and explain a substantial proportion of observed increasing flood trends at all sites (ranging between 28 and 93 percent). Residual trends highlight the role of endogenous catchment factors, especially data quality, changing hydrometric conditions and arterial drainage. By linking local flood discharge directly to GMST via causal chains, the framework quantifies catchment-specific flood sensitivity expressed as percentage change per degree of warming, enabling scaling to future warming levels. Results indicate that flood sensitivity per degree increase in GMST varies substantially across catchments, ranging from 8 to 18 percent per degree warming across the catchment sample. The approach provides an observation-based framework for flood attribution, leveraging established methods to bridge trend detection, process understanding, and climate scaling. Moreover, the approach can identify sensitive catchments, sentinel indices for monitoring floods and help better inform adaptation strategies. The approach is readily transferable to other catchments and hydrological extremes.