Historic disasters provide physically plausible stress tests of how today’s cities, infrastructure and insurance portfolios might behave under extreme conditions. Yet history cannot simply be replayed: rivers, cities and flood defences evolve, meaning the same event can have very different consequences over time. Here, we use JBA’s flood maps and catastrophe modelling tools to reconstruct the 1910 Paris flood under present-day conditions, illustrating how flood defences, climate change and future adaptation influence flood risk.
Past flood events are widely used to benchmark flood catastrophe models, stress-test portfolios and inform discussions of resilience. Their value lies in providing physically plausible scenarios for testing today’s assumptions.
Translating a historic event into present-day risk, however, is rarely straightforward. Changes in catchment hydrology, flood defences, urban form, asset values and vulnerability all influence how the same event would unfold today. Treating a historic flood footprint as though it had an unchanged probability or impact risks giving false precision.
The 1910 Paris flood illustrates this challenge well. But, more than a century later, the Seine basin, the city and the flood management system have all changed substantially, so the event cannot simply be replayed under present-day conditions.
The 1910 Great Flood of Paris
In January 1910, weeks of persistent rainfall across the Seine catchment culminated in one of Europe’s defining flood events. Water levels at the Pont d’Austerlitz reached 8.62 metres, inundating large parts of Paris. Around 14,000 buildings were flooded, while transport, power and other essential services were severely disrupted. Floodwater spread not only through streets but also through cellars, tunnels and sewers, exposing the vulnerability of the city’s interconnected infrastructure.
More than a century later, the flood remains an important benchmark for insurers, engineers and flood modellers. Much of its impact arose from the interaction between river flooding and the complex urban environment, rather than from flood extent alone. These underground pathways and infrastructure dependencies are still difficult to represent fully in catastrophe models, making the event a useful reminder of both the strengths and limitations of model-based views of urban flood risk.
What does a “1910 flood” mean today?
A direct recreation of the 1910 flood is neither possible nor particularly useful. The Seine basin has been altered by reservoirs and other flood-management measures; Paris has changed physically and economically; and the same river flow would not necessarily produce the same flood footprint or loss. Instead, we use the historical flood as a physically plausible benchmark for examining how changing hazard, protection and exposure influence its consequences today.
To do this, we combined the historic flood outline from the ALPAGE historical GIS dataset with JBA’s high-resolution fluvial flood hazard maps, based on today’s terrain and flood defences. This provides a consistent basis for comparing the influence of flood defences, climate change and future adaptation.
Modern defences change the picture
Figure 1 illustrates the influence of present-day flood defences on our reconstruction of the 1910 flood. Both panels use today’s terrain and hazard representation; the difference is whether present-day flood defences are included. Incorporating those defences substantially reduces the extent of inundation within central Paris, from around 19 km² to around 13 km².
Combining these reconstructed flood footprints with an indicative contemporary exposure dataset and JBA’s depth-damage relationships suggests that modelled losses are around 45% lower than they would be without today’s defences.
These figures are illustrative, but they highlight the importance of how flood defences are represented within catastrophe models. In this example, accounting for today’s protection substantially changes both the flood footprint and the resulting losses. Because relatively small changes in flood depth can translate into much larger changes in damage, assumptions about defence standards can materially influence loss estimates.
Future risk reflects both climate change and adaptation
Figure 2 explores how the reconstructed flood changes under warmer climatic conditions and with further investment in flood protection measures. We examined warming levels of 2°C and 4°C above pre-industrial conditions, alongside the planned Seine Bassée flood-storage scheme upstream of Paris.
Under warmer conditions, higher flows increase flood depths in some locations, leading to higher modelled losses. In the 4°C scenario, losses increase by around 9% relative to the present-day defended baseline. Applying the expected effect of the Seine Bassée scheme produces a reduction in losses of a similar order.
While not a portfolio forecast or a demonstration that adaptation simply offsets climate change, the analysis does illustrate how future flood risk depends on the combined effects of changing hazard, existing protection and future investment in resilience.
This has wider implications for climate-risk analysis. A climate-adjusted hazard view alone does not describe future loss. Decision-makers also need to consider how flood defences evolve, how exposure changes, and where important elements of the urban system remain outside the scope of catastrophe models. In many situations, these non-climatic changes may influence future losses just as much as changes in the hazard itself.
What does this mean for catastrophe modelling?
The Paris case highlights several broader lessons for catastrophe modelling.
Historic events are most valuable as stress tests. Their probability, footprint and consequences need to be translated into present-day conditions rather than carried forward unchanged. Used in this way, they provide physically plausible storylines against which portfolios, assumptions and model behaviour can be tested.
Flood defences are not simply background information. As the reconstruction indicates, assumptions about protection standards and their spatial variation can materially alter both flood extent and loss. Understanding how those assumptions are represented is therefore fundamental to interpreting model output.
Urban infrastructure remains an important source of uncertainty. Cellars, tunnels, utilities and transport systems helped transmit and amplify the 1910 flood, yet these pathways remain difficult to represent fully in property-focused catastrophe models. Scenario analysis and engineering judgement therefore continue to complement probabilistic modelling.
Probabilistic models and event-based storylines answer different questions. Probabilistic models support consistent estimates across many possible events, while historical reconstructions provide structured “what if?” scenarios that help explore the implications of different assumptions. Used together, they offer complementary perspectives for understanding uncertainty, testing resilience and informing decisions.
From historical reconstruction to practical decision support
Historical events remain valuable long after the floodwaters recede, providing physically plausible starting points for exploring how changing hazard, flood defences, climate and exposure might influence future risk. Used in this way, they help test assumptions, identify important sources of uncertainty and support more informed discussions about resilience with insurers, financial institutions and public authorities.
JBA’s present-day and future flood maps and catastrophe models for France and Europe provide a consistent framework for exploring questions of this kind. They can also support targeted scenario analysis of specific events, adaptation measures, portfolios and infrastructure dependencies. As simplified representations of a complex reality, these models cannot capture every aspect of flood risk, but they provide a consistent basis for comparing scenarios, making assumptions explicit and supporting informed decisions.
Stress testing portfolios against severe but plausible flood scenarios can help lenders evaluate mortgage and commercial property exposure, support insurers in underwriting and pricing risk, and help investors to assess the potential benefits of resilience and adaptation measures across their assets. Combined with probabilistic catastrophe models, these analyses offer complementary perspectives on how flood risk and its financial implications could change.
To discuss flood underwriting, portfolio stress testing or bespoke analysis in Paris and elsewhere, contact us.