Floods are among the climate change effects most likely to intensify, and they account for a disproportionate share of climate-related economic losses in the EU. Economic production is organized around complex, interconnected supply chains where sectors continuously buy from one another. Through buyer-supplier relationships, disruptions propagate to firms that were never directly exposed. Ignoring these inter-firm and inter-sectoral dependencies, therefore, risks underestimating the true economic burden of floods. This makes it important to understand not just how individual firms adapt, but how the network of interdependent firms as a whole regains its resilience after a shock.

While flood adaptation research has examined post-shock firm behaviors, these studies did not focus on how the network structure evolves because of them. The literature examines the changing network structure following a shock and the firm-level behaviors that influence it. However, it emphasizes the opportunity of researching what the restructuring of the network does to its long-term resilience. From this knowledge gap, the following research question is addressed: How do firm-level rewiring behaviors, including those exhibiting path dependency through firms’ memory of past shocks, reshape the inter-firm network structure and affect its resilience following a flood shock?

To address this gap, the EMERGO model developed within the NWO VIDI project “Social Tipping Points for Transformational Climate Change Adaptation” (grant number 191.015) was extended in three steps. First, a new sectoral interdependency structure was introduced between the capital goods, consumer goods, and consumer services sectors, capturing intermediate goods and services dependencies that were previously only implicit in the model (SQ1). Second, this was used to structure an explicit, dynamic inter-firm network in which supplier relationships evolve over time as firms enter, exit, and switch suppliers (SQ2). Third, several rewiring behaviors were implemented and tested: probabilistic rewiring based on competitiveness, rewiring toward market leaders, risk-averse rewiring based on suppliers’ past disruption history, and combinations of these. Each scenario was run 30 times and subjected to two sequential flood shocks, after which network-structure and resilience metrics were calculated and compared (SQ3, SQ4).

The key findings show two distinct patterns of network structure evolution: scenarios driven by competitiveness-based rewiring centralized gradually, while scenarios including market leader-based rewiring centralized rapidly early on. These structural differences mattered for resilience: the gradually centralizing scenarios showed better absorptive, adaptive, and restorative capacity after the first shock, while continuous rewiring improved resilience across all scenarios by the second shock. Firms’ memory of past shocks did not produce a visible difference in network structure, but it did improve restorative capacity under repeated shocks. This suggests resilience operates through two separate dimensions: a structural one, where the speed and pattern of network centralization shape resilience to the first shock, and a behavioral one, where memory of past disruptions improves recovery without changing the network structure itself.

These findings carry several policy implications. Excessive concentration around dominant suppliers, as seen in the pure Market Leader scenario, removes network redundancy and increases vulnerability to cascading failures, whereas performance-based rewiring, applied without such concentration, gradually strengthens resilience. Risk-averse, memory-based switching improves recovery but not initial shock absorption, indicating it should complement rather than replace other resilience measures. Because resilience effects differ by sector, policy should be sector-specific rather than uniform, and given the volatility observed in capital goods, sufficient inventory buffers appear especially important for capital-intensive sectors.

The study has several limitations. Attempts at sensitivity analysis revealed extreme sensitivity to the capital goods and intermediates utilization parameters. Moreover, the model showed no sensitivity to switching weight parameters and moderate sensitivity to supplier range parameters. Resilience could only reliably be assessed through inventory levels, meaning the analysis effectively measures inventory resilience rather than broader economic resilience. The strength of inter-sectoral dependencies modeled is also constrained to a relatively low, stable value, meaning the results reflect weaker interdependencies than may exist in practice. The capital goods sub-network’s resilience metrics could not be meaningfully interpreted due to high volatility.

Based on these findings, policymakers could discourage rapid, excessive supplier concentration around market leaders and instead encourage performance-based supplier selection, which produces a more distributed and resilient structure. Improving the availability of flood-risk information at the industry level could also support better-informed supplier decisions. Since resilience benefits differ across sectors, a uniform resilience policy is unlikely to be effective.

Future research should explore longer sequences of repeated shocks, the role of memory at the individual firm level, and less radical implementations of market leader preferences. Combining risk-averse rewiring with explicit firm-level adaptation actions would offer a more realistic representation of path dependency. Applying the full set of resilience metrics proposed by Bruckler et al. (2024) would give a more complete picture of network resilience. Finally, future research could include switching behavior from the supplier’s perspective and examine whether the model’s shock mechanism generalizes to other climate-related hazards beyond flooding. ... Read more