Development of a new integrated flood resilience model using machine learning with GIS-based multi-criteria decision analysis

Flood resilience assessment is an important step for any community as it gives the actual scenario of its ability to resist and recover from flood disasters. However, operationalising and quantifying resilience is still a challenge. In Pakistan, very limited research has been done to assess community resilience to floods. The present study proposed an integrated flood resilience model called the capacity-based flood resilience model (CapFlooR-M), which is based on, machine learning (ML), geographical information system (GIS), remote sensing (RS), and the analytical hierarchy process (AHP). The CapFlooR-M incorporates four main components-flood hazard susceptibility (Is), coping capacity (Cc), adaptive capacity (Ac), and transformative capacity (Tc). Random Forest (RF) and Support Vector Machine (SVM) models were used to create a flood susceptibility map, and the AHP was used to compute the relative scores of core capacities, such as Cc, Ac, and Tc and their respective maps were generated. Finally, the susceptibility map was integrated with Cc, Ac, and Tc maps via overlay analysis in GIS to develop a flood resilience map. The overall results reveal that the northwestern and southwestern parts (36.64%; 505 km2) of the study area have moderate to very high resilience, while the central and southeastern parts (63.46%; 877 km2) have very low to low resilience. The findings of this novel approach can support policymakers, land use planners, and other relevant stakeholders to build resilience against flood hazards.

Automated summary

Flood resilience assessment is crucial for understanding a community's ability to withstand and recover from flood disasters. However, quantifying and operationalizing resilience remains challenging. This study proposes a novel flood resilience model, CapFlooR-M, which integrates machine learning, GIS, RS, and AHP. The model incorporates different components such as flood hazard susceptibility, coping capacity, adaptive capacity, and transformative capacity. By using RF and SVM models, a susceptibility map is created, and AHP is utilized to compute the relative scores of core capacities. The integration of these maps with overlay analysis in GIS produces a flood resilience map. The findings of this study provide valuable insights for policymakers and planners in building resilience against flood hazards.