Anthropogenic sinkholes are rapid localized depressions caused by collapse above man-made underground cavities and are a growing urban hazard, amplified by ageing utilities and intense rainfall. Effective risk management requires maps identifying where susceptibility is highest and where people and assets are exposed. Here, risk is expressed as a normalized screening-level index (0–1), not an annual probability; the hazard term (H) is a susceptibility proxy. Rome is underlain by a dense, partly abandoned network of excavated cavities (pozzolana quarries, catacombs, galleries) whose degradation, with sewer leakage and flash-flooding, has contributed to increasing sinkholes. From the ISPRA inventory, we selected 1834 occurrences (1960–2023); given heterogeneous sources, results are interpreted as inventory-conditioned susceptibility. Events were integrated with ∼2800 mapped cavities and nine geological and anthropogenic predictors. Spatio-temporal analyses indicate post-2010 acceleration and clustering near cavities, sewer collectors, and flood-prone areas. An XGBoost workflow produced a high-resolution Sinkhole Susceptibility Map with District-based spatial cross-validation performance of ROC-AUC = 0.915 ± 0.04. SHAP interpretation shows susceptibility is dominated by cavity density/proximity and sewer-network variables, with secondary contributions from flood-prone areas. To translate hazard into impact (R = H × V × E), susceptibility was combined with building vulnerability and resident population at census-tract scale. The two highest risk classes include 1521 tracts (∼23%), with more than 400,000 residents (∼34%), and about €127 billion exposure, concentrated in Districts 1–2 and 5–7. The workflow supports screening-level prioritisation and is designed for local retraining and updates as new data become available.
Integrating susceptibility, vulnerability and exposure for screening level sinkhole risk mapping in Metropolitan Rome (Italy)
Ciotoli G.
;Ruggiero L.
;Mori F.;Di Salvo C.;Varone C.;Moscatelli M.;
2026
Abstract
Anthropogenic sinkholes are rapid localized depressions caused by collapse above man-made underground cavities and are a growing urban hazard, amplified by ageing utilities and intense rainfall. Effective risk management requires maps identifying where susceptibility is highest and where people and assets are exposed. Here, risk is expressed as a normalized screening-level index (0–1), not an annual probability; the hazard term (H) is a susceptibility proxy. Rome is underlain by a dense, partly abandoned network of excavated cavities (pozzolana quarries, catacombs, galleries) whose degradation, with sewer leakage and flash-flooding, has contributed to increasing sinkholes. From the ISPRA inventory, we selected 1834 occurrences (1960–2023); given heterogeneous sources, results are interpreted as inventory-conditioned susceptibility. Events were integrated with ∼2800 mapped cavities and nine geological and anthropogenic predictors. Spatio-temporal analyses indicate post-2010 acceleration and clustering near cavities, sewer collectors, and flood-prone areas. An XGBoost workflow produced a high-resolution Sinkhole Susceptibility Map with District-based spatial cross-validation performance of ROC-AUC = 0.915 ± 0.04. SHAP interpretation shows susceptibility is dominated by cavity density/proximity and sewer-network variables, with secondary contributions from flood-prone areas. To translate hazard into impact (R = H × V × E), susceptibility was combined with building vulnerability and resident population at census-tract scale. The two highest risk classes include 1521 tracts (∼23%), with more than 400,000 residents (∼34%), and about €127 billion exposure, concentrated in Districts 1–2 and 5–7. The workflow supports screening-level prioritisation and is designed for local retraining and updates as new data become available.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
