This methodological diagnostic tests whether broad building descriptors support retrofit-class shortlists beyond popularity. We mapped 2623 New York City Local Law 87 audit rows to six classes with five decision states, trained on 2019–2022, validated on 2023, and tested on unseen 2024 property groups. Under natural recorded-label visibility, state-aware and equal-information global policies both achieved an observed-positive Recall@3 of 0.832. The state model returned the same Top-3 set for 99.7% of test rows; all evaluated natural-condition neural models missed every recorded insulation and window-upgrade positive. Light-emitting diode (LED) lighting and heating, ventilation, and air conditioning (HVAC) controls comprised 277 of 394 test positives, explaining why high recall did not demonstrate personalization. In a secondary synthetic 10% visibility stress test, excluding unlabelled entries from negative supervision improved recall over naive binary cross-entropy by 0.279; the natural state–naive difference was 0.002 (95% confidence interval (CI) [−0.007, 0.012]). Post hoc capacity, ontology, status-precedence, and stopping sensitivities did not establish reliable superiority over popularity. The benchmark diagnoses label misspecification and nearly constant shortlists.

Extreme precipitation events have grown more common as a result of global climate change, and conventional static hazard assessments find it difficult to account for the dynamic progression of flood disasters. This study considers extreme precipitation factors for different return times and creates different extreme precipitation scenarios based on multiyear historical precipitation data and actual storm events. The study proposes a method for the dynamic assessment of regional flood hazard that takes extreme rainfall scenarios into account by simulating the dynamic flood inundation processes under each scenario using the Accumulated Runoff and Flood Estimation Model (AccRo v.1.0), iterative flow accumulation, and hydrological calculations. A dynamic assessment and zoning of flood hazards was carried out in Laiyuan County, Hebei Province. The results reveal that high-hazard zones coincide with the distribution of historically badly damaged townships, concentrated in the river valley plains along the Juma River. The results show that spatial patterns are simultaneously influenced by precipitation, terrain, and the river network. In the temporal dimension, under Scenario 3, the superimposition of the 50-year return period daily maximum rainfall at the 12th hour increased the high-hazard area by approximately 110% compared with that at the 11th hour. In addition, the non-uniform multi-peak rainfall pattern in Scenario 4 represented the rise, peak, and recession stages of the flood process. A combined assessment of water depth and flow velocity can effectively distinguish between two disaster-causing modes—deep water with low flow velocity and shallow water with high flow velocity—thereby addressing the underestimation of hazard in transition zones associated with the use of water depth as a single indicator.