An Integrated Methodological Framework with Historical Validation for Flood and Debris-Flow Hazard Zoning in Ungauged Andean Catchments

Flash floods and debris flows in Andean–Amazonian mountain catchments rank among the most destructive natural hazards, and the headwater catchments where they originate almost invariably lack hydrometric records, placing them within the classical problem of Predictions in Ungauged Basins (PUB). Hazard maps generated for these environments from open data share a critical weakness: they are released without empirical validation against observed events, so their credibility rests on unverified assumptions. This study proposes OF-VAL (Open-data Flood Hazard Framework with Historical VALidation), a six-stage framework based entirely on open-access data whose distinguishing feature is a validation gate: hazard products are released only if the model, forced with an open-data reconstruction of a documented historical event, reproduces the observed impact footprint above a predefined threshold (F≥0.50). The framework is demonstrated in the ungauged Sicre–Ipal microcatchment (Cusco, Peru; 19.41 km2), using as benchmark the debris flow of 21 December 2018 documented by INGEMMET and delineated on 0.7 m PeruSAT-1 imagery. A process-class compatibility rule (R1.4) screens the benchmark before gate execution: the event’s documented genesis as a landslide-dam outburst, together with four concordant open-data estimates of the event rainfall (6.3 and 9.7 mm day−1 at the two governing SENAMHI stations, 7.5 mm at the PISCOpd cell and 18.2 mm in GPM IMERG, all an order of magnitude below debris-flow triggering thresholds), establishes that the flow was not pluvially generated; the reconstruction executed with the largest of these estimates wets at most 5.6 ha against a 15.5 ha deposit, so the fit index is bounded at F≤0.36 and gate G4a is declared not executable with this benchmark. During revision, a volume-consistency audit of the design-storm chain revealed that extending the fitted Bell–IDF relationships to the 6 h design duration produced storm volumes exceeding the 24 h quantiles by 48–129%; the design hydrology was therefore rebuilt with the volume-restricted Dyck–Peschke parameterization (P6h=0.7071P24h) over the two station series satisfying all qualification rules (Quillabamba and Machu Picchu). Three further rules were added when the simulations were audited: runoff-volume consistency with the closed-form SCS excess (R4.4) and an outlet boundary at the channel invert (R4.5), which detected and corrected two defects of the previous runs, and the declaration of the fraction of runoff delivered to the outlet (R4.6). Under the corrected scenarios (return periods of 25–140 years), reported as non-validated diagnostic products, only 3.3–6.9 mm of the 61–76 mm areal design depths become excess runoff, and 70–80% of that excess is retained in closed depressions of the unconditioned 12.5 m terrain; the critical scenario wets 8.6 ha on the computational mesh, of which 1.2 ha reach FEMA High hazard or above and 2.9 ha are unsafe for people, with 0.25 ha wetted inside the settlement footprints of Sicre, Tablahuasi, and Pedregal. A gate taxonomy distinguishing blocking, release, and recording gates, and a rule-by-rule compliance table, document what was specified, what was executed, and the disposition of every deviation, including the numerical controls that the demonstration itself failed. The framework is reproducible and transferable to other ungauged catchments with documented events, and its planning-level products complement, rather than replace, site-specific topographic and geotechnical investigations.