Investigation of Uncertainties in 2D Dam Break Simulations by Monte Carlo Simulations
16th International Congress on Advances in Civil Engineering (ACE2026), Trabzon, Türkiye, 12 - 14 Ekim 2026, (Yayınlanmadı)
- Yayın Türü: Bildiri / Yayınlanmadı
- Basıldığı Şehir: Trabzon
- Basıldığı Ülke: Türkiye
- Orta Doğu Teknik Üniversitesi Adresli: Evet
Özet
Manning's roughness coefficient is a key input in two-dimensional (2D) dam-break flood modelling, yet it cannot be measured directly and is estimated from terrain classification. Deterministic models assign a single fixed value, which does not reflect this estimation uncertainty and can produce overconfident predictions. A Monte Carlo framework is applied to dam-break modelling with GeoClaw, a 2D nonlinear shallow-water solver with adaptive mesh refinement. The 1959 Malpasset dam failure, documented by post-event police high-water marks, is used as the test case. Two roughness configurations (a spatially uniform random coefficient and a region-wise variable field over thirteen sub-regions) and two priors (uniform and log-normal) are combined into four cases, each with fifty Latin-Hypercube realizations. Maximum water-surface elevation, depth and arrival time are extracted at the surveyed points and valley gauges, and a probabilistic inundation map is built from each ensemble. Results show that the inundation footprint is geometry-limited and nearly insensitive to roughness, whereas local depths, peak elevations and especially the wave arrival time are sensitive, the arrival spread growing from seconds near the dam to minutes downstream, broadly consistent with documented field records. The region-wise variable field gives a small improvement in coverage (24–29%) and median error, and the uniform prior outperforms the log-normal one. The framework thus quantifies how roughness uncertainty propagates into the predicted flood and identifies the modelling choices that most affect the depth and timing estimates.