Effects of record extension and statistical methods on dam reliability – insights from the Stronie Śląskie failure

Abstract

This paper investigates the effects of methodological choices and the extension of hydrological records on extreme discharge estimates and dam safety, drawing on the 2024 failure of the century-old Stronie Śląskie dam in Poland. It is based on long-term precipitation and discharge records from eight precipitation and seven gauging stations in the Biała Lądecka River basin (in the Sudetes Mountains) and neighbouring catchments. To isolate the effects of record extension, three periods ending in 1984, 2023 and 2024 are analysed. The series were tested for homogeneity, independence, stationarity and trends using a set of non-parametric tests. Extreme-value distributions were fitted by maximum likelihood following two sets of Polish guidelines (including the so-called event alternative method). Discharges at the ungauged dam cross-section were estimated by regional regression and extrapolation. No statistically significant monotonic trends were detected in hydrological data. Nevertheless, extending the series – by 40 years or by a single flood event – led to substantial increases in design discharge for low exceedance probabilities, and wider confidence intervals. The choice of probability distribution can change control flood estimates several-fold and shift the dam’s overtopping probability from levels typical of high-consequence structures to those below the lowest reliability class. The results indicate that periodic hydrological re-evaluation of existing dams is essential and that clearly defined methods for estimating extreme flows are needed to establish consistent safety targets and responsibilities and prompt timely spillway upgrades.