Increasing global trends in time series of annual maximum daily streamflow (AMX) raise the concern that the safety of dams and other sensitive structures is compromised. There is no defined methodology to estimate the design flood (DF) under non-stationarity; thus, the objective of this work is to evaluate the behavior of the hydrological risk of Brazilian dams due to the non-stationary nature of the AMX time series and the implications of the non-stationary nature of the AMX time series in the design of new dams. For this, the hydrological risk of 108 AMX time series was evaluated, comparing the time intervals between 1954–1984 and 1954–2014. A case study was also executed, where the DF was estimated in a non-stationary time series. The generalized distribution of extreme values (GEV) was applied in the time series analyses. The results indicate that the hydrological risk of Brazilian dams increased, and safety may have been reduced. Regarding the ranking of models, the use of physical covariates in the estimate of the DF makes the estimates more reliable. Finally, although significant trends are good indicators, they alone do not guarantee a reduction or increase in risk. It was also observed that using non-stationary models is less important than updating the estimates with newly observed data.

年度最大日流量（AMX）的时间序列中的全球趋势不断增长，引起了人们的担忧，即大坝和其他敏感结构的安全性受到损害。没有固定的方法来估计非平稳状态下的设计洪水（DF）；因此，这项工作的目的是评估由于AMX时间序列的非平稳性质以及新设计中AMX时间序列的非平稳性质所带来的巴西大坝水文风险行为。水坝。为此，对108个AMX时间序列的水文风险进行了评估，比较了1954–1984年和1954–2014年之间的时间间隔。还执行了一个案例研究，其中以非平稳时间序列估算DF。在时间序列分析中应用了极值的广义分布（GEV）。结果表明，巴西大坝的水文风险增加，安全性可能降低。关于模型的排名，在DF的估计中使用物理协变量会使估计更可靠。最后，尽管重大趋势是好的指标，但仅凭它们并不能保证风险的降低或增加。还观察到，使用非平稳模型比使用新观察到的数据更新估算值不那么重要。