<p>Extreme hydrological events, such as floods, pose considerable risks to infrastructure and society. Multivariate hydrological frequency analysis (HFA) is essential for quantifying risks, through the quantile, taking into account the dependence between variables such as flood peak and volume. However, the assumption of stationarity commonly used in multivariate HFA is increasingly being questioned. Consequently, modeling of extreme hydrological events becomes more complicated but also more realistic. In this sense, the present work focuses on studying quantiles in a multivariate and non-stationary framework. In the later, a quantile is a collection of curves with the same level of risk. For practical reasons, points (combinations) on the quantile curve must be selected. Two selection methods, namely the Most Likely Design Realization (MLDR) method and the alpha method, have been extended to the non-stationary context. Various scenarios were developed to illustrate the behavior of the selected combinations on quantile curves. The obtained results show that non-stationarity leads to more dynamic and complex quantile combinations, influenced by factors such as climate change. In addition, to enable practitioners to benefit from these advanced but realistic approaches, we have proposed a graphical tool providing an overview covering quantile curves and selected combinations, the univariate quantiles of each variable, and the trend of their dependence.</p>