<p>The development of forecasting and prevention tools for spring floods, especially in the context of global climate change, has become a critical focus in risk management. In eastern Canada, this issue is particularly important for the Saint John River watershed in New Brunswick (NB), which has experienced significant flooding events. This study aims to create a prediction model for spring flood events by utilizing historical weather and climate data through statistical models that incorporate covariates for extreme events. Hydrometric data from two stations (Fort-Kent and Mactaquac) were analyzed, alongside meteorological data from nearby stations (Edmundson and Fredericton), to assess the hydrological behavior both upstream and downstream within the watershed. Spring flood events are represented by the annual maximum flow observed between March and June during the analysis period. The research applies the GEV-B-Splines method (Generalized Extreme Value with B-Splines functions), extended to account for the combined effects of covariates. The selected models for both stations include three key variables, emphasizing the significance of total precipitation and teleconnection indices. This study proposes an extreme event model based on multivariate meteorological and teleconnection indices, tailored to the specific sites and informed by an analysis of the hydrological cycle. The approach could be further refined by incorporating more detailed information on the interannual variability of the snow regime.</p>