The connections between streams and aquifers can be spatially variable and uncertain due to heterogeneity in geology and topography. During drought seasons, farming activities may induce critical peak pumping rates to supply irrigation water needs for crops. This may lead to increased concerns about reduction in baseflow and adverse impacts upon riverine ecosystems. Quantitative management of the groundwater is a required component in this particular human-nature system to evaluate the tradeoffs between irrigation agriculture and the ecosystems requirements. Forecast of the impact of pumping on river-aquifer exchange depends upon spatially variable hydrogeological parameters, as well as temporally uncertain streamflow. In this study a novel component - systems reliability analysis framework is developed to assess risk. Physical parameters uncertainty is studied in light of the Glover-Balmer (analytical) and MODFLOW (numerical) models, while temporal random streamflow is modeled as a Markov process. Reliability methods have been developed in the aerospace industry and extensively applied in structural engineering, but have only seen limited use in water resources so far.