Accurately forecasting long-term streamflow and reservoir inflow is a significant challenge due to the extended forecasting horizon, uncertainties from climate change, and a complex array of hydrometeorological influences. To tackle these issues, we developed explainable AI methods equipped with uncertainty quantification (UQ). Our approach consists of three key components. Firstly, we designed an advanced AI model that synthesizes diverse data, including meteorological, satellite, geospatial, in-situ measurements, and streamflow records, and incorporates future hydrometeorological conditions for systematic modeling. Secondly, we introduced a computationally efficient yet robust UQ technique to accurately quantify forecast uncertainties at each lead time step. Lastly, we employed various explanation methods to understand the influence of hydrometeorological factors on streamflow forecasts. Our methodology has been applied across various catchments and reservoirs in the CONUS, demonstrating accurate and reliable forecasts at a series of time horizons.