Applications of process‐based models (PBM) for predictions are confounded by multiple uncertainties and computational burdens, resulting in appreciable errors. A novel modeling framework combining a high‐fidelity PBM with surrogate and machine learning (ML) models is developed to tackle these challenges and applied for streamflow prediction. A surrogate model permits high computational efficiency of a PBM solution at a minimum loss of its accuracy. A novel probabilistic ML model partitions the PBM‐surrogate prediction errors into reducible and irreducible types, quantifying their distributions that arise due to both explicitly perceived uncertainties (such as parametric) or those that are entirely hidden to the modeler (not included or unexpected). Using this approach, we demonstrate a substantial improvement of streamflow predictive accuracy for a case study urbanized watershed. Such a framework provides an efficient solution combining the strengths of high‐fidelity and physics‐agnostic models for a wide range of prediction problems in geosciences. Plain Language Summary This study proposes a new framework that combines three different modeling techniques to make flood forecasting more accurate. The framework combines the strengths of (a) complex models (or process‐based models, PBMs) based on our understanding of relevant processes that can reproduce measurable quantities; (b) simpler models that are designed to mimic PBM's solutions—known as surrogate models—and make predictions within a few seconds; and (c) machine learning models that can detect relationships among variables using only data, improve the accuracy of prediction, and provide estimates of prediction uncertainty. The framework is tested in an urbanized watershed and shows a significant improvement in both computational efficiency and accuracy of streamflow prediction. Ultimately, the proposed framework is a novel powerful solution that combines the latest advances in different types of modeling approaches to solve prediction problems in geosciences. Its adaptability and efficiency make it suitable for a wide range of situations. Key Points While PBMs are physics‐based, the complexity of uncertainties and the high computational burden have limited their utility for predictions The developed novel framework integrates process‐based models, surrogate, and machine learning (ML) models to predict ensemble flood attributes with error quantification A novel probabilistic ML model partitions the errors into reducible and irreducible types, also quantifying their distributions