Mathematical models are powerful tools and find very wide applications in almost all disciplines. Even the best models only mimic real-world systems, and thus incur uncertainties from multiple sources. For instance, measurement uncertainties exist because of errors in measurements of state variables, while structural uncertainties result from errors in the mathematical representation of actual processes. Further, parametric uncertainties arise from both measurement and structural uncertainties, and because of limited information in the available data in many instances. Uncertainty quantification in hydrologic models poses several challenges including (i) sparse, heterogeneous and high-dimensional data, (ii) lack of good algorithms for contending with uncertainties, and (iii) inability to separate the various forms of uncertainty through standard calibration exercises. Dr. Govindaraju will share his experiences from some collaborative efforts at assessing uncertainty and limits of predictability in hydrologic applications.