Overview

Uncertainty Analysis (an Uncertainty Trial in HEC-HMS) is a simulation type designed to quantify how uncertainty in model inputs and parameters propagates to uncertainty in computed results. Rather than producing a single deterministic hydrograph, an Uncertainty Trial runs a collection of simulations by repeatedly sampling uncertain variables (for example: precipitation adjustments, loss/transform/baseflow parameters, and initial conditions) from user-defined probability distributions.

The output is a collection of results that can be summarized as probability bands, exceedance probabilities, and statistics (such as median and percentile hydrographs) at locations of interest. This supports more transparent communication of confidence and risk—especially when decisions depend on ranges of possible outcomes rather than a single “best estimate.”

An Uncertainty Trial is composed of a Basin Model, Meteorologic Model, Control Specifications (time window), and an Uncertainty configuration (uncertain variables, distributions, and sampling settings). Results are viewable in the Watershed Explorer after compute and can be compared against deterministic runs and (when available) observed data.

Important note: Hydrologic models are incapable of perfectly simulating the physical watershed response. A quantitative assessment of hydrologic model uncertainty is a key component of determining the reliability of the model predictions. Uncertainty assessment is a critical component of risk-based engineering methodologies. 


Use Cases

  1. Decision support and risk communication

    • Create inflow hydrograph confidence bands for reservoir studies, operational planning, and emergency action planning.
    • Communicate the probability of exceeding thresholds (peak flow, stage, pool, volume).

  2. Forecast sensitivity and uncertainty

    • Explore the influence of uncertainty in precipitation magnitude/placement and key hydrologic parameters on forecast outcomes.
    • Support ensemble-style interpretation of results even when forcing data are not provided as a full meteorologic ensemble.

  3. Calibration confidence / parameter equifinality

    • After calibration, translate plausible parameter ranges into uncertainty bands on flow timing, peak, and volume.
    • Identify situations where multiple parameter sets produce similarly good fits but diverge for extremes.

  4. Sensitivity screening and data prioritization

    • Determine which uncertain inputs/parameters dominate output uncertainty.
    • Focus data collection and refinement on items that most reduce uncertainty (for example: precipitation bias/scaling, soil/loss parameters, initial conditions).

  5. Hazard/design studies with uncertainty statements

    • Provide decision-makers with ranges and exceedance probabilities to complement deterministic design runs (with appropriate engineering judgment and policy context).

Best Practices and Gotchas

  • Start with the biggest drivers

    • Precipitation uncertainty often dominates hydrograph uncertainty. If you include broad precipitation uncertainty, it can mask parameter effects.
    • Consider staged trials: parameter-only, precipitation-only, then combined.

  • Use physically defensible bounds and distributions

    • The uncertainty engine will sample whatever you specify. Unrealistic bounds create unrealistic results.
    • Base ranges on calibration results, literature, watershed characteristics, and/or measurement uncertainty.

  • Be intentional about correlation and shared factors

    • Some uncertainties should vary together (for example: consistent precipitation scaling across the basin, or similar parameter behavior across similar subbasins).
    • If correlation is ignored, results can become statistically “possible” but hydrologically inconsistent.

  • Use enough samples for stable statistics

    • Small collections of computed results can yield noisy percentile bands and unstable exceedance estimates.
    • Increase sample size until key metrics (median peak, 90% band, threshold exceedance probability) stabilize.

  • Summarize in decision-relevant metrics

    • Percentile hydrographs are useful, but decisions often require metrics like:
      • Probability of exceeding a peak/volume threshold
      • Confidence interval on time-to-peak / threshold crossing time window
      • Exceedance probability over a time window

  • Uncertainty results reflect assumptions

    • Uncertianty results are only as valid as the uncertainty model you define (ranges, distributions, dependence).
    • Always sanity-check individual results traces and compare against observed variability where observations exist.

Documentation & Resources

📖 User’s Manual

Step-by-step guidance for setting up uncertainty analyses and interpreting results.

📐Technical Reference Manual

Conceptual and mathematical background, plus the uncertainty methods available in HEC-HMS.

🛠️ Tutorials & Guides

Worked examples for applying the Uncertainty Analysis compute option and interpreting results.

This document was drafted from the HEC-HMS User's Manual (HMSUM), Technical Reference Manual (HMSTRM), and Tutorials & Guides (HMSGUIDES) Confluence spaces.