STRUCTURACONNECTED DIGITAL ENGINEERING
Sign inStart free
STRUCTURA ACADEMIC · LESSON AREA

Water, hydrology and climate-data forecasting

07 · Water, hydrology and climate-data forecasting · AI for Civil Engineering

Course review
StandardInternational professional-learning synthesis · ABET 2026–27 · IEA GAPC v4 · ASCE ethics and AI responsibility · NIST AI RMF 1.0
Source3 source files
Review stateTechnical and publication gates pending
LEARNING OUTCOMES

After this chapter, you should be able to

  • Freeze station, parameter, unit and qualifier metadata.
  • Create persistence and seasonal baselines.
  • Use rolling-origin evaluation.
  • Report event and subgroup errors.
  • Separate educational forecasting from operational warning authority.

Engineering context and methodSource §Lesson 07 · Engineering context and method · GOV-01 · GOV-02 · GOV-03 · RES-05 · DATA-04

Hydrologic records are tied to stations, parameters, rating methods, qualifiers, revisions and time zones. Freeze the exact station/time/parameter extract and preserve provisional or estimated flags. Evaluate in chronological order and compare with persistence. Extreme events and regime shifts matter more than a comfortable overall average; a forecast does not itself establish a warning or operational release decision.

Verified worked exampleSource §Lesson 07 · Verified worked example · GOV-01 · GOV-02 · GOV-03 · RES-05 · DATA-04

WORKED EXAMPLE

Persistence versus model at one lead time

Previous flow is 42 m3/s, observed future flow is 50 m3/s and model forecast is 47 m3/s.

  1. Persistence

    |42 − 50|

    8 m3/s error
  2. Model

    |47 − 50|

    3 m3/s error
  3. Gain

    8 − 3

    5 m3/s at this point

Result. The model is better at this single point, but event coverage, uncertainty, lead time and operational procedure determine usefulness.

Practical lab · 6 h lesson effortSource §Lesson 07 · Practical lab · 6 h lesson effort · GOV-01 · GOV-02 · GOV-03 · RES-05 · DATA-04

  • Freeze an exact USGS station, parameter, time range and qualifier schema or use the approved synthetic fallback.
  • Build persistence and simple seasonal baselines.
  • Evaluate with rolling origins, MAE by flow band and event-focused plots.
  • Write a forecast card specifying lead time, uncertainty and no-warning boundary.

Failure modes to investigateSource §Lesson 07 · Failure modes to investigate · GOV-01 · GOV-02 · GOV-03 · RES-05 · DATA-04

  • Random time split.
  • Provisional values silently treated as final.
  • Station relocation or rating changes ignored.
  • Only average-flow performance reported.
  • Forecast described as an official warning.

Knowledge checksSource §Lesson 07 · Knowledge checks · GOV-01 · GOV-02 · GOV-03 · RES-05 · DATA-04

Five review questions and answer rationales
QuestionAnswer rationale
Why preserve qualifiers?They identify provisional, estimated or otherwise qualified observations.
What is persistence?A transparent baseline that carries the latest observation forward.
Why rolling origin?It evaluates repeated forecasts using only information available at each origin.
Can low MAE prove flood-warning fitness?No; extremes, lead time, uncertainty, reliability and authorised procedures govern.
What must an operational forecast define?Data feed, timing, uncertainty, thresholds, monitoring, fallback, owner and authority.
Use these as formative checks. Technical and editorial review remain pending.

Key points

  • Start from the accountable engineering decision and its consequence.
  • Compare against a transparent non-AI baseline.
  • Validate on a split that represents intended use and retain human authority.

Source references recorded by the supplied chapter

  • USGS Water Data APIs and service documentation.
  • Peer-reviewed ML-in-hydrology review listed as RES-05.
  • NIST AI RMF 1.0.