2026년 한국수자원학회 학술발표회 원태상 기념강연(단국대학교 강부식 교수)

Scientific concepts / discoveries / nature phenomena

1) Hydrology “scale” problem (multi-scale behavior)

• Hydrologic and climatic phenomena can appear to have different attributes depending on the observational/analysis spatial scale (and the “window” used to view them).
• For future projections, physical climate model outputs should not be accepted mechanically; they require statistical verification and filtering to judge whether projected changes are significant and credible.

2) Downscaling to connect GCM climate to local hydrologic models

• To couple global climate models (GCMs) with local hydrological (rainfall/runoff) models, the talk emphasizes downscaling because the scale of atmospheric/climate outputs does not match the scale required by hydrology models.
• Downscaling types mentioned:
  • Dynamic downscaling
    • Uses numerical solutions of physical atmospheric equations (temperature, air motion, pressure, precipitation) on nested grids.
    • Has limitations due to physical/numerical constraints.
  • Statistical downscaling
    • Includes:
      • Regression interpolation
      • Stochastic random cascade models (fractal/self-similar approach)

3) Stochastic “random cascade” downscaling model (fractal/turbulent rainfall structure)

• Core premise: fine-scale environmental patterns show fractal characteristics—repeating self-similarly across scales (infinite-like repetition).
• Developed approach combines:
  • A regression AR component, and
  • A random cascade component.
• A key distinguishing scale (~32 km) switches modeling behavior:
  • ≥ 32 km: use an autoregressive (correlation-preserving) approach
  • < 32 km: combine with random cascade reflecting self-similarity/fractality
• The model maintains correlation structures while downscaling:
  • Temporal serial correlation (same location through time)
  • Spatial correlation (neighboring locations at the same time)
• A parameter called β (beta) is highlighted:
  • Represents the scale-variant “rainy fraction” / precipitation area ratio (fraction of area with rainfall above a threshold, depending on averaging window).
  • Treated as scale-invariant in the model to reproduce observed scale-variation of rainy fraction efficiently.
• Evaluation/verification uses statistical indicators (not purely physical ones), including:
  • Mass balance
  • Variability
  • Discretization effects
  • Clustering degree
  • Spatiotemporal correlations

4) Hydrologic application: runoff forecasting using downscaled climate inputs

• Second part: feed downscaled GCM precipitation into a hydrologic runoff model to forecast future runoff.
• Hydrology model referenced:
  • HMS (Hydrologic Modeling System) flood runoff model
• Testbed / basin:
  • South Platte / South Flat River basin (upstream, near Colorado State University)
  • Basin context: large (tens of thousands km²), with relevance to the Missouri River system (Mississippi/Missouri mentioned).

5) Climate-change impacts on precipitation, evaporation, and runoff extremes

Main findings described (using an IPCC scenario; SRES B2 mentioned):

• Temperature increases confirmed (including stronger contextual comparison between regions).
• Average precipitation changes were reported as relatively modest, but variability and extremes changed more.
• Evapotranspiration/evaporation plays a dominant role:
  • Even when precipitation doesn’t change much, runoff and especially peak runoff change substantially.
• Statistical hypothesis testing to decide whether mean changes are “significant”:
  • Behrens–Fisher test (for unequal variances)
  • An additional step: remove outliers and re-test
• Conclusion emphasized:
  • Decreased runoff is attributed more to increased evaporation/evapotranspiration, not to major precipitation decreases.
  • Peak flows were described as more sensitive to spatial/process uncertainties than precipitation.

6) Nature phenomenon / hydrologic hazard example: dry-region flash flooding from convection + orography

• Example event: July 1997 flood in Colorado (Fort Collins area).
• Despite being a generally dry region, intense flash flooding occurred.
• Mechanism (from cited paper analysis):
  • In strong summer radiant heating, heated air rises over Rocky Mountains terrain, producing heavy convective showers.
  • Major damage occurred over short periods, including a flood affecting a university library basement.

7) Related contemporary research theme: salinity/brine damage linked to evaporation and agricultural return flow

• Research mentioned (Colorado Water Center on CSU site) focused on:
  • Saltwater/brine damage in an inland region (not coastal).
• Proposed cause:
  • Agricultural return water (containing organic matter and trace metals like lead) mixes into rivers.
  • Low precipitation + high evaporation concentrates salts → crop damage.
• The speaker noted their earlier (25–26 years prior) forecast that evaporation would dominate runoff behavior in such regions, later corroborated by ongoing research.

---

Methods / methodology outline (as described)

1. Determine a scale threshold

   • Use moment-based analysis (moments up to 4th multipliers mentioned) to find a break around ~32 km.

2. Construct the downscaling model

   • For ≥ 32 km: apply a correlation-preserving autoregressive structure.
   • For < 32 km: apply a random cascade (fractal/self-similar) downscaling.
   • Combine both into an overall downscaled precipitation field.

3. Include scale-variance with β

   • Use β to reproduce the scale dependence of rainy fraction / precipitation area ratio.

4. Validate downscaling

   • Check:
     • Mass balance
     • Variability
     • Clustering
     • Spatiotemporal correlations
   • Use statistical indicators appropriate for a mathematical stochastic model.

5. Hydrologic forecasting

   • Apply downscaled precipitation to HMS (flood/runoff model).
   • Compare early/late periods (e.g., 1st vs 2nd decades around 2100) and compute changes in:
     • Mean runoff / peak runoff
     • Variability / extremes

6. Statistical significance testing

   • Use Behrens–Fisher test (unequal variances).
   • Repeat after outlier removal to confirm which changes are significant.

---

Researchers / sources featured (mentioned at end or in the subtitles)

Individuals

• Jeong Il-mun (preparatory committee chair; Korea Institute of Construction Technology)
• Professor Won Tae-san (honoree; referenced as “Professor Jeongam Won Tae-sang / Won Tae-san of Sujahan Korea”)
• Kang Bu-sik (speaker; Dankook University)
• Yoo Cheol-sang (President of the Korean Mathematical Society; also thanked)
• Jung Il-moon (chair of preparatory committee)
• Fred Upden
• Sheriff
• Smith Richardson
• Behrens–Fisher (named statistical method; not a person in subtitles, but refers to the method)

Organizations / data sources / models / scenarios

• Korea Institute of Construction Technology
• Korean Water Resources Association / Korean Society of Water Resources
• Korean River Association
• Colorado State University (CSU)
• Seoul National University
• GHRC (GCM data provider mentioned)
• NEXRAD radar data (United States-wide radar mentioned)
• IPCC / SRES B2 (climate scenario; relatively mild carbon emission scenario)
• HMS (Hydrologic Modeling System) flood runoff model
• ASC Journal of Hydrogen Engineering (journal mentioned)
• Colorado Water Center (CSU) (referenced for related salinity research)
• CSU (via “South Flat River basin near the CSU” context)

Note: subtitle text includes garbled names/spellings due to auto-generation (e.g., “A Waterlies Research” and “Orb Hydrology,” likely intended as specific paper/journal titles not perfectly rendered).