AI-Hydro Interactive Module · Capstone

A Basin's Water Story

This is where the course converges. You've traced the water cycle, delineated a watershed, read a hydrograph, separated baseflow, fetched real USGS data, and computed TWI. Now you put it all together — characterise one basin end-to-end and tell its water story.

Pick the basin you studied in Module 5 (or try a new one) · Run all three cells · Read the story · Export your report
Capstone synthesis Prerequisites: M4 (baseflow separation) · M5 (real data) · M6 (terrain/TWI). Every concept from Modules 1–6 surfaces here — the six flux arrows, the watershed divide, the hydrograph limbs, the Lyne–Hollick BFI, the USGS record, and the TWI wetness map.

Seven-step capstone — click to jump

📍
Choose Basin
📈
Record
🔬
Signatures
🏔️
Terrain
📖
Story
✓
Check
🚀
Next

1 · Choose your basin

Select the basin you explored in Module 5, or pick a different one. The same four teaching gauges are available — each with a distinct hydrological character. All three kernel cells below will update when you change your selection.

🌊
01013500
Fish River, Fort Kent, ME
Flood-prone · mixed rainfall-snowmelt · BFI ≈ 0.55
❄️
14305500
Siletz River, Siletz, OR
Pacific NW snowmelt · wet winter regime · BFI ≈ 0.62
🏜️
09484000
Rillito Creek, Tucson, AZ
Flashy arid · monsoon spikes · BFI ≈ 0.08
🌿
02472000
Bogue Chitto, Bush, LA
Groundwater-dominated · flat regime · BFI ≈ 0.82

2 · The streamflow record

This cell synthesises 10 years of daily discharge shaped to your basin's hydrological character and computes key statistics. Run it first — the signatures cell below builds on these results.

python 10-year daily streamflow record + flood frequency

    

3 · Hydrological signatures & flow duration curve

The Lyne–Hollick BFI and flashiness index compress the whole record into two numbers. The Flow Duration Curve (FDC) shows what fraction of time any given discharge is exceeded — a steep FDC means flashy; a flat one means steady. Run the cell, then compare the FDC shape to the client-side plot rendered below.

python BFI · flashiness index · FDC · seasonal amplitude

    

Flow Duration Curve — interactive (client-side)

Updates instantly when you change basin — no kernel needed. Uses aihydro.plot (Plotly).

Steep slope = flashy basin · Flat slope = steady groundwater-dominated basin

4 · The shape of the land

TWI tells you where the landscape stays wet. Each basin type has a characteristic wetness distribution. Run the cell to see the synthetic TWI grid for your basin and interpret where saturation-excess runoff is most likely to occur.

python Synthetic TWI grid · wetness distribution · classification

    

5 · Your basin's water story

Step through the six modules — each arrow from Module 1's water cycle appears here, annotated with what your basin's numbers say about it. Use the timeline controls or click the dots.

1 / 6
💧 Module 1 · The Water Cycle

A water molecule evaporates from the ocean, condenses in clouds, falls as precipitation over your basin, and begins its journey.

For Fish River, ME, much of the annual precipitation (~1 000 mm/yr) arrives as rain and snow. The six-flux budget: P ≈ ET + R + ΔS constrains everything that follows.

Key concept re-usedP = ET + R + ΔS. The mean Q you just computed is R expressed in m³/s.
🏔️ Module 2 · The Watershed

Every drop that falls inside the Fish River drainage divide eventually passes through gauge 01013500. Topography routes it there via D8 flow paths.

The watershed boundary is an invisible wall drawn by elevation — the continental divide concept applied at sub-basin scale. Flow paths converge in valleys; ridges are the divides.

Key concept re-usedContributing area (a) from Module 2 feeds directly into the TWI formula: TWI = ln(a/tan β).
📈 Module 3 · Reading the Hydrograph

The multi-year record you just plotted is a sequence of storm hydrographs. In Fish River, ME you'll see steep rising limbs in spring (snowmelt) and scattered winter peaks (rain-on-snow events).

The flashiness of the rising limb depends on catchment storage. A high BFI basin (like Bogue Chitto) barely shows individual storm peaks — groundwater smooths them out.

Key concept re-usedLag time and recession shape are encoded in the flashiness index and the FDC slope you computed in §3.
🌊 Module 4 · Baseflow Separation

The Lyne–Hollick filter extracted the slow groundwater component from your record. Your basin's BFI is ≈ 0.55 — meaning roughly 55% of all discharge comes from stored groundwater, not direct storm runoff.

A BFI above 0.7 indicates a basin where soil or aquifer storage dominates (e.g. Bogue Chitto, LA). A BFI below 0.2 signals rapid direct runoff with minimal storage (e.g. Rillito Creek, AZ).

Key concept re-usedα = 0.925 (3 passes) is the standard Lyne–Hollick default — same filter you used in Module 4.
📡 Module 5 · Real Streamflow Data

In Module 5 you found gauge 01013500 on the USGS NWIS map. Here you've extended that record to 10 years to see inter-annual variability — wet and dry years, flood years and drought years.

The annual regime (bar chart in §2) is the most compact summary of how your basin behaves across seasons — a fingerprint unique to its climate and landscape.

Key concept re-usedP95 flood threshold from Module 5 reappears in §2 to count flood days in the longer record.
🌱 Module 6 · DEM → TWI

The TWI grid from §4 tells you where saturation-excess runoff is most likely. Your basin's mean TWI of ≈ 8.2 places it in the moderately wet class.

This completes the spatial picture: the hydrograph tells you when flow is high; the TWI tells you where the landscape was already saturated when it happened.

Full synthesisTogether — BFI + FDC + regime + TWI — these four numbers tell your basin's water story from precipitation to sea.
python Basin characterisation report — export to text

    
AI-Hydro agent action To export a reproducible session with real data: in chat, ask "Export session for gauge 01013500 with all signatures and TWI." The agent calls export_session and returns a JSON artefact you can share or reload in a future session.

6 · Capstone checkpoint

These three questions draw on the full course — Modules 1 through 6. Answer all correctly to complete the course.

Rillito Creek (AZ) has a BFI of about 0.08. Bogue Chitto (LA) has a BFI of about 0.82. Which basin would show a steeper Flow Duration Curve (FDC)?

The water balance equation is P ≈ ET + R + ΔS. If a basin receives 900 mm/yr of precipitation, evapotranspires 450 mm/yr, and has negligible long-term storage change (ΔS ≈ 0), what is the expected mean annual runoff R?

You compute TWI for two basins: Basin A (forested humid Appalachian; mean TWI = 9.1) and Basin B (semi-arid desert wash; mean TWI = 5.3). Which basin would you expect to have more saturation-excess overland flow during a moderate storm?

7 · Where to go next

You've completed the AI-Hydro flagship course — the six flux arrows, seven modules, and every interactive primitive in the platform. Here's what you can do from here:

🏪

Module Marketplace

Browse community modules on flood frequency, recession curve analysis, PET methods, CAMELS benchmarking, and more — one-click install.

AI-Hydro: Browse Modules

🔬

Deeper Skills

Try the baseflow-separation deep-dive skill (multiple filter methods, BFI comparison, CAMELS benchmark) or flood-frequency-analysis (GEV, return periods, confidence intervals).

🤝

Contribute a Module

This course is the canonical reference example at github.com/AI-Hydro/Modules. Fork it, add your own module, and submit a pull request — your work reaches every AI-Hydro user.

The bigger picture AI-Hydro is an open platform: the same HTML module format, the same Python kernel, and the same window.aihydro primitives you used here power every community module. If you can author HTML + Python, you can build a module that teaches hydrological science — and share it with every student who has AI-Hydro installed.

References

  1. Addor, N., Newman, A.J., Mizukami, N. & Clark, M.P. (2017). The CAMELS data set: catchment attributes and meteorology for large-sample studies. Hydrology and Earth System Sciences, 21, 5293–5313. DOI: 10.5194/hess-21-5293-2017
  2. Lyne, V. & Hollick, M. (1979). Stochastic time-variable rainfall-runoff modelling. Institute of Engineers Australia National Conference, Publ. 79/10, 89–93.
  3. Beven, K.J. & Kirkby, M.J. (1979). A physically based, variable contributing area model of basin hydrology. Hydrological Sciences Bulletin, 24(1), 43–69. DOI: 10.1080/02626667909491834
  4. Dingman, S.L. (2015). Physical Hydrology, 3rd ed. Waveland Press.
Module
capstone-basin-story · v0.2.0 · CC-BY-4.0
Authors
Mohammad Galib, Purdue University
Synthetic data
CAMELS-shaped synthetic profiles for 4 curated gauges; no real network calls required.
Curated gauges
01013500 (ME), 14305500 (OR), 09484000 (AZ), 02472000 (LA).
FDC plot
Client-side via aihydro.plot (Plotly); no kernel needed.
Environment
Python (numpy, matplotlib) · Leaflet map via bridge · Plotly via bridge.
CC-BY-4.0