Validation of Hydrological Models Using Stable Isotopes

1
Validation of Hydrological Models Using Stable
Isotopes

• Tricia Stadnyk1
• St.Amour, Natalie2 Kouwen, Nicholas1 Edwards,
  Tom2 Pietroniro, Alain3 Gibson, John4
• 1 Civil Engineering and 2Earth Sciences,
  University of Waterloo, Waterloo, ON, N2L 3G1,
  Canada
• 3 NHRC, Environment Canada, 11 Innovation Blvd,
  Saskatoon, SK, Canada
• 4 National Water Research Institute, Water
  Climate Impacts Research Centre, Victoria, BC,
  Canada

2
Overview

• Objectives
• Modelling Background
• Hydrological model
• Tracer module
• Study Site
• Fort Simpson, NWT
• Results
• Wetland connectivity
• Early break-through
• 1997 Simulation
• Conclusions
• Future Work

3
Objectives

• To validate the partitioning of water in a
  hydrological model using d18O and d2H isotopes
• Sources
• Flow paths
• Water balance
• Model isotopic fractionations within a
  hydrological model framework
• Quantification of flow paths
• Water cycling and budgets

4
Hydrological Model

• The WATFLOOD Hydrological Model
• Developed over the past 30 years by Dr. Nicholas
  Kouwen
• Primary application is flood forecasting and
  flood studies
• Long time sequences for climate studies
• Model very large and small areas
• 1,700,000km2 for the MacKenzie River
• 20km2 for Tri-Creek, Alberta
• Optimal use of gridded data sources
• Land cover, DEMs, Radar data
• Universally applicable parameter sets
• Quick turn around time

5
Hydrological Modelling
6
Tracer Module Components
Tracer 0 Baseflow separation
Tracer 1 Sub-basin separation
Tracer 2 Land-cover separation
Tracer 3 Rain-on-stream tracer
Tracer 4 Flow-type separation - surface -
interflow - baseflow
Tracer 5 Snow-melt as a fn(flow-type) - surface
surface melt - interflow melt drainage -
baseflow interflow melt drainage
Tracer 6 Glacial Melt - surface - interflow -
baseflow
7
E.g. Baseflow Tracer Model

• Add tracer to groundwater inflow to stream
• Mass IN Conc qlz Dt
• Calculate mass of tracer leaving stream reach
  (iterative) ? ROUTING
• Tracer storage balance in the reach
• S2 S1 (In Out) / 2
• Calculate tracer concentration in the reach
• Tracer STORED / Channel Storage
• Calculate tracer mass leaving the reach
• Mass OUT ConcQstreamDt ConcEvap Dt
• - corrected for evaporative losses to preserve
  mass
• Mass OUT Mass IN (for next reach)

8
Study Site
http//atlas.gc.ca/site/english/maps
http//www.fortsimpson.com/

• Near to community of Fort Simpson, NWT
• Confluence of MacKenzie and Liard Rivers, Canada
• Inter-relationship via spring melt
• 5 river basins studied
• Ranging from 202 - 2,050km2
• April ? August 1997 to 1999

http//www.fortsimpson.com/
9
Basin Delineation

• Jean-Marie R.
• 1,310 km2
• Martin R.
• 2,050 km2
• Birch R.
• 542 km2
• Blackstone R.
• 1,390 km2
• Scotty R.
• 202 km2

The Lone rain gauge
10
Landcover Data

• Semi-permafrost
• Wetland dominated
• 65 organic peat
• Most vegetation is transitional
• NW-W ? relief
• Martin steepest
• Scotty flattest

Red/Orangemixed/decid Greenconifer Yellowtransi
tional Light bluewetland Dark bluewater
Töyrä, Jessika, 1997
11
Results

• Accuracy of simulation determined by
• Simulated Q ?? Measured Q
• Nash-Sutcliffe coefficient (R2) and Dv
• Modelled GW ? GW-d18O and GW-d2H
• Proportionality plots
• Isotopes identified problems with wetland
  connectivity to channel
• Hydraulically connected v. disconnected
• Early break-through problems
• Implementation of a dispersion coefficient

12
Wetland Connectivity

• E.g. Birch river 25 total area as wetlands
  (LandSAT)

?? About 10 of the wetland coverage is directly
interacting with the channel
13
Early Break-Through

• Flow Component gt Total Flow
• Result of instant mixing with large grid sizes

• Added retardation coefficient to disperse tracer
  movement
• Based on Pechlet number
• Constrains mass concentrations to a maximum of
  1.0 in any given time step
• Excess is put into stream at a later time step

14
1997 Simulation

• Reasonable fit
• R2 from 0.9 (Jean-Marie) ? 0.5 (Blackstone)
• Dv from 1 (Birch) ? 33 (Blackstone)
• Missed summer event
• Rain gauge error

15
BaseflowSeparation

• Baseflow is consistently under-estimated
• Too much dispersion (?)
• Wetland connectivity not constant
• Baseflow does not include water in wetlands
• Isotopically, it might!

16
Total Flow Separations

• S(GWIFSW) ? Simulated flow
• Water in wetlands
• Rain on stream/wetlands

17
Flow Separations

• Reasonable results
• No SW or IF isotopes for comparison
• No early break-through
• S(GWSWIF) ? Simulated Q
• (but not exactly)
• Wetland water
• Rain
• Dispersive error ?

18
Conclusions

• Reliable precipitation is critical in producing
  reasonable streamflows (duh!)
• At least 1 rain gauge per basin

• Wetland hydraulic connectivity
• Use isotopes to help quantify the degree of
  connectivity
• Identify any seasonal interactions and releases
  (i.e. Deltas)

• Accounting for dispersive effects
• Simulate dispersive mixing to avoid early
  break-through
• Allows use of large grid sizes and simplified
  routing

• Incorporation of isotopes are invaluable for
  hydrologic modelling!!
• Identification of model errors or inaccuracies
• Validation of modelled flow paths
• How much water? From where?
• Snowmelt quantity and timing
• Evaporative losses

19
Future Work

• New study sites for modelling
• More isotope data required
• Good meteorological data
• Model isotopic fractionation
• Focus on evaporation in WATFLOOD
• Coupling with WATCLASS (energy balance model)
• Wetland connectivity estimation
• Based on LandSAT imagery

20
Acknowledgements

• Natural Sciences and Engineering Research Council
  (NSERC) for funding
• University of Waterloo, Graduate Student
  Association for travel assistance
• Departments of Civil Engineering and Earth
  Sciences, University of Waterloo
• Dr. Bill Annable for his guidance suggestions