Part III: WATFLOOD Modelling of the Mackenzie

1
Hydrological Modellingwith WATFLOOD
http//www.watflood.ca
2
Aknowledgements

• Don McMullen, MNR (ON) hydrometeorologist
  (retired) 1972 (Inspiration)
• Alex Harrington (programmer) Jack Gorrie and
  Greig Garland (radar) John Donald (snow) Todd
  Neff (evaporation) Frank Seglinieks (mapmaker)
  Bob McKillop(wetlands) Trish Stadnyk(wetlands
  tracers) Allyson Bingeman (sensitivity
  analysis) Shari Carlaw (soil moisture
  validation) Luis Leon (water quality) NSERC
  (funding for program development)
• Ric Soulis (Applications)

3
Modelling objectives for WATFLOOD

• Flood forecasting and flood studies
• Continuous modelling
• Ability to model very large as well as small
  areas
• Ability to optimally use gridded data sources
  e.g.. Land cover, DEMs, NWP model output, Radar
  data
• Universally applicable parameter set
• Quick turn around
• Ability to model a wide variety of landscapes

4
Topics

• Overview
• Grouped Response Units (GRUs)
• Hydrological model (Vertical water balance)
• Dealing with snow
• Channel Lake Routing model
• Meteorological Forcing requirements
• Coupling to Weather Models
• Model validation
• Long time series

5
Overview
6
Modularity

• Modularity separate programming units for
• domain configuration for modelling (Ensim, Topaz,
  Mapmaker)
• event generation (Menus)
• point data to distributed data conversion for
  meteorological inputs (eg. Thiessen, distance
  weighting, user supplied)
• output visualization (e.g. Ensim, Grapher,
  Surfer, Excel, etc.)
• statistical analysis of output

7
Interfacing with other models

• Gridded model allows 1 to 1 matching of runoff
  units to meteorological driving data from NWM
• Gridded surface model allows 1 to 1 matching of
  recharge to groundwater model such as MODFLOW
• Computed river inflows can be accumulated on a
  reach by reach basis for input to an internal
  Lake routing module or be written to a file in a
  format compatible with routing models such as
  DWOPER or FLOW-1D
• Grid outflow computed with any model can be
  routed with WATROUTE (a subset of WATFLOOD Code)

8
Scaling/Domain Size

• WATFLOOD has been used with grid sizes from 1 to
  25 km and for watershed areas from 15 to
  1,700,000 km2
• WATFLOOD is not sensitive to grid size as long as
  there are a sufficient number of grids to
  maintain the integrity of the drainage system and
  preserve the variability in the meteorological
  data
• Regional model models multiple watersheds
  (cannot properly calibrate model with one or two
  flow stations)

9
IP3
Map created by Jackie Bronson
10
GRU Method

• Grouped Response Units

11
Nelson House - Thompson LANDSAT
BOREAS Northern Study Area This area is near
Thompson, Manitoba, Canada. This image shows a
number of areas burned at different times in the
past
12
Ontario LANDSAT
Uncorrected Landsat image of Southern
Ontario, Canada. The transition from the farm
region in the south to the forested Canadian
Shield is shown This and the previous images
show the wide variety of physiographic areas
that need to modelled with hydrological models.
Ideally, one model can perform this task on such
a variety of terrain. The grouping of
hydrological units facilitates this goal.
13
WATFLOODs Main Features

• Grouped response units (GRUs) will lead to
  universal parameter set
• Gridded model
• optimal use of remotely sensed data
• optimal use of numerical weather data
• optimal use of 1,2 and 3D display facilities
  (e.g. ENSIM)
• In WATFLOOD we ignore connectivity at the small
  scale

14
Elmira (S. Ont.) LANDSAT

• 10 km grid
• 100 km2 area receives equal meteorological input
• so group all areas with similar hydrological
  characteristics within a grid for 6 hydrological
  computations/grid
• some people model each pixel or each field
  separately - ok for science, not operations (104
  computations/grid)

15
The number of classes are reduced to a
manageable number, in this case to 5
hydrologically significant classes.
16
Watflood Theory
17
Parameters are for land cover classes A, B, C
D Parameters do not change with percentage of
each land cover (i.e. no grid or watershed
parameters) Each grid is represented by a
watershed with its own cover allocation.
18
Hydrological Model
19
Hydrological features

• 3 zone model
• surface
• upper zone (saturated, varying depth) (UZS)
• unsaturated zone (API dependent soil moisture)
• saturated lower zone (LZS)
• Hargreaves, Priestley-Taylor or climatic
  evaporation
• Green-Ampt infiltration, Hortonian runoff model
• Andersons snow model with cover based snow cover
  depletion curves
• Glacier melt (crude)

20
Water System Modeling
Evapotranspiration
Interception
Depression Storage
Wetting Front
Unsaturated Zone
Saturated Zone
Channel Flow
21
Overland Flow
H
d
Qs
Depression Storage (DS)
d - effective depth for surface flow
From Mannings Formula
Where x is a function of the grid size and
slope R3 is an optimized parameter
22
Schematic of the Infiltration Process
Intermediate Zone Storage (IZ) (Unsaturated)
23

• Hill slope for overland flow in the Saguenay area
• Slope needs to be represented in the model

24
Contour Density
Internal Slope

• 14 contours cross the diagonal line
• Line length is equal to the length of the sides
  of the grid
• Provides approximation of the localized relief
  and actual overland slope

Actual Slope
Average Slope (no good)
25
Interflow

• DUZ REC (UZS-RETN) Si where
• DUZ hourly depth of upper zone storage
  released as interflow in mm
• REC a dimensionless discharge coefficient
  (optimized)
• UZS water accumulation in the upper zone
  region in mm
• RETN retention (soil depth field capacity)
• Si internal slope (land surface slope)
• This is an important physiographic feature
  that controls surface flow interflow (before
  the water enters a stream)

26
Snow melt
When there is a mean snow depth of 10 cm, 100
of the ground will be snow covered For 5 cm,
50 will be covered in this example

• As snow melts, bare areas appear and grow
  larger
• Snow can not melt in bare areas
• Use a relationship between mean snow depth
  snow covered area
• A lookup table (bsnm.sdc) is used to enter the
  relationship
• Each land cover class has its own SDC

27
Channel Lake Routing
28
Routing features

• Storage routing (center difference KW solution
  with variable time steps to satisfy Courant
  criteria everywhere)
• Coupled stream-wetland routing model
• Lake routing, reservoir operating rules
  (diversions)
• Overbank flow (with different resistance
  coefficients)
• River, Lake and groundwater initialization based
  on recession curve of observed hydrographs.

29
Typical upland stream in the Saguenay Area
Bras Hamel at Highway 381 Channel section 15ft.
by 3ft. Longitude 48o-10'-24'' Latitude
70o-51'-10'' Direction South-West
30
Channel characteristics need to be quantified
for proper hydraulic routing in a hydrological
model. There are geomorphological
relationships for natural channels. In the case
shown here, we have to rely on engineering
relationships
Ticino River near Bellinzona, Sw.
31
Channel Cross-Section - Drainage Area Relationship
XA a(DA)b
32
Assumed Channel Section

• Fieldwork is still required to confirm assumed
  section (we now assume a rectangular section)

33
Channel Flow

• Bankfull Area is a function of the drainage area
  (from field surveys)
• Routing SIMPLE or STORAGE routing throughout
• GRID by GRID
• Highest grid to lowest

34
Wetland Model
35
Sapochi LANDSAT Classification
Wetland Coverage is approximately 7
36
Sapochi Field Site
Low flow period
37
Modelling Wetland Coverage
WRR, 35(4)
38
Dupuit Discharge Model
WRR, 35(4)
39
Finite Difference Model
40
Simplified Model
41
Modelled Hydrologic Processes
42
WATFLOOD-Wetland Mathematics
Wetland-Channel interaction is governed by the
Dupuit-Forchheimer discharge formula
Wetland inflow is defined by
Lateral channel inflow contributing to channel
inflow (I)
WATFLOOD storage routing
Channel Outflow
Channel Inflow
43
(No Transcript)
44
South Tabacco Creek Near Morden, Manitoba
45
(No Transcript)
46
Adaptation

• On the fly updating of snow pack
• On the fly updating of land cover
• Precip. smearing (convert 24 hr. precip. to a
  distribution)
• Precipitation Adjustment Factors (PAF) (to
  correct for bias in precip. data.)
• Unsaturated zone updating using API

47
Special Features

• Interfaced with ENSIM-HYDROLOGIC
• Chaining events unlimited simulation length
• Automatic soil moisture initialization for flood
  forecasting
• Grid shifting for ensemble forecasting
• Forecast mode for training purposes (in
  remission)
• Hooke Jeeves pattern search optimization
• Stage output for flood warning

48
EnsimHydrologic

• Developed by the Canadian Hydraulics Centre CHC
• Funded by Environment Canada
• Pre post processor

49
ENSIM-HYDROLOGIC GIS for hydrological models
50
Pre Processor S. Ontario DEM with automatic
watershed channel generation
51
Automatic basin file creation 10 km grid for S.
Ontario showing the Grand, Saugeen, Thames,
Maitland and Nottawasago Watersheds. Stream
channels, watershed boundaries, drainage
directions are generated automatically by
ENSIM-HYDROLOGIC
52
Post Processor Animate and/or plot And state
variable
53
Attribute selection for WFO_SPEC table
71 AttributeCount 1 ReportingTimeStep
Hours 1 1 Temperature 1 2 Precipitation 1 3
Lower Zone Storage class 1 4 Ground water
discharge m3/s 1 5 Grid Runoff 1 6 Grid
Outflow 1 7 Weighted SWE 1 8 Wetland Depth 1
9 Channel Depth 1 10 Wetland Storage in m3 1
11 Wetland Outflow in m3/ ... ... ... 50
additional attributes
... ... 1 62 Recharge mm Class 1 1 63 Recharge
mm Class 2 1 64 Recharge mm Class 3 1 65
Recharge mm Class 4 1 66 Recharge mm Class 5 1
67 Recharge mm (snow) Class 1 1 68 Recharge mm
(snow) Class 2 1 69 Recharge mm (snow) Class
3 1 70 Recharge mm (snow) Class 4 1 71
Recharge mm (snow) Class 5
54
Great Lakes and Ottawa River Model Environment
Canada Alain Pietroniro (Watershed setup) Pierre
Pellerin (Synoptic NWM data) Champa Neal (Flow
data) Nick Kouwen (Assembly)
55
Percent Coniferous Forest Source USGS GLOBAL
LAND COVER CHARACTERISTICS DATA BASE
56
Percent Crops Source USGS GLOBAL LAND COVER
CHARACTERISTICS DATA BASE
57

• Meteorological Data
• 6 hour Synoptic data for initial setup for
  October 2000 August 2003
• 3 hour GEM (Global Environmental Model) data for
  July August 2003

58
(No Transcript)
59

• Movie clip is an example of distributed Synoptic
  Data
• (Note the moving Bulls eyes)

60
(No Transcript)
61

• Next movie clip is for July 2003 using GEM data
• (GEM is Canadas operational weather forcasting
  model)

62
GEM data
63
Animation of Snow Cover (SWE in mm)
64
(No Transcript)
65
Animation of Grid Outflow
66
(No Transcript)
67
Flow stations
68
Computed hydrographs for 115 Sub-Watersheds
400-13500 km2
69
(No Transcript)
70
Lake Routing
Superior
St. Marys R.
Ottawa R.
GB
Michigan
Huron
St. Lawrence R.
Ontario
St. Clair R.
Niagara R.
Detroit R.
Erie
71
Lake Routing Rules (natural state) St. Marys
River Q 824.7(SUP-181.43)1.5 St. Clair
River Q 82.2((MHUSTC)/2-166.98)1.87(MHU-ST
C)0.36 Detroit River Q 28.8(STC-164.91)2.2
8(STC-ERI)0.305 Niagara River Q
558.3(ERI-169.86)1.60 St. Lawrence River Q
555.823(Oswego-0.0014(Year-1985)-69.474)1.5
72
Great Lakes water levels Oct. 2000 Sept. 2003

• Precipitation and temperatures from synoptic
  stations
• Lake evaporation from long-term average
• Storage-discharge equations from pre-project
  conditions

73
Ensemble Forecast South Nation River

• August 2003 not much happened then
• Example does show the concept of ensemble
  forecasts clearly
• We can do this for every watershed in the Great
  Lakes basin simultaneously
• The idea is that this method gives initial
  warning
• Operators would do their own runs and
  interpretations and also use their own data in
  addition

74
GEM forecast 16 trials with different randomly
generated initial and boundary conditions
75
(No Transcript)
76
(No Transcript)
77
GEM Forecast
78
(No Transcript)
79
(No Transcript)
80
Lake Ontario

• Example of ensemble forecast
• August 2003 not much happened here
• We do this for all lakes in the Great Lakes basin
  simultaneously
• We could do this for all of Canada!!

81
(No Transcript)
82
Summary Great Lakes

• Great tools are required to model large areas
  such as the Great Lakes Ottawa River basin.
• Pre-processor set up watershed files
• Post-processor debugging visualization
• GRUs ensure vastly different hydrological units
  are represented appropriately at the large scale
• Gridded model
• Efficient ingestion of gridded data DEM, Land
  cover, meteorological data (radar, numerical
  weather models)

83
Process Validation Where possible, time series
of state variable are compared to observed data.
84
Plots are used to check general principles are
ok. Plot of UZS and LZS Plots of snow
covered area, snow water equivalent and snow pack
heat deficit. Plots of cumulative
precipitation, evaporation and runoff.
85
(No Transcript)
86
Evaporation comparison for the BOREAS SSA-OBS
Tower Site
87
Evaporation comparison for the BOREAS NSA-OBS
Tower Site
88
Model verification

• E.G. Baseflow has been compared to isotope
  analysis of streamflow sources
• All other model components have been similarly
  verified

89
WATFLOOD Application

• Hydrologic studies for Dam Safety on the Columbia
  River System
• Study initiated by Bill Chin
• Contributors
• Maurice Danard, Wuben Luo, Allyson Bingeman,
  Frank Seglenieks, Ric Soulis

90
Columbia River LANDSAT Composite
91
70 pixels wide Each pixel 90m You will note
that there is an infinite variety of
classifications. Increase the number of
classifications allowed, and you will find that
side-by-side pixels are all different. Keith
Beven has an interesting article where he talks
about the uniqueness of place. An examination
of pixels will show that each one is unique
92
4 km grid generated by ENSIM-Hydrologic
93
(No Transcript)
94
(No Transcript)
95

• Adjusting rainfall (really??)
• First run showed substantial north-south bias
• Generated an gridded error field
• Created Precipitation Adjustment Factors (PAF)
• Adjusted precip with PAFs to remove bias

96
Comparison of observed SWE to modelled SWE for
for the Columbia River basin.
97
Comparison of snow pillow data and WATFLOOD/SPL
SWE estimates (Janet Wong)
98
(No Transcript)
99
(No Transcript)
100
(No Transcript)
101
Principle of geological weighing lysimetersby
Garth van der Kamp amd Saul Marin NHRC, Env.
Can.

• Based on the theory of soil mechanics
• Changes of mechanical load near the ground
  surface (e.g. buildings, embankments, trains) are
  transmitted instantaneously to underlying
  formations and are evidenced by changes of
  groundwater pressure.
• If the formation is confined by a
  low-permeability layer changes of groundwater
  pressure will reflect changes of vertical water
  balance due to hydrological processes such as
  snow accumulation, rainfall and evapotranspiration

102
Duck Lake SK observation wells - observed water
levels since 1964

103
DuckLake 2 1964-2005 Vertical water
balanceObservation well record vs Watflood
model results
The model follows the seasonal variation, but
not the low frequency phase (long memory)

104
DuckLake 2 1964-1970 Vertical water
balanceObservation well record vs Watflood
model results

105
Coupling to other models
106

• WATFLOOD is designed to be linked to other
  models
• Numerical Weather Models Groundwater models
  gridded input and output data. No interpolation
  needed if grids coincide.
• Inflow collected by user specified reaches for
  use in
• Dynamic Wave models,
• Lake models
• reservoir operational models
• diversions

107
Some applications

• Canadian participation in the Mesoscale Alpine
  Project (MAP) Special Observing Period (SOP)
• Sept. 15 - Nov. 15, 1999
• European Alps
• Produce daily flow forecasts for Ammer, Toce
• Ticino rivers.
• Robert Benoit, Pierre Pellerin, Stephane
  Chaimberland, Jayson Innes Kouwen.

108
MC2 6hr precip 0600Z 09/26/00Case study MC2
precipitation as input to WATFLOOD model
109
Watflood river stage 0600Z 26/09/99
In this image, the rivers are shown schematically
and the colours show the state of the river
relative to bankfull flow. The segments shown
in red are predicted locations where floodplains
are flooded, roads may be covered with water. The
time of the image is shown also.
110
MC2 6hr precip. 1200Z 09/26/99
111
Watflood river state 1200Z 26/09/99
112
Watflood river state 1800Z 26/09/99
113
Watflood river state 0000Z 27/09/99
114
Watflood river state 1200Z 27/09/99
Approximately one day after the heaviest
rainfall, the small tributaries have drained and
the water is still in the main rivers.
115
This is a continuous simulation using MC2 data as
input to WATFLOOD. It shows that so far, MC2
rainfall predictions are good.
116
This slide shows what would happen if
MC2 predicts precipitation in the wrong place. Of
course, it may not be reasonable to shift
orographically driven precipitation to the other
side of a mountain.
117
MAP (Fall 1999) Computed flows compared to
observed flows for the Danube River in Germany
Austria Met data from the high resolution MC2
Numerical Weather Model MC2 WATFLOOD
3 km grid
118
Mackenzie River Model A work in progress
2000 km X 2000 km
119
No Wetlands
120
With Wetlands
121
Note Lakes
No Wetlands
122
With Wetlands
123
Summary

• WATFLOOD is a versatile hydrological modelling
  system capable of modelling most of the Canadian
  landscape (still need a lot of work on permafrost
  regions)
• WATFLOOD has been validated in many different
  ways, for very different regional applications,
  over a long period of time.

124

• Grouped Response Units (GRU) and grid system take
  care of distributed land cover and meteorological
  input
• GRUs make WATFLOOD insensitive to scale because
  individual pixel locations of land cover units
  are inmaterial
• Because parameters are tied to land cover and
  apply to any physiographically similar landscape,
  parameters are universally applicable (sort of).

125
IP3
Map created by Jackie Bronson EC