Great Lakes Forecasting System

1
Great Lakes Forecasting System
Keith W. Bedford, Philip Chu and David
Welsh Department of Civil and Environmental
Engineering and Geodetic Science The Ohio State
University
2
Great Lakes Forecasting System

• A program of research and testing whose goal is
  to implement an operational system in each of the
  Great Lakes for making six hour forecasts of
  currents, temperatures, wind waves, water levels
  and other associated physical data

3
SYSTEM GOAL
A General System for Coastal Forecasting
Containing 1. Sufficiently credible, complex,
and integrated models which reflect the
effective processes controlling the information
desired by the consumer 2. Integrated
atmospheric coupling necessary to resolve the
momentum and heat fluxes with enough spatial and
temporal resolution to portray the effect of
storms 3. An operational framework for routine
prediction, dissemination of products, and
verification and 4. An observation network
capable of supporting 2. and 3.
4
SYSTEM ATTRIBUTES

• Modular System Architecture
• Image-Based User Friendly System Interface
• Full Exploitation of Three-Dimensional Model
  Technology
• and Supercomputering Availability
• Forecast Resolution Permits Accurate Prediction
  in Shallow,
• Nearshore Regions as well as Deeper Water
• System is Based Upon Existing Data and Networks
  which are
• Fully Integrated and Utilized in the Forecast
• Products/Output in Form of Maps for Wide
  Distribution to
• Consumers and/or Other Data Bases

5
SYSTEM USES

• Hazard Forecasting, Warning and Reduction
• Enhancement of Commercial and Recreational
• Activity
• Scenario Testing, Design and Risk Assessment
• Resource Preservation Activities

6
EXAMPLE USES

• Hazard Forecasting, Warning and Reduction
• Improved nearshore wave forecasts
• Water level/surges - harbor traffic, shore
  erosion
• and floods
• Spill trajectory
• Inlet sill blockage and mouth barring status
• Shore erosion/ bluff failure
• Man overboard, search and rescue
• Casualty Investigations

7
EXAMPLE USES

• Enhancement of Commercial and Recreational
  Activity
• Currents and thermocline data for recreational
• fisherman
• Upwelling and downwelling data for improved
  siting
• of artificial reefs
• Improved withdrawal of freshwater of known
• temperature and clarity for commercial, utility
  and
• industrial uses
• Real time data for underway water withdrawal by
• commercial ships

8
EXAMPLE USES

• Scenario Testing, Design and Risk Assessment
• Shore erosion/change resulting from proposed
• modification
• Effect of dredging disposal on harbors and
• openwater sites
• Siting of outfalls and withdrawals
• Long term water level impacts on erosion and
• residential and commercial activities
• Effects of global climate change on the lakes
  and the
• regional economy

9
EXAMPLE USES

• Resource Preservation Activities
• Improvement management of wetlands, parks and
• beaches
• Increased efficiency of fish stocking
• Siting of artificial reefs
• Tributary loading estimates and areas of concern
• Water quality event disruptions and planning
• Adaptive or timed placement of dredging material
• Wetlands water level

10
WHATS FORECASTED

• First Phase
• Full three-dimensional temperature velocity
  field
• Water Levels
• Wind waves, both nearshore and deepwater
• Tributary effects
• Second Phase
• Sediment transport
• Third Phase
• Watershed coupling

11
FORECAST TRANSMISSION AVAILABILITY

• GLFS Website at http//superior.eng.ohio-state.ed
  u
• 1.2 million hits in 1999
• Transmissions to Boats Underway via Cellular
• Telephone, Satellite or Radio Telephone
• Regulatory Agency/Government Access Through
• Regional NOAA-Coastwatch Network

12
HOW IS THE SYSTEM USED

• Hindcast
• From events in the archives
• Nowcast
• An estimate of the present condition
• Forecast
• Nowcast - Two day forecast
• Scenario Testing
• Designer event-based upon either probability of
• event occurrence(archives) or construct your own
• event

13
PHYSICAL INTERACTIONS
Global Circulation Model Input
ATMOSPHERE
evaporation
precipitation
shear stress, heat flux, precipitation
heat, moisture, roughness
precipitation
WATERSHED TRIBUTARY RUNOFF
GROUND WATER
LAKE RESPONSE
erosion, sediment transport
wave erosion
slope failure
erosion
ACTS
COASTAL MARGIN
14
REQUIRED INPUT DATA

• Circulation, Temperature, Water Levels, Wind
  Waves
• Weather Data - Wind Direction Speed, Dew
  Point,
• Temperature and Cloud Cover - NWS/AES
• Water Levels - NOS - Independent Evaluation Data
• Tributary Inflows - USGS
• Water Surface Temperature and Heat Fluxes
• NOAA AVHRR Satellite - Independent Evaluation
  Data
• GOES Satellite

15
GREAT LAKES FORECASTING SYSTEM

• Mellor-Blumberg(1987) Circulation Model
• Dynamic variables free surface, 3-D velocity,
• Temperature, turbulent KE, turbulence macroscale
• Sigma coordinates
• Mode-splitting time step procedure
• Vectorized Fortran code - Cray Y-MP
• Typical parameters for the Great Lakes
• 2 km horizontal resolution (209 X 57)
• 11 vertical intervals
• External time step 20 seconds
• Internal time step 10 minutes
• Cray Y-MP time 4 minutes/day

16
GREAT LAKES FORECASTING SYSTEM
INPUT S OBJECTIVE
INITIALIZATION MODEL OUTPUTS
ANALYSIS
METEOROLOGICAL
3D Water Current Field
SFC WX Observations (NWS Marine)
Spatial Temporal Interpolation
3D Water Temperature Field
SFC Wind Field (NWS)
3D Primitive Equation Numerical Model 2D Wave
Forecasting 2D Tributary Model
Calculate SFC Wind Stress Sensible Heat Flux
SFC Water Level Field
WATER
SFC Water Temp. (Marine WX Stns)
Wind Waves (deep shallow)
Spatial Temporal Interpolation
SFC Water Temp. (AVHRR)
3D Sediment Field
Evaluation Skill Tests
Water Levels
Shore Erosion Potential (Future)
Tributary Inflows
Bathymetry
17
NOWCAST - THE PRESENT CONDITION OF THE LAKE
Day I-1
Measured Water levels AVHRR Satellite
Data Cleveland, NMC Weather Data Wind Fields,
Cleveland NWS Cloud Cover Cleveland Marine Dew
Point Forecast Center Surface Air Temperature
Evaluate Day I-1 Nowcast
Day I
Input Data Day I-1
Model Prediction (Up to Day I)
To Forecast Module
1
Day I
Product Distribution (Day I)
Evaluate Day I Nowcast
Day I1
18
FORECAST
1
Wind Fields NWS/Penn St. Mesoscale Cloud Cover,
Dew Point, Air Temp. NWS Water Temp. Persistence
Forecast Input Data
Day I
Forecast Prediction Day I - III
Product Distribution
Measured Water Levels AVHRR Satellite
Data Cleveland, NMC, AFOS Data
Forecast Evaluation
19
GLFS SYSTEM INTEGRATION

• DATA INPUT
• 17 Canadian and U.S. Weather Service Stations
• 9 Canadian and U.S. Coast Guard weather stations
• 14 Canadian and U.S.water level gauges
• 6 U.S. streamflow gauges
• 17 Ships at sea
• 4 GOES-8, AVHRR satellite channels
• 4 U.S and Canadian wave rider and thermistor
  buoys

20
GLFS SYSTEM INTEGRATION

• MODELS
• 1 parametric wave prediction model
• 1 spectral wave prediction model
• 4 two-dimensional flow and tributary models
• 3 three-dimensional flow and transport models
• 3 mesoscale atmospheric prediction codes
• 2 heat flux and wind shear codes

21
GLFS SYSTEM INTEGRATION

• Computers
• 1 CRAY T3E at Ohio Supercomputer Center
• 1 CRAY at National Center for Environmental
  Prediction(NCEP)
• 7 Silicon Graphics Unix workstations (OSU)
• 3 Linux Workstations (OSU)
• 3 Pentium class PC (OSU)
• Personnel
• 2 Project directors
• 4 research Scientists
• 5 Co-principal investigators
• 10 graduate Students
• Finances
• 21 Contracts from 16 sources

22
GLFS SUPPORT DISTRIBUTION

• NASA - ARPA - ACTS
• NOAA -
• Coastal Ocean Program
• OAR, NOS
• ERL, NWS
• NOAA - Ohio Sea Grant (3)
• CRAY Research Inc. (3)
• Camp Dresser and McKee
• Havens and Emerson
• Northeast Ohio Regional Sewer District
• ODonnell Fund
• NASA - Center for Commercialization of Space
• USGS - Water Resources Center (2)
• USEPA
• Ohio Supercomputer Center
• Office of Research Ohio State University
• Army Corps of Engineers
• Silicon Graphics Inc.
• National Weather Service (tent.)

23
IMPACTS

• Used as proof of concept for the Coastal
  Forecasting System Initiative
• Allowed placement of Cleveland CSO remediation
  plans to the placed on a joint probabilistic
  basis i.e. sewershed - Lake Erie coupling.
• Full impact of investment in models and data
  collection realized.
• First generation of marine forecasters now being
  trained in our graduate program.
• Enhanced field research activities by EPA, USAC,
  and various universities dynamic site allocation.

24
1998 Lake Michigan Plume Event
25
RESEARCH PROBLEM

• The near shore circulation/transport is too
  complex for solution by typical disaggregated
  uncoupled models
• SOLUTION
• 1. Full coupling of all models through physics,
  computing algorithms and data passing
• 2. A parallel structure is required

26
Models Used (COMAPS)

• CH3D-SED marine circulation and sediment
  transport model
• WAM wind-wave model
• WCBL combined wave-current bottom boundary layer
  model
• collectively, the COastal MArine Prediction
  System (COMAPS)

27
WAM Wind-Wave Model (WAM development group)

• wave energy conservation equation solved for each
  component in a frequency-direction spectrum
• spherical (longitude-latitude) grid
• explicit (upwind) finite differencing
• source/sink terms for wind input,
  wave-waveinteraction, whitecapping, and bottom
  friction
• current-induced propagation and refraction
• depth-induced refraction

28
CH3D circulation model (U.S. Army Corps W.E.S.)

• conservation equations in 3-d for mass, momentum,
  energy, and salinity
• predicts 3-d currents and 2-d water elevations
• accounts for winds, surface heat flux, tides,
  density variations, rivers, and Coriolis force
• curvilinear, horizontal grid and sigma-stretched
  (terrain following) vertical grid
• mixed explicit and implicit finite differencing
• mode-splitting strategy

29
SED sediment transport model (Iowa Institute of
Hydraulic Research)

• conservation equations for bedload sediment and
  suspended sediment
• equations applied to multiple user-specified
  sediment size classes
• source/sink terms for erosion and deposition
• SED is completely integrated in the CH3D-SED
  model
• called during CH3D internal mode time-step
• implicit finite differencing

30
WCBL model

• combined Wave Current Boundary Layer model
• based on Glenn and Grant (1987), Grant and Madsen
  (1982), and Keen and Glenn (1998)
• written by Ohio State Univ. researchers
• nonlinear interaction of wave and current
  boundary layers needed because
• WAM does not account for sediment transport and
  uses a constant bottom friction coefficient
• CH3D-SED does not account for waves and uses a
  simplistic parameterization for bottom shear
  velocity

31
Coupling Strategy and Physics

• Coupling at the atmospheric boundary layer
• CH3D sends WAM surface current and elevation
  arrays
• WAM sends CH3D radiation stress and wave stress
  arrays
• MPI calls between master processes

32
Parallel Codes

• Parallel CH3D-SED uses the MPI library
• 1-d domain decomposition
• WCBL called by each CH3D-SED process
• CH3D-SED domain decomposition retained
• Parallel WAM uses the OpenMP library
• loop level parallelism
• Coupled runs use SGI ORIGIN 2000 platform

33
Parameter exchange in COMAPS
twave , trad Us , H
U(s1)
U cw
Ub Ab
H rb
fcw
H r
U (s1)
rs , Dn , D50 , Cbn
U cw
34
Lake Michigan Domain decomposition
35 rows
37 rows
blocks generated for a 4 CPU parallel run
35 rows
38 rows
Depth contours for 4-km grid
35
Communication schematic
MP_SET_NUMTHREADS3 mpirun -np 2 ch3dsed -np 1
wam
WAM
CH3D-SED slave
CH3D-SED master
i
ii
iii
Call WCBL
Call WCBL
36
CH3D/WAM performance larger grid
Wallclock hours

• ERDC O2K
• Lake Michigan
• 2 km grid
• (133 x 252)
• 24-hr. run
• 1 x CH3D
• no. of WAM
• threads varied

number of WAM threads
37
CH3D performance larger grid
Wallclock hours

• ERDC T3E
• Lake Michigan
• 2 km grid
• (133 x 252)
• 24-hr. run
• no. of CPUs
• varied

number of processors
38
Diagnostic Component Analysis

• To determine which physical component are the
  most important in developing the resulting plume

39
Sediment particle analysis

• Analyze the Sediment Particle Histories and
  Trajectories During Plume Events
• Map the Initial Distribution of the Bottom
  Sediments, e.g

40
Sediment particle analysis

• Define Other Sediment Sources Required for the
  Analysis
• Identify the Sediment Sources Using Particle
  Tagging
• Model Results will give the Means of Determining
  the Horizontal Transport and Entrainment/Resuspens
  ion Rates in the Lake

41
Examples
Examples from a Similar Work in Lake Erie