integrated modeling of the muskegon river ecosystem: a new approach to integrated risk assessment...
Post on 22-Dec-2015
218 views
TRANSCRIPT
Integrated Modeling of the Muskegon River Ecosystem: A New Approach to Integrated Risk Assessment for Great Lakes
WatershedsMichael Wiley1, Bryan Pijanowski2, Paul Richards1, Catherine Riseng1,
David Hynman4, Ed Rutherford3, and, John Koches5
Funded by the Great Lakes Fisheries TrustA product of the Muskegon Watershed Research Partnership
1School of Natural Resources and Environment, University of Michigan2 Department of Forestry and Natural Resources, Pudue University,
3 Institute for Fisheries Res., Michigan Department of Natural Resources4 Department of Geology, Michigan State University, 5Annis Center Grand Valley State University
Muskegon Watershed Muskegon Watershed ResearchResearch
PartnershipPartnership
The vision:The vision:
Collaborative,Collaborative,Integrated,Integrated,
Relevant Science Relevant Science for a better futurefor a better future
http://www.mwrp.net
~4% Lake Michigan’s ‘shed (2870 sq miles )~5% Lake Michigan’s Q (2404 cfs )
Design features:
Start and End with Stakeholders Questions2D spatial org by river channel unit: VSEC or [NHD arcs]Time represented by “frames” in Landscape trajectoryCollection (Integration) of many relevant Models Variable time scales are OK
Objective: Developing forecasting tools for Ecosystem Management in Great Lakes Tributaries
Watershed Stakeholders’
Questions
Managementscenario
evaluations
EcologicalInventory &Assessment
MREMSIntegrated modeling
Muskegon River Ecological Modeling System
2000,2002
2001-2003
2006
2007
2001-2005
Model Predicts TypeLTM 2 Land Use change Neural net
MODFLOW Groundwater flow Simulation
MRI_DARCY Groundwater upwelling GIS
HEC-HMS Surface water flows Simulation
MRI_FDUR Surface water flow frequencies RegressionSystem
HEC-RAS Surface water hydraulics Simulation
GWLF Surface dissolved loads Simulation
MRI_LOADS Surface dissolved loads Regression
Regional Assessment Models
All taxa Sensitive taxa EPT Index Algal Index
Fish/insect diversityFish/insect diversityEPT taxa/ Sensitive fishAlgal Index
RegressionRegressionRegressionRegression
Bioenergetic IB Models Steelhead Salmon Walleye
Growth rateandsurvivorship
Simulation Simulation Simulation
Standing Stock ModelsSport fishesTotal fishesSensitive fishesTotal AlgaeFilter-feedersGrazing inverts
Kg/hec total massKg/hec total massKg/hec total massg/m2
g/m2
g/m2
RegressionSEM1
SEM1
SEM1
SEM1
SEM1
MREMS Components
MREMS Directory Structure
MREMS is really a data sharing protocol and directory structurefor a collection of interacting models
All participating models store spatially referenced output into specific year and landscape scenario directories.
The use of redundant models in MREMS allows us to cross validate results and use weight of evidence
arguments in resource risk assessment.
In this figure, output from 2 very different groundwater models, MRI-DARCY and MODFLOW, are compared. Note correspondence between predicted loading to surface systems (light blue areas on the right with redder areas on the left).
MRI-DARCY MODFLOWsurface loadingrecharge
surface loadingrecharge
Climate
Reach Hydrology
Reach Hydraulics
Local hydraulics and substratum
Fish growth & mortality
Hec_HMSSMA/ MODFLOW
Hec_RAS
Steelhead IBM
hours ~x00 km2
decades ~ x00 km2
weeks ~x000 km2
days ~x km2
days ~x m2
days x cm2
Landscape
HistoricalDaily 1980-2000
LTM2 Neural Net
River Segment ID @ Time Frame yr=2020
t = 1 day
t = 0=fixed per run
Surfacet = 1 hr GW t = 1 day
t = .1 day
t = 1 day
t = 1 day
Climate
Reach Hydrology
Reach Hydraulics
Local hydraulics and substratum
Fish growth & mortality
Landscape
Modeling to forecast Modeling to understand
All modeling output is organized spatially using MRI-VSEC valley segment Map [Valley Segment Ecological Classification Units, Seelbach et al. 1997]
VSEC channel reach units constitute the 2D organization of the model “Ecosystem”
Individual Model output is always organized by VSEC unit
Historicalreconstruction
Air Photointerpretation
1830 1978 2020 2040
Neural Netprojection
Neural Netprojection
Historical data sets augmented by neural net predictions provide a temporal framework
Increasing the hidden layers from 1 to 2 increased model performance significantly.On average, one hidden layer correctly predicted around 50% of the cells to transition;the best 2 hidden layer model predicted 79% correctly. (which reflects a 50% increase in model performance!)
Future Landuse change in MREMSis handled by an enhanced version (LTM2)
of Pijanowski et al.’s Land Transformation Model
Pijanowski, B.C., D. G. Brown, G. Manik and B. Shellito (2002a) Using Artificial Neural Networks and GIS to Forecast Land Use Changes:
A Land Transformation Model. Computers, Environment and Urban Systems. 26, 6:553-575.
grams SRP /day @ Q10
.005 – 6.5 6.5 – 13 13 – 19 19 – 26 26 – 32 32 – 39 39 - 45
HydrologicModel
LoadingModel
1830
1978
2040
VSECFramework
Landcover maps (year)Data Source
Survey notes
Aerial photos
LTM prediction
VSECFramework
1830
1978
2040
SRP Load maps
1830 1978 2040
Database queries Output mapping
Figure 2. An illustration, from the current Muskegon River study, of our method for linking valley scale ecologicalclassification (VSEC) units to landscape models. A: Sample sites are used to represent the entire VSEC unit theyoccur in, based on the mapping objective of ecological homogeneity B: VSEC unit ID # is used to geo-reference andquery the associated catchment, buffers, site databases etc. C: Query results are used as inputs for regional models ofrelevant processes as illustrated here for soluble phosphate load. All segments are processed simultaneously in a matrixmodeling environment. Once modeling is completed predicted results are mapped back into the GIS using the VSECspatial framework. Coupled to changing input data sources on landcover distributions, this process can generate bothforecasts and hindcasts of ecological status.
A. B.
C.
Sample site
Unit catchment
VSEC unitUnit channel buffer
Together the landscape trajectoryand VSEC unitstructure provides MREMSan explicit time x spaceFramework for linking diverse MREMS componentmodels
MWRP Stakeholders’ WorkshopExamples of selected Modeling queries
What is the effect of different rates of urban development? What is the effect of with differing lot size constraints?Effects of Minimum Setbacks for new construction from surface water edge?How and where is channel erosion being affected by development?What is effect of Great Lake water level changes on channel erosion and deposition?How do headwater and main stem dam operations affect ecological integrity?How do wetland losses & urbanization affect river hydrology and fish?
Figure 6 - Modeled hydrographs for Cedar Creek using observed 1998 and LTM projected 2040 landcover scenarios. Precipitation and temperature patterns, and all other variables held constant. Days are arbitrary simulation dates.
MREMS can be used to evaluate effects of alternate land use patterns
1978
2040
1830
Cedar Creek
Table 2 . Example of multiple ecological responses predicted by MREMS in preliminary runs for a “Fast Growth” scenario. Change rates for a 1998 to 2040 time frame comparison. Site hydro
% DD 1 Channel2
Response % SedLoad
3 %TDS4 Fish
spp. loss
Cedar Creek -13 % aggrade +26 % +32% 3-4 Brooks Creek -22 % aggrade +72 % +20% 1-2 Main River @ Evart 0 % No change +1% +20% 2-3 Main River @ Reedsburg 0 % No change +6 % +3% 0-1
1 %DD: Percent change in Dominant Discharge (determines the size of the equilibrium channel); product of HEC_HMS run and empirical load model. 2 Channel response: expected response based on %DD 3 %SL: Percent increase in average daily sediment load [tonnes/day] 4 %TDS: Percent change in median Total Dissolved Solids concentration (ppm)
Mega-Model Runs target the entire watershed and provide a time-dependent context for understanding ourCurrent conditions, identifying risks that lie ahead, and a testing ground for alternate Management
Scenarios.
What will we do with the Model?
•August 24 2002 @ Annis Center•Representatives from 13 stakeholder organizations and 9 of the project PIs met •To develop scenarios to be evaluated with theMuskegon Watershed Mega model
•Goal was to develop 3-4 scenarios in each of 3 areas (land use, hydrology, sedimentation/ersosion)
•Land use management scenarios (12)•Hydrologic management scenarios (13)•Sediment management scenarios (12)
Muskegon Watershed Research PartnershipModeling Endpoints Workshop
What’s next on the MREMS What’s next on the MREMS agenda?agenda?
• Lower river hydrology and fisheries models Lower river hydrology and fisheries models completed by end of 2005completed by end of 2005
• Stake-holder scenario modeling completed Stake-holder scenario modeling completed by summer 2006by summer 2006
• Final report out to Stakeholders winter Final report out to Stakeholders winter 20072007
X
X