You are on page 1of 56

10

th
ANNUAL CLEAN RIVERS, CLEAN LAKES CONFERENCE
May 1, 2014

#217803
NOAA Sectoral Applied Research Program Grant
2
Study to Assess Potential Mid-Century Climate Change Effects
NOAA SARP Project Team
Sandra McLellan, Ph.D.: Professor, University of Wisconsin-Milwaukee,
School of Freshwater Sciences

Hector Bravo, Ph.D.: Professor, University of Wisconsin-Milwaukee,
Department of Civil and Environmental Engineering

Daniel Talarczyk, P.E., RLS: Ph.D. Candidate, University of Wisconsin-
Milwaukee, Department of Civil and Environmental Engineering

David Lorenz, Ph.D., Assistant Scientist, University of Wisconsin-Madison,
Center for Climatic Research

Jonathan Butcher, P.H., Ph.D.: Director of Modeling, Tetra Tech

Kevin Kratt: Director, Great Lakes Water Resources Projects, Tetra Tech

Michael Hahn, P.E., P.H.: Chief Environmental Engineer, Southeastern
Wisconsin Regional Planning Commission
Study Report: Climate Change
Risks and Impacts on Urban
Coastal Water Resources in the
Great Lakes, October 29, 2013

http://www.sewrpc.org/SEWRPC
Files/Environment/Rainfall/1-24-
14-SARP-FinalReport.pdf

Applied watershed water quality
models developed under the
2007 MMSD/WDNR/SEWRPC
Water Quality Initiative
SEWRPC Water Quality Model
Used for facility planning

Predicts
Fecal coliforms (FC)
TSS
Nutrients
Cu
BOD
Lake Pathogen Model
Funded by NOAA Oceans
and Human Health Initiative

Shows fate and transport of
bacteria in the nearshore; i.e.,
to beaches and water intakes
Todays Presentation
How climate change effects
were represented in the
models
Possible climate change effects
on instream and Lake Michigan
water quality
Value of modeling in
representing complex
interactions in natural systems
Current limitations on ability to
represent climate change in
models
Areas for future study


Downscaling Climate Data
Begin with Global Climate Model (GCM) "forced" with projected future
concentrations of atmospheric greenhouse gases.

GCM's solve fluid dynamical, chemical, and/or biological equations that
are either derived directly from physical laws (e.g. Newton's law) or
constructed by more empirical means.

Resolution of GCM's quite coarse (100-400km)
(Wikipedia)
Downscaling Climate Data
Downscaling: estimate local-scale (or small-scale) surface weather from
regional-scale (or large-scale) atmospheric variables that are provided by
GCMs.

Types of downscaling:

1. Dynamical downscaling: imbed higher resolution regional climate
model in GCM

1. Statistical downscaling: use observed relationships between regional-
scale and local-scale weather to predict local weather from regional
scale output from GCM
Differences in precipitation at different scales:

Unresolved boundary conditions (i.e. topography and lakes)
Leads to systematic bias in regions of strong topography or in the vicinity of the
Great Lakes.
Unresolved physical processes (i.e. small-scale thunderstorms/clouds)
Source of systematic bias
But significant portion is "noise" (i.e. GCM may simulate a coarse-scale version
of the observations very well, it simply cannot simulate how coarse-scale is
distributed locally)
For local impact studies, nature of local distribution important.

Regional-scale precipitation distributed evenly in space =>
relatively low variance & weak extremes
Regional-scale precipitation concentrated in a few locations =>
relatively high variance & strong extremes

(all else being equal of course)
Downscaling Climate Data
Need to characterize the
signal AND the noise

Signal: y = a
1
x + a
0
Downscaling Climate Data
Need to characterize the
signal AND the noise

Signal: y = a
1
x + a
0

Noise: Normal
distribution with
constant variance
(assumption of least
squares linear
regression) and mean:
a
1
x + a
0


If one neglects the noise,
the variance is
underestimated by a
factor of 1-r
2
, where r is
the correlation of your
fit.

Our statistical downscaling methodology
Predict Probability Density Function (PDF) of precipitation, temperature,
etc. given the regional-scale predicted by GCM.

The PDF is NOT constant but varies in space AND time (daily) conditioned
on state of GCM

Generalize linear least squares regression (i.e. conditional Normal
distribution) to arbitrary distributions (essential for precipitation)

We also characterize and simulate realistic co-variability in space, time
and between variables (i.e. multi-dimensional PDF)

Variables: precipitation, maximum and minimum daily temperature, dew
point temperature (i.e. moisture), vector wind.
Our statistical downscaling methodology
We predict PDF. How do we get a normal time series of values?

1. Draw random numbers from the PDFs to generate a possible realization
of the local scale that is consistent with the regional scale in the GCM.

2. Alternatively, use mean PDFs in present and future to map events in
present climate to their analog under future climate change. In other
words, map n
th
percentile in present to n
th
percentile in future.
Precipitation Example
Temperature Projections (2046-2065)
Milwaukee, WI:

Winter: mean temperature increases by 3.8C (6.9F).
From -4.8 to -1.0C (23.4 to 30.3F)
Summer: mean temperature increases by 2.8C (5.1F).
From 21.0 to 23.9C (69.8 to 74.9F)

Future Analog:
Precipitation Projections (2046-2065)
Uncertainty in Model Projections
Spread in model projections across 13 climate models:

Uncertainty in Model Projections
Spread in model projections across 13 climate models:

Selection of Climate Scenarios
Interested in Changes in Larger Precipitation depths associated with
combined and sanitary sewer overflows

For precipitation, the models are most consistent in Spring, therefore we
focus on the effect of changes in larger precipitation events in Spring.

Model Selection Metric: Change in probability of precipitation greater
than 3.0 cm (1.2 inches) in February-May. Use the 10
th
and 90
th
percentile.
Watershed Models
Models originally developed for the
Water Quality Initiative (SEWRPC
2007 Regional Water Quality
Management Plan Update and MMSD
2020 Facilities Plan)
Comprehensive models developed
based on best available data
Rigorous calibration and validation
and independent review by a
modeling committee
Hourly output available for 14
parameters at 682 modeling
subwatersheds for a 10 year period
(10 billion+ data points)
Modeling Processes
Watershed Models (continued)
Comprehensive modeling
system allowed for a
regional watershed
perspective to evaluate
facility improvements and
water quality management
Key pollutant sources
Attainment of water quality
standards
Response to management
activities
1975
Rural-
Agricultural
Runoff
21%
CSO's
49%
Urban-Non-
Agricultural
Runoff
23%
WWTP
5%
SSO's
2%
2000
Rural-
Agricultural
Runoff
21%
CSO's
7%
Urban-Non-
Agricultural
Runoff
68%
WWTP
2%
SSO's
2%
Greater Milwaukee
Watersheds Fecal
Coliform Loadings
Industrial
Discharge
0%
Industrial
Discharge
0%
Estimated Pollutant Reduction over 25-Year Period About 50 Percent
CONCLUSION: Focus on abating stormwater runoff pollution
Climate Scenarios
2020 population and land use
Baseline Conditions
1988 through 1997 climate data
Future Conditions
Selected based on 3.0 cm (1.2
inches) spring rainfall thresholds
associated with CSO and SSO
events
Best case had the least events
of 3.0 cm or greater
Worst case had the most
events of 3.0 cm or greater

Climate Variable
Baseline
(1988
1997)
Future
Best
Future -
Worst
Precipitation (in/yr) 32.5 33.2 33.4
Average Temperature (F) 47.7 53.3 56.4
Potential Evapo-
transpiration (in/yr)
30.4 37.5 42.1
Modeling Results
Significant decreases in
annual flow are predicted
Most annual pollutant
loads also predicted to
decrease
Results for sediment vary
Increased frequency of large
spring rainfall events results
in more channel erosion
which in some cases offsets
reduced upland loading


Modeling Results (continued)
Modeling Results (continued)
Predicted future changes in
annual mean and median
concentrations of pollutants
are small
Both best case and worst
case scenarios can result in
slight improvement or slight
degradation
Result depends on the
balance between changes in
load and flow


Effects of Stomatal Closure
Important effect of CO
2

fertilization is increased
stomatal closure
Plants do not need to transpire
as much water to obtain the
CO
2
they need for growth
Can potentially counterbalance
predicted increases in
temperature and potential
evapotranspiration

Effects of Stomatal Closure (continued)
Refined the Menomonee River
model to account for effects of
stomatal closure
Adjusted model parameter that
affects monthly plant
transpiration
Results indicate a small
increase in total flows
Total future flows remain less
than under baseline
+2.6%
+2.8%
Effects of Stomatal Closure (continued)
Flow or concentration changes relative to initially-modeled climate
change conditions without CO2 adjustment:
Average annual flow:
Fecal Coliform Bacteria:
Dissolved Oxygen: Essentially unchanged
Total Phosphorus:
Total Nitrogen:
Total Suspended Solids:
Total Copper: Unchanged
But the general conclusions regarding the direction of change between
current and estimated future climate conditions was the same (e.g., an
initially-modeled decrease in flow or concentration remained a decrease
with CO
2
adjustment)


Development of hydrodynamic and transport
model
Field data and model validation
Relation between tributary flows and bacteria
concentration
Analysis of climate change effects
Lake Michigan Modeling Component
Nested model
domain
Station Map
Hydrodynamic and transport model
The POM-based hydrodynamic model was expanded to
include a bacteria transport module.
Bacteria transport module simulates the processes of
advection, dispersion or mixing, bacteria fall through the
water column, light-dependent inactivation rate, and
base mortality.
The model is online at:
http://e320-lx01.ceas.uwm.edu/index.html


Measured (blue) and
modeled (red) specific
conductivity at
stations GC, SG and HB
for the 4/25
5/25/2008 period.
Field data and model validation
Field data and model validation

Measured (open circles)
and modeled (continuous
lines) fecal coliform
(CFU/100mL) at stations
MG, SG, NG and HB in June
and July 2008.
Relation between tributary flows and
bacteria concentration
Simultaneously-measured hourly streamflow (dashed line) and fecal
coliform concentration (continuous line) at the Milwaukee River
mouth between June 2009 and October 2011.
Measured (continuous line) and estimated (dotted line)
logarithm of fecal coliform concentration.
Important scientific questions :
1) the representation of physically correct climate change
scenarios to study the impacts on tributary flows and
bacteria loads, circulation and transport in Lake Michigan
coastal waters,
2) the selection of simulations periods, and
3) addressing uncertainty in climate change predictions.
Analysis of climate change effects
Climate change scenarios were developed using the
arguments that bacteria loads to Lake Michigan are most
sensitive to the spring season, and transport in coastal
waters is most sensitive to changes in wind speed and
direction.

Uncertainty in climate change predictions was dealt with
by using the climate projections that yielded the 10th and
90th percentile changes in spring-season wind speed at
the Milwaukee Airport station to define the worst-case and
best-case climate change scenarios, respectively.
Analysis Assumptions
Location of 11 ASOS stations and NBDC buoys
March-May average wind speed for station KMKE, for the baseline
period projected to 2046-2065 climate conditions by 13 models.
Baseline scenario Climate change scenario
March-May 2005 Worst case: model
cccma_cgcm3_1 projection for
2005 yielded second highest
(approximately 10th percentile)
March-May average change in
wind speed for station KMKE
March-May 2011 Best case: model
mri_cgcm2_3_2a projection for
2011 yielded second lowest
(approximately 90
th
percentile)
March-May average change in
wind speed for station KMKE

The whole-lake model and the nested model were run for 1990 using
concurrent meteorological forcing over the watershed and the lake,
and both the baseline and projected watershed loads estimated by
SEWRPC/ Tetra Tech.
No climate-change projection for meteorological forcing over the
lake could be developed for that year. The model results illustrate
the range of fecal coliform concentration that can exist at relevant
locations near Milwaukee.
The transport of baseline and projected fecal coliform at relevant
sites showed negligible effect of using baseline or projected loads for
the same lake hydrodynamics.

Effect of climate change on watershed loads under the same
lake hydrodynamics
Calculated fecal coliform concentrations (CFU/100 mL) during
March-May 1990 at the Milwaukee River mouth (left), sites MG, SG
and NG right). a) Baseline loads and b) Projected loads.
Baseline loads
Projected loads
The model was used to predict hydrodynamic conditions
and fecal coliform concentrations for the baseline and
projected worst case (2005) and best case (2011) climate
change conditions.
Effect of climate change on
hydrodynamics and bacteria transport
Predicted number of hours with fecal coliform concentration larger than 1,000
CFU/100 mL at relevant locations, for 2005 baseline and worst-case scenario.
Worst-Case Scenario
Station Baseline
condition
Worst-case
scenario
Main Gap (MG) 121 201
North Gap (NG) 156 223
South Gap (SG) 86 43
South Shore
Beach (SSB)
129 58
Bradford Beach
(BB)
35 74
Linnwood
Intake (LI)
0 0
Howard Avenue
Intake (HA)
0 0
Predicted number of hours with fecal coliform concentration larger than
1,000 CFU/100 mL at relevant locations, for 2011 baseline and best-case
scenario.
Best-Case Scenario
Station Baseline
condition
Best-case
scenario
Main Gap (MG) 334 321
North Gap (NG) 111 98
South Gap (SG) 206 227
South Shore
Beach (SSB)
164 142
Bradford Beach
(BB)
0 0
Linnwood
Intake (LI)
0 0
Howard Avenue
Intake (HA)
0 3
The changes in fecal coliform transport are explained by
changes in current vector fields (time average, at each
model cell, of the difference between projected current
vectors minus baseline current vectors) under climate
change conditions.
Model-predicted currents for baseline and worst-case
(best-case) scenario showed that the change in average
currents is mostly northward (southward), so the
predictions indicate more days with concentration higher
than the threshold at locations north (south) of the mouth
of the Milwaukee River.
Conclusions
SEWRPC Water Quality Model
Many uses:

Facility planning
Predict water quality with land use
TMDL development
Mapping flood plains
Lake Pathogen Model
Deterministic model:

Pathogen delivery to beaches
and water intakes

Examine impacts of CSOs
and stormwater
Lake Pathogen Model
Deterministic model:

Pathogen delivery to beaches
and water intakes

Examine impacts of CSOs
and stormwater
Challenges in this effort
Climate models have uncertainty: which model to use?
worst and best case scenario adopted

Temperature and runoff intricately linked: how do we
account variables not specifically addressed in climate
projections?
plant response modeled

Time series for each effort not continuous, needed
additional variables like wind direction
leverage other modeling efforts
Summary

Our ability to downscale GCMs has increased with more
variables. Temperature, precipitation, dew point
temperature (i.e. moisture) and wind vector. There is a
proposal to downscale the new CMIP5

Increased rain does not necessarily equal increased FC
loads. Continued research needed to evaluate impacts of
C0
2
concentrations on plants and the water cycle

Wind is a major driver of lake currents; while loads may
change, where it is distributed may also change.
What will we do with this information

Create more sophisticated tools to include a climate
component in planning (water resource managers)

Evaluate hypothetical scenarios (risk vs. costs)

Increase our understanding of drivers of water quality

Set the bar: Our region is ahead of the curve for
incorporating climate change predictions into planning


Milwaukee Working group
Wisconsin Initiative on Climate Change Impacts (WICCI)
is a consortium of scientists, natural resource managers
and stakeholders that look at adaptation strategies

Milwaukee as an urban area has unique challenges due
to infrastructure, population density and location on
Lake Michigan
Acknowledgements
NOAA Sectoral Applied Research Program
MMSD for initial funding of SSO/CSO project

Collaborations and contributors:
Ron Printz (SEWRPC)
Joe Boxhorn (SEWRPC)
Elizabeth Sauer (GLWI)
Deb Dila (GLWI)

You might also like