USE OF MODELS IN MANAGEMENT Generality Realism Precision

An innovative modeling approach for simulating
hypoxia/anoxia in estuarine ecosystems
Mark J. Brush
James N. Kremer
Scott W. Nixon
with contributions from:
John Brawley
Nicole Goebel
Jamie Vaudrey
ASFA SEARCH FOR "ECOSYSTEM MODEL"
ERSEM I
(1995)
Rigler & Peters
(1995)
45
NUMBER
PUBLICATIONS
PUBLICATIONS
OFOF
NUMBER
40
35
30
Also:
Reckhow (1994 & others)
Håkanson (1995, 2004)
Hofmann & Lascara (1998)
Pace (2001)
Duarte et al. (2003)
Fulton et al. (2003)
Kremer
& Nixon
(1978)
25
Baretta &
Ruardij
(1988)
20
15
Steele
(1974)
Riley
(1946, 1947)
Odum
(1983)
ERSEM II
(1997)
10
5
Chesapeake Bay Model
0
1940
1950
1960
1970
1980
1990
Odum
(1994)
2000
USE OF MODELS IN MANAGEMENT
Trade-off between realism & predictability:
Generality
Precision
Loss of
utility at
lowest
complexity?
Realism
Increasing complexity / realism
R. Levins (1966, 1968)
Phytoplankton
Primary
Production
Published Gmax Functions
1971-1998
10.0
8.0
Brush et al.
(2002)
elevated
Eppley
-1
max,,dd-1
Gmax
G
6.0
Eppley
Curve
4.0
2.0
0.0
0
5
10
15
20
o
TEMPERATURE,
TEMPERATURE, CoC
25
30
Duarte et al. (2003)
“The Limits to Models
in Ecology”
Can we find a
middle ground?
Generality
Question:
Can a simplified
eutrophication model be
useful as a heuristic and
management tool?
• Parsimony Principle
• Ockam's Razor
Precision
R. Levins
(1966, 1968)
Empirical
“Stressor-Response”
Models
Realism
Complex,
Mechanistic
Systems
Models
Estuarine Eutrophication Model
Estuarine Eutrophication Model
* Need to accurately model
both states and rates
Phyto
Production
Pelagic
Respiration
Macro
Metabolism
Denitrification
Phytoplankton
Primary
Production
Light x Biomass (“BZI”) Models
Pd = *Chl*Zp*PAR + 
… capped by available nutrients
Cole & Cloern (1987) MEPS v. 36
Brush et al. (2002) MEPS v. 238
Water Column
Respiration
Rd = *e kT*Chl10
Source
Nixon & Oviatt (1973)
Turner (1978)
Nowicki (1983)
Holligan et al. (1984)
Jensen et al. (1990)
Iriarte et al. (1991)
Sampou & Kemp (1994)
Smith & Kemp (1995)
Fourqurean et al. (1997)
Caffrey et al. (1998)
Moncoiffe et al. (2000)
MERL (Brush, unpublished)
Location
PCR = f of:
Bissel Cove, RI
Georgia creek
Potter Pond, RI
English Channel
Roskilde Fjord
North Sea
Chesapeake Bay
Chesapeake Bay
Tomales Bay, CA
San Francisco Bay, CA
Ria de Vigo
MERL mesocosms, RI
T
T
T
Chl
Chl
Chl
T
T
T, Chl
Chl
T, Chl
T, Chl, P, N
Carbon Flux to
Sediments &
Benthic Respiration
Nixon (1981)
Estuaries and Nutrients
The Humana Press
Csed = 0.25*Pd
Rsed = *e kT
Denitrification
Nixon et al.
(1996)
Biogeochemistry 35(1)
DENIT = Nload*f(RT)
Empirical Functions
• Robust, data-driven, & apply across several systems - ideal when
mechanistic formulations are insufficient or poorly constrained.
• Reduce model complexity by integrating multiple processes
(which are often poorly constrained) into simplified, bulk functions.
• Produce output we can measure and test.
• Excellent tools for model validation.
… a hybrid, empirical-mechanistic
approach
Greenwich Bay Eutrophication Model
Greenwich Bay, RI
(Avg Z = 3 m)
Surface Phytoplankton
Lower West
Passage Chl-a
Surface DIN
Bottom O2
Bottom O2 with Forced Maximum Chlorophyll a
original run
max chl
* Need to accurately model
both states and rates
Rate Processes
Mid-Bay: Daily P & R
Mid-Bay: Sediment Carbon
1.5
3
-2
1.0
gCm
2
0.5
1
0
0.0
J F M A M J
J A S O N D
J F M A M J J A S O N D
Lower Bay: Water Column Respiration
Annual Primary Production
4
-1
g C m-2 y-1
3
MERL fcn of T,
Chl, NPP
-2
Observed: 281 – 326
Modeled:
306
g O2 m d
-2
gCm d
-1
4
2
1
In the absence of flux measurements 
model
0
J
F
M
A
M
J
J
A
S
O
N
D
System-Level Validation:
Nutrient Reduction Scenarios
Keller (1988)
Nixon et al. (2001)
Nixon et al. (1996)
Generality
Precision
Empirical Models
A Simplified, Hybrid
Empirical-Mechanistic
Systems Model
R. Levins
(1966, 1968)
Realism
Multiple, parallel modeling
approaches, e.g.:
• Latour, Brush & Bonzek (2003)
• Scavia et al. (2003)
• Borsuk et al. (2002, 2004)
Complex,
Mechanistic
Systems
Models
Oviatt et al.
Models for Hypoxia Applied in
Narragansett Bay
NOAA Coastal Hypoxia
Research Program
Parameter
C:Chl
mBZI0
Chl tavg
ƒNPPSED
wtrclm Rƒ0
wtrclm RƒQ10
R tavg
Values
Chl-a
DIN
DIP
30, 60
± 20%
± 20%
0.15, 0.35
± 20%
± 20%
± 20%
37
14
10
12
11
16
12
12
17
12
12
11
12
O2
System
Py
18
16
10
11
Full 3D
resolution in
ROMS:
Nutrient Reduction Scenarios
Bottom O2, mg/L
14
12
0% watershed N,P
10
8
6
4
2
0
14
12
J
F
M A M
J
J
A
S
O
N
D
0% Narr. Bay N,P
10
8
6
4
2
0
14
12
J
F
M A M
J
J
A
S
O
N
D
0% Narr. Bay N,P
& saturating O2
10
8
6
4
2
0
J
F
M A M
J
J
A
S
O
N
D
PROVIDENCE
RIVER
LOWER
NARRAGANSETT
BAY
Scope for Improvement: Pre-Colonial Inputs
Bottom O2
Nixon (1997)
Estuaries 20(2)
Effect of Macroalgal Decomposition
Bottom O2
Bottom O2
Effect of Macroalgal Decomposition
Resultant O2, mg/L
8
6
4
2
0
0
1E+08 2E+08 3E+08 4E+08
Area, m2
Stochastic Simulation
Kremer (1983)
Surface Chl-a, mg/m3
70
60
50
40
30
20
10
0
J
F
M
A
M J
J
A S
Bottom O2, mg/L
O
N
D
12
10
8
Parameter
6
C:Chl
mBZI0
Chl tavg
ƒNPPSED
wtrclm Rƒ0
wtrclm RƒQ10
Atavg S O
R
4
2
0
J
F
M
A
M
J
J
Values
30, 60
± 20%
± 20%
Bottom
0.15, O
0.35
2
± 20%
± 20%
N D± 20%
Chl-a
DIN
DIP
37
14
10
12
11
16
12
12
17
12
12
11
12
O2
System
Py
18
16
10
11
Acknowledgements
James N. Kremer
Scott W. Nixon
John Brawley
Nicole Goebel
Jamie Vaudrey
Dr. Brush’s wardrobe provided by:
Bay St. Louis Kmart