Transcript prediction_GOAmodel_Gaichas

GOA Retrospective analysis
Model use: hypothesis testing
• The system, the stories, and the “data”
• The model: Elseas; like Ecosim but more
flexible for our purposes
• The simple and clear hypotheses: what
drives species trends in the GOA?
–
–
–
–
It’s
It’s
It’s
It’s
fishing
climate (the PDO)
everyone eating shrimp
complicated…
Walleye pollock, Theragra chalcogramma
Adult
diet
shrimp
euphausiids
4,000,000
stock assessment
trawl survey
Juvenile diet
2,000,000
copepods
1,000,000
euphausiids
2000
year
1990
1980
1970
0
1960
biomass (t)
3,000,000
Pacific cod, Gadus macrocephalus
Adult diet
pollock
stock assessment
800,000
shrimp
trawl survey
Juvenile diet
600,000
400,000
shrimp
200,000
benthic amphipods
year
2000
1990
1980
1970
0
1960
biomass (t)
bairdi
Pacific halibut, Hippoglossus stenolepis
Adult diet
pollock
stock assessment
800,000
trawl survey
600,000
400,000
shrimp
200,000
year
2000
1990
1980
1970
0
1960
biomass (t)
Juvenile diet
hermit
crabs
Arrowtooth flounder, Atherestes stomias
Adult diet
pollock
2,000,000
capelin
stock assessment
trawl survey
1,500,000
1,000,000
capelin
500,000
euphausiids
year
2000
1990
1980
1970
0
1960
biomass (t)
Juvenile diet
4,000,000
Pollock
P. cod
Arrowtooth
Halibut
1990-1993
snapshot
2,000,000
1,000,000
year
2000
1990
1980
1970
0
1960
biomass (t)
3,000,000
Mass balance to dynamic simulation
Bioenergetics
and mass accounting
Population rates
(total mortality is
key)
M2
GE
M0
B
P/B
Q/B
DC
EE
Catch
BA
q
Vul
(Bstart)
Equilibrium built here, perturbed here
Alternate stable states possible??
Modeling
Recruitment
– A delay-difference equation with juveniles divided into
monthly pools:
• Fixed age at recruitment
• Adjustable relationship between food intake and fecundity
– Knife edge recruitment to fishery, spawning, and
ontogenetic diet switch.
– Spawning biomass is not directly comparable to stock
assessments (because stock assessments vary).
Model structure: alternative myths
Predator Production/Biomass
• “Surplus” production (compensation) is an absolute
requirement for sustainable single-species fishing. As
biomass decreases, production per biomass must increase.
• This can happen in more than one way…
1
P/B (Age-structured von Bertalanffy)
P/B (Ecosim foraging risk)
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
1
2
3
Predator Biomass
4
5
6
Alternative myths II
Predator growth efficiency
• In von Bertalanffy (vB) models (MSVPA, single species), fishing
compensation comes from increasing growth rates (conversion
efficiency) of relatively younger fish in a fished population.
• In Ecosim, compensation comes from increased per-capita
consumption: all at the expense of other species.
0.6
Growth efficiency (Age-structured von Bertalanffy)
Growth efficiency (Ecosim foraging risk)
0.5
0.4
0.3
0.2
0.1
0
0
1
2
3
4
5
6
Predator Biomass
• By definition, in Ecosim there is no true energetic “surplus,” it all
comes from other species. Conversely, in vB models there is no
“bottom up control.”
Forcing alone (no fitting
of Ecosim parameters)
in the Northern
California Current
(Field 2004):
This run is forced by NPZ
output time series (19671998) and fishing mortality
derived from catches and
stock assessments
Forcing: fishing only (big 4)
Forcing (Fishery)
Fit to Catch
Fit to Biomass
Fitting with fishing only, but add 60’s POP fishery
Forcing (Fishery)
Fit to
Catch
Fit to
Biomass
Fitting using fishing only—all GOA time series
Fitting using fishing, pollock recruitment—all series
Fitting using fishing and all recruitment—all series
Summary…
• Can’t explain system dynamics (species trends)
– with fishing alone (unlike in other, more heavily fished
systems)
– with simple climate (PDO) forcing of primary
production
• Reproducing “known” groundfish dynamics
– OK when forcing with stock assessment “data”
• Recruitment variability dominates this system?
Predictive Process: predict, communicate, use
• Between Prediction and Use
– What ought to be predicted?
– How are predictions actually used?
• Between Prediction and Communication
– What does the prediction mean in operational terms?
– How reliable is the prediction, and how is uncertainty
conveyed?
• Between Use and Communication
– What information is needed by the decision maker?
– What content or form of communication leads to the
desired response?
Predictive Potential
• Single Species Stock Assessment Model
– Unknown parameters fit using data, updated annually
– Predict direct effects of fishing on target populations
– Quantitative prediction, 1-2 years out
• Ecosystem Model
– Predict direct effects of fishing on nontarget species
– Predict indirect effects of fishing mediated by trophic
interactions
– Predict consequences of ecosystem changes not
related to fishing, therefore beyond our control
– Qualitative predictions, must incorporate uncertainty
Data requirements in a simple food web
Biomass (B)
Population growth rate or
Production (P/B)
Consumption (Q/B)
Diet comp (DC)
For ALL groups!!
Alternative: solve for B
assuming a fixed
proportion of production
is used in the system:
“top down balance”
Too complex—uncertainty overwhelms?
• Each systematically added group adds constraints as well as
data requirements, does one outweigh the other?
GOA data pedigree
Base arrowtooth trajectory
baseGOA_1000 Arrowtooth_Adu
25
tons/km^2
20
15
10
5
0
2004
2014
2024
year
2034
2044
2054
1000%
-200%
Transient kille
Sleeper sharks
Salmon sharks
Toothed whales
Stellar sea lio
Other rockfish
Skates
Seals
Pacific halibut
Macrouridae
Arrowtooth fl.
Pisc. birds
Pacific cod
Spiny dogfish
Sculpins
Cephalopods
Baleen whales
Sablefish
Shortraker/roug
Thornyheads
pollock (all)
other demersal
POP/northern/du
Jellyfish
SMALL
Pacific herring
Salmon
osmeridae
Hexagrammidae
Zoarcidae
sandlance
bathypelagics
C. bairdi
C. opilio
King crab
Shrimp
EPIFAUNA
LARGE ZOOP
INFAUNA
Benthic Amph.
Copepods
Ocean productio
Phytoplankton
Results: “Base trophic uncertainty”
• Bars show 95% confidence interval for year-50 biomasses in accepted
ecosystems; symbols show varied assumptions of functional responses
• Limited confidence of exactly where system will be in 50 years, but
patterns do emerge...
baseStd
med
800%
600%
400%
200%
0%
Predicting trophically mediated fishing
effects (and level of control in a system):
Try to fish out arrowtooth?
• What effect would a “magic” arrowtooth
reduction have?
• What might a real increase in targeting of
arrowtooth look like?
• Different tradeoffs…
Fish out arrowtooth “magically”
magicATF_FisM_1000 Arrowtooth_Adu
12
10
tons/km^2
8
6
4
2
0
2004
2014
2024
year
2034
2044
2054
Scenario difference from base
magicATF_FisM_1000lessBase Arrowtooth_Adu
0
2004
2014
2024
2034
Percent change
-20
-40
-60
-80
-100
-120
year
2044
2054
-1
Pandalidae
Shortspine Thorns_Juv
Fish Larvae
NP shrimp
Dusky Rock
N. Fur. Seal_Juv
Mysid
N. Fur. Seal_Adu
W. Pollock_Juv
Offal
Capelin
Pacific Grenadier
Giant Grenadier
Shortraker Rock
Dover Sole
Rougheye Rock
Other Macruids
Scypho Jellies
Prickle squish deep
Sea Star
Central S.S.L._Adu
Other sculpins
Other Sebastes
Central S.S.L._Juv
Herring_Juv
Salmon Sharks
P. Halibut_Adu
Greenlings
P. Cod_Adu
West S.S.L_Adu
West S.S.L_Juv
Eelpouts
Resident seals
Arrowtooth_Juv
Sleeper Sharks
Atka_Adu
Sablefish_Adu
Arrowtooth_Adu
W. Pollock_Adu
Herring_Adu
Fish out arrowtooth “magically”
(F on arrowtooth increases with no bycatch)
5
Median
4
3
2
1
0
-1
Sea Star
4
Bathyraja Aleutica (Aleutian skate)
5
Central S.S.L._Adu
Central S.S.L._Juv
Giant Grenadier
Shortspine Thorns_Juv
Other Macruids
Other Sebastes
Salmon Sharks
P. Halibut_Adu
Herring_Juv
King Crab
P. Cod_Adu
Raja binoculata (Big skate)
Greenlings
Rougheye Rock
Shortraker Rock
Bathyraja interupta (Bering skate)
Prickle squish deep
West S.S.L_Juv
West S.S.L_Adu
Rex Sole
Offal
Raja rhina (Longnosed skate)
Bathyraja maculata (Whiteblotched)
Resident seals
Eelpouts
Arrowtooth_Juv
Shortspine Thorns_Adu
Sleeper Sharks
S. Rock sole
Sablefish_Adu
Misc. Flatfish
Atka_Adu
Dover Sole
N. Rock sole
Arrowtooth_Adu
W. Pollock_Adu
Discards
Herring_Adu
Fish out arrowtooth “realistically”
(increase flatfish fishery q for arrowtooth)
Median
3
2
1
0
Predicting fishing effects on nontarget species
• Can we use knowledge of some system components to
learn about effects of fishing on nontarget species?
• Apply the same method to “small” Gulf of Alaska model…
• Perturbations are new: stop fishing, increase fishing on
all, increase target fishing to MSY levels for major
groundfish
-200%
200%
100%
0%
-400%
Transient kille
Sleeper sharks
Salmon sharks
Toothed whales
Stellar sea lio
Other rockfish
Skates
Seals
Pacific halibut
Macrouridae
Arrowtooth fl.
Pisc. birds
Pacific cod
Spiny dogfish
Sculpins
Cephalopods
Baleen whales
Sablefish
Shortraker/roug
Thornyheads
pollock (all)
other demersal
POP/northern/du
Jellyfish
SMALL
Pacific herring
Salmon
osmeridae
Hexagrammidae
Zoarcidae
sandlance
bathypelagics
C. bairdi
C. opilio
King crab
Shrimp
EPIFAUNA
LARGE ZOOP
INFAUNA
Benthic Amph.
Copepods
Ocean productio
Phytoplankton
1000%
Transient kille
Sleeper sharks
Salmon sharks
Toothed whales
Stellar sea lio
Other rockfish
Skates
Seals
Pacific halibut
Macrouridae
Arrowtooth fl.
Pisc. birds
Pacific cod
Spiny dogfish
Sculpins
Cephalopods
Baleen whales
Sablefish
Shortraker/roug
Thornyheads
pollock (all)
other demersal
POP/northern/du
Jellyfish
SMALL
Pacific herring
Salmon
osmeridae
Hexagrammidae
Zoarcidae
sandlance
bathypelagics
C. bairdi
C. opilio
King crab
Shrimp
EPIFAUNA
LARGE ZOOP
INFAUNA
Benthic Amph.
Copepods
Ocean productio
Phytoplankton
No fishing (top), 2xF (bottom)
noFlessBase
median
800%
600%
400%
200%
0%
2FlessBase
median
-100%
-200%
-300%
-50%
Species affected by a 10% increase in W. Pollock production
Species affected by a 10% increase in W. Pollock production
EBS
Shortspine
Thorns_Juv
Grenadiers
Misc. fish shallow
Greenlings
Shortspine Thorns
FH. Sole
Shortraker Rock
Rougheye Rock
-30%
Offal
-20%
Steller Sea Lion_Juv
-10%
Resident seals
0%
Steller Sea Lion
10%
P. Halibut
20%
W. Pollock_Juv
40%
W. Pollock
30%
Percent change in equilibruim production
50%
P. Cod
Steller Sea
Lion_Juv
Steller Sea Lion
Kamchatka fl._Juv
Herring
Wintering seals
W. Pollock
Kamchatka fl.
Resident Killers
Shortraker Rock
Other skates
Sablefish
Offal
P. Halibut
Alaska skate
Percent change in equilibrium production
Predicting effects beyond our control
• Changes in species or group production
• Evaluate system structure, relative predictability
50%
40%
GOA
30%
20%
10%
0%
-10%
-20%
-30%
-40%
-40%
-50%
Conclusions
• Predictive potential?
– Most powerful when considering uncertainty
– Error bars incorporate both data quality and
predictability
– Direction of change a robust indicator
– The GOA and the EBS may have different levels of
predictive potential—useful information for management
• Implications for policy
– Keep active policy options for changing fishing mortality
– Explore new policy options for preparing for the
unexpected (system change will happen)
Discussion: What controls recruitment variability?
• Ideas:
– The difference between the single species models’
recruitment predictions and the ecosystem model’s
may reflect the effect of predation
– So, these models can measure the proportion of
recruitment variability due to trophic effects
• Next step:
– Fit to series of diet composition to identify prey
switching, quantify mortality due to predation
– Time series of low trophic level production would
help—output from NPZ model as in NCC
Discussion: When does fishing matter?
• Is there a threshold where dynamics switch from
“recruitment dominated” to “fishing dominated”?
– How much fishing, and on whom?
– Is threshold dependent on system characteristics?
• The tradeoff:
– Cross the line, and you can explain dynamics
– Stay below it but live with low predictive power
– Either way you may have less fish!!
• The policy implications…