An agent-based model of fisher behavior and fish ... · • Reproduce fish population dynamics and fishing fleet behavioral dynamics – allow them to interact across space and time

An agent-based model of fisher behavior and fish

population dynamics in the Gulf of Mexico How does fisher behavior affect stock assessment?

Steve Saul, Ph.D.

Assistant Professor

ASU

2

Objectives

• Uncertainty in fisheries and human behavior. • West Florida Shelf agent-based simulation

model configuration and parameterization • Model results. • Stock assessment of simulation model results

and comparison with “known” simulation system dynamics.

• Results, discussion, and implications.

• Future research.

3

Fisher Behavior: Key Source of

Uncertainty in Fisheries

• Resource users sometimes respond to policies in an unintended way.

• Though uncertainty is widely accepted as a problem in fisheries management, attention has only focused on scientific uncertainty around the status of exploited resources.

• Uncertainty due to human responses to management has received much less attention.

• Human behavior dictates the spatial and temporal locations of fishery-dependent observations, often used to infer abundance trends in assessment models.

(See Fulton, et al. 2011 for review)

4

Catch per Unit Effort

Catch per unit effort (CPUE) is used as fishery dependent observations to determine a relative measure of population abundance if the following relation holds:

i

i

ii Nq

f

CU

Where U is the catch per unit of fishing effort, f and q is the catchability coefficient, the proportion of the population captured by a unit of fishing effort

5

Sources of Variation in CPUE

Characteristics of animal population

• Density (=abundance) • Spatial & temporal differences

• Behavior of animals • Aggregation behavior • Competition for gear space or bait • Physiological condition

Fishing power of fishing unit

• Gear characteristics • Vessel characteristics • Fishermen’s skills

CPUE is a stochastic process

• Environmental conditions

• Affecting gear performance • Affecting animal behavior

Economic/management drivers

• Regulated catches • Reporting accuracy • Market value of target species

6

Ability to use CPUE as index of abundance depends on being able to adjust for (i.e. remove) the impact on catch rates of changes over time from factors other than stock abundance

CPUE Standardization

Relative

abundance

• Spatially • Temporally

CPUE Observations

Other sources

of variation

7

Hypothesis Testing Through Simulation

• Reproduce fish population dynamics and fishing fleet behavioral dynamics – allow them to interact across space and time under realistic state conditions.

• Compare metrics derived from simulated fishing fleet observations, with the biological dynamics simulated in the system.

• Agent-based modeling well suited to simulate complex social and biological dynamics and their interactions.

• Fish and fishing vessels each modeled as individual agents.

8

Process of Study Fishing Fleets and Fish Modeled

9

-88 -87 -86 -85 -84 -83 -82 -81

2425

2627

2829

3031

Longitude

Latit

ude

Simulation Model Spatial Partitions20m Depth Strata

Stratum 1Stratum 2

Stratum 3

Stratum 4

Stratum 5

Stratum 6

Stratum 7Stratum 8

Stratum 9

Stratum 10

1 minute by 1 minute (1.8 km2) Grid

Simulation Model Spatial Partitions

20 Meter Depth Strata Shown 20 Meter Depth Strata Shown

Simulation Spatial Definitions

• Grouped Strata (): 3 groupings of Strata

• Strata (): areas of equal latitude where Florida’s coast runs north-south, areas of equal longitude where the coast runs east-west.

• Depth Sub-strata (): 20m isobaths

• Grids (): cells one minute latitude by one minute longitude

10

-84.0 -83.8 -83.6 -83.4 -83.2 -83.0

27.0

27.2

27.4

27.6

27.8

28.0

Red Grouper

Distributed With Spatial Autocorrelation

Spatial Distribution of Abundance

012345678

20 40 60 80 100

Sta

nd

ard

ize

d C

PU

E

Depth Strata (Meters)

c) Red Grouper

Grids 1 and 2 Grids 3 - 7 Grids 8 -10Strata Strata Strata

0.0 1.0 2.0 3.0

0.0

0.2

0.4

Distance (km)

Sem

ivar

ianc

e

a) Gag Grouper

0.0 1.0 2.0 3.0

0.0

0.2

0.4

Distance (km)

Sem

ivar

ianc

e

b) Mutton Snapper

0.0 1.0 2.0 3.0

0.0

0.2

0.4

Distance (km)

Sem

ivar

ianc

e

c) Red Grouper

(Saul et al. 2013. Modeling the spatial distribution of

commercially important reef fishes on the West Florida

Shelf. Fisheries Research 143: 12-20)

11

Validation

0 1 2 3 4 5

2000

4000

6000

8000

Range (km)

Sem

ivar

ianc

e Area 1Area 2Area 3Area 4Area 5Area 6Area 7Area 8Area 9Area 10Area 11

Variogram Fits to 11 Areas

12

Biased Random Walk Migration

Algorithm

Mortality

• Develop pathway to simulate movement of an individual fish to offshore habitat.

• Important part of reef fish life history.

• Fishing effort removes migrating individuals in transit to offshore habitats.

• Establish an age structured population that is properly distributed across space and time.

(Saul et al. 2013. An Individual-Based Model of

Ontogenetic Migration in Reef Fish Using a Biased

Random Walk. Trans. Am. Fish. Soc. 141: 1439-1452)

13

Biased Random Walk Migration

Algorithm

14

Tagging Simulation Results

Empirical Linear Speed

Grid Cells/Day

Freq

uenc

y

0.0 1.0 2.0 3.0

010

030

0

α = 0.7β = 0.2

Simulated Linear Speed

Grid Cells/Day

Freq

uenc

y

0.0 1.0 2.0 3.00

500

1500

α = 0.9β = 0.1

(Wilcoxon rank sum test: P=0.3764)

Red Grouper Example Results

15

Tagging Simulation Results

-85.0 -84.0 -83.0 -82.0

26.0

27.0

28.0

29.0

Longitude

Latit

ude

Empirical Migrating Fish

Red Grouper

-85.0 -84.0 -83.0 -82.0

26.0

27.0

28.0

29.0

LongitudeLa

titud

e

Simulated Migrating Fish

Red Grouper

Comparison of 350 randomly selected empirical and simulated red grouper recaptures from off the coast of

Mote Marine Lab (location of half the releases)

16

Tagging Validation

1 2 3 4 5 6

0.0

0.2

0.4

0.6

0.8

1.0

Age (years)

Pro

babi

lity

Mat

ure

a) Maturity at Age b) Simulated Total Migration Time

Migration Time (years)

Freq

uenc

y0 1 2 3 4 5

050

0015

000

1 3 5 7 9 11 14

Sed. AdultsMigratingSed. Juv.

c) Age and Stage of Tagging

Age (years)

Freq

uenc

y0

500

1500

2500

1 3 5 7 9 11 14

Sed. AdultsMigrating

d) Simulated Recapture Age

Age (years)

Freq

uenc

y0

200

600

1000

1 3 5 7 9 11 14

DiscardsCatch

e) Commercial Catch at Age

Age (years)

Per

cent

05

1015

1 3 5 7 9 12 14

f) Simulated Age at Arrival

Age (years)

Per

cent

010

2030

4050

17

Fish Population Parameterization

• Life history, starting age structure, and recruitment parameters taken from most recent stock assessment reports.

• Spatial distribution of fish abundance validated against the spatial distribution of fish habitat characteristics.

• Migration algorithm validated using tagging data.

18

10 20 30 40 50

0.00

0.04

0.08

Wind Speed (knots)

Prob

abilit

y

Regulated Fisheries OpenRegulated Fisheries Closed

Trip Choice HL: Panhandle

Wind SpeedFisher Decision-Making

• Discrete choice models developed for three decisions: participation, site choice, and trip termination.

• Questionnaire completed by vessel captains determined factors to test in model.

• Panel dataset developed to fit models using commercial logbook data, weather, fish price, vessel characteristics, expected revenue, fuel price, and regulations.

-88 -87 -86 -85 -84 -83 -82 -81

2425

2627

2829

3031

Longitude

Latit

ude

-88 -87 -86 -85 -84 -83 -82 -81

2425

2627

2829

3031

Fishing PortsFishing Sites

Fishing Ports and Sites

Gulf of Mexico

(Saul et al. Accepted. Modeling the decision making

behavior of fishers in the reef fish fishery on the West

Coast of Florida. Human Dimensions of Wildlife.)

19

Fisher Decision and CPUE Analysis

20 Year

Projection From 2005/2006

State of Fishery

(pre-IFQ)

When to fish

Where to fish

Return to port

Simulated Logbook Data

20

Model Validation

Source: Strelcheck and Farmer (2012),

Personal Communication,

NOAA Southeast Regional Office

21

CPUE Standardization

22

Catch at Size

Length Distribution of Catch

Length (mm)

Den

sity

0 200 400 600 800

0.00

00.

002

0.00

4Red Grouper Year 20

Length Distribution of Population

Length (mm)

Den

sity

0 200 400 600 800

0.00

00.

001

0.00

20.

003

Red Grouper Year 20

0

0.2

0.4

0.6

0.8

1

0 200 400 600 800

Sele

ctiv

ity

Length (mm)

Red Grouper HL SelectivityUsed in Saul Simulation Model as Derived from SEDAR 9

Selectivity

Size Limit

23

0

10,000

20,000

30,000

40,000

50,000

60,000

70,000

80,000

0 4 8 12 16 20

Met

ric

Ton

s

Simulation Year

Biomass Over Time: Red Grouper

Simulation

0

5

10

15

20

25

30

35

40

45

50

0 4 8 12 16 20

Mill

ion

s o

f Fi

sh

Simulation Year

Numbers Over Time: Red Grouper

Simulation

Findings with Stock Synthesis

24

0

10,000

20,000

30,000

40,000

50,000

60,000

70,000

80,000

0 4 8 12 16 20

Met

ric

Ton

s

Simulation Year


Simulation Base Model

0

5

10

15

20

25

30

35

40

45

50

0 4 8 12 16 20

Mill

ion

s o

f Fi

sh

Simulation Year


Simulation Base Model


25

0

10,000

20,000

30,000

40,000

50,000

60,000

70,000

80,000

0 4 8 12 16 20

Met

ric

Ton

s

Simulation Year


Simulation Base Model Perfect CPUE, init F 1/3 SEDAR

0

5

10

15

20

25

30

35

40

45

50

0 4 8 12 16 20

Mill

ion

s o

f Fi

sh

Simulation Year


Simulation Base Model Perfect CPUE, init F 1/3 SEDAR


26

0

5

10

15

20

25

30

35

40

45

50

0 4 8 12 16 20

Mill

ion

s o

f Fi

sh

Simulation Year


Simulation Base ModelPerfect CPUE Perfect CPUE, est. init. FPerfect CPUE, init F half SEDAR Perfect CPUE, SEDAR init FPerfect CPUE, init F 1/3 SEDAR

0

10,000

20,000

30,000

40,000

50,000

60,000

70,000

80,000

0 4 8 12 16 20

Met

ric

Ton

s

Simulation Year


Simulation Base ModelPerfect CPUE Perfect CPUE, est. init. FPerfect CPUE, init F half SEDAR Perfect CPUE, SEDAR init FPerfect CPUE, init F 1/3 SEDAR


27


0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

0.0 1.0 2.0 3.0 4.0 5.0

Dep

leti

on

(SS

Bcu

rre

nt/

SSB

msy

-pro

xy)

Harvest Rate (Fcurrent/FMSY-Proxy)

Control Rule Plot: Simulated Red GrouperMSYproxy = SPR30%

Base Model, MSYReferenceBase Model, SPRReferencePerfect Index, MSYReferencePerfect Index, SPRReferenceSEDAR 12

SEDAR 12 Update

SEDAR 42

OFL

MSST

28


0

5

10

15

20

25

30

0 5 10 15 20 25 30

Mill

ion

s o

f P

ou

nd

s

Simuilation Year

Red Grouper MSY Based Projections

SS Yield Base Model - OFLBase Model - OY Base Model - FcurrentPerfect CPUE - OFL Perfect CPUE - OYPerfect Index - Fcurrent

29

0

5

10

15

20

25

30

0 5 10 15 20 25 30

Mill

ion

s o

f P

ou

nd

s

Simuilation Year

Red Grouper SPR30% (MSY Proxy) Based Projections

SS Yield Base Model - OFLBase Model - OY Base Model - FcurrentPerfect CPUE - OFL Perfect CPUE - OYPerfect CPUE - Fcurrent


30

• Direction and magnitude of CPUE index matters and drives biomass trend estimation in stock assessment models.

• An index that perfectly fits biomass produces results that estimates nearly identical biomass trends.

• Accurately anchoring fleet fishing mortality at the beginning of the time series in stock assessment models is critical to obtaining properly scaled results.

Results and Implications

31

• Historical data such as landings can be useful in tuning the early phases of a fishery.

• Runs with higher fishing mortality at the start underestimated spawning stock biomass, and therefore estimated less biomass and numbers of fish, and vice versa.

Results and Implications

U.S. Red Grouper Fleet Historical Catch

012345678

1880

1885

1890

1895

1900

1905

1910

1915

1920

1925

1930

1935

1940

1945

1950

1955

1960

Year

Millio

ns

of

Po

un

ds

West Florida Shelf Campeche Bank Other Gulf of Mexico Locations

32

Spatial Distributions of Fishing Effort

and Biomass

33

Results

• Fishing effort spatially and temporally affected by weather: less fishery dependent observations during times and at locations of high wind.

• Trip duration sometimes limited by hold size (filled vessel to capacity) which suggests effort saturation effect on CPUE.

• Spatial and temporal fishing effort distributions driven by closures indicate stock level catchability cannot be constant in space and time.

• Local depletion - spatial distribution of fishing effort and fish populations not uniform. —People tend to have their fishing sites that they know

work and only exploratory fish a fraction of the time if at all (survey responses suggest 25%).

34

Fisher Behavior and Oil Spill

35

Where did the fishing effort go?

36

Full Model

Evaluate socioeconomic and fisheries assessment and

management affects

Source: The Sea Around Us Project, University of British Columbia

37

Acknowledgements

This research was made possible in part by a grant from The Gulf of Mexico Research Initiative, and in part by CIMAS at the University of Miami, NOAA Cooperative agreement NA17RJ1226, the NOAA/Sea Grant Population Dynamics Fellowship Grant Award NA080AR4170763, and an International Light Tackle Tournament Association Scholarship. Data for this project will be made publicly available through the Gulf of Mexico Research Initiative Information & Data Cooperative (GRIIDC) at https://data.gulfresearchinitiative.org.

38

“Managing fish is managing people”

A.P. Bell and Starfish Market Abrahm’s Seafood Ariel Seafood Buddy Gandy Seafood Cox Seafood Fish Busterz (Madeira Beach) Holiday Seafood Jenson Tuna Little Manatee Fish House Madeira Beach Seafood Sammy’s Seafood Save On Seafood Water Street Seafood

Glen Brooks David Krebs Jason de la Cruz Bobby Spathe

-Ray Hilborn, 2007

Gulf Fisherman’s Association Southern Offshore Fishing Assn. Reef Fish Shareholder’s Alliance

39

Documents

An agent-based model of fisher behavior and fish ... · • Reproduce fish population dynamics and fishing fleet behavioral dynamics – allow them to interact across space and time