Scientific Uncertainty in New England Groundfish Assessments and - - PowerPoint PPT Presentation

Scientific Uncertainty in New England Groundfish Assessments and its Impact of Successful Catch Advice

John Wiedenmann & Olaf Jensen Rutgers University

Objectives

• 4 Phases to the overall project

– Phase 1 & 2 aimed at identifying how catch advice performed and sources of error – Phase 3, & 4 seek to test alternative measures and how they would have performed

What role has scientific uncertainty played in the lack of recovery of many groundfish stocks?

Low Scientific Uncertainty High Scientific Uncertainty Scientific and Management (Implementation) Uncertainty

Data Sources

• Stock assessments

– Catch time series, estimates of F, N, SSB, etc. (most recent assessment) – Historical estimates of estimated F, R, SSB, rho, from all assessments (GARM 1 – present)

• Amendments and Frameworks

– Target catches (TACs or ABCs) – Target F

• Projection files

– Projection assumptions – Uncertainty in estimates – Target F (sometimes)

• For most stocks in most years, catches have

been below the target, yet fishing mortality rates have been well above the target

• What is causing this? Potential factors include:

– Overestimation of terminal abundance – Below expected recruitment – Improper catch / F assumptions in the projection bridge years – Changing weights / selectivity at age

Biomass and Recruitment Estimates Across Repeated Stock Assessments

Relative Error in Biomass Estimates RE = (Historical - Updated)/Updated

Positive values mean historical estimates were higher Negative values mean Historical estimates were lower Red line used to mark 0

Mean = Median = Mean = Median = Relative Error Aggregated Across Stocks, Using Only the Terminal Estimates

RE in Terminal Stock Biomass Recruitment GB Cod 0.55 0.62 GOM Cod 1.16 0.83 GB Haddock 0.06 0.26 GOM Haddock

• 0.18

0.84 GB Yellowtail 2.38 0.94 SNE/MA Yellowtail 0.34

• 0.56

CC/GOM Yellowtail 0.82 0.06 Plaice 0.64 0.10 Witch 1.66 1.84 GB Winter 0.25 0.50 SNE/MA Winter

• 0.24

0.03 Redfish

• 0.02

0.00 White Hake 0.45 0.20 Pollock 0.29 0.66

Can Assessment Diagnostics Predict Accuracy?

• Do we see greater error in terminal estimates from

assessments that had a large retrospective pattern(Mohn’s ρηο) or larger C.V. in terminal estimates?

NO: Retrospective Error and the Terminal C.V. are Poor Predictors of Estimation Error

Does Error Increase Over Time?

Data-management lag = years between catch target and terminal year in assessment e.g., catches in 2005 based on the 2001 GARM 1 estimate = 4 year lag

Recruitment Since 2004 Has Been Below Historical Levels for Most Stocks

• Update historical projections with information

from the most recent assessment

– Analogous to the work by Liz Brooks & Chris Legault, but using more recent information – GOM cod, GB cod, GB yellowtail flounder, Witch flounder, SNE / MA winter flounder, and pollock

GARM 1 GARM 2 GARM 3 Updated

SSB Recruitment Target Catch

GOM Cod

SNE / MA Winter Flounder

GARM 1 GARM 2 GARM 3 Updated

SSB Recruitment Target Catch

Relative error in projected catch Assessment Updated Updated Updated Updated Updated Stock Basis Years Base N F R Wc s GARM 1 2004-2005 0.70 0.95 0.35 0.48 0.24 0.66 GB Cod GARM 2 2006-2009 3.00 0.60 2.01 2.69 2.60 3.04 GARM 3 2010-2011 1.03 0.40 0.46 0.92 0.91 1.13 GARM 1 2004-2005 1.12 0.53 0.71 1.03 0.85 1.11 GOM Cod GARM 2 2006-2009 1.95 0.38 1.87 1.78 1.60 1.70 GARM 3 2010-2013 2.40 0.47 1.76 2.18 1.97 2.75 GARM 1 2004-2005 6.96 0.57 4.70 6.96 6.31 6.29 Witch GARM 2 2006-2009 4.36 0.34 2.28 4.36 4.26 4.14 GARM 3 2010 2.24 0.43 1.66 2.22 2.20 1.87 GARM 1 2004-2005 0.55 0.36 0.46 0.21 0.61 0.40 SNE/MA GARM 2 2006-2009 0.42 0.44 0.29 0.23 0.34 0.34 Winter GARM 3 2010-2012

• 0.08

0.59

• 0.31
• 0.10
• 0.17
• 0.10

GARM 1 2004-2005* 6.10 1.53 4.08 5.15 5.68 5.82 GB GARM 2 2006-2009*

• Yellowtail

GARM 3 2010-2012* 6.89 0.79 5.02 5.53 5.55 6.55 Pollock SAW 50 2011-2014 0.07

• 0.04

0.12 0.07 0.02 0.12

R2 = 0.55; p<0.001 Relative Error in Terminal Biomass Estimates Relative Error in Achieving Target F

Conclusions

• Catch advice has been over-estimated for the majority
• f groundfish stocks
• Overestimation of terminal abundance greatly

contributed, but declining recruitment also played a role

• Current work focusing on how alternative methods

would have performed

– Control rules – Modified projection inputs – Gradual changes to target catch (e.g. + / - 20%)

• john.wiedenmann@gmail.com

Assumed Recruitment Model in the Projections