Fish stock assessment - what is it and how does it work? Dankert W. - - PowerPoint PPT Presentation

Fish stock assessment

• what is it and how does it

work?

Dankert W. Skagen

What do we want to know?

Abundance and exploitation Present

• Present abundance: That is what we can fish on
• Present exploitation: That is how managmenet works

and past

We need the experience to:

• Know how the stock responds to exploitation
• Know about fluctuations in stock productivity

What do we know?

Use the data that we have:

• Catch statistics
• Survey observations

For both we want the tonnes converted to numbers at age = numbers by year class For that, we need samples to tell:

• How many fish is there per ton
• How are these fish distributed on year classes

What can catch data tell us?

numbers by year class The number caught must have been out there!

• Within a year class, sum the numbers caught over the years.

These fish must have been there.

• There must have been more, because
• Some have died from other causes ('Natural mortality')
• There are still some left in the sea

The abundance in a year class in a year in the past is the sum of:

• Numbers caught from the year class later on
• Those lost by natural mortality – which can be added by simple

means.

• Those still present, which we cannot infer from this accounting.

1 2 3 4 5 6 7 8 9 10 100000 1000000 10000000 100000000

Blue whiting 1999 year class

Stock number

Canum Sum canum M included Incl still present Incl still present Incl still present

Age Stock numner (log scale)

The effect of the number still left is mostly

• n the recent

years

Stock number estimate from different sources

More on what catch data tell us

• If there is little fish left (old year class), we know the

stock numbers backwards in time

• Then we also know how fast the year class has been

reduced.

• The rate of disappearance can be expressed as

percent reduction per year or total mortality which is reduction rate relative to abundance. They are equivalent, but mortality is more mathematically tractable.

• The total mortality is the sum of an assumed natural

mortality and the fishing mortality

• The catch data do not tell us how much is left at

present.

Survey data – why and how

What the stock looks like now compared to the past.

• Survey data are usually treated as relative measures
• f abundance.
• Abundance in absolute terms is theoretically possible,

but usually not reliable in practise.

• Compare survey data in the past with catch derived

abundance to calibrate the survey

• Then we can use the recent surveys to tell how much

is left of each year class.

Use of information in brief

Catch data tell us how much fish there must have been in the past, but do not tell what we have at present. Survey data tell what we have now compared to the past That's it!

Is it that simple?

Basically yes, but there are some additional points.

• Use of the survey data to estimate present state
• Models for catch data
• How to fit a model to the data
• Estimating incoming year classes
• The role of data quality

Finding recent stock abundance with survey data

Calibration:

• Survey index at age =

Calibration factor * Stock number

• Find calibration factor (catchability) that gives the

best fit of the survey to the stock numbers.

• Most of the influence on catchability is from the past

Fit each year class to the calibrated survey data:

• Find the present amount that starts a history

closest to the survey. That is how the present is estimated.

If signals are conflicting, the result will be a compromise.

The effect of the number still left is mostly

• n the recent

years 'Wagging the tail'

Fitting to survey data

Stock numbers from catches and natural mortality +contribution from the remaining is fitted to calibrated survey data.

1 2 3 4 5 6 7 8 9 10 100000 1000000 10000000 100000000 1000000000

Blue whiting 1999 year class

Stock number fit to surveys

Best fit Too low Too high Survey 1 Survey 2 Survey 3

Age Stock numner (log scale)

Note that the earliest survey data have little impact now, the stock numbers are bound by the catch data and cannot reach the survey value

Models for catch data

• Either: Use the catch data as they are, count them

backwards and add natural mortality.This is the VPA in the classical sense.

• Or a separable model: Make assumptions about

fishing mortality and numbers left, derive expected catches and find the assumptions that give a best fit to the observed catches. Used for Blue whiting

The assumption is that fishing mortality is separable: It is the product of an age factor (selection) and a year factor.

Each approach has pros and cons, but if the data are good and the seelection stable, the result is largely the same.

Separable model for fishing mortalities

Expected catches are derived from these mortalities and modelled stock numbers

Selection = 0.13 0.16 0.4 0.74 0.88 0.98 0.91 1 1 1 Year factor 1 2 3 4 5 6 7 8 9 10 0.69 2000 0.09 0.11 0.27 0.51 0.6 0.67 0.63 0.69 0.69 0.69 0.61 2001 0.08 0.1 0.24 0.46 0.54 0.6 0.56 0.61 0.61 0.61 0.58 2002 0.07 0.09 0.23 0.43 0.51 0.57 0.53 0.58 0.58 0.58 0.65 2003 0.08 0.1 0.26 0.49 0.57 0.64 0.59 0.65 0.65 0.65 0.72 2004 0.09 0.12 0.29 0.54 0.64 0.71 0.66 0.72 0.72 0.72 0.61 2005 0.08 0.1 0.24 0.45 0.54 0.6 0.56 0.61 0.61 0.61 0.53 2006 0.07 0.08 0.21 0.39 0.46 0.51 0.48 0.53 0.53 0.53 0.56 2007 0.07 0.09 0.22 0.41 0.49 0.55 0.51 0.56 0.56 0.56 0.61 2008 0.08 0.1 0.24 0.45 0.53 0.6 0.55 0.61 0.61 0.61 0.51 2009 0.06 0.08 0.2 0.38 0.45 0.5 0.47 0.51 0.51 0.51

Pros and cons with separable models

Advantages

• More statistically

satisfactory

• Less sensitive to noise

in the catch data

• Can cope with missing

data (to some extent)

• Catches at young age

can inform about recruitment. Disadvantages

• Misleading if selection

changes

• Often large deviations

from a fixed selection at young ages that are caught just

• ccasionally.

Fitting models to data

We have seen some examples of unknowns that we have to assume, which we call model parameters. In a separable model these are:

• Selection at age
• Yearly fishing mortality levels
• Stock number in the last year for each year class
• Catchabilities for the surveys
• Natural mortality (usually just has to be assumed)

We want values of these that leads to a best possible fit to the observations we have. The criteria for model fit are based

• n statistical theory.

There are several variants of criteria and several methods for finding the best parameters. There are limitations to which parmeters can be estimated with the information at hand.

The incoming year classes

Important because they dominate the stock in the coming years. Difficult because the data are sparse. Opportunities:

• Survey data – recruitment surveys
• Catch data with a separable model.

Can change very much next year if new data have a different messsage

1 2 3 4 100000 1000000 10000000

Blue whiting 2007 year class

Stock number fit to surveys and separable model

Best fit Too low Too high Survey 2 Survey 3 Sep model

Age Stock numner (log scale)

Data quality – crucial factors:

Catches:

• Catch statistics
• Sampling for age reading, individual weights and maturities

Typical problems:

• Misreporting
• Unaccounted discards
• Non-random sampling
• Age reading

Surveys:

• Consistency
• Coverage (right place at right time – every year)
• Sampling for age composition – trawl hauls representative

for the stock Typical problems

• Year, weather, vessel, equipment, interpretation, migrating fish,

partial coverage

• Trawling on registrations, depth stratification, different ages

in different areas.

Quality of the data – impact on assessment

Catch data

• Underreporting:

Underestimate of the stock

• Wrong age

distribution:

Wrong estimates of mortality Conflicts with survey data

Survey data

• Random noise. Not

so critical if moderate

• Year effects:

Strong impact for several years. Problem with the blue whiting surveys.

That's it. Any questions?