Transcript Slide 1

Data collection for stock assessment
and fishery management
FMSP Stock Assessment Tools
Training Workshop
Bangladesh
19th - 25th September 2005
Purpose of talk
To present the data needs of the different FMSP stock assessment tools,
and give some guidelines on data collection for stock assessment
Limited coverage in Section 3.2 of FAO Fish. Tech. Pap. 487
Other references provided
Note – there are no special data collection methods or forms for LFDA,
CEDA and Yield – general methods apply
Special forms are provided for collecting interview data for ParFish
See Section 3.2 of FTP 487, and ParFish toolkit
Content
Why collect data?
Useful references on data collection
-----------------------------------------------------------------------------------Data commonly used in stock assessments (e.g. using FMSP tools)
Data needs of the different SA approaches and FMSP tools
-----------------------------------------------------------------------------------Data collection methods (C/E, LF, biological, ParFish)
System design
Sampling design
Data forms
Database systems
Why collect data?
1. To provide information needed by policy makers and to meet
international reporting obligations
2. To evaluate likely benefits & costs of alternative management
options (long term, strategic SAs)
3. To monitor if you are achieving your goals and provide short term
management advice (tactical SAs)
Useful sources of information on data collection
•
•
•
•
•
Bazigos (1974) – Survey design, maths etc (for inland waters)
Caddy and Bazigos (1985) – Adaptation for manpower limited situations
FAO (1997) – General guidelines on data collection for management
FAO (1998) – New guidelines for objectives-based data collection
FMSP project R8462 – Guidance for developing data collection
systems for co-managed fisheries (developed with Bangladesh DOF, to
be published soon as an FAO FTP – see www.fmsp.org.uk)
• Hoggarth et al (2005) – Specific data needs of FMSP Fish Stock
Assessment Tools (Section 3.2)
• Sparre & Venema (1998) – Survey designs for collection of LF data
(Chapter 7)
• Stamatopoulos (2002) – Simple guide to survey designs for C/E data
References
• Bazigos, G.P. (1974). The design of fisheries statistical surveys. Inland
waters. FAO Fisheries Technical Paper, 133.
• Caddy, J.F. and Bazigos, G.P. (1985). Practical guidelines for statistical
monitoring of fisheries in manpower limited situations. FAO Fisheries
Technical Paper, 257.
• FAO. 1997. Fisheries management. FAO Technical Guidelines for
Responsible Fisheries No. 4. Rome, FAO. 82 pp.
• FAO. 1998. Guidelines for the routine collection of capture fishery data.
Prepared at the FAO/DANIDA Expert Consultation. Bangkok, Thailand, 1830 May 1998. FAO Fish. Tech. Pap. 382. Rome, FAO. 113 pp.
• Hoggarth et al, 2005. Stock Assessment for Fishery Management – A
Framework Guide to the use of the FMSP Fish Stock Assessment Tools.
FAO Fish. Tech. Pap. 487.
• Sparre, P. & Venema, S.C. 1998. Introduction to tropical fish stock
assessment. Part 1. Manual (Rev. 2). FAO Fish. Tech. Pap. 306.1. Rome,
FAO. 407 pp.
• Stamatopoulos, C. 2002. Sample based fishery surveys. A technical
handbook. FAO Fish. Tech. Pap. 425. Rome, FAO. 132 pp.
http://www.fao.org/DOCREP/004/Y2790E/Y2790E00.HTM
Data commonly used in Stock Assessment
Catch, effort and abundance
•
•
Use catch data to examine fishery ‘phase’, increasing or decreasing? (see next)
Use CPUE or other abundance estimate as index of stock size (explain why
catches are increasing or decreasing), and as input to biomass dynamic models
Catch composition (Age frequency / length frequency)
•
•
•
•
•
Use to estimate (growth and) mortality rate as the indicator of fishing pressure
Intensive sampling also allows construction of a stock-recruitment relationship
based on VPA methods (used to avoid recruitment overfishing)
Can also estimate size-selectivity of gears – important for setting mesh size rules
Age frequencies better if fish can be aged (see FTP 487 chapter 10) – gives
better estimates of growth and mortality rates
Note catches may also need to be subdivided by species, in a multi-species or
ecosystem management situation
Biological data
•
•
Analytical models also need inputs on size at maturity, fecundity at size, weight
at length
Information on seasonality of spawning, feeding, growth and recruitment can also
be useful for guiding the use of closed seasons in the fishery
(see FTP 487 Sec 3.2)
Use of catch data to show fishery ‘phase’
Generalised
Fishery
Development
Model
Mature
Senescent
Landings
(see Grainger
& Garcia,
1996;
and Hoggarth
et al, Chapter
14)
Developing
Time
Grainger, R.J.R. & Garcia, S.M. 1996. Chronicles of marine fishery landings (1950-1994): Trend
analysis and fisheries potential. FAO Fisheries Technical Paper. 359. Rome, FAO. 51 pp.
Data needs of the different approaches / tools
Analytical approach (LFDA / Yield) (See FTP 487, Tables 4.1 & 4.3)
• Catch composition data (either from length frequency data – LF, or
ageing studies)
• Biological data (e.g. size at maturity)
• Management advice can be produced from just one seasons’ sampling
(e.g. from a short time-series sample of LF and biological data)
•
But note some reference points also need long-term Stock-Recruit relationship
Biomass dynamic approach (CEDA)
• Multi-year time series of catch and effort data, or catch data with a
secondary index of abundance (e.g. from a survey)
ParFish approach
• Uses C/E and/or abundance data as with the other biomass dynamic
models
• Due to Bayesian formulation, can also add other sources of information
to improve the analysis, e.g. where few or no C/E data are available,
and to ‘tune’ the outputs to local users preferences
Data inputs of the FMSP analytical approaches
(e.g. Yield) for estimating reference points
• Use with LFDA or others to estimate growth parameters and F indicator
• Note that different model options within the software require different inputs
Notation
Inputs - Ecological
VB Assymptotic length
VB Growth rate / curvature parameter
VB Age 'at zero length'
Length / Weight parameters
Natural Mortality
Ambient temperature (for Pauly M equation)
Mean length (or age) at maturity
Length at maturity as a proportion of L∞
Spawning seasons
Stock-Recruit Relationship (form and parmeters)
Density Dependence in SRR (B&H steepness)
Inter-annual variability in recruitment
Inputs - Management controls
Fishing mortality (simulate a range)
Mean length (or age) at first capture
Length at first capture as a proportion of L∞
Fishing seasons
L∞
K
t0
a, b
M
T
lm (tm)
Lm
B&H ‘invariant’ methods
Constant
With SRR
recruitment
(Yes)
Yes
(Yes)
Yes
Equil.
YPR
Yield
Equil.
Yield
Trans.
Yield
Yes
Yes
Yes
Yes
Yes
(Yes)
Yes
Yes
Yes
Yes
Yes
Yes
(Yes)
Yes
Yes
Yes
Yes
Yes
Yes
(Yes)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
h
Yes
Yes
lc (tc)
Lc
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Table 4.3 of FTP 487
Data inputs of CEDA software
• ‘No recruitment’ model for a short (e.g. 1 yr) time series;
• DRP models for longer data sets (e.g. several years with good contrast)
Inputs – Ecological
Natural Mortality
Time lag between spawning and recruitment
Index of recruitment (annual)
Initial proportion (fraction exploited at start of data set)
Pella-Tomlinson shape parameter
Inputs – Fishery
Total catches by time period (in weight)
Total catches by time period (in numbers)
Mean weight of individual fish 1
Fishing effort applicable to total or partial catches 2
Partial catches by time period (in weight) 3
Partial catches by time period (in numbers) 3
Abundance index (proportional to biomass only) 4
Variance of abundance index 5
Outputs - Parameters etc
Initial population size (numbers)
Carrying capacity (unexploited population size)
Population growth rate
Catchability coefficient
Outputs - Indicators
Population size (numbers) at time t
Equil. Biomass at time t
Outputs - Reference points
Maximum Sustainable Yield
Replacement Yield
F giving MSY 6
Notation
No
Recr.
Index
Recr.
DRP
Const
Recr
M
Yes
Yes
Yes
Yes
Yes
Yes
Yes
(Yes)
Yes
Yes
(Yes)
Yes
Yes
(Yes)
Yes
Yes
Yes
Yes
DRP
Schaefer
DRP
Fox
DRP
Pella
Toml.
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
(Yes)
Yes
Yes
(Yes)
Yes
Yes
(Yes)
Yes
Yes
(Yes)
(Yes)
(Yes)
(Yes)
(Yes)
(Yes)
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
z
N1
K
r
q
Yes
Yes
Yes
Yes
Yes
Yes
Nt
Bt
Yes
Yes
Yes
MSY
RY
FMSY
Yes
FTP
487
Table
4.8
Data needs for FMSP and other methods for estimating indicators
Data
Catch
Abundance
(CPUE /
survey)
Length
Freq
Age
Freq
Biological
Intermediary method(s)
Indicators
Method
Method
Indicators
Grainger & Garcia
Fishery phase
(D/M/S)
Parameters
Myr TS
Myr TS
Myr TS
1yr TS
SS
Average
Myr TS
Average
Myr TS
(for tuning)
Notes:
SS
Myr TS
Myr TS
Mat., Wt/age
CEDA/ PFSA
r, K, q
CEDA / ParFish
Bt
LFDA
Growth
LFDA
Feq
Pauly
M
Spreadsheet
Age-L Key
Catch curve
Feq
Spreadsheet
Growth
Pauly
M
LFDA / FiSAT
Growth
FiSAT LBCA
Neq, Feq at length
Pauly
M
VPA
Nta, Fta, Sel.
Spreadsheet
SSBt
Myr = multi-year, TS = time series of data, SS = single sample, Mat. = maturity at age data, Sel. = selectivity pattern by age
Fishery phases: D = developing, M = mature, S = senescent (see Section 3.4.1). Subscripts: t = year, a = age, eq = equilibrium.
FiSAT LBCA = length based cohort analysis ('pseudocohorts' method can also estimate Nt etc from time series data for fast growing species)
Chapter 5, Table 5.1
Data needs of FMSP reference point estimators
Data
Catch
Abundance
(CPUE / survey)
Length
Freq
Age
Freq
SS
Biological
Resource
area
Mat. at size
Intermediary method(s)
Reference points
Method
Para-meters
Method
Spreadsheet
Lm50
Empirical models
MSY, fMSY (approx.)
CEDA
r, K, q
CEDA / PFSA
MSY, BMSY, FMSY
LFDA
Growth
R7040
YPReq/BPR0, FMSY*
(Pauly)
M
R7040
Yieldeq/B0, FMSY
Yield
Fmax, F0.1, F0.x
Yield
F%SPR
Yield
Fmax, F0.1, F0.x
Yes
Myr TS
Myr TS
1yr TS
Sel., Mat.
DDRecr.
1yr TS
Sel.
Sel.
Growth
Pauly
M
Spreadsheet
Growth
Pauly
M
Yield
F%SPR
VPA
Nta, SRR
Yield
FMSY, Fcrash
R An.var.
Yield
Ftransient
Mat.
Myr TS
Myr TS
Myr TS
(for tuning)
Notes:
Myr TS
Mat., Wt/age
Size-based (approx.)
LFDA
Mat.
SS
Reference point
Myr = multi-year, TS = time series of data, SS = single sample, Mat. = maturity at age data, Sel. = selectivity pattern by age or size
DDRecr. = density dependence in SRR (Beverton & Holt 'steepness'), FMSY* = FMSY assuming constant recruitment
R An. Var. = annual variability in recruitment (coefficient of variation)
Chapter 5, Table 5.2
General Guidance on Data Collection
Collection of catch and effort data
• First priority of data collection system
• Usually either sampled at landing sites, or by completion of log books
• Port samples raised to total catches within administrative and practical
strata using frame survey and other data (see below)
• May also use observers on board vessels, e.g. where problems with
bycatches/discards may otherwise be unrecorded
• May need to use General Linear Models (GLM) to standardize data for
the fishery as a whole, where more than one gear type is used, or
where fishing effort patterns have changed over time
Estimation of stock abundance
(from C/E or abundance survey data)
• Needed as input for biomass dynamic models, which estimate the MSY
of the fishery based on a time series of catch and abundance data
• Can obtain index of abundance from CPUE of the commercial fishery
data,
• …. or from ‘fishery independent’ surveys using a standard survey track
and gear type every year (e.g. ‘swept area’ surveys)
• Commercial data may be easier to obtain, but may be biased by
changes in catchability over time and by spatial factors in fishing
patterns
• Fishery independent surveys should be less biased but may be
expensive (and may also be less accurate – i.e. have higher variance).
Using effort units to ensure constant catchability, q
• Biomass dynamic methods assume constant catchability, q, over the
time period of the data set, i.e. that CPUE = q Biomass
• Only works if effort, E is a good measure of the effect of fishing on the
stock, i.e. that the fishing mortality rate is proportional to effort, F = q E
• Effort measure may need to include number, power and size of vessels
and gears, and the time actually fished (actively for some gears, or as
soaktime for others)
• e.g. for trawl fishery – boat-hours (for a standardized boat power and
size)
• See other examples on next slide for Bangladesh inland fishing gears
(and see FAO, 1998 Annex 2 for general categories)
• For multi-species, multi-gear fisheries, measuring effort is complicated,
so sometimes use a standard fisherman-day or a fisherman-year as an
effort measure. In this case, q clearly not constant, so take care.
Examples of effort units for Bangladesh inland fishery
Gear Type
Effort unit
Passive Filters / Barriers
Active Filters / Seines
Open water (many hauls per day)
Restricted water (few hauls per year),
including Bangladesh Katha FADs
Drag / Push Nets
Barrier trap-hours1
Lift Nets
Cast Nets
1
2
Net-hour
Site2
Net-hour
Net-hour
Gill Nets
Net-hour
Net-metre-hour
Portable Traps
Hooks (long lines, individual hooks etc)
Trap-hour
Hook-hour
Dewatering
Kuas (fish pits)
Hand fishing on floodplains
Spears
Site2
Man-hour
Spearing-hour
Effort units including hours are either soakhours for unattended gears (e.g. traps,
barriers) or active fishing hours for attended gears (e.g. push nets, spears)
eg per katha/kua in Bangladesh. This effort measure takes no account of the
different water areas fished.
Length or age frequency data?
• Age frequency data most useful where ageing is possible, as mortality
rates are estimated directly and will be more accurate
• With length-based methods, length samples are converted to age using
a fitted growth curve, and then mortality rates are estimated, so
accuracy is reduced by the extra step needed
• Length-based methods needed for analytical methods for species which
can’t be aged (e.g. crustacea, some tropical stocks)
– But note alternative option to use biomass dynamic approach for
these species also (using only C/E data, not catch compositions)
Using length frequency data
• Can estimate growth rates fairly easily from small samples (see LFDA
tutorial)
• … but good estimates of mortality rates need data on the total LF of the
catch in the fishery, so you need to sample over the full year, and for
different gear types and locations, and raise according to the relative
catches etc (see Sparre and Venema, 1998).
• Best results from length-based methods will be obtained by sampling
gears that have small mesh sizes, and therefore give a wide size range
of fish in the sample (may use special small-mesh sampling gears)
• A growth curve can sometimes be fitted through a single sample, but a
short time series of samples will increase confidence that a real curve
has been detected (see e.g. in next slide)
• Very difficult to use with gears such as gill nets as gear selectivity
biases the samples too much
• Be careful in the case of migratory stocks – see Sparre & Venema
(1998) Chapter 11
Example length frequency data (LFDA analysis)
Grow th Curve = Non Seasonal. Linf = 150.00. K = 1.00. Tzero = -0.80.
Length Class
150
100
50
0
0
1
2
Sample Timing
3
4
Collection of biological data
• Used as inputs in analytical models (e.g. Yield, YPR)
• E.g. maturity (by size and season), weight at length etc
• Can be sampled every few years, unless conditions change significantly
between years
• May be incorporated into LF sampling, as a second step for some
randomly selected fish …
or as special surveys, depending on needs.
Collection of interview data for ParFish
Stock assessment interviews
Use of other information as ‘priors’
Preference interviews
See ParFish toolkit etc
Data collection for empirical SA methods
‘Empirical’ approach compares benefits of alternative management
approaches as measured by indicators of the various fishery
objectives
Useful where resources can be sub-divided into small units, with
different management practices tried in each (e.g. in separate lakes,
or co-management units in floodplain fisheries)
But can be confounded by other differences between units, so best to
have large number of ‘sample’ units to compare
Need to measure both inputs to the system (management practices,
other possible influences) and outputs from the system (catches,
biodiversity, social benefits etc)
Analyse system influences using GLM or Bayesian Network
approaches (see Hoggarth et al, 2005, Chapter 14, based on FMSP
Project R7834, and guidelines of FMSP Project R8462)
Data collection system design
Focus on need to provide information for management feedback on
goals and objectives… (see next slide)
• If goal is to maintain biodiversity, sampling must take multi-species
approach
• If management options include gear controls or restrictions, sampling
must take multi-gear approach
• If fishery is single-species, single gear, nice and simple!
Decide data collection needs, depending on which management
measures and stock assessment tools are intended
• See selection procedure in new guidebook
Design institutional arrangements depending on degree of comanagement
• Note need to involve collaborators in decision making (both on system
design and management regulations), not just collecting data
Examples of other data to collect (depending on
goals, objectives and fishery type / situation)
Biological
• e.g. discards by species and gear type
Ecological
• impacts of fishing gears on habitats etc
Social
• Numbers employed by fishery / fleet / gear
• Numbers of processors, supply chain etc
• Subdivision of stats by age group and gender
Economic
• Fishers incomes, costs, profits (by gear / fleet type)
See FAO (1997, Tables 1-3) and FAO (1998) for detailed data to collect, and
use of these data in formulating policy, making management plans and in
monitoring performance.
Template for selecting SA tools and data collection…
(see new guidebooks from Project R8468)
Operational
Objectives
Possible
management
standards
and
measures
Relevant
tools
Input
requirements
What
precautionary
advice can be
given
immediately
(using existing
data)?
What resources
are needed to
collect additional
data to complete
stock
assessment?
Costs and
benefits of
different tools
and data
collection needs
… Add more rows as necessary to cover all the operational objectives and tool options
Data collection systems for different fisheries
Industrial and artisanal
• Large scale fisheries vs small scale – differences in value affect
management options and data collection needs
Freshwater and marine systems
• Vary in degree of containment and controllability and in need for
collaboration with ‘shared stocks’ (both with neighbouring countries at
sea, and with neighbouring co-management units in river fisheries)
• Most important to work on ‘unit stocks’ – may be more sub-divisible in
inland fisheries than marine – but note ‘blackfish’ and ‘whitefish’ issues
Sampling design
• Applies to both C/E and LF data
• Draw thematic map to show fishery areas, vessel harbours and landing
sites, and potential sampling points (see next slide)
• Need frame survey to give numbers of fishing units, both for sample
design and to raise total catches (see data form in Iraq project doc?)
• Estimate total catch by basic formula
• CPUE estimated by landings survey – average CPUE by gear type,
species etc
• Four options for estimating effort given on next pages (see
Stamatopoulos, 2002 for details)
Example of
thematic map
Caddy & Bazigos,
1985
Frame survey
Inventory of fishing industry used to raise total catches from sample
data
Repeat every few years (depending on rate of change)
Includes following data:
•
•
•
•
Size and area distribution of fishing sites (map)
Number of fishing boats (by type)
Number and type of fishers (full-time, part-time, migratory etc)
Number and type of fishing gears owned
(See Bazigos, 1974 for details)
Estimating fishing effort – Option 1.
Complete enumeration (census)
Best result, but often impossible
Estimating fishing effort – Option 2.
Census in space, sampling in time (Second best)
•
•
•
where:
AverE is the average fishing effort in boat-days over the sample days.
A is a raising factor expressing total number of days of fishing activities during
the month, i.e. it is calculated each month.
Estimating fishing effort – Option 3.
Census in time, sampling in space (third best)
•
•
•
where:
AverF is the average fishing effort exerted by a single fishing unit during the
month and is associated only with the sampling locations from which data have
been collected.
F is a raising factor expressing the total number of fishing units that are
potentially operating at all fishing sites (i.e. the overall geographical stratum).
Estimating fishing effort – Option 4.
Sampling in both space and time
(most common, economical for data collection, but least robust)
•
•
•
•
where:
BAC is the Boat Activity Coefficient, expressing the probability that any boat (= fishing unit)
will be active (= fishing) on any day during the month.
F is a raising factor expressing the total number of fishing units that are potentially operating
at all fishing sites (i.e. the overall geographical stratum as in Option 3).
A is a raising time factor expressing total number of days with fishing activities during the
month (as in Option 2).
Combination of surveys to estimate total catch
For Option 4. Sampling in both space and time
(See Stamatopoulos, 2002 for when to use Options 1, 2, 3 or 4)
Survey stratification
Stratification is used to:
• give more homogeneous target populations, which will provide lower
variances in the estimates;
• categorize the data population in order to respond to specific user
needs (e.g. by different gear types, if different management regulations
are being considered for each one);
• allow reporting for different administrative, functional or other
categories.
• See FAO, 1998 Table 5.1 for examples of stratification variables: spatial,
time, fishers/processors etc, vessel and gear types, social
(communities, ages etc), environment (habitats), seasons…
Data forms to use - Examples
• Caddy & Bazigos (1985) – manpower limited situations
• Bazigos (1974) – survey designs for inland waters
• Stamatopoulos (2002) – simple examples for Options 1-4 above, and
compatible with FAO ARTFISH database
• Note need for standardization of data – e.g. using consistent codes for
species names, and for allowing data exchange between adjacent
states for shared stocks
• See FAO, 1998 for general guidance on designing data sheets
Database system
Need to keep good records, check for data entry errors and backups
etc
Simple spreadsheets may be OK for small scale local units?
Need proper relational database for larger fisheries with strata etc
• e.g. ARTFISH (download from FAO web site)
• e.g. PISCES generic database produced by FMSP Project R7042
(www.fmsp.org.uk)
• e.g. CARIFIS database?
• Use/development of existing FRSS systems in Bangladesh?
Discussion Questions
• What data do you have already that could be used for stock
assessment? In FRSS or in local management units?
• How are catch and effort estimated? How reliable are the data?
• What effort measures are used for different gear types? Would they
provide useful, unbiased indicators of abundance?
• Do you have any routine survey data that could provide abundance
estimates (time series)
• Do you collect length frequency or age frequency data? How often?
For what species, what sample sizes?
• Do you have biological data needed for analytical methods?
• How could national (e.g. FRSS) and local management data
collection systems be integrated, especially in inland fisheries? Are
any data collected in both?
• What other data should be collected, besides C/E, LF, biological?