Chesapeake Bay Fishery-Independent Multispecies Survey (CHESFIMS)

1
Chesapeake Bay Fishery-Independent Multispecies
Survey(CHESFIMS)

• T. J. Miller1, M. C. Christman3, E. D. Houde1, A.
  F. Sharov2, J. H. Volstad4, K. Curti1, D.
  Loewensteiner1, B. Muffley2, and D. Sams3

1. CBL UMCES Solomons, MD 20668 3. Biometry Program UMCP College Park, MD 20742
2. Fisheries Service MDNR Annapolis, MD 21401 4. Versar Corp Columbia, MD 21405
2
Toward ecosystem-based management

• There is no substitute for good monitoring
  programs of fished species and of key interacting
  species. Modeling evolves from and depends on
  monitoring results, and management depends upon
  an understanding of the status and trends of
  stocks. Fishery-independent surveys to monitor
  resources and obtain biological data, if
  instituted and coordinated throughout the bay,
  would help improve management.

Executive summary of Multispecies Management
workshop report. Houde et al. 1998
3
CHESFIMS Objectives

• Conduct a baywide survey of the bentho-pelagic
  fish community, focusing on young (juveniles, and
  yearling) fishes in the mainstem of Chesapeake
  Bay.
• Design and implement a complementary survey of
  the bentho-pelagic fish community in the
  extensive shoal habitats (lt 5 m depth) in the
  mainstem of Chesapeake Bay.
• Conduct pilot surveys of the pelagic fish
  community in key tributaries and in the mainstem
  to generate sampling statistics that will of use
  in subsequent design improvements.
• Determine trophic interactions among key
  components of the pelagic fish community, and
  examine the implication of the relationships
  uncovered in empirical studies using bioenergetic
  modeling.
• Conduct statistical analyses of existing and new
  data to optimize the complemented pelagic survey
  with respect to consistency and accuracy of key
  parameters.

4
CHESFIMS

• 3 components
• Baywide, broadscale midwater trawl survey.
• Complex design involving fixed transects and
  random stations within three strata (upper, mid
  and lower Bay)
• Samples depths gt 5m, using an 18 m2-midwater
  trawl (6 mm cod end) fished in 10 equal depth
  bins from surface to bottom.
• Builds on existing 1995 2000 NSF-sponsored
  survey (TIES).
• Regional, shoal survey.
• Stratified random sample currently involving 9
  strata.
• Samples depths lt 5m, using a 16 otter trawl
  towed for 6 min.
• Complements and extends existing MDNR and VIMS
  surveys.
• Statistical evaluation.
• Analysis of alternative survey designs to
  optimize final survey design.
• Application of spatial statistical models as to
  develop Baywide abundances.

5
Broadscale catch summaries
6
2002 Catch summary
(/tow)
/tow
1 10 50 100 500 1000 5000 10000 15000
Spring
Summer
Autumn
7
Interannual comparisons
2001
2002
Autumn
Spring
Summer
8
Biomass time series
Bay anchovy
Croaker
White perch
Spr Sum Aut
9
Shoal survey

• Survey conducted three times during 2002,
  involving 9 strata
• Sampling conducted with a 16 otter trawl
  deployed lt 5 m depth
• 7,365 fish sampled
• Less than 50 of 2001 catch, despite increased
  effort

10
Shoal catch summaries
11
Shoal catches
61.02
12
CHESFIMS as a single species monitoring tool

• Multispecies surveys can also provide single
  species indices
• Calibration with existing single species surveys
  is important

CHESFIMS
13
Diet characterization

• Sample size adequacy determined for key species
  by region
• Diets of random samples of preserved fish
  quantified
• Biological characteristics of fish and prey items
  determined
• Evidence in diet data of substantial variation
  spatially and temporally in key species

14
Croaker diets

• Spatial variation evident
• Feeding incidence
• Prey types
• Prey importance
• In mid-Bay during summer 60 by weight of
  croaker diets is comprised of fish, principally
  anchovy

15
White perch diets

• Temporal variation apparent
• Some prey items consistently represented
  throughout year
• Some prey appear dominant only in certain periods
  of the year
• Quantifying allometric influences

16
Evaluation of Survey Efficiency

• Use the design effect and effective sample
  size to measure the efficiency of a specific
  survey design
• Estimates under simple random sampling are used
  as benchmarks for comparison
• Applied to mean CPUE as an example

17
The design effect

• The design effect is the ratio of variances
  from complex and simple random sampling
• Neff is the number of stations that would be
  required under simple random sampling to achieve
  the same precision obtained with the complex
  design
• The effective sample size is estimated as

18
CHESFIMS 2002 Design effects for mean CPUE,
19
Design comparison

• Estimates of mean CPUE, precision and design
  effects suggest that the stratified random survey
  is more effective than the transect sampling for
  most species
• This suggest that CPUE from stations within
  transects on average are more similar than hauls
  from different transects.
• The relative abundances of weakfish were more
  effectively estimated from the transect survey
  during all seasons (deff lt 1)
• This indicates that the CPUE by station within
  transects exhibit minimal intra-cluster
  correlation for this species The designated
  strata did not substantially reduce variability
  in CPUE between stations
• The designated strata did not substantially
  reduce variability in CPUE between stations
• Alternative stratification should be explored in
  future surveys, possibly based on
  post-stratification analysis of exisiting survey
  data

20
Design recommendations

• The design efficiency
• Depends on the underlying spatial distribution of
  the target species
• May vary with the season, and over time
• Depends on survey cost (e.g., stations that can
  be sampled per day)
• Model-based estimators that incorporate spatial
  auto-correlation could potentially improve the
  effective sample size

21
Spatial modeling

• Abundane estimation comparison of techniques
• Design-based using the actual sampling design
  (correct design-based approach)
• Design-based assuming the stations had been
  selected according to simple random sampling (a
  very typical but incorrect approach)
• Model-based using a geostatistical spatial
  autocorrelation model (a model-based method)

22
Which method is best?

• The design-based approach is best if one wishes
  minimal assumptions to be made and wishes the
  procedure to be data independent (i.e.
  methodology is NOT data-driven).
• Iif spatial autocorrelation is present, and it is
  correctly modeled,then kriging-based estimates
  will be better
• Spatial modeling provides additional insights
  including inferences regarding distribution
  patterns of species not available in design-based
  methods.

23
Conclusions

• Baywide, broadscale midwater trawl and regional
  shoal surveys
• Provide a basis for estimating time series of
  abundances (mean ? SE) of individual species, of
  species guilds and of the overall fish community
• Provide data on the biological characteristics of
  the survey catch
• Provide inferences regarding the distribution of
  individual species, guilds and of the fish
  community
• Dietary analyses quantify trophic relationships
  within the fish community
• Revealing spatially and temporally variable
  patterns
• Statistical evaluation.
• Compares alternative survey designs to optimize
  final survey design.
• Single survey design will not be optimal for all
  species
• Applies spatially-explicit statistical models as
  to estimate baywide abundances and distributions
  for species for which the design is not optimal

24
2003

• Full field season (broadscale and shoal survey)
• Continued effort on dietary analyses (broadscale
  and shoal)
• Statistical analysis of
• Multispecies patterns
• Abundance
• Distribution
• Biological characteristics
• Efficiency and adequacy of alternative sampling
  designs
• Integration of multiple surveys
• Correlations with commercial landings

CHESFIMS on the web at hjort.cbl.umces.edu/chesfim
s.html