PPTX - PICES WG27 North Pacific Climate Variability and Change

http://rwcatalog.neaq.org/Terms.aspx
Erin Meyer-Gutbrod - Cornell University
Dr. Andrew Pershing – Gulf of Maine Research Institute
Dr. Charles Greene - Cornell University
Right Whale
Feeding and
Breeding Zones
 RWs spend Spring, Summer
and Fall feeding in the Gulf
of Maine and off the Scotian
Shelf
 Pregnant cows migrate to
waters off Georgia and
Florida to calve during the
winter months
Adapted from E. Paul Oberlander, Woods Hole
Oceanographic Institution Graphics
Critically endangered: fewer
than 500 individuals
Chief Prey Group:
Calanus finmarchicus
 Abundance levels respond to climate forcing from the
Arctic and within the North Atlantic basin, and are
correlated with the NAO (Greene et al.)
 Due to the high energetic demands of pregnancy and
nursing, right whale vulnerabilities to prey limitations
are most likely to be manifest in the reproduction
cycle
Continuous Plankton Recorder
 Data collected from silk
screens dragged behind
ships of opportunity
between Boston and
Cape Sable, Nova Scotia
 CPR data sorted by
region and by bimonthly
time period to create a
high resolution, long
term data set
Calving rates driven by prey
abundance
 High calving rates in 1980s
and 2000s driven by high
copepod abundances
 Low rates in 1990s driven
by fewer copepods
3-Stage Reproduction Model
 A calving cycle takes at least 3
years: 1 year in each stage
 2 Probabilities must be
computed:
Transition between
Recovery and Pregnancy Φ21
2. Transition between
Pregnancy and Nursing a
calf Φ32
1.
Calculating Transition Probabilities
 Probabilities ϕ21 and ϕ32 are modeled as logistic regressions
to constrain them between 0 and 1
 The logit is a linear combination of an intercept and a suite
of prey terms which are multiplied by coefficients
 In the prey-independent model, the logit will only contain
an intercept
Projecting Right Whale Population
 Insert transition probabilities into a demographic
population matrix:
 Multiply the population matrix by a vector of
individuals categorized into reproductive stages to
determine the distribution of individuals in the
following time step:
Model Optimization
 This is a big job.
 Using R and AD Model Builder we test each of the
possible CPR data sets by season and geographical
region.
 Look for combinations of CPR data sets that yield
accurate calving series.
 Find the set of parameters (intercepts and coefficients
inside the two logits) that best fit the model to the
observed time series of calf births.
3 Calving Models:
Independent model with no prey dependency –
transition probabilities are constant through time
2. Prey-dependent model using only yearly average
values from the Continuous Plankton Recorder
3. Prey-dependent model using bimonthly values from
the Continuous Plankton Recorder
1.
Parameters of the 3 models
Comparing 2 calving models: a simple
independent model and a prey-dependent model
 Prey-independent
model follows
general trend of
increased calving
 Prey-dependent
model captures the
dips and spikes in
reproduction that
are caused by
environmental
conditions
Viable cow distribution
 Total number of viable cows increases steadily
 Cows spend more time in the “resting” stage
 After a period of low prey abundance, cows wait in the
“resting” stage until a year of high prey abundance leads to
a spike in pregnancies.
 Given average prey availability,
cows are very likely to get
pregnant
 Great declines in calf production
are explained by the connection
between low prey abundance
and low pregnancy rates
 Chance of successful delivery is
much lower than chance of
conception given any set of prey
conditions
 This may result from the
increased nutritional
requirements for pregnant cows.
What’s next?
 Fit prey-driven reproduction model separately for each
decade to look for different patterns between the three
decadal regimes
 Choose a general model that is robust enough to withstand
regime shifts
 Reproductive model can be nested into a complete
demographic model to project future population growth
 This model can be used to project right whale extinction
risk under varying prey conditions related to climate
change and / or varying rates of anthropogenic mortalities
Thanks to:
 Cornell University
 National Defense Science and Engineering Graduate




fellowship
Office of Naval Research
Atkinson Center for a Sustainable Future
Gulf of Maine Research Institute
New England Aquarium