Marbled Murrelet Population Viability: results of a PVA workshop

Introduction

A population viability workshop for the Marbled Murrelet was held on 22-23
November 1996 at Lewis and Clark College in Portland, Oregon.  This document
presents the results of this workshop pertaining to the population viability
analysis (PVA) model that will be developed for the Marbled Murrelet.

First we summarize the general structure of the model and the background on its
implementation.  Then we discuss various aspects of the model.  These
discussions reflect the comments and suggestions by the panel of advisors and
by the observers present at the workshop, as interpreted by Applied
Biomathematics.  Finally we summarize the format in which the results of the
PVA will be presented.

General model structure

The PVA model will be a stage-structured, stochastic metapopulation model.  The
model will be parameterized with published and (where available) unpublished
data on the demography, population trends and habitat relationships of the
species.

The model will be implemented on the software RAMAS/GIS (Akcakaya 1995)
developed at Applied Biomathematics.  This program has recently been used to
develop models for the California Gnatcatcher in Orange County (Akcakaya and
Atwood 1997), Helmeted Honeyeater in Australia (Akcakaya et al. 1995), Northern
Spotted Owl in the Pacific Northwest, and Red-cockaded Woodpecker in Louisiana;
reviews of the program have appeared in Conservation Biology (Kingston 1995)
and Quarterly Review of Biology (Boyce 1996).

Stage matrix and initial abundance

The within-population dynamics of the model will be described by a stage
matrix, whose elements are the vital rates (survival rates and fecundity) of
juveniles (0-12 months), subadults (12-24 months) and adults (24+ months).  The
vital rates estimated by Beissinger (1995) provide a good starting point for
the stage matrix:

           Juv.    SubAd.    Adult
  Juv:    0.0000   0.0000   0.1050
SubAd:    0.6125   0.0000   0.0000
Adult:    0.0000   0.7770   0.8750

The parameters (vital rates) in this matrix should be fine-tuned with data on
the recent population trends.  A range of numbers should be used for each vital
rate.

The initial abundance of the population should be based on data from off-shore
counts and in-land detections.  The initial distribution of individuals to
stages (juvenile, subadult, adult) probably will not affect the results very
much.  A low proportion (e.g., 6%) may be used for juveniles to reflect the
off-shore counts.

Density dependence

The population growth of Marbled Murrelets may be limited by food and/or
nesting sites.  The observed population changes may reflect a deterministic
decline to extinction (systemic pressure), a decline to an equilibrium
abundance (carrying capacity) or to a ceiling that is lower than the current
abundance, or fluctuations without a significant trend.

Density dependence should be modeled with several models, including
ceiling-type with a deterministic decline, ceiling-type with a
deterministically stationary population, and contest (Beverton-Holt) type with
a stable carrying capacity.  In the latter case, the maximum rate of population
growth at low densities (when there are no density effects) must also be
specified.  This should be based on an assumption of a maximum of 1 chick
fledged per nest.

The contest (Beverton-Holt) type density dependence may arise when the
limitation ("ceiling") acts only on the breeding population, and the adults in
excess of this ceiling become non-breeders, decreasing the average number of
chicks per adult in the population.  In contrast, the ceiling model assumes
that all stages are limited; individuals in excess of the ceiling are assumed
to be dead.

Another type of density dependence involves Allee effects which may arise from
the negative effects of low abundances on the genetics and social structure of
the population.  Such effects may be modeled by a local (population-specific)
extinction threshold, below which the population is assumed to be extinct.

Stochasticity

There are no data on the temporal (year-to-year) variation in survival rates or
fecundities of Marbled Murrelet populations.  Such data may exist for other
alcids; if so, they should be analyzed to estimate the coefficients of
variation in vital rates.  In addition to environmental variations, the model
will incorporate demographic stochasticity in survival, reproduction and
dispersal (Akcakaya 1991).

Rare and extreme changes in vital rates ("catastrophes") may occur due to fires
on land, and due to oil spills off-shore.  The frequency of such events should
be calculated from historical data and incorporated into the model.  The
spatial structure of the metapopulation (see below) may interact with the
effect of catastrophes (Akcakaya and Baur 1996), so the catastrophe estimates
should be revised for the model with multiple populations.

Spatial structure

The model should investigate the effect of logging (see below) both on the
local population ("bioregion") and on the metapopulation of the Marbled
Murrelet in California, Oregon and Washington.

The spatial structure of the metapopulation model (the location and size of
each population) will be decided in consultation with Fish and Wildlife
scientists, and should be based on the distribution of suitable habitat.  This
consultation may take place in a workshop setting where relevant data and
software are brought together by the participants.

Effect of logging

The PVA model will be used to analyze the viability of the Marbled Murrelet
under three options:  logging in all Pacific Lumber Company land ("full
logging"), logging in part of the Pacific Lumber Company land as specified in a
proposed agreement ("partial logging"), and no logging.  In all three cases,
the effect of logging will be modeled as a decrease in the carrying capacity
(or ceiling) of the model, in proportion to the decrease in total habitat
suitability (i.e., decrease in available habitat, weighted by the suitability
of the habitat).  This effect should be quantified with an analysis of the GIS
data on the habitat characteristics.

Logging may also effect the vital rates, for example, through increased edge
effects that may cause an increase in predator densities.  This effect should
be studied with data from an experimental study in Washington that uses
artificial nests to measure predator pressure in different habitat types and
configurations.  This effect may be especially important in the full logging
option, but may be unimportant or negligible in the partial logging case.  This
is because the partial logging will target small fragments of suitable habitat,
and leave larger patches relatively unimpacted.

The effect of the three options should be considered within the context of
similar impacts on Marbled Murrelet habitat in other parts of the bioregion, as
well as in other regions in California, Oregon and Washington.  These
cumulative effects should be analyzed with data on the planned and proposed
habitat alterations, in consultation with scientists from the U.S Fish and
Wildlife Service and the California Department of Fish and Game.

Results

The results of the PVA will be expressed as increases in the risk of decline
with partial or full logging, from the risk of decline with no logging.  The
results of the analysis, including risk results, will contain error due to lack
of data to calculate the model parameters precisely.  Our ignorance about the
parameters will be reflected in a range of values rather than a single value
for each parameter (Ferson and Ginzburg 1996; Ginzburg and Goldwasser 1997).
The comparison of the three options mentioned above will be made for the high
and the low value of each parameter.  In addition, these comparisons will be
made under various assumptions of the model, such as with and without the other
populations in California, Oregon and Washington; or with and without
cumulative effects.  The degree of uncertainty will be reported with each risk
result or comparison.

Two parameters must be specified for the presentation of these results: the
decline threshold (amount of decline) and a time horizon (number of years for
which to make the prediction).  Results should be presented for two different
numerical values of each of these parameters.  One numerical value should be
fixed (such as 50 years for the time horizon, and 90% decline for the amount of
decline).  The other numerical value should be specific to the comparison,
giving the result for the threshold and time horizon for which the change in
risk was maximum.  This is necessary because it may be impossible to find the
amount of decline and time horizon appropriate for all cases.  To illustrate
this consider that all models will predict a zero impact for a very short time
horizon (we can be almost certain that the species will not be extinct tomorrow
or next week, regardless of how much logging occurs), and a zero impact for a
very long time horizon (fossil record suggests that many species extant today
will be extinct within a few million years, regardless of whether they are
impacted or not).  In models with a lot of uncertainty, these extremes are much
closer to each other, prediction a zero risk of extinction (with or without
simulated impact) for 1-5 years and near 100% risk of extinction (with or
without simulated impact) for 200-500 years.  In summary, case-specific time
horizons and decline thresholds, in addition to fixed ones, will allow the
selection of the appropriate criteria for assessment.

References

Akcakaya, H.R. 1991. A method for simulating demographic stochasticity.
   Ecological Modelling 54:133-136.

Akcakaya, H.R. 1995.  RAMAS/GIS: Linking Landscape Data with Population
   Viability Analysis (ver 2.0t).  Applied Biomathematics, Setauket, New York.

Akcakaya, H.R. and J.L. Atwood. 1997.  A habitat-based metapopulation model of
   the California Gnatcatcher.  Conservation Biology 11 (in press).

Akcakaya, H.R., and B. Baur. 1996.  Effects of population subdivision and
   catastrophes on the persistence of a land snail metapopulation. Oecologia
   105:475-483.

Akcakaya, H.R., M.A. McCarthy, and J. Pearce. 1995.  Linking landscape data
   with population viability analysis: management options for the helmeted
   honeyeater.  Biological Conservation 73:169-176.

Beissinger, S.B. 1995.  Population trends of the Marbled Murrelet projected
   from demographic analysis.  Pages 385-393 in Ralph, C.J., G.L. Hunt, M.G.
   Raphael, and J.F. Piatt (eds.), Ecology and Conservation of the Marbled
   Murrelet. U.S. Forest Service Pacific Southwest Research Station, Albany,
   CA.

Boyce, M. 1996. Review of RAMAS/GIS. Quarterly Review of Biology 71:167-168.

Ferson, S. and L.R. Ginzburg. 1996. Different methods are needed to propagate
   ignorance and variability. Reliability Engineering and Systems Safety 51 (in
   press).

Ginzburg, L.R. and L. Goldwasser. 1997. Variability and measurement error in
   extinction risk analysis: the northern spotted owl on the Olympic
   peninsula.  In: Ferson, S. (ed.) Quantitative Methods for Conservation
   Biology. Springer-Verlag, New York (in press).

Kingston, T. 1995. Valuable modeling tool: RAMAS/GIS: Linking Landscape Data
   with Population Viability Analysis.  Conservation Biology 9:966-968.