Discover Life in America

Michael Huston - 17 December, 1998

ATBI Inventory Design

From: Michael Huston 
To: Chuck.Parker@nps.gov, Keith.Langdon@nps.gov, pswhite@unc.edu
cc: alley@discoverlife.org, jody@discoverlife.org
Subject: ATBI Inventory Design
Date: Thu, 17 Dec 98 10:58:56 -0500

Hi, I hope that this message can be delivered to Chuck, Keith, Peter,
John, and
others involved with Inventory Design today while the meeting is still going.
If not, it's no big deal, since we can discuss these issues further later.
        This is basically my take on how our discussions can be organized and
synthesized a bit.  I think that some version of these issues should be
included

in the "User's Handbook" that Peter is planning.

                Best wishes, Michael


Inventory Design Issues
for the Great Smoky Mountains National Park
All Taxa Biotic Inventory


        In order to meet the objectives outlined in the mission statement
of the

ATBI, the inventory must be designed and organized to provide estimates of the
uncertainty of all products based on the inventory data.  The primary ATBI
products will be species lists and estimated species richness for selected
taxonomic groups for specific locations ranging in size from approximately 1
square meter to the entire park and maps of the estimated distribution and
abundance for all species based on "habitat predictor variables" derived from
observed occurrences of the species.

       The inventory design should address the following criteria:

        1) Sample locations should be amenable to multiple types of
stratification.  No single stratification will be able to address all
management

issues and scientific questions.  Consequently, sites should be selected, and
ppropriate characterization data obtained, so that a variety of different
tratifications can be created (e.g., by elevation, vegetation type,
disturbance

istory, geological substrate, landform location, etc.)

        2)  Sample locations should be natural units with naturally defined
boundaries, except for the small intensive sample areas (e.g., 1.0 or 0.1
ha, or

1 m2) which should have permanently marked artificial boundaries.

        3)  Sample areas should be nested over multiple spatial scales.
Advantages of this include: 1) ease of access; 2) calculation of species/area
relationships; 3) estimates of heterogeneity and predictability of
patterns;  4)

estimation of large-scale diversity and distribution patterns based on
subsamples of area.

        4)  Sample design should allow replicate measurements for
estimation of
means, variance, and similarity, and for testing predictive models.

        5)  Sites should be accessible for intensive inventory efforts.
Inventory methods should be refined and tested in areas that are readily
accessible in order to plan sampling in remote areas to be as  efficient as
possible.

        6)  Sample designs should facilitate development of methods to predict
spatial distribution and abundance patterns under current conditions and for
altered future conditions resulting from management actions or environmental
change.

        7) Inventory design, site selection, and estimation/extrapolation
should be supplemented by remote sensing analyses.



        Inventory Methods

        The wide variety of taxa, as well as the range of spatial scales over
which information is needed, requires that a variety of different sampling
methods must be used.  Many methods will be taxon-specific, although many taxa
will require a variety of methods to meet all information needs.  While the
specific methods must be determined by the taxonomic specialists, it is
possible

to classify sampling methods into a variety of general types, each with
its own
advantages, disadvantages, and information requirements.

        1)  Fixed area samples

                These include permanent (or temporary) vegetation plots and
subplots, small mammal trapping areas, bird point counts, and some aquatic
methods such as seining, emergence traps, and electo-shocking.   Results
can be
expressed quantitatively in terms of abundance per unit area, and all species,
and most individuals can be sampled.  Detailed environmental information
can be
obtained for the sample areas, which are generally selected to be homogeneous
and representative of specific conditions (e.g., positions along a gradient),
but may be located randomly depending on sampling objectives.  All of the
following methods can be used with fixed area samples.  However, the
impacts of
multiple sampling efforts on fixed area plots should be carefully managed and
minimized.

        2)  Point samples of points

                These samples (e.g.,  soil samples, plant tissue samples)
produce data that apply to a specific small area that can be precisely located
and whose environmental properties should be quantified in great detail.
These
samples often require substantial post-collection processing effort (e.g.,
chemical analysis, species sorting and identification.

        3)  Point samples of areas

                These samples are collected at a specific point location, but
actually sample a much large area surrounding the point (e.g., blacklights,
Malaise traps, drift nets, drift fences).  For these  samples it is
essential to

determine the approximate area that is sampled, as well as the properties
of the

area sampled, which may be quite different from the properties of the sample
location.  For example, a blacklight trap placed on a heath ridge overlooked a
valley of cove forest will attract insects from the entire area within the
line-of-sight from the light, which will vary depending on the orientation of
the light in relation to the landscape.  Some of these methods require
substantial post-collection processing effort (sorting and identification).

        4)  Location search or transect samples

        These samples typically focus on specific taxonomic groups and are
designed to determine presence, absence, and relative abundance of known
species.  Taxonomic specialists may focus on habitats where particular
organisms

are most likely to be found, or survey predetermined study areas to
determine if

the species occur there.   This will be the primary method used for many
taxonomic groups, and can provide a variety of types of important
information. A
variety of approaches can be used to provide estimates of relative
abundance of
organisms, typically based on area sampled (e.g., transect length) or
amount of
time spent sampling, or number of "units" sampled (e.g., leaves sampled for
galls, logs overturned for salamanders).   Information should be provided that
would allow the sample area to be relocated.  Such location data could include
coordinates, distance along road or trail,  location on a topo map PLUS
specific

habitat information, such as general vegetation type, dominant plant
species, or

permanent landmarks.

        5)  Chance or serendipity samples

        These samples provide useful information only if the location can be
specified (to some known degree of precision).   This is a primary means of
discovering unknown or unexpected species.


Proposed Inventory Sampling Design

        Based on written comments of Parker, Langdon, and White, as well as
discussion at the December 1998 ATBI meeting, a nested sampling design
could be
used to coordinate data collection and analysis, as well as to prioritize
inventory efforts.  White suggested using nested areas within a subset of the
20+ major watersheds draining the park.  One approach could be to select 5
or 6
watersheds (draining from the main ridge to the park boundary) that encompass
the major biotic variability in the park.  These watershed would be the
focus of

most effort over the first half of the project (although parkwide sampling
would

continue for specific taxa and for specific projects).  During the second half
of the project, a second set of watersheds "matched" to the initial set
would be

selected and used to test predictive models (estimates) of species
richness and
distributions based on data collected in the 6 "focal"  watersheds.  In
addition

to improving estimation methods, the second set of watersheds would provide
additional "hard" data and observations on the biodiversity of the park.

        The nested sampling design could include:

        1) Fixed area permanent plots 0.1 to 1.0 ha in size, with subplots
down
to 1 square meter or less for quantitative samples of herbaceous vegetation,
forest floor invertebrates, etc.  These plots should be designed to allow
sampling for many different taxa and physical properties, without damaging
them.

 Plot locations and the total number of plots would be selected to
optimize/maximize their value for estimating the properties of a larger
area by
chosing specific locations in relation to major gradients in the area to be
estimated.

        2) Small watersheds of 10-50 ha (drained by first order stream).
Taxonomic richness and species distributions in these watersheds would be
characterized using fixed area permanent plots and/or other methods described
above (2,3,4,5).  Different combinations of methods could be used for
"replicate" watersheds to develop and test estimation methods.   Several small
watersheds (3 or 4 "types" plus replicates of each type) would be selected to
characterize the major environments/ habitat types / and variability within a
larger watershed.

        3)  Large watersheds of 10 to 100 square kilometers would be
selected to

encompass the major variability in geological substrate, land use history, and
climate within the park.  These watersheds would be characterized on the basis
of the small watersheds described above, as well as additional sampling
(mainly
type 4 described above) to test the predictive power of the small
watershed data

and develop a species list, richness estimates, and distribution estimates
with
known uncertainty for the entire large watershed.  Unique components of large
watersheds, such as high order streams and their riparian zones, would require
quantitative sampling (e.g., fix area plots) as well as other methods.
        The initial set of 6 large watersheds could be used  for developing
estimation methods for the rest of the park landscape.  These methods could be
tested in "matched" large watersheds that would be sampled during the second
half of the project.

        With this inventory design, a logical series of estimation procedures
could be developed and applied, leading to better estimates with known
uncertainty for the entire surface area of the park.  This approach would
allow
refinement of inventory methods to avoid oversampling and increase the
efficiency of inventory efforts.  In addition, it would provide a globally
unique dataset on spatial patterns of biodiversity, which could help resolve
many contentious scientific issues.


Discover Life in America | Science | Inventory Design | Huston - 17 December, 1998