Peter White - 3 September, 1998
Inventory Draft Plan
Date: Thu, 3 Sep 1998 10:52:50 -0400
From: "WHITE, PETER S."
To: keith_langdon@nps.gov, pick@pick.uga.edu, wfharris@utk.edu,
chuck_parker@nps.gov
Subject: Inventory Plan
Dear Pick, Keith, Frank, and Chuck:
Keith asked that I send a straw man outline for an inventory plan as a
central framework within ATBI. Below I present a draft. It is longer
than it need be and definitely rough--just treat it as a starting point.
Pick and Keith, your idea was to share this with a small group and then
use if for discussion when I come to the Smokies in two weeks. Feel free
to forward this on to others that should be involved now.
Peter
=========================================================================
Principles for an Inventory Strategy: a central thesis for a biological
inventory of the park's landscape
=========================================================================
I. Rationale:
This framework has the following objectives:
1. To Organize the diverse environments and historical influences of the
Park so that collectors can array their sampling effort efficiently
around the variables known to influence species distributions.
2. To Provide an ecological and management context for the inventory
work to maximize the usefulness of the information collected in building
understanding of the park's ecosystems, detecting change, and better
managing the biota against unwarranted loss of diversity.
3. To Establish, at least for some portion of the work, an ability to
recover species-effort, species-time, species-individual, and/or
species-area curves so that we can:
(a) develop certain and repeatible data for small scales (as we go to
finer units of resolution, we have more certainty in how many species are
present) and
(b) make more informed judgements of how complete a given checklist is.
In accomplishing these objectives, we provide a new model for how
classical biological inventory should be done.
At the various meetings I have attended (Gatlinburg, White House, and
Cullowhee) there has been general enthusiasm but lingering doubt.
The enthusiasm has centered on the pure joy of producing the first
complete inventory of biota in North America in a very rich landscape,
thereby demonstrating how important taxonomic knowledge is to human kind,
and the excitement of including school children, the public, and
scientists in a collaborative effort.
The lingering doubt centers on the last two of the objectives above:
namely, that the information will not be useful regardless of how
complete it is , that we will have no idea how complete the information
is, and that the project either can't be done because of the
impossibility of finding every last species or because it will be
prohibitively expensive.
The framework I propose counters these lingering doubts. It establishes
a central framework for insuring that the information is useful, it
develops error bounds for how complete the knowledge is (and develops
narrower bounds at smaller units of sampling effort), and allows
forecasts of how much effort (or dollars) are needed to develop knowledge
at various levels of resolution.
II. A note on freedom of choice, the individuality and evolution of
methods and approaches for each discipline, and the amount of time,
money, and effort that different investigators bring to the project
This framework in no way removes the freedom of choice for investigators
in terms of where they go in the park, what methods they use, and how
much time, money, and effort they expend.
At one level ATBI should just be a project that advertises and
facilitates the work of any and all investigators who apply for a
collecting permit. The investigator might be an international expert
spending two days in the Smokies as part of a tour of North America. The
investigator might be leading a week long collecting expedition of
graduate students or undergraduates. The investigator may be an MS
student doing a checklist of organisms for several watersheds or a
particular habitat. The investigator may be a grad student monographing
a particular group of organisms in the Smokies or across the southern
Appalachians.
I believe that ATBI should welcome with open arms and encourage any and
all taxonomists to participate in the effort over the years. If
possible, ATBI should help find housing or camping permits, should supply
maps and even GPS units, and should provide access to what others are
already doing in the park (to avoid unnecessary dupication of effort and
maximize the gain of new information). In return, ATBI should be assured
of getting copies of information collected and indexes of the location
and identities of museum specimens. I will argue below that ATBI could
go further by providing information on the park's environments and
history (a "Users Guide" to collecting in the Smokies) that will help
investigators decide where and when to collect and may be able to channel
the effort so that it will be of maximum benefit to ATBI itself. The
"Users Guide" would present guidelines whose use would be voluntary on
the part of the investigators. As developed below, the guidelines would
include (a) a stratification of the variables that inflluence the
distribution of biota in the park, (b) locations and maps of suggested
landscape reference areas for collecting, with biodiversity reference
points located within those, (c) suggestions for the species-effort,
species-time, species-area, and species-individual sampling, and (d)
requests for basic information about each specimen collected or
observation made. The last of these (d) is not covered in this draft--it
is for the TWGS to devise. The framework described here would insure,
however, that (d) included information on the ecosystem type,
environmental and historical context, and spatial context of the
observations. We also need to make sure that the temporal context of the
specimen can be recovered--not just date, but if the specimen was part of
a particular expedition or series of specimens collected.
But, at the very least, ATBI should be part of a welcoming spirit with
regard to taxonomists willing to observe in the Smokies--be it for a
major research effort or a single day on a single stream in Cades Cove.
We should facilitate the events of observation and we should capture
these for the overall data base.
Besides this welcoming, recruitment, and facilitation, though, ATBI might
want to fund or collaborate in funding for some portions of the
inventory. As a result, ATBI can play a role in specifying more
completely how, when, and where the inventory was done.
Keith has suggested levels of knowledge:
(1) discovering what species are in the Park,
(2) ascertaining distributions (several levels of resolution are possible
here), and
(3) collecting other natural history information like relative abundance,
association of species, and phenology.
We generally want to do as much of (1) as possible, but as we proceed
from (1) to (2) and then (3), we will probably have to target specific
groups for more detailed information and we should develop the funding
that will guarantee that the target groups make it all the way to (3).
In a similar way, I see the following as a series of guidelines:
(1) Users Guide: what factors control the distribution of species in the
park
(2) Recommended areas for collecting: places to go for representative
pieces of the park's landscape for collecting, chosen to represent the
factors described in the User's Guide within information on
accessibility, guidelines on assessing the impact of collecting, and
(3) Controlled inventory: places and methods for developing detailed
knowledge and links to monitoring.
Just as with Keith's list, I see (1) and (2) being useful to everyone,
regardless of funding source and regardless of how many person-hours they
spend in the park. I see (3) as applying only in some cases--and
probably requiring that ATBI itself find the funding or collaborate on
the funding to make sure that it happens for target groups. Even if we
funded individuals to work in the Landscape Reference Areas, however, we
still ought to give them some portion of the time to wander at
will--using human intuition to search out novel areas and document those.
We could even test whether human intuition is effective in discovering
interesting unique biodiversity hotspots for particular taxa.]
III. Organizing principles for biological inventory
The following is the draft I sent last December. I'll embed new comments
in brackets "[ ]".
In the rest of this message I will step down a series of spatial scales
from Landscape Reference Areas to Biodiversity Reference Points to some
thoughts on fixing the observational windows at the points (I have
sometimes called these Ecological Observatories to capture the excitement
that people feel about astronomical observatories). My idea is that,
while much of the inventory can be unfettered, the park and researchers
would both benefit from the establishment of the template--a series of
fixed control points for inventory. These would also have substantial
benefit to the park in other areas (e.g., mapping vegetation and
monitoring ecological change).
1. Principle 1: Stratification by history and environment
It is clear to everyone--even researchers working on their own--that any
collecting or sampling has to be organized against the environmental and
historical variation in the park. That is, that we can stratify the
large and diverse park landscape by community type, elevation, slope
aspect, slope shape, geology, history (2nd growth, old growth, burn
scars), and geography. You also mentioned this at the workshop and I
know it is on Keith's mind, too. This stratification would also allow
us to compare observations as a function of the variables--e.g., logged
vs. unlogged sites at a given elevation and topographic class.
[The stratification variables also are correlated with primary
productivity o the ecosystems and thus also help with the overall
ecological context. A notebook that desscribed the variation and how it
might be stratified for inventory is what I've called the "User's Guide
to the Smokies Landscape"; this notebook could also present
recommendations for Landscape Reference Areas that represent the
landscape types of the Smokies--that is recommended places to visit that
will cover the stratification parameters presented--see the next
principle.]
So the first principle is to use a GIS to organize the
environmental/historical variation--to create stratification cells within
which to sample or collect. We have a copy of the GRSM GIS here at
Chapel Hill and could run alternative scenarios for such a
stratification. We have done an exercise like this for TNC and the Park
in conjunction with the TNC vegetation mapping efforts now underway.
Principle 2: Landscape Reference Areas for Gradients, Spatial Context,
and Spatial Dependence
[Volunteer taxonomists, in addition to having the User's Guide as
background, might welcome suggestions for accessible places that
represent the park's environmental variation--thus they might very well
organize their work around covering the park's elevation range,
disturbance history range, geological range through the User's Guide, but
they might also welcome recommendations of specific trails, places, and
stream segments that represented that variation. If we develop funding,
then we might require visits to Landscape Reference Areas even if we also
allowed free wandering for some portion of the effort. Landscape
Reference Areas contain representative points of the stratification
blocks--for example, spruce-fir forest at 6,000 ft--but they also array
these along explicit spatial gradients, so that collecting along a
gradient transect would also present the opportunity to document species
presence along gradients.]
[Landscape Reference Areas would incorporate Representative Stream
Segments; if this were judged too much of a compromise, then I'd
recommend separate Landscape Reference Areas and Reference Stream
Segments.]
I would argue the following: for practical reasons, scientific reasons,
and management reasons, the points that represent the cells in the
stratification matrix should not be randomly and independently
selected--rather they should be clustered in space in areas Landscape
Reference Areas (within those areas they could be randomly chosen).
Here is my rationale:
1. Practicality: Hiking to a 6,000 ft, old growth cell, on a
N-facing ridge takes time. If a sample for a 6,000 ft old growth N-facing
slope in a cove is nearby, then two cells can be hit on the same trip.
Also, the Landscape Reference Areas can be documented in terms of information
that will useful to researchers when they analyze their results--e.g.,
community types, vegetation maps, disturbance histories. This will also
aid producing precise label data for specimens.
2. Science: Everywhere you go in the park you are on gradients
of elevation and moisture. One of the dominant features of the park is
that as you walk, you are shifting your position relative to important
variables. There is a turn over of species along these gradients. If
the Biodiversity Reference Points are clustered within Landscape Reference
Areas (rather than being each randomly selected in the landscape), we will
be able to look explicitly at spatial gradients. There are other spatially
dependent processes of interest-- for example, dispersal or a species that
moves regularly between two different kinds of habitat patches during daily
or seasonal movements.
3. Management: Climate change and other sources of biological
change are also likely to cause shifts along gradients. There are other
spatially dependent processes of potential interest such as spread of an
exotic pest. Also, some monitoring efforts will be expensive--e.g., air
pollution monitoring--so that it would be better to intensify the sampling
where that monitoring is taking place.
Gradient response can be reassembled from random points, of course, but I
think that we gain by being able to look at real spatial gradients rather
than reassembled ones.
My point for now is that we should blocks of the park landscape as part
of the template. These would represent the Park's distinctive landscape
types.
How big would the Landscape Reference Areas be? Well, now I am
speculating--and I would want to run some scenarios on the GIS--but let
me say for now that they might be no smaller than 1 square kilometer and
maybe as large as 10 square kilometers (3.1 km on a side). Each would
incorporate local gradients and allow access to a variety of habtiat
types--a variety of cells of the stratification matrix. They would be
sizeable chunks of the park's landscape. They would include reference
stretches of the park's rivers and streams.
How many distinctive landscape types are in the park and How many
Landscape Reference Areas would be selected? A guestimate: 25-50.
There would be replication within this total, for example: 3 Landscape
Reference Areas in the distinctive western pine landscape west of Cades
Cove (these would incorporate hemlock and a variety of hardwood types)
and 3 Landscape Reference Areas centered on parts of the spruce-fir high
ridge (these would include beech gaps, northern hardwoods, cliffs, heath
balds, and other habitats).
3. Principle 3: Biodiversity Reference Points
[The kind of ecosystem at the locale of collecting is the ecological
"address" of the species and specimen--and may be more important than
latitude, longitude, and date. The points would be both terrestrial and
aquatic.]
Within these Landscape Reference Areas, Biodiversity Reference Points
would be chosen. These would be fixed points within the Landscape
blocks. A permanently marked and monumented vegetation plot would be
established at each Biodiversity Reference Area. Plot size could be
fixed at 1 ha or could vary (0.1 ha in low diversity, high density
vegetation like fir and pine forests and 1.0 ha in high diversity, low
density vegetation like old growth cove forests).
How many Biodiveristy Areas within each Landscape Reference Area? Again,
a guestimate (with the idea that we use GIS to run scenarios): 10-50.
We now have 250-2500 fixed points as an inventory template (if we
multiply the Landscape Reference Areas times the Biodviersity Reference
Areas).
Not all collectors would have to visit each Riodiveristy Reference Area.
Some collectors might concentrate limited sample effort only at one end
of the variation (e.g., high elevation spruce-fir landscape reference
areas), whereas others might select some random subset of the
Biodiveristy Reference Areas for inventory. But we would have created a
monumented set of fixed reference points across the park's landscape.
Within the Landscape Reference Areas and Biodiversity Reference Points,
all collectors would be encouraged (and in fact recruited) to overlap
their work. They would also be given encouragement to wander throughout
the park and to wander between the Biodiveristy Reference Areas within
the Landscape Reference Areas. But there would be several advantages to
collecting specifically on the Plots: the GIS and field instruments
would be used to supply environmental data, the history would be
documented and known (logging, fire), they could trade ideas and samples
with other taxonomists working in the same areas, we could supply a
unique identifier for collecting locales, simplifying the writing of
specimen labels, some work could be done by volunteers or technicians
when the PhDs were back at their home institutions, and the inventory
could be repeated at different seasons and in different years or even
different decades--the fixed reference points allow the inventory to be
repeated with more precision than is usual in taxonomic surveys.
We would also have established the explicit ecological conceptual model
discussed at the workshop.
I should point out that the collectors need not be limited to the
vegetation plot: bird people might traverse a larger area around the
plot. Destructive sampling areas, parallel to the permanent plots, might
have to be established for those who do destructive sampling (e.g., roots
for soil fungi). And I also add a note for aquatic researchers: the
Landscape Reference Areas would also contain stream sections, as well as
terrestrial vegetaiton plots as fixed Reference Plots.
We could also have remote sensing images for each reference area--and we
could map the communities at a series of spatial scales for those
reference areas. Although TNC is currently working on a detailed
vegetation map for two quadrangles, extending the map to the whole park
is more uncertain. Extending the map will depend heavily on interpreting
aerial photos, whereas TNC is also analyzing on the ground plot data.
Only some of the information needed for the TNC classification system is
visible from the top of the vegetation. Also the scale at which you map
influences the mapping units and the variation you depict on the map
(e.g., the shape of the spruce-fir ecotone varies with map scale). By
contrast with this uncertainty, the Landscape Reference Areas and
Biodiversity Reference Points could be mapped at a series of scales.
They would serve as on the ground control points for the whole park map.
I add one more justification for fixed reference areas: the bigger the
area and the more diverse and cryptic a group, the less confidence we
have in estimates of "completeness" for a checklist. Even though we will
be generating checklists for the whole park, we will be uncertain how
good these are. By contrast, we can achieve more certainty on the
reference areas: using species-samplng effort curves and rarefaction
(but see also next section), we can be more precise about what we know
based on observational effort. This is similar to the work you presented
in Gatlinburg on species richness versus the number of individuals
collected and examined. We can more fully report our confidence in a
checklist for the Biodiversity Reference Points than for the park as a
whole. So these fixed points become valuable, stand alone, indexes of
diversity. I think this is our only hope of achieving "stopping rules"
as well. For really cryptic taxa like some fungi, nematodes, and soil
bacteria, working just on a set of Biodiversity Reference Points may be
the only feasible way to get a start.
[Impact of collecting would be documented for biodiversity reference
points in order to make sure that incompatible collecting activities and
prior collecting doesn't bias results; several replicate biodiversity
refrence points would have to be available.]
4. Principle 4: Scale dependence, grain, extent, and the species-area,
species-effort, and species-individual curves
[We could add species-time to the list above, as one way of measuring
observation effort. Keith reports that Fred Coile does his spider work
by timing observations.]
"Scale" tends to make people become glassy eyed, so let me step through
this as simply as I can (the underlying idea is a very simple one):
Let's divide "scale" into two components: grain and extent. Grain size
and extent influence the number of species observed. We can think of
gain and extent in either spatial or temporal terms.
Grain is the unit of resolution--e.g., what constitutes a single
observation. Grain size is the size of a quadrat you make a species list
for. Or the size of a pixel in a GIS or remote sensing. Or the duration
in time of an observation on what birds are singing or the duration of
time an insect trap is open and collecting insects.
Extent is the spatial area or total time span over which a series of
observations are made.
Both grain and extent influence how many species you observe or collect.
I usually think in terms of quadrats and space, but I'll use a temporal
example to emphasize generality. For example, with 24 hours of total
effort to observe birds, you'd get different numbers of birds if you
observed one hour on one day for 24 weeks or one hour on one day for 24
days. The two observations have the same grain (one hour), but different
extents (24 weeks versus 24 days). The same 24 hours could have been
split up other ways--for example, 2 hours every other day for 24 days.
Now the grain size has changed (2 hours versus 1 hour), but the extent is
still 24 days.
Obviously, the number of species you collect is a function of both the
grain and extent of the observations. Increase in extent usually builds
species number faster than increases in grain--increase in spatial or
temporal extent means that you are encountering more environmental
conditions, more weather changes, more seasonality, or more dispersal
events.
If we collect species data at a series of different grain sizes and
extents (as we did in the American Naturalist paper), we can draw curves
that represent the change in species number with change in grain and/or
extent. Those curves depict the "scale dependence" of the
observations--how the observations change with scale.
Are you still with me? At this point I argue that it is the shape of the
scale dependence curve that is important.
At a one square meter scale, the coastal plain savannah in NC is richer
in plant species than a tropical rain forest, but at a square kilometer
scale, the tropical rain forest is richer. There are a lot of small
plants packed in a square meter in the savannah and tropical trees are
large--not many individuals fit in a square meter! On the other hand,
each one square meter across the savannah is rather similar to the
others--spatial extent does not contribute as much to the species list as
it does in the tropical rain forest.
At some grain size between one square meter and one square kilometer, the
tropical rain forest and the savannah will appear equal in species
number.
My conclusion is that what is different about the savannah and the rain
forest is not the difference in species number at an arbitrary vertical
slice through the species-grain curve, but rather what is interesting is
that the species-grain curves are different! I think we should be
comparing the scale dependence curves and not the total number of species
as some arbitrary grain size.
I would make the same argument in time: if global warming or an exotic
species invasion has an effect on species richness, I'd like to compare
the full curves. I don't think that we know how diversity will change at
different scales and I don't think there will ever be an ideal scale,
even for one taxonomic group. I think that the curves are useful in
extrapolating from samples to the potential total species number, but I
think they have an even more important use in describing the way species
number is distributed. I think the curve is fundamental to being able to
detect change.
This is all the more a problem when sampling methods are not sufficiently
described. This is why I argued that collectors need to describe their
observation window in has much detail as possible--even if it consists of
simply the date, duration and spatial extent of their wanderings.
So, at the Biodiversity Reference Points, I would establish a nested set
of boxes within which species lists would be made. This is incorpoated
in our NC Vegetation Survey plots. The strict nesting of boxes of
different sizes allows us to draw the species-grain curves I've
suggested. Not all inventory is done on quadrats or plots--so for other
researchers I would simply argue for the specification of the analogs for
grain and extent in their samples. For some this would be quantitative
(light traps set out on a spatial grid, with speciments collected at
fixed intervals) and others qualitiative (a report on a week spent
looking for macrofungi on a series of trails).
[I also argue that the smaller the unit of effort (a square meter vs. a
hectare; an hour vs. a day), the more exact and complete the knowledge
(species presence), all else being equal. In 50 years scientists may
repeat the ATBI to see how much is changed, but they will be most certain
about change at smaller scales along gradients than they will for the
park as a whole.]
===========================================================
Peter S. White email: pswhite@unc.edu
Department of Biology -- CB# 3280
University of North Carolina at Chapel Hill
Chapel Hill, NC 27599-3280 USA
Biology Phone: 919-962-6939 Biology FAX: 919-962-1625
NCBG Phone: 919-962-0522 NCBG FAX: 919-962-3531
Home Phone: 919-967-4926
Web information: www.unc.edu/depts/biology/white.html
===========================================================