I have received several interesting responses from readers and I agree with
what they all have to say about sampling philosophy. I agree that relevance,
consistency, and context are all important, but I haven't even started on those
potential problems with this protocol. My first question concerns whether the
protocol insures that data are accurate and precise without regard to whether
they are relevant. However, I think I may not have clearly stated my question
regarding this. The protocol being used does measure cover for all plant
species (herbaceous and woody), based on point data along a transect, i.e., if
50 of 100 points touches species x, then species x has 50% cover. I have no
problem with that part of the protocol, although it is very time consuming to
gather only 100 points of data, the data can be collected subjectively with
little observer bias. My main concern is with the second part of the protocol,
where woody species density is determined. In that protocol, woody plants
rooted outside a 100m x 6m plot are being tallied, when any part of the plant
overtops the plot. It seems to me that including those plants would
over-estimate woody species density, since density is the number of plants per
unit area (with area defined by the plot boundary). For example, if 50 stems
(trunks) of white oak are rooted in the 600m2 plot, there are 833 stems/ha
(50*10,000/600). However, if an additional 25 stems rooted outside the plot are
included, then density would be 1,252 stems/ha (75*10,000/600). So, even if
this protocol is applied consistently, the answer will be consistently wrong
and fraught with sampling error due to the difficulty of large, tall, far way
trees being consistently included in counts.
Including plants rooted outside the plot when determining cover (proportion
ground covered) would not be problematic, but cover isn't being measured in the
second part of the protocol.
-----Original Message-----
From: Ecological Society of America: grants, jobs, news
[mailto:[email protected]] On Behalf Of Wayne Tyson
Sent: Sunday, June 02, 2013 11:04 AM
To: [email protected]
Subject: Re: [ECOLOG-L] plot sampling for density
To RELEVANCE, RELEVANCE, RELAVANCE, I suppose should have added "CONTEXT,
CONTEXT, CONTEXT, and I thought I make it clear the PURPOSE, PURPOSE, PURPOSE.
IF, and only if, one is restricting one's questions to some subset of the
whole, one can simplify (actually reduce) the kinds and numbers of factors
observed and recorded.
As to edge-effect "error," I would add CONSISTENCY, CONSISTENCY, CONSISTENCY. How much does it
matter whether or not something is counted that is partly "in" and partly "out" as long
as it is consistently done and is an actual stated part of the protocol? These are ESTIMATES, they are
SAMPLES, and nothing more. Now, how GOOD a sample, how truly representative it is of the whole or how well it
reflects variations and changes--that's a matter for clarity in the research design and its claimed results
and the confidence in conclusions--find THAT in most research designs.
How many study results, how many conclusions, should be replicated by others
and carefully examined by other researchers before we place confidence in them.
And how many should be thrown out or done it the first place?
WT
----- Original Message -----
From: "David M. Lawrence" <[email protected]>
To: <[email protected]>
Sent: Sunday, June 02, 2013 5:17 AM
Subject: Re: [ECOLOG-L] plot sampling for density
If you're working in the mixed-grass prairie of Oklahoma or the Chihuahuan
Desert of southern Arizona, Wayne, I would agree that one should include
grasses and other small-stature species -- because they are the dominant
vegetation in the area.
In forests, though, trees are typically the dominants in terms of stature,
biomass, consumption or capture of resources such as light, etc., so that
those other species may not need to be measured. (That is not to say
their presence is irrelevant, just that it may not be as important to
those trying to characterize the system.)
Rick, in most of the work I've done in forests, we use tenth-hectare plots
and count, measure the DBH of all trees rooted inside the plot, and count
saplings (using a pre-determined size criteria) within the plot overall or
within smaller subplots.
If your metric is based on canopy cover, however, then what you are asked
to do sounds reasonable, even though I agree with you that it would be
fraught with error. I have enough time standing straight up --
I usually stand and walk with a persistent list -- and things get really
screwy when I've bent my neck to look vertical rather than horizontal.
The method you are asked to use may be based on of those good theoretical
ideas gone bad. I remember inhaling when I saw some scholarship that
circular plots are better than rectangular plots because they minimized
edge error -- by minimizing the amount of edge per area of plot. It
sounded great until I got in the field and had to figure out where the
edge was when deciding what was in or out. It's hard to do when your edge
is a continuous curve. I've been committed to rectangular plots from then
on.
Dave
On 6/2/2013 12:55 AM, Wayne Tyson wrote:
Rick/Ecolog:
I think you are absolutely right to question these procedures; in fact
your post set off so many bells in what's left of my mind that it looked
like the Fourth of July and New Year's combined! I hesitate to make any
remarks at all unless there is enough interest to get into a lot of
specifics, and I'm sure there would be a lot of disagreement amongst
subscribers, as "sampling" procedures are the Holy Grail for a lot of
people.
You do not want comments on anything but density, and I think you are
right on this too, but I again hesitate to comment because it would take
a long time and exchange of emails to resolve the sticky issues that will
arise. Suffice it to say that a lot of "protocols" have, if any,
foundations that are highly suspect to me; hence, I share your instincts
(which, I suspect, are only a sample of the entire thicket of stickery
issues.
I believe that it is simply lazy, if not downright fraudulent, to leave
"minor" (grasses, cryptobiotic communities and species) out of most
studies--unless the purposes of the study are stated up front to be
looking only for data on a limited fragment of the ecosystem. Apart from
that, I believe that there is often little value to any survey data
without long-term replication that would reveal at least interesting
changes and trends, but how many actually do that. And when they do, what
is actually DONE with the data? All too often it seems to me to be in the
realm of "employment act" stuff . . .
I will say only this for now: RELEVANCE, RELEVANCE, RELEVANCE!
WT
PS: I ran many, many miles of transects during my brief Farce Service
"career" back in the last century, and I think we produced some very good
data and mapping, with little relevant "error." I re-visited my old
stomping grounds a few years ago, and was kindly allowed to poke around
in the storeroom of the Supervisor's Office, where I found our old field
notes and maps (apparently undisturbed, and probably un-analyzed).
As to density--like "cover," I don't think it reveals much except when
done over time like your case, it might provide interesting data on the
maximum productive (carrying) capacity potential for the vegetation being
studied, as well as recovery times following perturbations like fire and
logging.
----- Original Message ----- From: "Rheinhardt, Rick"
<[email protected]>
To: <[email protected]>
Sent: Saturday, June 01, 2013 4:43 PM
Subject: [ECOLOG-L] plot sampling for density
I recently started re-sampling vegetation in fixed plots on a U.S. Army
base. This base, and I presume many others in the U.S., use a standard
protocol for collecting vegetation data. There are many plots on this
base that are marked with benchmarks so that they can be re-sampled at
intervals of years to decades. I presume that the data obtained from
these plots will be used to monitor vegetation changes (structure and
composition) through time. One aspect of the sampling protocol is a
straightforward line-intercept method: a 100-m tape line is laid out in a
straight line and the height and species of all plants that touch a
vertical rod are recorded at 1-m intervals along the line. This method
provides a fairly objective measure for cover but cannot provide any
information on density. A second aspect of the protocol is designed to
obtain density data for woody species > 1-m tall. This protocol
essentially involves delimiting a 100-m-long plot using a range pole to
determine the width of the plot, walking along the 100-m tape line from
one end to the other, and recording woody plants, by height category and
species, within the pre-determined horizontal distance delineated using
the range pole. Usually, the predetermined distance (plot width) is 6 m,
which delineates a 600-m2 plot (6 m x 100 m). Horizontal distance can be
altered, based on perceived stem density.
The question I would like to submit to LISTSERV subscribers concerns the
methods used to apply the density sampling protocol. In all the plot work
I have done, I have always recorded only woody plants (stems in the
vegetation ecology vernacular) that are rooted within (or mostly within)
the fixed plot of interest. That is, plants rooted outside the plot, but
with canopy overhanging the plot boundaries, are not counted. However,
the protocol we have been asked to apply involves also recording plants
whose canopies overtop the plot even though they are rooted outside the
plot. I believe that counting plants rooted outside the plots severely
compromises both the accuracy and precision of the data, i.e., accuracy
is compromised in that the plots are no longer of a fixed size, and
precision is compromised in that there is much room for observer error
when determining whether canopies from large far off trees are
overhanging the plot (because the observer has to be in the middle of the
plot to hold the ranging pole in place).
If we were measuring cover, then it would be immaterial whether a plant
were rooted inside or outside of a plot, since canopy overtopping the
plots would be the parameter of interest. Part of the confusion may be
due to the terminology used in explaining the protocol. The protocol says
that woody "stems" are to be recorded in the plot. To me, the term "stem"
refers to the main stem (trunk for a tree) that directly attaches to the
roots, but I think the term may have been misinterpreted to include
branches and secondary branches of plants.
My concern that the density data we collect will be a nightmare to
interpret, and worse, will not measure what it is intended to measure.
Unfortunately, in searching the web, searching papers, and even looking
through plant ecology texts, I have not found any guidance concerning
what plants should be counted in plot work (plants rooted outside vs.
insides of plots). Is this because protocol writers assume that everyone
knows how to do it? Could there be there a potential problem with density
data in the peer-review and/or gray literature? How much of a problem
could misapplied protocols be having on data collected by natural
resource programs? Should the word "stem" be defined every time it is
used in describing a protocol?
Rick Rheinhardt
ECU
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 10.0.1432 / Virus Database: 3184/5874 - Release Date: 06/01/13
--
------------------------------------------------------
David M. Lawrence | Home: (804) 559-9786
6467 Hanna Drive | Cell: (804) 305-5234
Mechanicsville, VA 23111 | Email: [email protected]
USA | http: http://fuzzo.com
------------------------------------------------------
"All drains lead to the ocean." -- Gill, Finding Nemo
"We have met the enemy and he is us." -- Pogo
"No trespassing
4/17 of a haiku" -- Richard Brautigan
-----
No virus found in this message.
Checked by AVG - www.avg.com
Version: 10.0.1432 / Virus Database: 3184/5876 - Release Date: 06/02/13
