Two methods for percent cover monitoring

Wed Oct 24 17:47:58 EDT 2001

Hi Victor and others,

We thought a lot about sampling units when designing the ReefBase Aquanaut
Method (the method is described on the ReefBase Version 3 CD-ROM). I'm
pleased to put down a summary of those thoughts here for anyone interested,
to initialize the discussion you suggest.

The first principle is that the sample design is generally more important
than the sample unit. Pielou suggests that the sample unit should generally
be larger than the focal objects included in the space one is sampling.
Thompson (Sampling, Wiley) mentions that little has been done on optimizing
sampling units, though a little work favors donuts. There is a lot of work
on "how long should one's transect be?" in the coral reef literature.
However, the analyses sometimes have led me to wonder if someone was
confusing sample units with samples. Thus, a "curve-off" of accumulated
sample units makes more sense than a "curve-off" in length of a single
sample unit. Note that a long transect can impel field workers to sample
areas you might not mean to include in your sampling universe, such as areas
beyond the reef substrate. Some people also seem to confuse sample design
with experimental design. The best books I have found on sample design are
those by Cochran (Sampling Techniques) and Thompson. The latter is more
recent, and has a lot on area sampling (like what most benthic ecologists
do), whereas most texts describe object sampling. Simple random, stratified
random, systematic, nested, clustered, adaptive and other strategies are all
legitimate and each has advantages and disadvantages. The confusion between
sample units and samples seems to stem to phytosociology (Braun-Blanquet,
Ellenburg, etc.), from which most of our benthic approaches originate, but
in which a single sample was often chosen to "represent" an otherwise
subjectively delineated ecological community. Even that has utility for some
kinds of classification and ordination studies, but it is not useful for
inferential statistical analyses -- those involving estimates of variance.

Also, one must distinguish between one-shot sampling and monitoring over
time. In the latter case, there is a general debate over whether one should
use the same spots, the same general sites or simply design a new sampling
pattern for each repetition. I am building simulation models to look into
aspects of that debate, as the answer is not at all clear to me.

Line transects have the advantage that they can be used horizontally on
vertical slopes with minimal difficulty. Quadrats often fall down. I used to
like area-based (as opposed to point-intercept) quadrats because we (various
colleagues and I) could use them rapidly, thus giving us a larger sample
size representing variance and mean over a larger area. However, we decided
that this was simply because we were using large divisions (typically nine
squares in a 1 sq. m quadrat). So, we decided that there was no reason that
a line transect could not be small (5 m) and that it could not be divided
into 50 cm or whatever as were the edges of a quadrat. That gave us
something handier than a quadrat (a cord marked with fishing weights that
could fit in a pocket), nearly as quick to use as a quadrat and not as
likely to weigh us down or fall off a wall. However, there is still room for
work on how the bigger swaths affect the estimations. For many patterns of
coral cover, presence/absence in a 50 cm swath does not vary much over large
areas than presence/absence in a 1 cm swath (i.e. measuring to the cm).
However, the larger unit would not quantify small non-coral spaces as well.
On the other hand, two people often vary in reading seagrass to the cm, but
usually read the same way at the 50 cm level. How either approach relates to
point-intercept transects or quadrats is even more puzzling.

For permanent transects, Jim Maragos has a nice system of drilling in
stainless steel bolts, and then resuspending a very tight line for each
sampling. There again, the longer the line, generally the more it will shift
among times. A shift of a few inches can vary the estimates enormously.

Kinzie III and Snyder had a good article in the 1978(?) UNESCO Coral Reef
Sampling Handbook. It showed in simulations that a low levels of coral cover
(as in their particular study areas in Hawaii), many sampling units were
equally bad (quadrats, line intercepts, point-intercepts). The nice thing
about a simulation is that you know what the "real" answer should be. The
errors in their estimates were very high. They concluded that one should
seek to use units that were small and rapidly deployed, and could be
scattered widely across a reef. I'm following up on their simulations, but
so far see nothing to change from that conclusion.

Cheers!

John

_________________________________________________________

John W. McManus, PhD
Director, National Center for Caribbean Coral Reef Research (NCORE)
Rosenstiel School of Marine and Atmospheric Sciences (RSMAS)
University of Miami, 4600 Rickenbacker Causeway
Miami, Florida 33149.
jmcmanus at rsmas.miami.edu
Tel. (305) 361-4814
Fax (305) 361-4600
www.ncoremiami.org

-----Original Message-----
From: owner-coral-list at coral.aoml.noaa.gov
[mailto:owner-coral-list at coral.aoml.noaa.gov]On Behalf Of Gomelyuk,
Victor
Sent: Tuesday, October 23, 2001 8:51 AM
To: coral-list at coral.aoml.noaa.gov
Subject: Two methods for percent cover monitoring

Dear colleagues,

I wonder if anyone would like to share with me and other coral-list memebers
ideas of advantages/disadvantages of using Line Intercept Transects and
Permanent Quadrats methods for coral environment monitoring. Refferences on
any literarure sources will be also appreciated.

Thank you.

Dr Victor E. Gomelyuk
Marine Scientist
Cobourg Marine Park
PO Box 496 PALMERSTON NT 0831 AUSTRALIA
phone 61 (08) 8979 0244
FAX 61 (08) 8979 0246

~~~~~~~
For directions on subscribing and unsubscribing to coral-list or the
digests, please visit www.coral.noaa.gov, click on Popular on the
menu bar, then click on Coral-List Listserver.