[Coral-List] Predictive modelling of coral distribution

Louis Celliers louis at ori.org.za
Wed Mar 3 01:44:16 EST 2004

Greetings, coral listers
South African coral communities are considered to be high-latitude
assemblages growing exclusively on submerged quaternary sandstone reefs at
the limit of coral distribution on the coast of Africa (27deg Lat).  Reefs
are conveniently grouped in "reef complexes" (North. Central and Southern),
all within a 150km stretch of the exposed coastline of Maputaland (long-axis
from N-S).  Reef areas vary between 6ha to as little as 0.1ha.  Total cover
ranges between 10 and 90 % with the majority of communities (35-55 % total
coral cover) being either soft coral dominated or mixed soft/hard ("typical"
WIO coral community, ~110 hard coral species ).  We have been surveying
these reefs for the last 5 years and we have been using primarily digital
video/still cameras to collect coral data to generating cover and
composition (i.e. mainly pattern).  I am comfortable with the sampling
method  (point intercept counts from high-res digital photography) but my
question relates to sampling design and data extraction and interpretation.
The results from our surveys are thus a number of sampling points spread out
across the reef in what we calculate to be a representative sampling design.
I apologise if my story is a bit long-winded but the background is
important:  My question/issues are thus:

1)   We have created a grid overlay each reef and I need to populate the
grid (each individual grid) with species composition, community type,
susceptibility indices etc and I have intuitively chosen a grid size of
100mx100m.  The intuition part of it is obviously going to be problematic
once this data has to be published.  What would be the rationale for
selecting a grid size.  I have been toying with the idea of using mean
longitudinal distance (E-W; mainly changes in depth, also traditional coral
zones) between sampling points as measure of resolution.  It is obvious from
the distribution of our sampling points that we favour sampled across the
reef (E-W) in "groups" from N to S; and it is clear that we interpret and
make coral community similarity assumptions along bathy lines (N-S). It is
then possible to calculate the vertical resolution (E-W) of our data and use
that as a justification for grid size.
2) Which brings me to my second point.  Obviously we have to fill the "gaps"
and develop some sort of a predictive model to generate and justify assiging
non-data blocks with predictive data because you can't show a manager a reef
area with only 30 grids populated with data.  Again I am comfortable with
our sample design but the rationale for linking between sampling points
remain too subjective (mainly based on bathymetry).  Can anyone possible
contribute any good references or experiences for
populating/calculating/modelling the gaps, assiging confidence to grids etc.
Thanks for actually reading this!

Dr Louis Celliers
Senior Scientist
Coral Reef Programme
Oceanographic Research Institute - Durban
Telephone: +27-31-3373536
Facsimile: +27-31-3372132

More information about the Coral-List mailing list