CHAPTER 3. M O S

CHAPTER 3. M O S
CHAPTER
In order
viewed
to achieve
from a basic
processes
than
to
of
basic
ity
were
in
only
the
have,
of the
therefore,
been
and
components.
sequence,
method
The
methods
method,
(Coetzee
over a ten year period
added
to
published
build
within
1975).
rather
that
that
the
The
was
of the
classification,
techniques
Braun-Blanquet
and refined
method
in terms of the components
sampling
selected
Braun-Blanquet
developed
aims, the Braun-Blanquet
perspective,
application·
elements
of
the stated
stratification,
view
describe,
3. METHODS
follow
on
the
flexibiltechniques
and improvements
versions
where
applic-
able.
The
basic
elements
familiarization
species
form with
of
species
cover-abundance
into 25 contiguous
The
Braun-Blanquet
method
are
to a formal or an informal
sampled
the two 10 x 10 m plots
only were
the
leading
presence,
classification
of
in
quadrats,
(rows)
values
and
samples
forming
in the second
2 x 2 m sampling
cover
vegetation
stratification,
estimation
(columns)
the matrix.
in matrix
Accordingly,
study area were
units, wherein
and
each divided
species
presence
recorded.
components
for convenience,
of the
stratification
be grouped
according
and
sampling
to preparatory
processes
can,
work and field
sampling.
24
3.1
PREPARATORY
Preparatory
commence.
WORK
work
This
probably
includes
the most
appropriate
pretation
achieve
methods
linear
Bureau,
sampling
will
Scale
Scale
determines
particular
study
can
is
aspect of a study because
selection
of
on the aim. Results
6e dependent
on the methods
of rainfall
annual
and the inter-
work also included
data,
of altitude
rainfall
selected,
obtained
from
and latitude,
could be estimated
to
dethe
so that
for each
the study area.
the
maximum
detail
and is dependent
that
can
be obtained
with
a
on the aims of the study.
Background
The plant
and
community
is a "group of plants
is distinguished
(Mueller-Dombois
plant.
It
agreed
generally
1956;
that
Mueller-Dombois
communities
can
be
sharing
by a particular
& Ellenberg
of the individual
plant
sampling
aim which
regressions
site within
field
well-defined
is dependent
of mean
3.1.1
(Poore
before
a clearly stated,
and a combination
interpolations
is
required
the aim. In this study, preparatory
Weather
ment
work
important
thereof,
termining
3.1.1.1
is that
1974)
plant
&
included
floristic
at any scale
in
form
1974)
larger
in
plant
environ-
composition"
smaller
communities
Ellenberg
a common
than
that
hierarchies
that
smaller
communities.
25
However,
Reitz
this
(in Poore
ecological
plant
to
inclusion
groups
community
include
depends
pose
of the
in a study
often
states
that
or vegetation
border
in a plant
nize
explicit
1956)
is
Scale,
In other
determines
units".
community.
the
This
and the
Which
than
words,
role in the stratification,
level
whole.
of other
complicate
community
to
recog-
by the
amount
of detail
should,
therefore,
sampling
Du
of heterogeneity
is determined
the
Scale
can often
plant
in turn,
scale.
rather
"a stand may have parts
recognition
on scale.
study.
partial
pur-
required
have
and classification
an
of
vegetation.
According
study determines
turn
is
determines
also
the working
the
related
accuracy
than
& Westfall
to Rutherford
and
user
a millimetre
the
a given
lines
requirements
which
Westfall
include
Each
& Westfall
not be closer
ii. not be closer
It follows
scale,
that
SMUA
area
&
Westfall
the
area
units
1986).
under
potential
study
(SMUA).
at working
of
scale
The
in
SMUA
cartographic
seldom
To these
of printed
less
may
a
be
maps.
is subdivided
containing
sampling
in a
scale which
precision
and shrinkage
each
required
number
sites
(Rutherford
should,
according
by
of
&
to
(1986):
than 1 rom to a stratified
than 2 rom to another
the width
is scale-dependent,
unit
considerations
stretch
stratified
can
1986).
Rutherford
i.
working
sampling
of a linear
(Rutherford
separating
SMUA's
mappable
practical
added the humidity-related,
At
or associated
smallest
to
(1986) the detail
being
unit's
sampling
of an unmappable
border;
and
site.
ecotone
or transition
less than 2 rom on a map at a given
work26
A unit
ing scale.
have
community
working
scale
status
to or broader
than
2 mm could,
at the
relevant
scale.
to the smallest
mappable
unit area
from the distance
and is given
equal
from sampling
The
therefore,
relationship
of
(SMUA) is derived
unit area to stratified
unit border
by:
x)
(-12~67-ln
y=e
= SMUA in
where
y
tion.
In their
account,
of MSDB
(minimum
and seD
(shortest
king scale
and
m>,
= scale,
Rutherford
sampling
cross
corresponds
x
& Westfall
distance
distance
scale will not improve
3.1.1.2
Methods
frac-
(1986) give examples
from a stratified
of a mappable
with detail,
working
as a representative
unit
unit).
border)
Because
any map enlargement
wor-
beyond
the
the detail.
applied
If the smallest
mappable
the relationship
between
unit area
(SMUA) is taken as a circle
SMUA and working
then
scale can be simplified,
as follows:
r = x/1 000
where
r
scale
as
metres
is also
=
radius
SMUA
a representative
of the
SMUA
the
minimum
unit border
SMUA
of
is, therefore,
fraction.
for a working
distance
and the SMUA,
and minimum
in metres,
distance
always
and
For
=
denominator
example,
the
scale of 1:50 ODD,
on the
ground
at this working
between
x
SMUA
between
scale.
radius,
1 rom on a map at any working
the
in
is 50 m. This
a stratified
The radius
and stratified
of
unit
of the
border
scale.
27
The
scale
the
detail
used
required
Agricultural
3.1.2
The
in this
regional
planning
(see section
Concise
Dictionary
a stand as "a growth
ially trees
by
is that
vegetation
corresponds
the
with
Department
of
4.1.1).
of the English
of plants
in sampling
it becomes
stands because
Language
in a particular
in a forest or a crop in a field".
into account
pling)
is 1: 250 000 which
area
New Collins
scale
for
Development
Stand
defines
study
area, espec-
The effect
at 1:250 000 scale
(1985)
of taking
(small scale sam-
increasingly
more difficult
to recognize
of increasing
heterogeneity
with decreas-
ing scale.
3.1.2.1
Background
Gabriel
& Talbot
abstract)
in
(1984) define a stand as follows:
aggregation
species
physiognomy,
condition
stands
to
we
sample
Werger
be
in such
selected
vegetation
should
in
(1974)
of which
yield
terms
it
or
and communities
vegetation".
a more
of
further
both
of more or less similar
composition,
distinguish
which
communities
of plants
from
are
(1974) states
spatial
adjacent
measure
a manner
are
further
that
that
communities.
aggregated
abstracted
"Stands
each
or less typical
states
that
composition
"Stands
which
and
are
and
Concrete
into
abstract
into a general
for sampling
is representative
description
(vs.
uniformity
arrangement
it is part and that each plant
floristic
"A concrete
sampled
should
of the
therein
of that vegetation
structure".
obviously
Werger
heteroge28
neous
therefore,
units
logically
information
community
floristic
should
which
be avoided
smaller
smaller
Furthermore,
than
than
the
a stand
parts
because
the
should
be
of two or more
do not con-
the two
or more
From these definitions
a relatively
community
sampling
and might,
they
can be used to describe
that a stand is generally
of vegetation,
composition,
to represent
types that they represent".
be inferred
not
or
be expected
or associations,
tribute
but
structure
in habitat,
plot
homogeneous
of which
with
unit
it is a part,
which
representative
it can
it is sampled.
of a community
and
not of an ecotone.
3.1.2.2
stand
Methods
area
previous
is not explicit
section
representative
hand,
a stand
of the
(SMUA) because
stand
should
community
definitions
be
given
sufficiently
it is to represent.
of scale
limitations,
where
in the
large
to be
On the
other
mappable
unit
it is intended
to
communities.
Werger
(1972)
itself,
in terms
of
but
in the
a stand should not be smaller than the smallest
area
map
applied
about
plot
area.
one
says
a
of species
half
However,
then the following
that
to
one
vegetation
stand
composition
and structure,
hectare
but
can
recommends
fully
would
in an area
a much
if stand area were to be equated
manifest
smaller
with the SMUA
apply:
1: 5 000 scale represents
a stand area of 78 m2;
1: 8 000 scale represents
a stand area of 201 m2;
1: 10 000 scale represents
a stand area of 314 m2;
29
1: 20 000 scale represents
a stand area of 1257 m2;
1: 50 000 scale represents
a stand area of 7854 m2; and
1: 250 000 scale represents
Equating
stand
account
when
area with .the SMUA permits
sampling
vegetation,
is then representative
fined
in section
The stand
tified
3.1.2.1
with
radius
a representative
is
a
function
increase
providing
to be taken
into
that the sampling
plot
defined
as a circular
area within
equal to the denominator
Thus,
scale
with decreasing
the criteria
de-
can still be applicable.
fraction.
of
scale
of the stand. Furthermore,
is, therefore,
unit
a stand area of 20 ha.
and
the minimum
stand
area
scale,
as
for a community
heterogeneity
scale. The stand radius
area is 250 m, representing
of the
a stra-
is
likely
to
for the main study
a stand area of about
20 ha.
3.1.3 Reconnaissance
A reconnaissance
prior
to sampling
3.1.3.1
and can have various
to Werger
known
in all
point
is
of the
study
area
objectives.
before
the study
by Mueller-Dombois
&
cover
established".
has
In
been
Ellenberg
clarified,
other
must
is started".
in stating that "Once the entitation
vegetation
essentially
(1974) "The area of investigation
its variety
supported
even further
the
inspection
Background
According
of
is the preliminary
words,
the
be well
This
(1974)
viewwho
go
or subdivisioning
communities
tentative
are
communities
30
should
be mapped
mapping
and
confirms
ever,
prior to field sampling.
classification
the mapping
what
details
Although
plant
systematically
tific
value
3.1.3.2
The
of
later
work
observations
of
during
give
can only
reconnoitring
This
entails
floristic
and camping
are
more
reconnaissance.
added
perspective,
increase
variation
study
area
over
sites,
the
if required.
roads,
area
is
throughout
scien-
terrain
the
variation,
permission
study
familiari-
from landowners
Routes
followed
for
during
to the main environmental
topography,
these
study
and environmental
should correspond
available
istic
the
travelling
the flora and to obtain
such as rainfall,
geology
and soil, where
in order to detect
gradients.
Data
the range
recorded
graposs-
of flor-
for the
main
units
as an
include:
estimation
of the number
indication
of
area.
can
researcher
How-
applied
purpose
ible, using
L
1977}
classification
of a reconnaissance.
the reconnaissance
dients
necessary
recorded
area to estimate
with
actions
the
of separate
to in this study.
inexperienced
{Tinley
familiarization.
zation
for the
the
in which
has been adhered
empathy
Methods
main
units,
is required
explicit
processes,
The principle
This
precise,
estimate
the
is
but
for
different
variation
required
an
colour
structural/floristic
of structural/floristic
in
for
plant
stratification
overestimate
allocation
units,
can
structural/floristic
communities
is
be
and
preferable
{section
units
which
by
the
need
to
3.1.4}.
estimated
in
an
The
study
not
be
under-
number
of
counting
the
int.ercept
the
31
reconnaissance
route.
to
Edwards
or
co-dominant
these
(19B3)
units
and
the
species
of
can
intercept
the
necessary.
Scale
nizing
the
those
stand
be
structural
floristic
each
can
for
be
stand,
with
a
either
side
of
the
buffer
the
unit
(see
a
route.
zone
This
to
section
be
to
3.1.4).
by
that
only
exceed
a
the
included
where
of
they
validation,
permits
equal
dominant
Borders
map,
account
units
according
the
stratum.
route
into
is
is
stratification
taken
stand,
component
on
structural/floristic
on
component
structural
indicated
route
radius
floristic
The
in
Aerial
recog4
minimum
stand
the
if
times
of
one
radius,
on
structural/
photographs
could
also be used for these estimations;
ii.
estimation
in
each
of species
by
the
of
sampling
counting
commensurate
in
in terms of species
structural/floristic
determination
done
richness,
with
unit.
unit
area
This
(section
the
number
of
plant
plant
height
and
spacing
structural/floristic
unit
specimen
of
is
would
per m2,
required
3.1.6).
species
such
result
This
in
that
in
for
is
an
area
any
area
a
similar
count;
iii.
voucher
plant
species.
tered
during
plant
species
required
iv.
collection
These
field
are
effectively
records
plants
sampling.
for field sampling
photographic
the
dominant
in
The
the
most
and
other
likely
ability
field
to
to
common
be
encoun-
identify
reduces
these
the
(see section 3.2.2); and,
and notes on any features commensurate
with the aims of the study can also be taken.
32
time
3.1.4
stratification
stratification
prior
to
is
the
process
vegetation
sampling
sampling
intensity
in
approach,
is far too
according
to the
a
low
of
preliminary
and
classification.
community,
using
for the community
sampling
units,
vegetation
without
mapping
Generally,
the
the
Braun-Blanquet
borders
to be defined
some form of stratifica-
tion.
3.1.4.1
Background
It can
be
Blanquet
can
argued
that
approach
be
is
recognized
field
ignores
the
spatial
ties,
is difficult
case,
some
for sampling
In stratified
vegetation
total
for those
sort
sample.
since
that
up
make
dancy
not
each
the
familiar
unit
with
map
the
plant
community
relationships
the vegetation
to which
the
Braun-
community
diagnosis.
between
communi-
the techniques
is required,
total
is divided
allocated
to
in any
that
vegetation
the
are
into preliminary
specific
can actually
is ensured
portions
improve
various
the
accuracy
vegetation
represented
of the
in the
of
units
sample
1958).
prior
at vegetation
are
sampling
it
(Freund & Williams
Such divisions
using
using
the vegetation.
Such
estimates
because
of vegetation
sampling
units
communities
necessary
the
however,
and
not
plant
in
This,
apply,
to map
to sampling
unit borders
also reduce the problem
of redun-
(Werger 1974), and ensure .an even
33
distribution
of sampling units in each vegetation
of vegetation
unit area.
stratification
aerial
as
photo
well
for random
mappable
geomorphology
the
to
and
experience,
factors
can
process,
although
the
usefulness
purposes.
usually
stratification,
The
include
suitable
following
aerial
where
integration
be
photo
factors
their
may
after.
interpretation
limits
only
different
is
subj ect
be
correspond
to
corre-
(environmental)
the
be implicit,
can
geology,
Furthermore,
non-vegetation
shown
the case,
as
of
on
classification
which
for
can
limit
stratification
reliably
with
used
those
for
of
the
units.
stratification
ery.
only
such
in vegetation,
and
correspondence
of
factors,
Vieual
can be based
as is often
for repeatability.
vegetation
Non-vegetation
vegetation
types.
discontinuities
between
sampling
of a study area,
non-vegetation
soil
detect
observer's
spondence
or systematic
interpretation
as
factors,
unit, irrespective
of vegetation
advantages
improved
of satellite
geographic
for processing,
article
can also be based
imagery
fidelity,
and suitability
describes
the methods
over
data
in
on satellite
aerial
applied
photography
a digitized
for small
imag-
form,
scale work.
in this
The
study.
34
Bothalia 16,2: 263-268 (1986)
A method for vegetation stratification using scale-related,
vegetation-enhanced satellite imagery
R. H. WESTFALL" and O. G. MALA
**
Keywords: Landsat, Munsell parameters, South Africa, stratification, Transvaal, vegetation
ABSTRACT
A method for visual vegetation stratification and pattern refinement, using scale-related, vegetation-enhanced
satellite imagery, is described. The method simplifies colour assignment, facilitates accurate vegetation mapping
and could lead 10 balanced floristic classifications.
UITTREKSEL
'n Metode vir visuele plantegroei-srratifikasie en patroon verbetering wat van skaalverwante, plantegroei-versterkte satellietbeelde gebruik maak, word beskryf. Die metode vereenvoudig kleurtoekenning en vergernaklik
akkurate plantegroei-kartering en kan lOt meer gebalanseerde floristiese klassifikasies lei.
INTRODUCTIOl
In vegetation sampling (Werger 1974) the subjective selection of sample sites, based on vegetation
homogeneity, can only be done effectively by an operator with considerable experience. In addition,
there is often a lack of repeatability in the methods
and a tendency to ignore vegetation dynamics by
only sampling those areas representative of 'good'
vegetation.
Stratified random sampling overcomes these problems and increases sampling efficiency by ensuring
adequate representation of subdivisions (Elliott
1983). Furthermore, in contrast to random or systematic sampling, the heterogeneity of vegetation, in
terms of possible number of communities, can be related to the number of stratified units. Stratification
also facilitates the avoidance of transitions which
generally do not contribute more information than
the adjacent communities (Werger 1974).
Stratification of vegetation prior to floristic sampling entails primarily the categorization of vegetation according to structural characteristics. The categories can be further refined according to the factors
apparently responsible for differentiating the strata.
The most important factor is usually taken to be topography but others such as geology, pedology, climate or combinations of the four may also be decisive.
Problems encountered with vegetation stratification when using aerial photographs include radial
distortions, altitude-related scale differences and often inconvenient scales, which do not facilitate precise vegetation mapping. The excessive detail present in aerial photographs can be potentially confusing and time consuming for stratification, especially
• Botanical Research Institute, Department of Agriculture and
Water Supply, Private Bag X101, Pretoria 000l.
•• National Physical Research Laboratories, CSIR, P.O. Box
395, Pretoria 0001.
for small-scale work. The use of small-scale, almost
orthographic satellite imagery for stratification overcomes these problems but introduces the problems
of pattern interference by factors such ·as soil, and
colour assignment where patterns are formed by unfamiliar colours and textures. This paper describes a
method for vegetation stratification, using satellite
imagery that can overcome the problems of pattern
interference and colour assignment and facilitates
later pattern refinement.
CONVENTIONAL USE OF LANDSAT DATA IN
VEGETATIO
STRATIFICATION
False colour images
The Landsat multi-spectral scanner (MSS) records
radiance from the earth's surface in four spectral
bands: 500--600,600-700,700-800 and 800-1100 nm,
usually referred to respectively as bands 4, 5, 6 and
7. These data are obtainable in image or digital
form. The ground resolution of the system is nominally 79 x 56 m which corresponds to a picture element of about 0,30 x 0,22 mm at a scale of
1:250000.
The MSS collects six lines of data simultaneously
using a set of six detectors for each spectral interval.
The mismatch in these sets is one of the major causes
of noise in MSS data, which is apparent as striping
with a periodic cycle of six lines at extreme radiometric enhancements.
Traditionally Landsat MSS data have been used in
mapping at scales between 1: 1 000 000 and
1:250 000 (under special circumstances up to
1:100 000 or even 1:50000 scale) in the form of false
colour images, i.e. bands 4, 5 and 7 displayed respectively as the colour primaries, blue, green and
red.
This type of display suffers from two disadvantages: firstly only three of the four bands can be displayed, with a possible loss of crucial information in
band 6. Secondly, because of the high degree of cor"
relation between bands (Table 1), a very limited re35
264
Borhalia
16,2 (1986)
gion of the available three-dimensional
colour space
is utilized: live vegetation
appears exclusively in
various shades of red, depending on its structure and
vigour.
tral differences in surface cover. Although the third
component normally contains considerably less variance than the second, this may be crucial information for stratification.
The fourth component contains predominantly
noise.
TABLE
For optimum results peA must be based on the
statistics of a (composite) subscene, approximately
equally representative
of the relevant floristic subdivisions, of the area to be stratified. The three principal components may be displayed in any combination of the colour primaries. However, particularly
in regions of large variations in overall brightness,
such as in areas of rough topography, this results in
multicoloured imagery which is difficult to interpret.
1. - Correlation
matrix of the radiance values of the
Landsat MSS data, in the four Landsat MSS bands, for the
study area
Band
4
5
6
7
4
5
6
7
1,0
0,88
0,71
0,59
1,00
0,62
0,47
1,00
0,94
1,00
Digital multispectral
classification
Digital multispectral
classification methods have
been used successfully in crop mapping. The methods have proved to be a problem in the stratification
of natural vegetation,
particularly under southern
African conditions. The reasons are mainly: rugged
topography
(crops are normally grown on level
fields), heterogeneity
of stands with the consequent
problems of selecting 'typical' training sites of sufficient size (i.e. several hectares in size) for the extraction of spectral signatures and the interference
of
soil reflectance caused by incomplete canopy cover.
Furthermore,
this method relies exclusively on
spectral characteristics
and cannot make use of the
contextural or contextual information consciously or
subconsciously available to the human interpreter.
'DIGITAL
E 'HANCEMENT
OF LANDSAT
DATA
In the alternative approach of enhancement of the
Landsat MSS data by digital image processing followed by visual interpretation,
the versatility of the
human analyst is assisted by means of quantitative
enhancement
of vegetation differences in stratification but particularly in pattern refinement.
Principal Component
Analysis
Principal component
analysis (peA) provides a
convenient
method of data compression
and removal of redundant correlation between bands (Lasserre et al. 1983). Experience has shown that typically more than 99% of the total variance in the data
is retained in the first three principal components
(Table 2), the maximum which can be accommodated in any colour display.
TABLE
2. - Variance in terms of proportional
eigenvalues for
the first four principal components
of the Landsat MSS
data for the study area
Principal
Component
First
Second
Third
Fourth
0,872
0,106
0,016
0,005
The first component represents shadow-enhanced
topography
and overall terrain brightness differences, whereas the second and third reflect the spec-
36
A much more informative
product can be obtained by displaying the first three components
respectively as the Munsell colour parameters, brightness, hue and saturation.
Display in the Munsell colour space
In order to generate a practicable colour display,
the components
representing
the Munsell colour
parameters
must be converted into the colour primaries, blue, green and red by computation,
special
care being taken that visual hue differences in the
display truly reflect numerical differences in the data
(Malan & Lamb 1985)". The first and third components are first contrast-stretched
to about 1-2% of
the data in maximum and minimum values. The distribution of the first component
is approximately
Gaussian (Figure 1) while a histogram equalization
stretch is applied to the second component (Figure
2). These stretch lookup tables must be based on the
statistics of the representative
subscene used for
peA.
In the final product the overall impression of terrain brightness of the original image is retained, thus
facilitating registration with overlays and later pattern refinement. The effect of the histogram equalization stretch of the hue component is to spread the
spectral differences of the cover over the complete
hue gamut (as modified in saturation by the third
component) as opposed to the limited range of hues
in the conventional
false colour representation
(compare Figures 3 & 4).
Filtering
When vegetation mapping is to be done at a scale
where the shortest cross distance of a mappable unit
is greater than a ground resolution of 79 x 56 m, the
original resolution of the Landsat MSS data could
lead to a product with too much small detail. Prior
smoothing of the data produces a scale-related product which facilitates stratification
and pattern refinement.
The minimum colour-stratified
area is given by
12,56 n mm2 with a shortest cross distance of 4 mm
(Rutherford
& Westfall 1986). The value of n is determined by the minimum number of samples required for possible floristic subdivision of a colourstratified area which is dependent on scale as related
to sample area and spacing as well as the area of the
colour-stratified
unit. For example, with n=4, full
resolution Landsat MSS data could therefore theo-
Bothalia
265
16,2 (1986)
2
Cumulative
histogram
(1)
0
e
(1)
~
~
:::J
0
0
0
0
>c
0
(1)
:::J
C'
(1)
~
(1)
0)
~
e
(1)
0
~
(1)
a.
0
255
0
Grey
FIGURE
1. Munsell
Histogram
brightness
showing the approximately
parameter.
2
level
Gau
ian di tribution
of the first prirtcipal component,
0
e
as the
Cumulative
histogram
I,
(1)
displayed
d
, I
(1)
~
~
:::J
0
0
0
I
-
,
0
,
I I
,
I I
i!
I
I,
II II
Ill,1.1
I
I I
1'1
I I I
1,1
v'_-1
1,1
II I I
II
, ill I
, I )II I
I
!
II
I
'll'lHldl'l'lllll
k11,Ihl
,I'lijil'
j 1111,.,11
,
II ,Iii
T'l
ill
Il~ ll'I'!'I!
I I \1I II I I ! I i I I j I i j !'~-1 I I I I i n
I
II
II III 1llll'IIlllii 111"1' I III '11'111
,
'11'
Illl
P
nu
I I jI I' jl I li i I I l'i 1;-1 !jl II I Ii I III I I'
1
>0
e
I
II
Iii
I II
i I j!
I I
i H I
I I
i"",1
I I III j I
I"
I i III Ii Iii
1
.
lit
'11
"I
1
,
1
I
'i i,
'lllll
I
1
,I
I,
1111
I jjj I\
1'1'II
m l! u I III Il .(1I I 'P II n 1 mu l! IP II q PI!
11l1/1.11!'j' u Ililli / d;-flt H! lilll hi n ldllhillhidII
II III u II II HI I I t '-II I III i I I u Ill! !I u nu n u HIiI hi. ,
~
H
C'
G)
~
1.111 I
II I I I ')
I I II
I
"
,I II I' I I, ),,1
!
II P W,II IlllI! II !-l-i '.j 1 ii',' !! ! 11111111 j I' 1111PI PI I!III II I 11!11! 1li
G)
Ill! ! I! I ! 1111 H I!! llllllllll
n HII Illl!! Id II
.,-11
h I 1I1l!! III II !II! II lil' 1111l11IL!I\l1!! Ijll!! II 1,1
h 111111\"1 In Illl
u ! II! q P n lj I P q n !ill
H Iill h I j!! 111111111 ! III i ! I!I II !!! II Ill!1I!lh II ill !!I1I!l
1'nlli Illtl IiII H I III'II
Ii Ii i L dIP m Ii ,uuu PIIW ,Ihl
pIli IIil
1'1II
H IJIll!
u II jll !I 1i H l n i! jl HltilldII IlJdI nI! Iiduhqdm IH!lnh Hn Pm
H
I 11111P II',' I it!14jlll
-
0)
~
'illll!
e
j
G)
0
~
(1)
a.
0
, i
IIII
t ..
\jlll
jl
I
II
,
III
j I I I
II
I Ii
Ii
j
t I
I ' , I lil
jl
il
I!
II
1lI
II I '
III
t,
llll Ii i III
I 'Ii jl
I
255
0
Grey level
FIGURE
2_ -
Equalized
histogram
of the second
principal
component,
displayed
as the Munsell
hue parameter.
37
Bothalia 16,2 (1986)
266
retically be used for vegetation stratification up to a
scale of 1: 10 000. However, PCA dramatically enhances noise in the data in the higher principal components, particularly the striping in Landsat MSS
data. The application of a median filter with a kernel
at least six lines wide smoothes the data and removes
striping effectively.
Consequently,
in practice, a
median filter with a minimum kernel size of 6 x 9
picture elements corresponding
to a square ground
resolution of about 24 ha, is used. This corresponds
to a maximum useful working scale of almost
1:50 000.
The (first) brightness component is excluded from
the filtering process, to retain fine detail in the representation
of topography, which assists in pattern
refinement as well as in the identification of ground
control points for registration with map overlays.
PATIERN
REFINEMENT
Pattern refinement refers to the process of modifying the stratified units, usually after sampling and
classification of the vegetation and prior to mapping
floristic units. This process includes the grouping together of similar, smaller, discreet areas and subdivision -of larger, uniform colour-stratified
areas by
contextual
comparison
if required,
with suitable,
simplified topographical,
geological, pedological or
meteorological overlays and sample-set classification
at the given working scale. The choice of any or all
of these overlays is determined by the range of variation exhibited by the overlays that can relate to the
structural variation. For example, smaller units of
differing colour can be grouped together on the basis
of floristics and soils or topography while larger, uniform units could be subdivided on the basis of floristics and geology or climate in order to refine the
patterns.
Minor inaccuracies can occur in the registration of
Landsat hard copy images and other maps at the
same scale used for overlays which are attributed to
differential stretch and shrinkage caused by fluctuating humidity as well as some obvious inaccuracies in
the maps. These errors can be compensated
for by
shifting local fit to achieve the maximum number of
registration
points rather than be compounded
by
maintaining fixed registration points. However, the
effect of stretch and shrinkage can be further reduced by the use of dimensionally stable transparencies of both satellite images and other maps, where
available.
RESULTS
The results of colour-stratification
using scale-related vegetation-enhanced
satellite
imagery
are
given for a portion of the Transvaal Waterberg in the
north-western
Transvaal
(Rutherford
& Westfall
1984) at 1 250 000 scale. The area has a highly diverse topography with a consequently high variation
in vegetation structure which is inferred from the variation in colour pattern. The vegetation is mainly
representative
of Sour and Sourish Mixed Bushveld
veld types (Acocks 1975). Structural heterogeneities
related to topographic diversity were indicative of
potentially small stratification units and the value of
n=4 was accordingly assigned. This corresponds to a
38
map area of 50 mm? which is equivalent to 320 ha.
The appropriate filter kernel size was therefore 32 x
22 picture elements.
A contrast-stretched
falsecolour MSS image (acquisition date March 1981) is
shown in Figure 3. A vegetation-enhanced
image, at
the original resolution of the same scene, is shown in
Figure 4, which illustrates the complexities of colour
assignment
exacerbated
by noise, particularly
six
line striping. A scale-related,
vegetation-enhanced
image at the hard copy scale of 1:250 000 of the same
scene is shown in Figure 5 with resultant simplification of colour pattern to form stratified units.
INTERPRET ATIO
A critical prerequisite for successful stratification
is the choice of the optimum data acquisition date
for maximum differentiation
between the structural
subdivisions. Because the four wettest months in the
study area are from November to February, the data
acquisition date of March ensured high vegetation
cover with minimal cloud interference.
The inputs required from the user are location of
study area, working scale and data acquisition dates
of scenes required. Primary colour-stratification
is
automated
after these inputs and is, therefore, objective and time and labour saving. Many researchers use topographic, geologic, pedologic or meteorologic maps for comparison with floristic units or simply to show the variation in these factors. Unless
working scale is standardized at the outset of a project, comparisons
are difficult. Overlays at a standardized scale are used for pattern refinement and
can also be used for later comparisons with floristic
units in addition to the enhanced
Landsat MSS
image. They can, therefore, serve dual purposes and
their effective use is increased.
Training sites would often be required for unsimplified images (Westfall & Malan in press), whereas
the use of scale-simplified images largely overcomes
the need for this training. Furthermore,
scale-related stratified units should improve floristic classifications by providing a balanced distribution of sample sites commensurate
with vegetation heterogeneity and the amount of detail required for a given
working scale. It should be pointed out that, like
PCA, the colour stratification
by vegetation
enhancement
is highly scene-dependent.
In practice,
however, this is not a great disadvantage
because
one Landsat scene covers 34 000 krn-'. Also multiples
of this size can be treated identically if they are contiguous images on one north-south Landsat swath.
The use of satellite images, which have better geometric fidelity than aerial photographs,
also simplifies the process of accurate vegetation
mapping,
especially where first-order stereo-restitution
instruments are not available. This ensures greater mapping precision when compared to base maps and facilitates the effective use of overlays for comparison
or pattern refinement.
The proposed methods also ensure objectivity because the primary colour-stratification
process is
computerized,
which produces repeatable stratification units. It is doubtful that subjectively stratified
units could be repeated by different workers. This
Bothaiia 16,2 (1986)
267
FIGURE 3. - A contrast-stretched, false-colour. multispectral-scanner (MSS) image
of a portion of the Transvaal Waterberg with limited range of hues. Scale
1:250000.
FIGURE 4. - A vegetation-enhanced image at the original resolution of the same
scene as shown in Figure 1
using the complete hue
gamut.
FIGURE 5. - A scale-related vegetation-enhanced image of
the same scene as shown in
Figure 2 showing simplification of colour pattern to
form primary stratified
units.
could affect the balance of the resultant floristic classifications and could be especially significant in comparisons over time.
CONCLUSIONS
The proposed methods of vegetation stratification
prior to' floristic sampling are objective, and timeand labour-saving. The process of colour assignment
to stratified units is simplified. The stratified units
are related to working scale and the structural heterogeneity present in the vegetation. Overlay comparison and accurate vegetation mapping are facilitated and more balanced floristic classifications can
be expected. Landsat data are generally more detailed than required for vegetation stratification at
scales smaller than 1:5000 and hence filtering is
necessary.
39
Bothalia 16,2 (1986)
268
ACKNOWLEDGEMENTS
The authors thank Dr J. C. Scheepers for comments and suggestions.
REFERENCES
ACOCKS, J. P. H. 1975. Veld types of South Africa. Memoirs of
the Botanical Survey of SOUlIlAfrica '0. 40.
ELLlOTI, J. M. 1983. Some methods for the statistical analysis of
samples of benthic invertebrates. Kendall, Wilson.
LASSERRE, M., MALAN, O. G. & TURNER, B. 1983. The
application of principal component analysis to Landsat
MSS data. Proceedings of seminar on Principal component analysis in the atmospheric and earth sciences, Pretoria 7-8 February 1983.
MALAN, O. G. & LAMB, A. D. 1985. Display of digital image
data; quantitative and oprimal. Proceedings of the third
40
South African symposium on Digital image processing,
Durban 22-23 July 1985.
RUTHERFORD, M. C. & WESTFALL, R. H. 1984. Sectors of
the Transvaal Province of South Africa. Bothalia 15:
294-295.
RUTHERFORD, M. C. & WESTFALL. R. H. 1986. Biomes of
southern Africa - an objective categorization. Memoirs of
the Botanical Survey of SOUlhAfrica No. 54.
\VERGER, M. J. A. 1974. On concepts and techniques applied in
the Ztirich-Mompellier method of vegetation survey. Bothalia 11: 309-323.
WESTFALL, R. H. & MALAJ"l",O. G. in press. A comparison of
vegetation units derived from vegetation-enhanced satellite
imagery wirh the vegetation units derived from the florisric
classification of the farm Groothoek, Thabazimbi District.
Proceedings of the symposium on Pattern recognition in
remote sensing and geophysics.
location
3.1.5 Stand
The
main
disadvantage
possibility
data,
of hidden
which
could
stratified
sampling
unit
random
the
localities
does
sampling
process
sampling
within
tical
tests
3.1.5.1
periodicities
sampling
is
(Freund·
preferred.
of ·stands within
reduce
observer
not necessarily
because
the
a vegetation
infer
number
unit
In
each
bias.
&
is
phenomena
in the
Williams
stratified
is generally
so
random
vegetation
Randomization
of
validity
repetitions
the
1958)
stratified
statistical
of
on normal
sampling
stand
on the
in vegetation
too few for statis-
distributions.
Background
co-ordinate
of stand
system
or a random
selection
is used,
number
the
and hence grid
states
that
unnecessary
sampling,
be
Whatever
minimum
not
done.
Apart
of stands.
3.1.5.2
Methods
applied
considerations
given
by a grid
using
random
system of co-ordinate
spacing
between
stand
because
Werger
(1974)
to be
does not appear
spacing
geometric
are
achieved
are selected
should be known,
stands
minimum
the
co-ordinates
appropriate
would
literature
is generally
generator.
size,
contiguous
work
the
localities
in which
sites,
From
systematic
or recurring
results
that are based
Randomization
tables
to
stratified
influence
location
is random,
of
recommended
from
to address
in section
as much
non-contiguous
the problem
3.1.1,.a
of
grid
41
l
a
a
1
3
2
4
...•
<,
/
1
5
I
I
,
\
2
I
/
~"
,- ...
I
3
f
\
\
"-
....
/
--
"-
I
4
\
f
I
\
\
5
,,
"-
.•. ..-
;'
/
/
--
-,
\
\
,
\
f
-,
f
-
-
I
./
4mm
t
-4mm--
FIGURE 3.1. - A 4 x 4 mm numbered grid for map overlay to determine possible stand sites at the grid intersections.
solid
circles represent
stands; and dotted circles represent buffer
zones around stands. Buffer zone radius = 2x stand radius =
maximum ecotone width = minimum distance between stands (from
Rutherford
& Westfall 1986).
42
size of 4 x 4 mm should
map
overlay
ensures
Stand
The
to
locate
a minimum
area
use
of
computer
minutes
of
one
grid,
of
their
and seconds,
to facilitate
it is ignored
also be ignored
selected
the
co-ordinates
stand
location
stand includes
not commensurate
and the following
stand
more
quired.
a stratified
stand
Observer
regarding
stand
regarding
stand
localities
bias
is,
location,
with
4
minutes
grid
and seconds
and
The
is selected.
grid
It
is for
than
following
heterogenscale,
Stands
can
Furthermore,
are
by
these
reasons
actually
re-
decision-making
may be necessary
article
stand co-ordinates
for field location
In
is less than one stand
reduced
co-ordinate
co-
degrees,
in the field.
obvious
some decisions
of random
grid
with the working
generated
therefore,
suitability.
for the generation
mm
are
border.
although
program
a
unit
3.1.
simplify
of
in the field if they are inaccessible.
from
stands.
can
selecting
as a
This
in Figure
scale,
for
may not be used if the circumference
diameter
between
is illustrated
to
scale,
co-ordinates.
diameter
numbers
conversion
then
random
for any working
random
if a randomly
from
stand
a standard
such as disturbance
that
sites
to the grid
eity
stands
at any given working
in relation
and
the field,
stand
spacing
generation
ordinates
be used,
conversion
describes
a
to be used
to
degrees,
of stands.
43
Bothalia 16,2: 269-271 (1986)
PHYTOLOC -
A RA~DOM-NUMBER GE1\ERATOR AND SAMPLE-SET LOCATION PROG RAM FOR
STRATIFIED RAl'mO\1 VEGETATION SAYIPLING
The stratified random method of egetation sarnpiing is objective and efficient, in terms of sampleset distributions, for floristic classifications (Westfall
& Malan 1986). However, the commonly used randam number tables and calculator-generated
randam numbers often require number abbreviation
. and manual recording of the numbers, which can be
time-consuming.
But these inconveniences are insignificant when compared
with the time taken to
measure the location of random sample sets and express their location in terms of the latitude and longitude co-ordinates of degrees, minutes and seconds.
The PHYTOLOC
program was developed to generate random numbers for sample set location in terms
of random co-ordinates and, in addition, to express
these co-ordinates as latitudes and longitudes in degrees, minutes and seconds, thereby sa ing considerable time and effort.
The program is written in Basic and runs on a
Sharp PC 1500 computer. Use is made of a consec-
can also be used on larger-scale maps for more precise field location of sample sites.
For valid categorization
and analysis of floristic
units, based on multivariate data, a minimum of four
sample sets are required, although a single sample
set is mappable at the given working scale. Co.n.sequently for a single floristic division of a stratified
unit a minimum of eight sample sets would be required. However, for statistical comparisons of univariate data such as biomass or number of taxa, sarnpie-set number should be proportional
to area (EIliott 1983). It is, therefore, suggested that a sampling
intensity of 2,5% of the potential sampling sites (i.e.
total number of co-ordinate interceptions)
within a
stratified unit should be maintained to ensure proportionality.
This is approximately
commensurate
with the relationship of study area to sample number
(Rutherford
& Westfall 1986) but modified by vegetation heterogeneity
in terms of number of stratified
units.
utively numbered 4 mm transparent grid map overlay which is related to any working scale, in terms of
sample set spacing and size (Rutherford & Westfall
1986). Grid overlay registration with a base map is
according to the zero co-ordinates
of the overlay
with the intersection of minimum latitude and longitude of the study area on the base map, as well as
with the zero x-axis of the overlay with the minimum
latitude of the base map. Inputs required for the program are:
(i) maximum latitude of the study area in decimal
A non-random
set of sample sites could, therefore, be required to fulfil the categorization
and
analysis requirements
for stratified units with le~s
than a total of 320 interception
points. The additional non-random sample sites can be selected objectively by:
f
(i) using best fit of additional samrle sets . or
areas equal to those of eight or less interception
Points (i.e. 100% sampling intensity), and
degrees;
(ii) minimum latitude of the study area in decimal
degrees;
(iii) minimum longitude of the study area in decimal degrees;
(iv) difference in millimetres, between minimum
and maximum latitudes, at the given working scale;
(v) mean distance in millimetres, between 1 minute longitudes at the minimum latitude;
(vi) mean distance in millimetres, between 1 minute longitudes at the maximum latitude;
. (vii) number of sample sets required, estimated
by 10 SU + (0,25 x 10 SU) where SU is the number
of stratified units. This should generally allow for
omissions due to transitions and proportionality;
(viii) upper limit (integer), within the study area,
of the x-axis of the grid overlay, and
(ix) upper limit (integer), within the study area,
of the y-axis of the grid overlay.
The program generates and prints random numbers for the x- and y-axes of the grid overlay and
computes and prints the equivalent values in degrees, minutes, seconds and decimal fractions of seconds for longitude and latitude respectively. Convergence of longitude is also taken into account. Each
set of co-ordinates, representing
a potential sample
site, is numbered
consecutively.
In addition the
means and standard deviations of the x- and y-arrays
are computed to show the statistical distribution of
potential sample sites. Co-ordinates are transferred
to the base map using the grid overlay and the printout can be used for field allocation of latitude and
longitude to the field data sheets. These co-ordinates
(ii) using additional random sample sets, to ensure representation
of vegetation variati~n, for ar~as
8
d 320
equal to those with between
an
In~erc~ptlon
points. (i.e. > 2,5% but < 100% sampling intensity).
Additional
random sample sets can be obtained
together with the relevant co-ordinates, if require~,
by the same procedure, but with each relevant ~tratlfied unit registered separately on the 4 mm gnd.
ACKNOWLEDGEMENTS
The author thanks Drs J. C. Scheepers
Ark for comments and suggestions.
REFERE
and H. van
ICES
ELLIOT, J. M. 1983. Some methods for the statistical analysis of
samples of benthic ill vertebrates. Kendall, Wilson.
RUTHERFORD, M. C. & WESTFALL, R. H. 1986. Biomes of
southern Africa - an objective categorization. Memoirs of
the Botanical Survey of SOUlh Africa No. 54.
WESTFALL, R. H. & MALAN, O. G. 1986. A method for vegetation stratification using scale-related, vegetation-enhanced satellite imagery. Bothalia 16: 263-268.
R. H. WESTFALL
44
3.1.6
The
Sampling
sampling
vegetation
a releve
unit area
unit
stand
is the vegetation
sample
and in the classification
& Ellenberg
(Mueller-Dombois
plot is used to sample
a vegetation
plot
phase
1974).
used
to
sample
is referred
Where
more
to as
than
stand then the sampling
a
one
unit is
a sub-plot.
3.1.6.1
Background
Werger
(1972),
concludes
that
has been
in
formulated
a suitable
further
suggests
which
sampling
a
unit
area
10 x
10 m;
vary
Coetzee
size
unit
on species
may be an extreme
statement
al.
for
sampling
unsuitable
much
richness
but
relationships,
for sampling
size
considerably.
et
species-area
the concept
unit
sampling
This
of
minimal-area
and regards
depends
can
review
no convincing
taining
hectare
his
(1976) used
for ascer-
than
and structure
example
size
vegetation.
smaller
in southern
For
unit
half
10 x 20 m
a
of stands.
Africa,
Coetzee
He
sampling
(1975)
for
used
herbaceous
"
plants
and
Theron
(1978)
Van
der
used
et
Meulen
used
sized
(1988)
(1985)
(1979)
used
used
units
for
woody
16 m2 for grasslands
used
10 x 20 m and variable
al.
trees
variable
plants;
Van
Rooyen
size for the woody
all
10 x 20 m for herbaceous
&
and 100 m2 for woodlands;
10 x 20 m;
10 x 20 m for
Bredenkamp
plants
plants
et
al.
component;
and
and
Le
(1981)
Westfall
Raux
et
al.
100 x 100 m for
and shrubs.
Such variation
in sampling
unit area can be expected
to complicate
45
comparisons
within
Furthermore,
20
m
recording
a
often
plot
was,
emphasized
for
difficulty
sampling
approach
species
stand
the
to
represent
than
that
a
adopted
sub-samples
presence
is less
communities
was experienced
therefore,
that
more
and between
in selecting
20
to
are
from different
ha
stand.
sample
based
only and although
than' 200 m2 the
which
could
be
a single
A
It
minimum
must
area
the total
sample
number
recorded
species
expected
from
10 x
sub-sampling
stands.
on
studies.
a
10 x
be
for
area
was
20
m
plot.
The
following
to determine
article
minimum
describes
sub-sample
the methods
applied
in this
study
area.
46
S.-Arr.
44
Tydskr.
Planlk.,
1987,53(1)
Predictive species - area relations and determination of subsample size for vegetation
sampling in the Transvaal Waterberg
R.H. Westfall*,
J.M. van Staden and M.D, Panagos
Botanical Research Institute, Department of Agriculture and Water Supply, Private Bag X101, Pretoria, 0001 Republic
of South Africa
Accepted 20 August 1986
An expression for predicting the number of species in a given area is described. Derivatives of this expression,
to increase sampling efficiency in a egetation stand, include a minimum of four separate subsamples; a maximum
number of subsamples when less than 10% increment in new species is achieved; and subsample size. It is also
suggested that species diversity in terms of species per unit area can be more consistent when derived from
this expression.
'n Uitdrukking om die aantal spesies vir 'n bepaalde gebied te voorspel word beskryf. Afleidings van hierdie
uitdrukking, om doeltreffendheid by die monsterneming van 'n plantegroeistand te verbeter, sluit die volgende
in: 'n minimum van vier afsonderlike submonsterpersele; 'n maksimum aantal submonsterpersele wanneer die
nuwe spesiesaanwas minder as 10% is; en submonsterperseelgrootte.
Dit word ook voorgestel dat spesies
diversiteit, ooreenkomstig spesies per eenheidoppervlakte, meer konsekwent kan wees wanneer dit van hierdie
uitdrukking afgelei word.
Keywords: sample size, savanna, species number, Transvaal, vegetation sampling
'To whom correspondence
should be addressed
Introduction
The vegetation ecology of the Trans aal Waterberg is currently
being investigated at a scale of 1:250 ()()()(Westfall, in prep.).
The smallest mappable unit area or vegetation stand, of which
a sample should be representative, is determined by scale
(Rutherford & Westfall 1986) and is 20 ha for this study. A
single sample of this dimension would be prohibitive both in
terms of cost and time. Random subsamples within the stand
offer an objective method of obtaining a representative
sample. The efficiency of the subsamples, in terms of plant
species recorded, can be determined if the total number of
plant species in the stand can be estimated.
The object of this study is to determine the optimum
subsample size, in terms of reduced effort and improved
efficiency, for sampling the vegetation of the Transvaal
Waterberg.
Methods
Vegetation analysis
Two observers, A and 8, counted species in eight samples
representing four very different vegetation types, namely
closed grassland (sample I); open woodland (sample 2); closed
woodland (samples 3 and 4) and unidorninant forest (samples
5 and 6). In samples I and 2 each observer repeated the
countings of the Other observer for the same area, each
without knowledge of the other's results. In each sample
cumulative plant species totals were recorded in 14 nested
rectangular subsamples, each with a width-to-length ratio of
1:2 (Figure 1). Subsamples size increased from 62,5 mm X
125 mrn (0,0078125 rrr') to 10 m x 20 m (200 nr') and the
length of each subsample was perpendicular to the contour.
In addition, both observers jointly counted the observable
species present in a 20-ha stand of which sample 2 was
representative. In this stand nested circular subsamples were
also used with radii of 6,3 m (128 11l2), 12,7 m (512 m '),
25,5 III (2 048 1112), 51 m (8 192 11l2), 102 m (32 768 rrr'),
204 111 (131 070 1112) and 252 111 (200 000 111).
Synthesis
The data in the form of species number for a given area were
tested for best fit with various curves (Parton & Innis 1972)
as well as linear relationships. The curve which best fitted the
data is described by the function:
f (x, a, b, c, d)
= (1
a
+ b/x) +
(I
c
+ d/x)
where a and c = the parameters which control the maximum
value of the function and band d = the parameters which
control the rate at which the function approaches its maximum
value (Parton & Innis 1972).
o parameters could be found with actual or derived values
similar to the computed values for parameters a, b, c & d.
This reduced the application potential of the function and it
was accordingly not further applied.
The linear regression which best fitted the data is described
by:
y =
e(lI/lm
+
Ine)
where y = number of species for a given area (x), III = slope
(rate of species increase for increasing area), x = area (nr'),
c = number of species in 1m2.
This linear regression was then used to derive the number
of species in 1 ~2 from a given area and to synthesize data
for illustrating species ~area relationships.
The validity of the relationships was tested by using data
pertaining to the first three reieves recorded in the Transvaal
Waterberg (Westfall, in prep.) which utilized these relationships.
Results
The number of species recorded by observers A and B for
the nested samples 1 to 6, with predicted values, are given
in Table I and Figure 2. The linear regressions of these
samples are illustrated in Figure 3. With the exception of
sample A6, the slopes are generally similar. Sample A6
represents a small stand of unidorninant Podocarpus latifolius
47
S. Afr.
1. \301., 1987,53(1)
- - - - - ---
----
- --
45
-----
---
----
----
----
- ---
- - - - - --I
,
type concerned. The values of c, derived from different areas
and corresponding derived species number are given in Figure
I
I
I
I
I
9m
4.
,
,,
The derived relationship between number of subsamples
and percentage new species increment is given in Figure 5 and
the relationship between subsample number and percentage
accumulated new species total is given in Figure 6.
. The results of the 20-ha sample together with predicted
values is given in Table 2.
I
I
I
I
I
7m
I
I
I
,,
Sm
I
I
I
,
I
I
,
I
I
I
3m
,
I
I
,
,
I
I
I
I
trn:
E
o
N
E
E
E
co
N
(0
,....
or-
E
,....E~
.r
,
:
E
N
,
E
(0
E
0
,....
E
.r
,....
E
co
,....
I
I
I
I
I
2m
I
I
,
I
I
,
I
I
4m
I
I
I
I
I
I
I
I
I
6m
I
,
I
I
I
8m
I
I
,,
I
,
10m
Figure 1 Layout of rectangular nested subsamples.
62,5 rnrn x 125 mm; 125 nun x 250 rom; YO rom
x 1000 mm; 1 m x 2 m; 2 m x 4 m; 3 m x
5111 X 10111;6 m x 12 m; 7 m x 14111;8 m x
and 10 111X 20 m.
Subsarnple sizes are:
x 500 rnm; 500 mm
6 01; 4 01 X 8 01;
16 m; 9111 x 18m;
(Thunb.) R. Br. Ex Mirb. in which forest-margin species
were recorded before the maximum sample size of 200 nl
was reached and can therefore be considered atypical. The
mean slope of the regressions, excluding sample A6, is 0,34
but is taken as 0,3 which was used for calculating the predicted
values in Tables 1 and 2. The mean correlation co-efficient
for the recorded and predicted values for the seven samples
(excluding sample A6) is R = 0,9860, whereas the mean
correlation co-efficient for the corresponding recorded and
predicted values using the curve function is R = 0,9819. The
difference between these correlation co-efficients is not significant. The correlation co-efficient for recorded and predicted values in the 20-ha sample is 0,9725 (Table 2).
The value of c is derived from' y = e(minr+ lOC) and is given
by
c
=
e' -
mlltr
+
Iny)
which is dependent on the species diversity of the vegetation
Discussion
The prediction of number of species for a given area has been
ested for a wide range of vegetation types with satisfactory
results. The larger difference between recorded and predicted
number of species for areas greater than 2 048 m2 (Table 2)
can be attributed to the difficulty in observing all species in
these areas. Destructive sampling would be necessary to ensure
the recording of all species. Both observers encountered
increasing difficulty in observing all species with increasing
sample size where sample size was greater than 20 m2. With
nested subsamples, however, species omitted in a particular
subsample were often recorded in a larger subsample. This
is illustrated by the similarity in totals obtained by observers
A and B for 200 m2 and the differences in totals for 18 m2
to 128 m2 in samples 1 and 2 (Table I).
The effect of sampling two highly distinct vegetation types
in a small sample is shown in Table 1 (A6) where a greater
proportion of new species was recorded from 8 m2 than in
the other samples. This is also illustrated in Figure 3 where
the slopes (m) of all samples except A6 are similar regardless
. of egetation type. Local heterogeneity in the 20-ha sample
did not influence the results as much as with the smaller
sample. This can be attributed to the logarithmic increase in
area used in the prediction expression where slope is not
significantly affected by a greater proportion of new species
in a large area. It is, therefore, essential for predictive purposes
that the vegetation should be relatively homogeneous for a
sampling area of at least 200 rrr'.
For a given vegetation type, the value of c in the expression
should be more or less constant when derived from different
areas and a horizontal line could, therefore, be expected when
these values are plotted against the area from which they are
derived. However, in Figure 4 the values of c increase to an
area equivalent to 2 m2 before remaining more or less constant. This increase can be attributed to the edge effect of
small· subsample size where the proportion of species intercepted by the subsample border to species within the subsample is greater than for larger subsamples. It is difficult
to record fractions of species present so that in practice
intercepted species are often ignored. Subsample size should,
therefore, be greater than or equal to 2 m2 in the vegetation
concerned to reduce this edge effect. Similarly the value of
c should also be derived from an area greater than or equal
to 2 m2 in the vegetation concerned which will also ensure
greater precision in comparability of species diversity in terms
of species per unit area.
The predicted increment of new species as a percentage of
the total number of species for increasing number of subsamples (Figure 5) has maximum inflexion for four subsamples or less or an increment of 8,3070,or greater for given
c values and subsample sizes. This indicates an optimum
efficiency of four subsamples for vegetation sampling or an
increment of 8,3070.With additional subsamples only relatively
small increments decreasing from 6,50/0 could be expected
(Figure 5). In practice this could be taken as an increment
48
S.-Afr.
46
Table 1 The number of species recorded by observers A and B for nested quadrats
values' in brackets and correlations with observed and predicted values
Area
(rrr')
.AI
BI
A2
B2
Tydskr.
°
B5
°
0,00781
2 (4)
2 (3)
1 (3)
1 (3)
0,03125
4 (5)
3 (4)
I (4)
3 (4)
0,12500
5 (8)
4 (6)
4 (6)
5 (6)
3 (5)
2 (8)
0,5‫סס‬00
9 (12)
8 (10)
8 (10)
9 (9)
7 (7)
8 (12)
(2)
0(3)
1987,53(1)
1 to 6 with the predicted
A4
B3
Plantk.,
°
°
°
(3)
I (5)
(0)
(I)
(I)
1 (I)
A6
°
°
°
(0)
(I)
(I)
1 (I)
2,0
17 (18)
15 (15)
13 (15)
14 (13)
10 (II)
14 (18)
1 (2)
8,0
28 (28)
22 (22)
22 (22)
20 (20)
17 (17)
27 (27)
3 (3)
3 (3)
18~0
37 (36)
22 (28)
26 (28)
24 (26)
19 (22)
34 (34)
3 (4)
8 (4)
32,0
38 (42)
27 (33)
34 (33)
26 (30)
25 (26)
36 (41)
3 (5)
8 (5)
50,0
72,0
44 (49)
34 (38)
38 (38)
35 (35)
28 (29)
40 (47)
4 (5)
12 (5)
46 (54)
35 (43)
46 (43)
39 (39)
35 (33)
49 (52)
4 (6)
16 (6)
98,0
50 (59)
40 (47)
49 (47)
41 (42)
39 (36)
52 (57)
6 (6)
24 (6)
128,0
53 (64)
48 (51)
52 (51)
43 (46)
47 (39)
57 (62)
8 (7)
26 (7)
162,0
55 (69)
55 (54)
54 (54)
49 (49)
51 (42)
60 (67)
9 (7)
30 (7)
200,0
57 (74)
57 (58)
55 (5 )
520)
55 (45)
61 (71)
9 (8)
33 (8)
Slope
Correlation
co-efficient
(R)
Value' of InC**
*Prediction
··C
according
= species number
1,199
1,022
0,945
1,002
0,791
1,044
0,773
0,207
0,9891
0,9894
0,9970
0,9971
0,9910
0,9948
0,9434
2,7085
2,4673
2,4673
2,3720
2,2095
2,6721
0,4749
0,9365
0,4749
e(mlnr
to: y =
2 (2)
~
Jne)
for 1 rrr' derived
where y = predicted
from
number
of species,
III
= 0,3 (slope),
x
= area (nr')
8 m2
y:bcr::
y
-4
Quadrat
o
-2
x
2
4
- <I ---':2~--':-O~'
6
2
4
~
x
A1
Quadrat
81
y'ke 'kL
y:kc :b::L
...
y
.
2
,
.'
,,
,
I
-4
Quadrat
-2
. .
2
o
I
x
2
4
:
,
,
~!
-4
6
Quadrat
A2
-2
o
x
2
4
6
2
4
6
82
v
-4
-2
o
x
2
4
6
Quadrat
Quadrat 83
<[,,'~,
-4
Quadrat
-2
-4
o
x
2
-2
0
x
A4
6
85
Figure 2 The relationship
between In area (nr') (x-axis) and In number of species (j-axis) showing observed numbers (dots) and predicted
values (regression lines). Generally, fewer species are observed in quadrats smaller than 2 m2 (broken lines) than predicted, which is attributed
to edge effect. Observer difficulty in recording all species is indicated by the trend of fewer observed species than predicted for the larger quadrat
sizes (greater
than 20 rrr').
49
S. Afr. J. noi., 1987.53(1)
47
6
100
Ii
.D 5
·E
~
90
c
.,
.s
4
~
3
a.
80
~
ct)
c
os
2
E
1
-
II>
1:
70
60
c
C>
·4
-3
-2
o
In area
-1
2
3
5
4
E
so
e
40
eo
6
(rn 2)
en
C>
·u
30
II>
Figure 3 The relationship between In species number and In area for
eight samples, showing the similarity in slope (m). Sample A6 is atypical.
a.
en
~
c
Z
Table 2 The number of species
recorded in nested subsamples with
predicted values in brackets (C
2,5147 derived from 128 m2)
53
82
120
165
199
213
274
2
(53)
(80)
(122)
(184)
(280)
(424)
(481)
9
<ii
8
1:
7
N
E
II>
.~
o
II>
a.
6
5
139
100 y
4
II>
'0
~
3
Y =
II>
.D
E
5
6
7
8
9
10
frequency). Too small a subsample size would not include
sufficient high frequency species for the relationship in Figure
5 to be valid as each subsample would include only a fraction
of the high frequency species resulting in a greater number
of new species for successive subsamples. In contrast, too large
a subsample size would include too many of the less frequent
species with the same result. The percentage difference in the
increment between subsamples 3 and 4 is 11,5 - 8,3 = 3,2%.
This difference should be reflected in a limited number of
new species. If the number- is only a fraction, then the
subsample size is too small and conversely, if too large then
the subsample is too large. For convenience and to allow for
some local vegetation variation 3,2070 difference can be taken
to represent 3 new species. This difference can also be
expressed as percentage accumulated totals between subsamples 4 and 3 where the first subsample is taken as 100%:
151,6-139 = 12,6% (Figure 6). The number of species (y)
in the first subsample with difference of 12,6% = 3 new
species between subsamples 3 and 4 is, therefore:
10
>
4
Figure 5 The relationship between number of subsamples and new
species increment expressed as a percentage of the accumulated total.
Subsample increments: I = 100OJo;2 = 18,8OJo;3 = II ,5OJo;4 = 8,3OJo;
5 = 6,5OJo; 6 = 5,3OJo;7 = 4,5OJo;8 = 3,9070;9 = 3,5OJo& 10 = 3,IOJo.
11
o
3
Number of subsamples
Correlation co-efficient (R) between recorded and
predicted values: R = 0,9725
~
10
0
'umber of species
128
512
2048
8 192
32768
131072
200 000
20
2
+ 3
3
0,126
:l
= 23,81 species
C
C
ca
II>
E
-5
-4
-3
-2
-1
0
2
3
4
5
6
In area (m2)
Figure 4 The relationship between In c values, derived from different
areas and corresponding In area showing the increase in c values to
2 m2 (In 2 = 0,693) representing sample edge effect.
of less than 10070 to allow for local heterogeneity. A minimum
of four subsamples is recommended because if an increment
of less than 10% is achieved with fewer subsamples local
disturbance could be indicated.
Species distribution in a stand varies from closely spaced
species (high frequency) to I species in the stand (low
If c = 12 species m - 2 then the area (x) corresponding to 23,81
species is:
x
=
=
(Jny-Ine)
e -'-~---'.111
9,82 m2
Subsample size is, therefore, 9,82 rrr'. This relationship also
takes the diversity of the vegetation into account because
subsample size increases as diversity (c value) decreases. The
subsample size is also within the range of: greater or equal
to 2 m2, for reduced edge effect; and less than 20 m2 for
observer efficiency, in the vegetation concerned in this study.
For less diverse vegetation, maximum subsample size could
be increased if required. To· ensure a width to length ratio
50
S.-Afr.
48
200
180
160
~
140
CO
"0
"C
s
120
100
CO
"3
E
80
:>
o
o
co
60
'"
.s
40
o
8.
'"
~
20
q)
Z
234
Number
5
of
6
7
8
9
10
subsamples
Figure 6 The relationship
between number of subsarnples and accumulated total of new species expressed as a percentage.
Subsarnple
accumulated
totals: I = 100070; 2 = 123,20:0; 3 = 139,0070; 4 =
151,6070; 5 = 162,10;'0; 6 = 171,10;0; 7
= 193,30;0 & 10 = 199,50;'0.
=
179,20;'0; 8
=
186,5Ojo; 9
of 1:2 for rectangularity of subsamples (Mueller-Dornbois &
Ellenberg 1974) the following convenient sizes can be used:
I m x 2 m (2 rrr')
I,S m x 3 m (4,S rrr')
2 m x 4 m (8 rrr')
2,S m x 5 m (12,5 m')
3 m x 6 m (I8 m2)
3,S m x 7 m (24,S rrr')
The nearest convenient size which is not less than 9,82 m2
is, therefore, 2,S m x 5 m (12,5 rrr'). In the first three releves
recorded in the investigation of the vegetation ecology of the
Transvaal Waterberg (Westfall, in prep.) using this subsample,
size increments of less than 10070 were reached with four
subsamples in two cases and with five subsamples in one case.
These results substantiate the validity of the subsample number
and area relationships. Species recorded in each reIeve aried
from 38 to 60 and sample areas covered varied from 50 [Q
62,S rn". This is a considerably higher number of species than
that recorded by Westfall (1981) in vegetation with similar
diversity using single 200 m2 sample areas. Separate random
subsamples, therefore, appear to increase efficiency. The
smaller area also facilitates cover estimation of the herbaceous
stratum. It is noteworthy that in all three releves (Westfall,
in prep.) only approximately 10070 of the total number of
Tydsk r , Plantk.,
1987,53(1)
species estimated to be present in each 20-ha stand was
recorded. Recording of all species within the stand would have
been prohibitive in terms of time and effort. It can, therefore,
be assumed that the remaining 90070 of the species have a very
low frequency and that the so-called 'infrequent species' in
phytosociological classifications are often not as infrequent
as assumed. The method proposed here is also objective in
. distinguishing species, which on a frequency basis, should be
more relevant in classifications. In randomly selected subsamples each species chance for inclusion in the subsamples
is proportionate to its frequency.
The alue of 0,3 for slope (171) can be refined with more
data for improved predictions especially for areas greater than
20 ha.
The expressions described in this paper, as well as those
describing area, sample number and scale relations (Rutherford & Westfall 1986) have, for convenience, been programmed for the SHARP PC ISOO computer. This program,
titled VEGFORM, is available from the first author on
request.
Conclusions
The advantages of expressing species diversity in terms of the
c value include:
(I) overcoming the edge effects of small quadrat size in certain
vegetation types; thereby
(2) improving comparability of species diversity when expressed as species per m2;
(3) associating subsample size with species diversity;
(4) determining number of subsamples required for sampling
a vegetation stand; and
(5) predicting the approximate number of plant species which
could occur within a vegetation stand.
Acknowledgements
The authors thank Dr J.e. Scheepers, Dr M.e. Rutherford,
Dr H_ van Ark and Mrs J. Schaap for assistance.
References
MUELLER-DO\IBOIS,
D. & ELLENBERG,
H. 1974. Aims and
methods of vegetation ecology. Wiley, New York.
PARTON,
w.r. & INNIS, G.S. 1972. Some graphs and their
functional
forms. U.S. International.Biological
Program,
Grassland
Biorne Project Technical Report No. 153: 1 -41.
RUTHERFORD,
M.C. & WESTFALL,
R.H. 1986. The Biomes
of southern Africa - An objective categorization.
Mem. bot.
Surv. S. Afr.
o. 54.
WESTFALL,
R.H. 1981. The plant ecology of the farm
Groothoek,
Thabazimbi
District. M. Sc. thesis, Univ. of
Pretoria.
WESTFALL,
R.H. (in prep.) The vegetation ecology of Sour
Bushveld in the Transvaal
Waterberg.
51
3.2 FIELD
SAMPLING
The purpose
of field
essing
logical
and
study. Because
subsistence
sampling
interpretation
field sampling
costs,
savings
in the field,
through
3.2.1 Sampling
unit
This
refers
vegetation
is to acquire
to
that
the
often incurs additional
improvements
location
and
achieve
can be effected
sampling
stand
will
data,
with
aims
procof
the
transport
and
by reducing
time
spent
in methodology.
and number
unit
not stand
location
(sample
plot)
location,
as described
within
a
in section
3.1. 5.
3.2.1.1
Background
Where
working
stand
can
quadrat
most
be
scale
efficient
the
carried
the
a single
they
plots
then
representativeness,
is
quadrat
location.
vegetation
stand
used for sampling
locations
(sampling
represent
scale-related
quadrat
units)
unless
out when the question
vegetation
selection
probably
However,
the stand,
by Werger
should
random
where
is far larger
will reduce the number
to be made. This is confirmed
"sampling
stands
of
quadrat
ization of sub-quadrat
that
on
scale-related
that
by
based
method
area of a single
decisions
such
represented
location,
area of the
is
then
than
the
the
the
random-
of observer
(1974) who states
be representative
or systematic
of representativeness
of
sampling
of
is
is by-passed."
52
north
/\
Contour
...................
(downslope)
2
r-------------------,
3
.
Sampling
unit
Contour
...................
(upslope)
FIGURE
53
'--------------------'
1 Reference corner
.
4
3.2. - Reference corner (1) of sampling unit in terms of
direction, for level terrain; and slope, for sloping
terrain.
In this
3.1.6)
study,
to
where
sample
difficult
to
2,5
a 20
x 5 m sampling
ha
determine
stand,
units
are
used
representativeness
and
random
sampling
point
in
field
(section
is
unit
extremely
placement
is
required.
3.2.1.2
The
Methods
stand
centre
intercepts
units
on
The
method
stand
can
duced
in
metres,
be
printed
by standardizing
re-location.
presented
corner
sides
then
with
by the sampling
vice
prior
to
stand
can
where
be
is the
versa.
field
corner
centre
is a con-
used
3.2).
south
to
A
distance
generate
is a
from
list
of
these
sampling,
to
save
can be
the
co-
time.
further
re-
corner.
This
in terms of sampling
unit
is that point
unit co-ordinates
is the
as
system based
unit reference
monitoring
(Figure
centre
the x co-ordinate
unit orientation
corner
grid
of sampling
stand
on the ground
re-
and is the left upslope
unit, when facing downslope,
downslope
reference
the
3.1.5.3)
or
the
A co-ordinate
co-ordinate
vegetation
of the sampling
the
y
Placement
by
unit.
on the sampling
The reference
orientated
a sampling
for sampling
can also facilitate
from
for this purpose,
the
represented
3.1.5).
simplest
(section
and
Decision-making
be
distance
of locating
bearing
ordinates
(section
can
and
PHYTOLOC
centre
the
the stand is circular.
co-ordinates
compass
a
stand
bearing
program
random
the
because
compass
venient
is
and co-ordinates
within
reference
applied
If the
west
with the longer
ground
corner,
is level,
when
facing
54
north,
with
the longer
3.2.2 Plant
Plant
sides orientated
identification
identification
field.
These
and verification
aids
are
can be portable
photostats
of
due north.
herbarium
often
used
by
field herbaria,
specimens
and
researchers
in
photographic
notes
on
the
records,
identifying
characteristics.
3.2.2.1
Background
Although
it
specific
and
reference
these
purposes,
is
are
The
identification
aids,
purposes,
cation
knowledge,
of standardization
states
used
studies
is
for
known.
if the
and inhibit
which
by
the
evident
for identifying
plants
for
identification
of
with
a
voucher
reliance
of
for
if the criteria
most
whereas,
can
example,
for
identifi-
to each field researcher.
in the character
are
This
of species
in published
of
area
material.
study,
the transfer
is also
study
for comparison
sterile
given
is unique
be repeatable
used
specimens
on fertile material,
on
a
a
Comparison
exacerbated
repeat
in
field
criteria
often
voucher
recorded
criteria
are
to
identifications
collect
such as floras,
identifications
monitoring
to
taxa
made
problem
efforts
plant
the
rarely
complicate
tion
practice
insufficient
given.
field
common
infra-specific
plants
specimen
not
is
A lack
and character
works.
Can vegeta-
used by researchers
for
in the field, are not made known?
55
Botanical Journal of the Linnran Society (1986), 92: 65-73
A new identification aid
combining features. of a polyclave
and an analytical key
ROBERT H. \\ ESTFALL,
}.tfICHAEL D. PANAGOS
H 'GH F. GLEN A.ND
Botanical Research Institute, Department of Agriculture and rVater Supply)
Private Bag XIOl) Pretoria 0001) South Africa
Received December 1984, accepted Jor publica/ion February 1985
WESTF.-\LL,
R. H., GLE?\', H. F. & P.~~AGOS,
xr. D., 1986. A new identification
aid
combining
features
of a polyclave
and an analytical
key .. -\ new identification aid combining
features of a polyclave and an analytical
'er is described. It is based on presence-or-absence
characters and is presented in the form of a matrix, with the characters in rows and the taxa in
columns. The PHYTOTAB
program package is used to order the matrix, in order to facilitate
identification. The method was used to construct an idcmification aid for vegetative material and
has wider taxonomic and teaching implications.
ADDITIO?\'.-\L
PHYTOTAB
KEY
\\'ORDS:-Computer-aided-identification
-
matrix
-
phytosociology
-
CO:-';TE:-';TS
Introduction
.
Existing identification
Analytical keys
Polyclavcs
Mcrhods
.
Results
Discussion.
Conclusions
Acknowledgements
References.
65
66
66
67
methods.
68
70
70
72
72
73
INTRODUCTIO?\,
During the course of work on a pilot study of the Transvaal \Vaterberg
(Westfall, 1981) the need for an objective method of site identification of the
numerous plant species arose. The use of field herbaria proved time-consuming
and cumbersome. Existing keys would have been of limited assistance because,
first, they often use evanescent characters at an early stage (e.g. requiring
flowers out of season) , and secondly, they are only available for a minority of the
0024-4074/86/010065
+ 10 S03.00/0
65
© 1986
The Linnean
Society of London
56
66
R. H. WESTFALL ET AL.
taxa present In this study area. \ hen a specimen is collected for later
identification by a reputable herbarium, the onus of identifying correctly the
same taxon at other sites rests entirely on the knowledge, memory and
experience of the field ecologist. During ph tosociological work, vegetative
characters are necessarily used for identification, whereas vegetative keys are
rarely available.
Bearing these problems in mind, we decided to investigate alternative
methods of identification, a subject which had been pursued by one of us in
some detail (Glen, 1974). Traditionally,
taxonomists have used one or another
form of dichotomous key for identification, and the alternative methods have
been gi\'en very little practical use. A selection of identification strategies is
discussed below.
EXISTI;\C lDE~TIFIC.\TIO;\
:-'!ETHODS
A wide variety of different manually operated identification systems has been
proposed for use in biology (Leenhouts, 1966), as well as a small but growing
number of computer-aided
systems (Morse, 1971; Pankhurst,
1978). These
include such apparently diverse systems as the 17th-century precursors of the
indented dichotomous key described by Voss (1952) and the computer-aided
system described by Morse (1971), in which the computer interrogates the
operator. All these techniques belong to one or other of two groups: analytical
keys, which have only one entry point, and polyclavesj this term, coined by
Duke (1969), is used here in the wider sense given (0 it by Morse, 1974), which
have many entry points. The method described later in this paper is unique in
being usable as both an analytical key and as a polyclave.
Analytical keys
Analytical keys are easily reproduced,
being printed documents,
and
relatively easily optimized, the theory of optimization having been first worked
ou t by Lamarck (1778). Possibly the most accessible review of these is by
Osborne (1963). A dichotomous key can be used with maximum efficiency if
each step splits the remaining group of taxa into two equal subgroups. In
general, a polychotornous step with N branches at any point in an analytical key
operates at maximum efficiency if it splits the group into 'IV equal subgroups.
This means that the mean path length from the start of the key to the level at
which an identification is achieved is minimized.
.
Attempts have been made (Morse, 1971, 1974) to weight different taxa
according to their relative commonness or rarity, so that common taxa are
keyed out with the shortest average path length, and to \V'eight characters so
that the easiest to observe are used preferentially. Taxon weighting would tend
to minimize the number of errors per unit time on the assumption that the key
will be used more often to identify common taxa than rarities. Character
weighting tends to minimize the chance of using a difficult character near the
start of the process, and so it minimizes the possibility of using the wrong major
section of the key, thus minimizing the number of errors per identification.
Analytical keys do have practical disadvantages,
despite their clear theory
and apparent simplicity. These are largely related to their single entry point and
57
A CO~IBI~ED POLYCL\VE
A~D .-\..."IALYTIC.-\LKEY
67
limited
number
of paths to the correct
identification.
The most frequently
encountered
problem is that it is very difficult to construct a key that will enable
one to identify a fragmentary
(sterile) specimen unambiguously
if it belongs to a
large group, for example the Mesembryanthernaceae.
Keys to the genera of this
family have been constructed
both by Bolus (1958) and by Herre & Volk in
Herre
(1971). Both of these require flowers and ripe fruit for certainty
in
identification,
yet both parts are seldom available at the same time.
Variable
taxa present a problem in the construction
of analytical
keys. If one
allows for the variation
by allowing the taxon in question to key out in more
than one place,
then one degrades
the overall
efficiency
of the key by
lengthening
it. If one allows for variation
by describing it where it occurs in the
key, then efficiency is degraded
by basing a dichotomy
on a property
which is
less than desirably clear, or by lowering the ease of observation.
The addition
of taxa to an existing
analytical
key is a relatively
severe
problem involving the rewriting of at least part of the key. If the additions are
more than minor, it is probably
most efficient to rev, rite the key from the
beginning,
if optimization
is to be retained.
Computer-constructed
keys can
achieve this end more easily, for example those of 110rse (1974) and Pankhurst
(1971).
Ease of observation
optimization
requires that a minimum
of characters
be
used at each step. In fact, the use of more than one character
in a step implies
that-either
the characters
used are not fully independent,
or that one or more of
those used are ·ariable. Rypka el at. (1967) ha ve shown that it is theoretically
possible to construct
a key to any number
(T) of taxa using not more than N
characters,
where N = log? T, rounded
up to the nearest integer.
In fact,
considerably
more characters
than this minimum
number are usually used, but
the number of characters
rarely exceeds the number of taxa. Therefore,
it may
happen
that a highly distinctive
and potentially
useful character,
for example
flower colour or an accurate locality label, is present on the specimen but is not
used in the key.
Polyclaues
A polyclave has as many entry points as there are characters,
and an effectively
infinite number
of pathways
from the start to identification
of the specimen.
Very little skill is necessary to identify a plant using a polyclave, but the speed
and accuracy
of identifications
improve significantly
with increasing
skill. In
general, it appears that the most efficient strategy for using a polyclave is at
each step to use the rarest character
that is available and still unused.
Sneath & Sokal (1973) divide polyclave
algorithms
into simultaneous
and
sequential.
The basic strategy of a simultaneous
polyclave is to calculate
the
value of one or another similarity function between the unknown specimen and
all taxa in the identification
matrix. The value of this function should be high
with only one taxon, to which the unknown
may be assigned if the value is
above a predetermined
threshold.
Alternatively,
taxa are considered
to be
regions in character
hyperspace,
and the locali ty of the unknown in this space
may be calculated.
The unknown is assigned to a particular
taxon if it falls into
that region of the space. Typical simultaneous
methods are those of Gyllenberg
(1965) and Lapage et al. (1970, 1973).
58
68
R. H. \\'ESTFALL ET AL.
In sequential
polyclaves,
characters
are considered
one at a time. In each
step, a subset of taxa having the character
in question in the same state as the
unknown
is extracted
from the complete
set, and compared
with the subset
resulting from the previous step. Those taxa common to both subsets form the
set of possible identifications
for the next step. \\ hen this step contains only one
taxon, the unknown has been identified.
A wide variety of such polyclaves has
been described
by Leenhouts
(1966). Essentially,
they fall into three groups:
printed,
card- and computer-operated,
although
the same polyclave
may be
converted
from one form to another.
Printed polyclaves may be even cheaper
to reproduce
than the equivalent
dichotomous
key, because they can be compressed
into less space. For example,
the key of :'leyer
(1969)
to the families
of flowering
plants
of SW
Africa/Narnibia,
a polyclave,
occupies about 220 lines (excluding
explanatory
figures), while the dichotomous
key by Merxrnueller
(1972) to the same taxa
plus 15 families of Pteridophyta
occupies about 1500 lines. However, a polycla ve
is frequently
more expensive to reproduce.
/\. form of polyclave on plastic cards has been described
(Leenhouts,
1966),
but polyclaves on paper cards are far commoner.
Most punch-card
polyclaves
fall into two groups: centre-punched
one card per character
state, and edgepunched
one card per taxon. The former kind is exemplified
by the keys of
Bianchi (1931) (the first punch-card
polyclave},
Hansen & Rahn (1969) and
Weber & Nelson (l972); the latter by Baker (1970), a key to the species of the
genus Erica.
Wi th the possible exception
of com pu ter-genera ted punch-card
polycla ves,
this kind of key is very expensive to reproduce,
as accurate punching dies must
be made for each card in the key. \\ ith computer-generated
keys, standard
80column cards are punched
with a standard
card punch, and the key can be
produced
for a realistic price. With the almost total extinction of the computer
card reader-punch
as a result of improved
magnetic
storage media, it is not
certain whether computer-generated
punch-card
keys will remain viable for any
length of lime.
Edge-punched
card keys can be operated
with a mechanical
card sorter or
with a skewer, but are more subject to damage than centre-punched
cards. On
the other hand, the latter do not usually have any space for annotation.
Computer-aided
identification
systems (automated
polyclaves) are many, for
example
those of Boughey,
Bridges & Ikeda (1968), Goodall (1968), Morse
(1974), Pankhurst
(1978) and the simultaneous
methods
mentioned
above.
Each has its own advantages
and disadvantages,
but at present none can be
used in the field. It is possible that in the future one or more may be made to
work on a portable personal computer.
Morse (1975) has given a particularly
lucid account of the computer-aided
systems available at that time. \Ve plan to
add to the system described here, routines enabling it to be used in the field on a
. micro-com pu tel".
:-'lETHODS
Eighty plant specImens
were collected
in the Transvaal
Waterberg
during
March
1983. The specimens
were numbered
and sent to the National
Herbarium,
Pretoria (PRE) for identification.
59
--------------~------.---~
Table I. A preliminary
vegetative key for some species of the Transvaal
Growth tips
C:haracters
+ pre-sent
• sometimes
Waterberg
Venation
present
"
~
o
~
>
>
'"
;;
E
o
o
.Q
.Q
1
'0
~
>
o
o
u
"
U
es
'iij
'0
"
IV
"
C
C
0
o >
C .c
~11
"
> ""
'" >
.....
u
"
0
Species
Trees/shrubs
Rhus teotoaictvs
Rhus sp. ct. R. dentets
Ozoroa oeoicutose
Rhus sp. ct. R. dentste
Rhus keettii
+
+
+
+ +
Cambretum erythrophyllum
Cambrerum zeyheri
Vangueria ;n{Busta
Vangueria inFausta
Cambretum zeyheri
Tapiphyllum parvifolium
Pseudolechnosty/is maprouneifolia
Vitex pooara
Scbrebere alata
Syzygium coradatum
$yzgyium coredeturn
Schrebera alata
Croton gratissim~5
Olea europaea subsp. africana
Syzygium cordatum
Syzygium corda rum
Cambrerum motte
Fadogia monticola
Mimusops zeyheri
Apodytes dimidiata subsp. atmtateis
Mundu/ea sericee
Rhoicissus digitata
Elephantorrhiza burke;
+
+
+
Ferns
Pel/aea calomelanos
Cheifanthes viridis
Cttettenttves
virdis
+
+
+
+
+
+
+
+
+
+
.+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
Hereropyxis nacalensis
- ~S3:igna
rtexstotnrs monopetalus
Turraea ob:usifolia
Rothmannia cepensis
Vangueria cyanescens
Forbs
Nidorella reseditotie subsp. resedifolia
Helichrysum kreussii
Oicoma anomale subsp. enamels
Oldenlandia neroecee
Zinnia peruviana
Agathisanthemum bojeri subsp. sustrete var. australe
Berterie oretoriensis
Penranisia angustifolia
Phyllanthus parvulus
Solanum panduriforme
Zornia milneana
Cleome maculara
Cassia quarrei
Scabiosa columba ria
Pavon;a cransvaalensis
Stachys natalensis va-. galpinii
Uppia rehmannii
Triumtette sonaeri
+
+
+
Vitex rebmsrmit
Climbers/creepers
Crvptotepts oblongifolia
Lando/phia capensis
Rhoicissus cridenrara
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+ +
+
rilutolhUII"hll
~'
01)1111 tlf jl.,lIldutlrHllHHJrI
o·
,
Rotlculodromous
Hyphodromous
~
~
n
rt
Brochidodromous
3-
Cra spcdodromous
rt
...,
Eucarnorodrornous
Asvrnmemcal
base
Asymmetrical
lamina
~
~
...,
Oblong
Narrow I/obiong
rt
Leaf form
<1>
U">
Obovate
U">
Ovalo
::;
...,
Elliptic
C
Cuneate
0"""
Lobate
U">
Decurrent
Q.
Leaf base
Cordate
:3
Obtuse
0"""
<1>
Acute
jA
Even pinnate
Multitolinte
Twice pinnate
o
...,
Leaf type
<1>
<1>
Odd pinnate
Digitately
'"0
trifoliate
<1>
...,
Simple
~U">
Opposite
Whorled
0'
...,
Leaf arrangement
Decussate
Alternate/scat
0"""
Y'
rered
Very dlffcrel"'ri
Colour
Similar
Texture
-I- -I-
+-+-+
:., +
+
+1 +1
-I-
-+- •••
+ .•. + +1 +
-I-
+
......
+ .•. +
+++++
++++++++++
+
-I. ""
+
·1
+
+
+
+
01'0)(
+
++
Acute
Eroso
Serrate
...
+
Dentate
...
.•....
novorurc
Leaf margins
Crunute
Lobed
+ + -+ -+ .•.
'1-
+1 + + .•. + +t
+ + +
-I-
+
-I-
+ +
-I- -+- •••
+
-I- -I-
+ + +
-I- .•..•.
-I
1-
+ +
-I-
1 .•.
-+-+++++++++++
-I- .•. -+-
+
-I- -I- -I-
+
.j
Entire
Amolexlcnut
+
-I-
~+ + +
Leaf
Obtuse
+
'
Mucronate
-I- +1-1-
-I-
+ + ++
Snssue
-I
",-1
+++++++++++++++++,+++++++++++++++++
Petiolate
(b>
...,
::l
U">
Truncate
Rounded
-I-
Leaf surfaces
Leaf attachment
.-\ CO:-"IBI;\ED POLYCLWE
·
.-\:\D .-\:\.-\LYTIC.-\L KEY
69
Characters
relevant to each specimen were observed in the field and recorded
on a field data sheet on which provision
is made for recording
the collector's
number, provisional name and locality or releve. If more than 60 characters
are
recorded additional
page(s) could be used.
The morphological
characters
used in the preliminary
key were taken from
various sources such as Henkel (1934), :\letcalfe
& Chalk (1979) and Tainton,
Bransby
& Booysen (1976) and condensed
into an illustrated
photocopied
booklet for field use. This proved
useful in maintaining
consistency
in the
allocation of characters.
The PHYTOTA.B
program
package
(\ \ estfall el al., 1982) was used to
construct
the matrix. The data from the field data sheets were transferred
to
encoding forms using the PHYTOTAB
input format where each character
was
regarded
as a releve. Taxa were grouped
into trees and shrubs, climbers and
creepers, forbs, ferns and graminoids
for convenience.
This would also allow for
each group to be 5ho\\"I1 on a separate
table when more data are collected.
Characters
can also form groups such as leaf-margin
patterns
and venation
patterns. The character
groups are sequenced
in descending
order of character
(group) frequency from right to left. Characters
not in groups are on the left of
the table in frequency
order (Tables I & 2). Spaces were inserted
between
characters
to facilitate vertical reading.
Taxa are sequenced within groups so that dichotomies
may be formed. That
is, taxa with similar characters
are grouped
together.
The dichotomies
are
continued
until they consist of a single line for each taxon, and are ended with a
vertical
blocking
line after the first unique
character
for each taxon when
compared
with the two adjacent
characters.
Blocking was done by hand. If a
taxon has two or more character
states that express possible variation
within a
character
group falling within the blocked area, the species may be repeated
to
facilitate use of the key. This is illustrated
in our example by Combretum reyheri
Sond., in which the leaf apice
may be either obtuse or mucronate,
and by
S)I<)gium cordatum Hcchst., in which the leaf variability
is so great as to require
that the same taxon appears four times (Table I).
To use the table, the specimen
to be identified
is compared
with the
characters
gi\'en for the group to which the specimen belongs, starting from the
right-hand
side of the table. The user determines
whether
the character
is
present or absent in the specimen.
If the character
is present in the state
described,
then the user follows the 'present'
pathway
indicated
by +. If the
character
is sometimes absent it is still shown as +. If it is always absent. then
the 'absent'
pathway,
indicated
by blanks, is followed. The next character
is
found to the left of the newly completed
one. This procedure
is repeated
until
the end of a block, indicated
by a vertical blocking line, is reached. The user
should now be able to look across to only one name on the left-hand side of the
table. This will gi\'e him the desired identification.
The blocked pathways
indicate
the minimum
characters
needed
for identification.
Additional
characters
outside the blocks may be used for corroboration.
Two other possible methods of using the table suggest themselves. However,
these methods should only be necessary
as a teaching aid, or if the specimen
lacks important
characters
or in a preliminary
version of the key. It must be
stressed that these methods are by no means obligatory.
First, it is possible to use the table as a form of tabular key (Newell, 1970).
61
70
R. H. "'ESTFALL ET AL.
A profile of the unknown specimen is obtained by noting the characters present,
in the order of the table, and this profile is compared with those of the taxa in
the table, one by one, until a matching.profile is found. This would be the name
of the unknown specimen. This method is probably too cumbersome to be of
practical value.
Another method of using the table would be as a classical pol yclave. The
unknown specimen is exarni ned for any conspicuous (or unusual, if one is
familiar with the table) character, and the taxa which display this character in
this state are determined. Another character is searched for, to see which of the
taxa found in the previous step display this character in the required state. The
specimen has been identified when only one taxon remains.
RESCLTS
The problems discussed in the introduction were overcome by devising a
method for constructing vegetative keys during the reconnaisance
phase of
phytosociological work. The method had to allow for increasing the number of
taxa as the work progressed. It had also to be simple to apply in the field. The
key presented here will operate ati factorily only in the geographical area
concerned; it is used in this paper solely as an example.
In Table 2, Schiracliyriuni sanguineum (Retz.) Alst. in the Graminoid group has
an open pathway indicating that this species does not key out with the available
characters. However, its identity can be inferred from the absence of flattened
stem bases which are presen t in Hyparrhenia filipendula (Hochst.) S ta pf var. pilosa
(Hochst.) Stapf.
DISCCSSIO:'-i
The matrix method of key construction described in this paper allows for
biological variation where more than one character state, such as different leaf
forms, may be present in the same taxon, without detracting
from the
effectiveness of the key. Characters which are periodically available for which
the population is variable may be used, as well as corroborative characters. The
right-to-Ieft method of use described above in the Methods may seem at first
sight to' be. idiosyncratic. This method is used because the phytosociological
example makes use of the species name file for both vegetation and taxonomic
classification. The PHYTOTAB
program package allows the user to print
species names to the right of the table as well as printing the matrix in mirror
image, thereby reading from left to right. Thus, in ordinary taxonomic usage
the table would be rearranged to read 'correctly'. A further advantage of this
package is that in the printout form, both horizontal and vertical lines may be
employed to improve legibility.
This form of key is unique in that it can be used as both a polyclave and an
analytical key. It combines the flexible strategy of the former with the ease of
reproduction of the latter to produce a very efficient identification system. One
advantage of polyciaves is that either taxa or characters can be added at will,
with minimal effort required, although the addition of taxa may result in
ambiguities in identifications, which may only become apparent after long use.
By using PHYTOTAB, both taxa and characters can be added to existing keys
62
Table 2. A preliminary
vegetative key [or some species of the Transvaal
Waterberg:
Graminoid
I
o
.~
s:
s
I
-3
~ I:.:;
VI
V)
0)
I i~
zI
~
ro
~
VI
~
(1J
"'0
~
Q)"O
ceo..
... w
E
~roEE
~~
~
E_nJ,.-
V)
VI
~
C
(I)
~
g
'6..
~
1J
(I)
Q)
••.•
cu
B
Q>-:J
>-.§ 9
.= 1J
~
~
-5 ~ ~ -g --g ~ ~
';::.S
-5
E ~ e e- "Cl
6 ~
E g ~
:J
:J
Q)
QJ
.9 ~ .J:. ro :J C 0 ~ 0 J::. ro ~ c s:
~u~I~~~u~C~~Z~I~~~CZ~rororo~Oococ~~roC~Z
~
~
e.
A
.c ~
E ro .~ .2
~
~
.D
.s
a :;
~ ~
.D
0
.D
E
~
Z
~
c
trt
rororo~
(0'0
s
0
0
r-
Em
cuQ,)Q)CU:J
(I)
0
~.t::..r::.-
~ ~ ~ ~ :6 ~
g
V
A
II
~ ~ ~
ro
1'0
ro
~
eru
:;
o
~
'm
~
s:
:"Q
0
.Ql
"'0
e E
e
5s: ~= roe
~"O
C
~
ro '0
~
C
~
:J
VI
~
0
s~~u
Aristide canescens subsp. canescells
Brachiaria nigropedara
Loudotie simp/ex
Tnemede rriandra
Digiraria erienthe subsp. erienttu:
Sera ria perennis
Andropogon schirensis var. angusrifo/ius
Eragrosris racemosa
Digiraria eriontbe subsp. srofonifcra
Cymbopogon excavarus
Hyperrhe/ia dissolute
Dihereropogon
ampfecrens
+
+
+
+
+...
+ .•.
+......
+
+
+ + +
+ .•.
+
+
.•. .•. .•.
.•. .•.
+
+
+
+
+
+
+
...",
+
+
+
+
+
.•.
+
+
.•.
+
.•.
+
+
+
+
+...
+
.•..•.
+
+ +
+
.•.
+
+
prolixus
Seraria sphace/ara var. sphscetets
Pogonarrhria squarrosa
Setaria ustifata
I"''''
1+
Panicum maximum
0>
c.;l
""\j
~
r
()
~
Asrrida aequig/umis
Hyparrhenia fi/ipendu/a var. pifosa
Schizechvrium sanguineum
Perosis parens
Arisrida spectsbilis
Trecbvondrs spicara
Rhynche/yrrum serifofium
Trisrachya biseriara
Cymbopogon
::::::
to
...
0
Q)
~
III
Vi:g.fj
Q)
Oo!l.l
ell .0
o
o
E
~ ~ 'x
~
~
:.:;
I I
>-
3
VI
.E§
VI
(I)
~
+
+
+
+
+
+ +
+
+ .•.
+ .•.
+
+
+
+'"
+
+
+
+
+
+
+
+
+
+
+ +
+
+ +
+ +
+ +
+ +
+ +
+
+
+
+
+
+...
+ +
.•. +
+ +
+ +
+ +
+ +
+ .•.
+ +
+
+ +
+
+ +
+
+ +
.•..•.
+
-I- +
.•.
~
~
d
.>%
>-
s;:
....;
0
>-
l'
'"
:<
72
R. H. WESTL\LL
ET AL.
at will, and ambiguities
will be revealed when the tables are blocked. Correcting
a key is also a straightforward
task.
It will be noted that the horizontal
length of the block for each taxon is
roughly proportional
to the path length that would be required in an analytical
key to identify that taxon. Two uses for this proportionality
(which, it should be
stressed, is no more than a rough approximation)
suggest themselves.
First, it
offers a check on the discriminatory
efficiency of the characters
chosen. If all the
blocks are long, then redundant
characters
which do not contribute
to any
identifications
are present. These can be removed, or placed where they are only
used as corroborative
characters.
Secondly,
the proportionality
can be used to
demonstrate
Lamarck's
principles
of efficient key construction
to students
elegantly
and graphically
(Lamarck
1778; Osborne,
1963). This form of key
seems to be the only one available which demonstrates
its effectiveness with field
data.
A further teaching application
of the 'PHYTOTAB-key'
may be mentioned.
\\ ith the falling cost of computer
usage (at least in terms of constant units), and
with the relative ease and accessibility
of use of the PHYTOTAB
package,
a
method
is presented
of allowing students
to learn the mysteries of good key
construction
in the most effective
possible manner,
by making
their own
mistakes in the course of constructing
their 0\ n keys to taxa they think they
know.
The PHYTOT AB program
package (Westfall et al., 1982) allows the user to
construct a key to which taxa and characters
may be added as and when more
specimens
are collected.
The keys can be subdivided
easily to form tables of
convenient
size for field use .. An advantage
with computer
printout is that keys
may be duplicated
conveniently
and cheaply.
COl\CLUSIO)iS
Our identification
aid has peculiar and significant advantages
in three areas:
in teaching, in reconnaisance
and in the field. Its suitability
as a teaching tool
stems from the way in which the printed
table, while being usable as a
polyclave,
displays explicitly
the structure
of an analytical
key. Because the
PHYTOTAB
program
package is relatively simple, quick and inexpensive
to
use, students can be given practical
experience
of key construction
using this
package. For the same reasons, 'PHYTOTAB-keys'
can be produced
repeatedly
during the course of a taxonomic
investigation.
Ambiguities
in the key at each
stage may be taken to indicate
areas requiring
further research, and the keys
produced
will help in field-work while the group is under revision. In the field,
use of keys made by this method will mean that characters
present only in fresh
material
can be used to improve
the accuracy
of determinations.
It may be
expected that the facility with which both taxa and characters
may be added to
the table will be found useful in both taxonomic
and ecological work.
:\CKl\io\\·LEDGE:--IENTS
We thank A. V. Hall, O. A. Leistner, R. \. Lubke, E. J Pankhurst
and J. C.
Scheepers for helpful suggestions.
Miss A. P. Backer and Mr J. F. van Blerk are
thanked for assistance in the early stages of this work. The staff of the National
Herbarium,
Pretoria identified
the specimens collected.
64
A CO~IBINED
POLYCLAVE
.·\ND ANALYTICAL
KEY
73
REFERENCES
BAKER, H. A., 1970. A key for the genus Erica using edge-punched
cards. Journal of South African Botany, 36:
151-156.
BIA~CHI,
.-\. T. J., 1931. Een nieuwe dererminatie-rnethode.
Tectono 2-1: 884-893.
BOLuS, H. ~1. L., 1958. Key to the genera of Mesembryaruhernaceae.
Solts on Mesembryanthemum and allied
Gmera, 2: 392-403.
BOuGHEY,
A. S., BRIDGES, K. w. & IKEDA, .-\. G., 1968. An automated biological identification
key .
.Ilustum of S)'Slemalic Biology, Unicersity of California, Irnne, Research Series, 2: i-xix, 1-35.
DuKE, j. A., 1969. On tropical tree seedlings. L Seeds, seedling systems and systematics. Annals of the
.llissouri Botanical Gardens, 56: 125-161.
GLEN,
H. F., 197+. 'A taxonomic
re\'islon of the subtribe
Gibbaeinae
(~Iesembryanthemaceae)'.
unpublished
:-'I.Sc. Thesis, University of Cape Town, Cape Town.
GOODALL,
D. W., 1968. Identification by computer. BioSciena, 18: 485-488.
GYLLE?\BERG,
H. G., 1965. A model for the computer identification of micro-organisms.
Journal of Cmeral
Xl irrobiology, 39: 401-405.
HA!\'SEN, B. & RAH:-.i, K., 1969. Determination
of anciosperm families by means ofa punched-card
system.
Donsk botamsk .·Irchiv, 26: 1-46 & 172 punched cards.
HE~ KEL, J. S., 193+. A jitld book of the uoody planls of Xalal and ,(ululand. Durban: Robinson.
HERRE, A. G.J., 1971. Genera of Mescmbryanthemaceae: Cape Town: Tafelberg.
LA~I.-\RCK, J. B. A. P. ~L de, 1778. FloTt Francoise, 3 eels. Paris: Imprimerie Royale.
L.-\P.-\GE, S. P., B.-\SCO~IB, S., \\'ILCOX,
\\'. R. & CCRTIS,
~1. .-\., 1970. Computer idenuficarion
of
bacteria. In A. Baillie & R. J. Gilbert (Eds), Aulomalion, Mechanisotion and Data Handling in .llicrobiologr
1-22. London: Academic Press.
L.-\PAGF" S. P., BASCO~IB, S., \\"ILCOX,
W. R. & CCRTIS,
~1. A., 1973. Idcnrificarion of bacteria by
computer: general aspects and perspectives. Journal of Genera! Xl urobiology, 77: 273-290.
LEE~HOCTS,
P. W., 1966. Keys in biology: a survey and a proposal of a new kind. Proceedings of tlre
Koninklijke .\'fdalandsche Akademie mn fIItlroschappe Amsterdam, Series C, 6'1: 571-596.
~IERX\luELLER,
H. 1972. Prodromus finer Flora <'OnSueduest Afrika. Lehre: Cramer.
~IETC:\LFE,
C. R. & CH.-\LK, L., 1979 .. 4nalom)' of the Dicotyledons, \'01. I, 2nd edition. Oxford: Oxford
University Press.
~IEYER, P. G., 1969. Eintuehrung in die Pfianreniccl! Suedicestafrikas.
\\'indhock:
S.\\·.A. \\'issenschafdiches
Ccscllschafr.
~[ORSE, L. E., 1971. Specimen ideruificarion and key construction with time-sharing computers. Taxol/,20:
269-282.
~!ORSE, L. E., 19i4. Computer programs for specimen identification,
key construction
and description
priming using taxonomic data matrices. Publications of the .Ifuseum, .I/ichigan Stale University, Biological Series,
5: 1-128.
~IORSE, L. E., 1975. Recent advances in the theory and practice of specimen identification.
In R. j.
Pankhurst
(Ed.), Biological identification irith Computers: II-52.
London: Academic
Press (Systematics
Association special volume, 7).
NE\\'ELL, L ~L, 1970. Construction and use of tabular keys. Panfic Insects, /2: 25-37.
OSBOR?\E, D. V., 1963. Some aspects of the theory of dichotomous keys .• \·ew Phytologist, 62: 144-160.
PA!\'KHuRST,
R. j., 1971. Botanical keys generated by computer. 1I"0Isonio,8: 357-368.
P.-\;-.iKHURST, R. J., 1978. Biologicat Identification. London: Edward Arnold.
RYPKA, B. \Y., CL-\PPER,
\Y. E., BOWE!',
I. G. & BABB, R., 1967 .. -\ model for the identification
of
bacteria. Journal of General .lficrobiology, 46: +07-+2+.
S!\'L\TH,
P. H .. A. & SOKAL, R. R., 1973. Numerical T'axonomy. San Francisco: Freeman.
T.-\I~TO!\', x. ~I., BR.-\i\SBY, D. L & BOOYSE:-.i, P. de \1.,1976. Common Veld and Pasture Grasses of Xotal.
Picterrnar irzburg: Shurer & Shooter.
.
VOSS, L. G., 1952. The history of keys and phylogeneuc
trees in biology. Journal of the Science Laboratory
Denison Unicersity 13: 1-25.
WEBER, \Y ... \. &-NELSON,
P. P., 1972. Random access kty 10 the gfllera ofColorado mosses. Boulder: Univcrsiry
of Colorado ~ Iuscurn.
WESTFALL,
R. H. 1981. 'The plant ecology of the farm Croorhoek, Thabazirnbi
disrricr'. unpublished
~I.Sc. Thesis, University of Pretoria, Pretoria.
WESTFALL,
R. H., DEDNA\I,
G., VAN ROOYEN,
N. & THEROi\',
G. K., 1982. PHYTOTAB-a
program package for Braun-Blanquet
tables. Vegllolio,49:
35-37.
65
3.2.2.2.1
Improvements
Subsequent
improvements
to the techniques
of plant
identification
are:
i.
ii.
an illustrated
booklet
cation
study
in this
a program
of plant characters
(Appendix
was written,
gram package
identification
ing to the required
are standardized
to small
trated
attributes.
booklet
groups
is easier
sequence,
a family
The slide
incorporates
of the slide
the specimen
It was
to apply
namely,
if char-
from large
in the illus-
key; and,
by means
A photographic
of the plant when the voucher
accord-
the same characters
identification
comparisons.
pro-
for plant
species
is also used
forms. Using
of plant
slide/plant
is taken
in their
to produce
verification
graphic
sequence
and encoding
species
for key production.
This sequence
it is also possible
iii. visual
This program
dichotomies
found that the character
acters
for the PHYTOTAB-PC
3.6), for sequencing
purposes.
for identifi-
I);
as an option
(section
used
colour
specimen
number.
slide
is collected.
After
film, the slides are not mounted
of photo-
development
but returned
to
the film cassette.
After
ette
a plant
with
viewer
is identified
the
relevant
for comparison.
drive
Konica
light
transmission
places
the
FS
specimen
of the matrix
number
The field viewer
1 camera
with
lens.
facilitates
The
is
inserted
mirror
key, the
into
cass-
a field
used here is an old motor-
a} a translucent
and b) a magnifier
standard
The motor-drive
by means
lens
perspex
for viewing
system
is
also
back
for
which
re-
removed.
slide location.
66
3.2.3
Species
In this
study
for a plant
cision
ative
species
species
suitable
species
Generally
ever,
cover
as in this
cover is visually
greater
precision
3.2.3.1
the sample
sampling
with
change
and
a pre-
determining
rel-
a visual
estimation
estimate
is achieved
unit area for cover
attributes
and not sampled.
is required,
through
should relate
of presence
How-
and cover
samp-
to cover.
can require
unit areas.
&
estimating
species
scale
eleven
with
classes
and
advocates
the
Ellenberg
(1974)
cover-abundance
classes,
the
Daubenrnire scale
a nine class
(1976),
following
however,
list
such
several
as
the
Braun-Blanquet
with
six
scales
for
Domin-Krajina
scale
classes.
with
Werger
seven
(1974)
scale.
states
that
a cover
scale
should
meet
the
separated.
It
requirements:
i.
the scale
ii.
its points
values
iii.
are required
cover
Background
Mueller-Dornbois
Londo
canopy
estimated
than
and cover
ling, then the sampling
different
cover
projected
composition.
study,
Therefore,
to total
and cover estimations
for monitoring
species
where
cover refers
cover
should be sufficiently
should be related
so that these
and
abundance
detailed;
to the actual
can be treated
should
be
cover
arithmetically;
strictly
67
is not
a single
iv.
be
the
If the aim
an eleven
to these
class
scale,
other
quired,
such
relative
species
inversely
proportional
should,
hand
be
monitoring
requirements
for monitoring
estimations
cover
pattern
should
crude.
to
to effort
required
for
cover
or species
then
even
Precision
proportional
be
cover
cover
achieved
then,
scale,
or
an eleven
cover
class
between
to
possibly,
could
data
to
is re-
determine
class
scale
and
size
meet
con-
scale.
determination
size
class
change.
five class
of cover
change
could
be considered.
is
is
directly
for such determinations.
determinations,
composition
in
cover
first
in a matrix
processing
monitoring
composition,
too
in
with repeatability,
such as the Domin-Krajina
for
therefore,
required,
commensurate
if arithmetic
as
features
as possible.
and the ideal might be a simplified
the
to
be as simple
aim of cover
is to detect
fuse pattern
likely
should
such dissimilar
scale; and,
of precision,
added
However,
On
quantitative
the symbols
The criterion
also
to combine
logical
A balance
and
for
effort
example,
requirements.
68
Bothalia 18,2: 289-291 (1988)
.Miscellaneous notes
VARIOUS AUTHORS
THE PLANT NUMBER SCALE AN IMPROvtD METHOD OF COYER ESTIMATION USING
VARIABLE-SIZED BELT TRANSECTS
INTRODUCTION
.The vegetation ecology of the Transvaal Water berg is
currently being investigated at a scale of 1:250 000
(Westfall in prep.). Vegetation structure is being analysed
according to Edwards (1983) using the cover meter
(Westfall & Panagos 1984) for cover determinations in
each height class. In the floristic analysis, individual
species cover is estimated by using the Dornin-Krajina
cover-abundance scale (Mueller-Dombois & Ellenberg
1974).
A comparison of recorded species cover, being the
.sum of the class midpoints according to the DominKrajina scale (Mueller-Dombois & Ellenberg 1974), with
the structural cover should result in the summed species
cover for a stand being: 1, greater than the cover of the
height class with the greatest cover, and 2, less than the
summed cover of all the height classes, provided that 3,
quadrat size is such as to include those species contributing significantly to the total cover of the stand.
In the vegetation being investigated, quadrat size is
generally commensurate with species richness (Westfall
et al. 1987). However, the summed estimated species
cover was often considerably less than the mean cover
for the height classes with the greatest cover for the
same stands. Overcompensation for the underestimation
of species cover often led to the summed estimated
species cover being considerably greater than the summed
cover of all the height classes for the same stands (Westfall in prep.). Clearly, improved species cover estimations
are required for species cover to have any relevance other
than an approximate indication of relative abundance.
In estimating species cover, according to Edwards
(1983), the observer is often inclined to ignore grasses
without inflorescences and to estimate from a static
position without taking plant size and distribution into
account. For example, a larger plant should require a
larger area to be observed than a smaller plant. Furthermore, although mean canopy diameter can be readily
estimated it is often far more difficult to estimate mean
distance apart in terms of mean canopy diameter because
of often highly irregular plant distribution.
To overcome these problems, the approach suggested
here is based on a simple estimate of area and a count of
the individuals of a species within the area.
METHODS
The cover of a species is given by Edwards (1983) for
hexagonal packing by
90,7
c=-(n+1)2
where c = percentage crown cover and n = the mean
number of crowndiameters by which the plant crowns
are separated.
Assuming hexagonal packing, the transect area, of
which the percentage crown cover is a proportion, is
given by:
sin 60° (n+1) 30D
where D = mean crown diameter and 30 = the value for
obtaining a minimum of 0,1 % cover. Cover of less than
0,1 % is not considered significant. Transect length is,
therefore, 30D and transect width is sin 60° (n+ 1). In
practice, transect width was taken as slightly less than
the average gap between plants within or nearest to the
sample quadrat plus the mean crown diameter. The
number of individuals of a species was then counted
within the transect. Only species occurring within the
sample quadrat were recorded and for each a count of
individuals within a transect commensurate with each
species spacing and size was made. Counts of individuals
did not include the first individual as the transect was
started adjacent to the first individual. This permitted a
cover of less than 0,1 % where no individuals were
counted. Transect width was never greater than the
length as this could have resulted in actual cover values
of less than 0,1 % being given higher cover values.
The mean number: of crown diameters (n) by which
the plant crowns are separated within each transect is
given by n = ~
where I = number of individuals
counted. Percentage crown cover can then be calculated
according to Edwards (1983). Table 1 shows number of
individuals counted, representation by a single character
symbol and percentage cover for recording purposes in
the field.
Vegetation structure was analysed using the cover
meter and the summed species cover was estimated with
both the Dornin-Krajina scale and the plant number scale
as outlined above for five vegetation stands represented
by 21 quadrats. The mean of the shortest and longest
cross distances of each crown was taken as the crown
diameter for each species and these distances were noted
in four categories (Edwards 1983) namely, forbs (herbs),
grasses, shrubs and trees. Class intervals were selected on
a basis of trial and error to give an approximately normal
distribution of occurrences within crown diameter class
intervals. All estimations were done by an independent
observer.
RESULTS
The results of the crown cover determinations
given in Table 2.
are
69
290
Bothalia 18,2 (1988)
of plant individuals counted with single
TABLE I.-Number
character symbol and percentage crown cover
No.
Symbol
% Crown cover
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
>31
+
1
2
3
4
5
6
7
8
9
A
B
0,00
0,10
0,40
0,91
1,61
2,52
3,63
4,94
6,45
8,18
10,08
12,20
14,51
17,03
19,75
22,68
25,80
29,12
32,65
36,38
40,31
44,44
48,78
53,31
58,05
62,99
68,13
73,47
79,10
84,76
90,70
96,85
100,00
C
D
E
F
G
H
I
J
K
L
M
N
0
P
Q
R
S
T
U
V
W
The Fibonacci sequence, where each number is the
sum of the preceding two numbers, provided the closest
resemblance to a normal distribution of occurrences
within crown diameter class intervals. This is illustrated
in Figure 1 with a frequency polygon with the class
intervals on a natural logarithmic scale to reduce the
effect of increasingly larger class intervals.
The class intervals used for mean crown diameters
according to the Fibonacci sequence, are shown in Table
3. Transect lengths were determined by the midpoints of
each class interval.
INTERPRETATION
In Table 2 estimations according to the Dornin-Krajina
cover-abundance scale are considerably lower than those
for the single height classes with the most cover. The
plant number scale, in contrast, yielded higher values
than that of the single height class with the most cover
and lower values than the cover of the combined height
classes for each releve, except releve 37 _This indicates a
greater precision in estimating cover when using the plant
number scale as opposed to the Dornin-Krajina scale.
According to Westfall et al, (1987), quadrat size
should have been larger for the vegetation type represented by releve 37, but this was not apparent using the
Dornin-Krajina scale at the time of sampling. However,
simple summation of the values obtained by the plant
number scale in the field indicated inadequate quadrat
size. It is far too time-consuming to verify quadrat size
for each quadrat according to Westfall et al. (1987)_ The
plant number scale together with a structural analysis of
the vegetation provides a simple means of verifying
adequacy of quadrat size.
In the frequency polygons (Figure I) the peaks to the
. left of the central troughs for forbs, grasses and shrubs
are caused by a relatively higher proportion of 0,2 m
diameter crowns. This can be attributed to the observer
rounding off crown diameters to 0,2 m some of which
should have fallen into the 0,211 to 0,34 class. If class
intervals had been known at the time of recording, it can
be expected that greater care would have been exercised
in measurements where crown diameters were close to
class borders. The troughs mentioned are, therefore,
considered to be a result of measurement inaccuracies
which could be overcome by using class intervals.
The use of standard transect lengths as illustrated in
Table 3 should simplify transect length determination
and provide for variability in crown size. A further advantage could be the simultaneous counting of individuals
of different species with similar crown diameters and
spacing to save time. It is also suggested that a simple
counter be used for recording number of individuals for
each species as marking paper for this purpose requires
stopping at each individual recorded.
It must be emphasized that the parameter determined
here is projected crown cover and not projected foliage
cover which is more species and age-dependent.
The class 'r' on the Braun-Blanquet scale and '+' on
the Dornin-Krajina scale (Mueller-Dombois & Ellenberg
1974) both with 'solitary, insignificant cover' are difficult
to determine. Species with a single occurrence in a sample
quadrat often have significant cover outside the quadrat.
If a stand is defined, as in this study, as 20 ha (Westfall
et al. 1987), it is impracticable to determine whether a
species is 'solitary' within that area. The concept of
'solitary' is relative to the area defined. In the plant
number scale used here the lowest cover class is less than
0,1 %, which seems better defined than 'solitary'.
In contrast to the Dornin-Krajina scale, the plant number scale has proportionately finer subdivisions at the
lower cover values of the scale. This is of significance in
the South African context with often high species richness characterized by many dominant species with
generally lower cover in contrast to the few dominant
species with higher cover often found in the relatively
impoverished European vegetation.
TABLE 2.-Percentage
crown cover in five vegetation stands
represented by releves 33 to 37
Percentage crown cover
Releve
no.
Single
height
class
Combined
height
classes
Domin-Krajina
scale
Plant number
scale
33
34
35
36
37
55
31
44
41
63
100
104
103
86
146
27
25
19
21
17
60
74
79
53
35
70
291
Bothalia 18,2 (1988)
U)
25
:s
20
::l
-ij
'5
.S
~
10
••>.,
TABLE 3.-Class intervals of crown diameters according to the
Fibonacci sequence for determining standard transect
lengths
Crown diameter
class interval (m)
Transect
length em)
0,001-0,01
0,011-0,02
0,021-0,03
0,'031-0,05
0,051-0,08
0,081-0,13
0,131-0,21
0,211-0,34
0,341-0,55
0,551-0,89
0,891-1,44
1,441-2,33
2,331-3,77
3,771-6,10
6,101-9,87
0,15
0,45
0,75
1,20
1,95
3,15
5,10
8,25
13,35
21,60
34,95
56,55
91,50
148,05
239,55
c
e:;
5
u
o
o
-6
-5
-4
Mean
-3
crown
-2
-1
diameter
0
2
3
In (m)
FIGURE I.-Frequency
polygons of occurrence of plants in
mean crown diameter class intervals for [orbs (.:..), grasses
(---),
shrubs (-'-'-)
and trees (-).
Class intervals
are according to the Fibonacci sequence on a In scale.
CONCLUSIO S
The projected crown cover determinations based on
area estimations and counting of individuals shows
improved precision compared to the Domin-Krajina
cover-abundance scale. Although this method is more
time-consuming than a purely visual estimation of cover,
the use of standard class intervals and a counter should
decrease the time required for cover determinations. The
method appears more suitable for the species-rich South
African vegetation than the traditional European coverabundance estimation scales. The method also provides a
means of verifying quadrat size adequacy.
ACKNOWLEDGEME~~S
The authors thank Dr J.C. Scheepers for comments
and suggestions.
REFERENCES
EDWARDS, D. 1983. A broad-scale structural classification of
vegetation for practical purposes. Bothalia 14: 705-712.
MUELLER-DOMBOIS, D. & ELLENBERG, H. 1974. Aims and
methods of vegetation ecology. Wiley, London.
WESTFALL, R.H. & PANAGOS, M.D. 1984. A cover meter for
canopy and basal cover estimations. Bothalia 15: 241244.
WESTFALL, R.H., VAN STADEN, 1.M. & PANAGOS, M.D.
1987. Predictive species area relations and determination
of subsarnple size for vegetation sampling in the Transvaal
Waterberg, South African Joul7U11of Botany 53: 44-48.
WESTFALL, R.H. in prep. The vegetation ecology of Sour Bushveld in the Transvaal Waterberg .•
R.H. WESTFALL and M.D. PANAGOS
MS. received: 1987.07.31.
71
TABLE
3.1.
Symbols
Crown diameter
for
crown
Symbol
0,001 - 0,010
A
0,011 - 0,020
B
0,021 - 0,030
C
0,031 - 0,050
D
0,051 - 0,080
E
0,081 - 0,130
F
0.131 - 0,210
G
0,211 - 0,340
H
0,341 - 0,550
I
0,551 - 0,890
J
0,891 - 1,440
K
1,441 - 2,330
L
2,331 - 3,770
M
-
6,100
N
6,101 - 9,870
0
3,771
72
(m)
used
diameter
classes
3.2.3.2.1.
Improvements
Subsequent
3.2.3.2
improvements
include
diameter
derived
from
the
the allocation
classes
with the cover
to
and
of symbols
recording
symbols
the
methods
of
between
in
section
(Table 3.1) to the crown
these
for each species.
relationship
described
class
symbols
The following
mean
cover
together
data can be
and mean
canopy
diameter:
mean
cover
(%);
ii . mean
crown
diameter
i.
iii. individuals
(m);
per hectare;
iv.
square
v.
mean
spacing;
vi.
mean
canopy
radius
vii. mean
canopy
to canopy
It was
metres
found,
stituted
forbs,
if half
within
ii. shrubs
to centre
(m);
(m); and,
gap
(m).
of these methods,
that the pre-
cover estimations
could be improved
considerably,
strata,
if rooting
overhang
is, therefore,
grasses
within
a sampling
in the sampling
unit
and trees:
In this way vegetation
sub-
applied:
and dwarf
the sampling
was
unit.
shrubs:
species
presence
or more than half of the stem or tuft base
of the canopy
presence
centre
field testing
for canopy
The following
i.
plant
during
cision of species
for the higher
per individual;
unit;
species
overhangs
structure
recorded
occurs
and,
presence
recorded
the sampling
is taken
if any part
unit.
into account
for sampling
and cover.
73
3.2.4
Floristic
Floristic
trial
data
plant
cording
3.2.2.
Gibbs
data recording
is taken
life
of
and
plant
3.2.4.1
nomenclature
et al.
species
&
and
presence
is
to terres-
area.
described
in this
work
is
The
in
re-
section
according
to
(1985, 1987).
optimum
results
that
3.2.4.2
Methods
be
should
that
apart
other
aspects
are
quantity,
purpose
recorded,
be
suggest
dispersion,
To this is added vitality
the
of
the
the
study
minimum
considered,
can be derived
taking
from
often
morphoor vigour
will
often
requirements
into
for
account
the
from those that are recorded.
applied
It follows
from
data
should
which
species
Although
should
aspects
as
or layering.
sociability.
what
(1974)
cover-abundance
such
structure
dictate
Ellenberg
and
recording
logical
in a given
data
used
relating
Background
Mueller-Dombois
worth
its distribution,
identification
Taxonomic
Russell
here to mean the data
the
previous
be
recorded
sections
that
in the
the minimum
field,
for this
floristic
study,
are
the following:
i. species
ber.
name
The
abbreviation
number
strokes
than
ture
used;
is
a
is
name
the
or voucher
preferred
because
abbreviation
plant
specimen
if
identification
it
direct
key
reference
requires
computer
can
serve
numfewer
key-
data
cap-
as
num-
a
74
ber
control
control
to
different
erence
whereas
ensure
that
species;
the
to
the
entification
ii.
plant
reference
Although
use
of
overall
data
correction,
used
for different
iii. crown
high,
D: Dwarf
than
by
is taken
over
not
improve
inclusion
be
then
it
numbers,
the
of
quicker
known,
same
for
some
can
is
separate
used
least
which
using
example,
a
than
is
especially
and,
(section 3.2.3).
form symbol
(from Edwards
1983 and Ruther-
as follows:
single-stemmed
plants
over 2 m high or
plants
from 1 to 5 m
5 m high;
woody,
are rooted,
multi-stemmed
when
less than or equal to 2 m high;
woody
are rooted,
or graminoid
resemble
or partly woody,
plants
less
plants
to the
family
belonging
such as Cyperaceae
and Restionaceae
and
non-graminoid
species
ed. Those occurring
plants
grasses;
are rooted,
Not all plant
herbaceous
occurring
herbaceous
in a sampling
plants.
unit
in seed form and those occurring
id-
can be recordunderground
num-
suggested
abbreviation
3.2.3);
ref-
possible.
into account;
(section
a
voucher
1 m high;
Poaceae
F: Forbs
for
sYmbol
at
can
are
cover class sYmbol
or single-stemmed
G: Grasses
which
saved
is
checklist
species
be
woody,
are rooted,
shrubs
a
1986) is also recorded,
multi-stemmed
keys
abbreviations
can
class
are rooted,
S: Shrubs
abbreviation
necessitates
in
the
require
programmatic
numbers
species,
the growth
ford & Westfall
and
where
canopy
diameter
In addition,
T: Trees
time
would
identification
if
that
same
number
name
initially,
percentage
the
accuracy;
specimen
bers,
abbreviations
at
75
if
is
the time of sampling,
larly, sterile
plant material
cult to identify.
significance
taxonomic
high.
cover
Furthermore,
likely
and
recording
criterion
seldom
The
voucher
as
that
identifiable
species
availability
It
than
is
of
for
within
the often-quoted
that
this
"total
low
their
might
cover
species.
plants
The
identifiable
suggested
a
diagnostic
occurrence
diffi-
although
species,
with
obtained.
Simi-
have a low ecological
material
be
is also often
endemic
doubtful
value
to identify.
concerned
plants
of a plant
be
and
It
with
is
high
criterion
for
the
a sampling
unit
is,
material
is
a
floristic
at the
more
time
realistic
recording"
which
classification,
how-
can be true.
minimum
ever,
as, for example,
cannot
sampling.
community
unidentifiable
of
the
with a low cover
plant
occurrence
therefore,
of
the
significance
are
are difficult
Such plants would generally
within
occurrence,
seldom
for example,
are
floristic
initially,
specimen
following
article
data
requirements
releve
reference
describes
number,
number
for
species
name
abbreviation
and cover percentage
a field data capture
symbol.
program
or
The
for these
data.
76
Bothalia
I~. -' 8: -1: 7-19-750 (Ige.-)
I'HYTOCAP.
A FIELD·DATA
CAPTURE
PROGRA~1 FOR THE I'HYTOTAB
Manual field recording of floristic data, for phyrosociological
studies, entails re-encoding
data on
computer encoding forms, transfer to magnetic tape
and finally loading onto the mainframe computer for
multi-variate
analyses. The time taken from re-e ncoding to access on the mainframe computer can be
from one to six weeks. Classification of data while
fieldwork is in progress, therefore, becomes impracticable in the summer-rainfall
areas because fieldwork generally takes place in the relatively short
growth period and the delays in computer
access
would limit fieldwork considerably if data were classified during this period. A serious disadvantage
of
classifying vegetation only after completion of fieldwork is that vegetation units are often either undersampled,
resulting in invalid syntaxa,
or oversampled resulting in wasted labour and expense.
Furthermore,
the potential for errors is increased by
re-encoding the data.
These problems were overcome by using computerized field-data capture where data is recorded directly onto a hand-held computer and loaded onto
the mainframe computer on return from the field.
Multivariate analysis techniques used are the PHYTOTAB program (Westfall et al., 1982), which are
compatible
with both the DECORANA
(Hill,
1979a) and TWINSPA
T (Hill,
1979b) programs.
Preliminary classification using either PHYT020
or
TWINSPAN
can then be available within a day of
returning from fieldwork.
The system used for field-data capture is the Sharp
PC 1 500 computer with an additional 8K expansion
memory module, printer/cassette
interface, cassette
recorder and programmable
RS232C interface. The
program PHYTOCAP,
written in BASIC, is used
for recording floristic data in the PHYTOT AB
(Westfall et al., 1982) format. The program features
include automatic line number allocation,
sample
and/or subsarnple numbers,
alphanumeric
species
codes, cover-abundance
values, data pertinent to individual species, data display, data printing in two
formats, halting and continuing program execution,
line editing, saving data to tape and loading data
from tape. Furthermore,
data input is verified for
errors such as sample number length, species omission and cover-abundance
omission. The user is also
informed when five lines of memory are left. Thecapacity of the computer is 70 lines which is approximately 14 releves or samples with 40 species per
sample, which is generally more than adequate for
the floristic data recorded in one day.
. Field procedure includes quadrat location, quadrat demarcation,
floristic sampling, voucher specimen collection and environmental
parameter sampling. The Sharp PC 1 500 computer is used for floristic sampling. Species for which voucher specimens
exist are input as a four-letter genus code and a
three-letter species code. Species for which voucher
specimens are required are tagged with pre-numbered, specimen number, adhesive address labels
PROGRA~1 PACKAGE
and input as a left-justified specimen number. Specimen collection is effected after completion of floristic data recording. This process ensures a smooth
flow of data input and reduces the possibility of
species
being overlooked.
Environmental
parameters are recorded directly on a field data sheet. It
is envisaged that a second Sharp PC 1 500 be used
for recording environmental
parameters
and as a
standby machine. Memory capacity currently precludes the use of a single machine for both floristic
and environmental
data.
After a day's recording the computer is attached
to the printer/cassette
interface, which remains in
the vehicle. A printout of the floristic data is obtained of each sample for stapling to the field data
sheets which form the hardcopy for eventual permanent safe-keeping
at the Botanical Research Institute, Pretoria. The data are then transferred to C 15
cassettes, when the computer can be cleared for the
following data set.
Loading data to the Burroughs B 7 900 mainframe
is effected by means of a Burroughs ET 1 100 terminal, RS232C interface and a transfer program
called DATATRAN
written by S. J. Crafford. Data
are read from the cassette and simultaneously
transmitted at 300 Baud.
The advantages of this system of data recording
include the cost-saving production
of preliminary
classifications for optimum sampling as well as the
labour- and cost-saving of not having to re-encode
data. The potential for errors is also reduced by the
reduction
in data handling.
Documentation
and
taped copies of the program are available from the
author. Please forward a blank C 15 cassette for
copying.
ACKNOWLEDGEMENTS
The author thanks Dr J. C. Scheepers for comments and suggestions and Mr S. J. Crafford for
writing the program DATA TRAN.
REFERENCES
HILL
M. 0 .. 1979a. DECORANA.
A FORTRAN
program for
detrended correspondence
analysis and reciprocal averaging.
Unpublished report. Ecology and Systematics. Cornell University, Ithaca, New York. pp. 1-30.
HILL, M. 0., 1979b. TWINSPAN.
A FORTRAN
program for
arranging multi-variate data in all ordered ovo- way cable by
classification of the individuals
and attributes. Unpublished
. report. Ecology and Systematics, Cornell University, Ithaca,
New York. pp. 1-48 .
WESTFALL. R. H .. DEDNAM. G., VAN ROOYEN, N. &
THERO . G. K.. 1982. PHYTOTAB. A program package
for Braun-Blanquet
tables. Vegetatio 49: 35-37.
R. H. WESTFALL'
• Botanical Research
Supply. Private
Water
Institute,
Bag XI~)).
Department
of Agriculture
Pretoria 0001.
&
77
..
3.2.4.3.1
Improvements
A program
PHYTOFORM
the sampling
eve
stand.
unit
one with
3.2.5
Habitat
which
the
lists
a
computer
the environment
floristic
single
species
on the mean
reducing
list
for
th~
each
is the same as for the origi-
can have two
data
sampling
Werger
(1974)
is empirically
states
sets,
namely,
unit data.
c~rrespond
with patterns
Group
Vegetation
minimum
"In the
that
Ecology
and
and
community
plant
abiotic,
and which
communities.
Ztlrich-Montpellier
patterns
in
in the environment".
environmental
correlation
between
both biotic
can
These
in the field.
that
determined
data,
of a plant
differences
recorded
Background
ronment
to
are the environmental
3.2.5.1
gested
is based
for the stand, thereby
data and the other with
are usually
for
conversion
rel-
data
describe
explain
The
recorded
the host
releve
data
stand.
converts
for each stand to a single
to the host computer
Thus
Habitat
for the Sharp PC 1500 in BASIC
(sub-plots),
species
Transfer
data.
data
each
for each species
sampling
nal
unit data
representing
cover
written
in South
parameters
prediction
of
floristic
approach
composition
The National
Africa,
for
floristic
Working
example,
for vegetation
it
and
potential.
sugenviThese
are:
i.
sample
number;
78
ii.
sample
iii.
date;
iv.
altitude
v.
aspect
of sample
vi.
slope
inclination
vii.
lithostratigraphy;
viii.
percentage
ix.
geomorphology;
x.
vegetation
xi.
land use and management.
The
first
site
is
an
then
mental
does
size
factor
not meet
example,
habitat
soil
of terms
if plant
factors,
the
to
of
to
communities
statistical
and
if the
However,
need
certain
clarification.
vegetation
is
described
facilitate
all
in
could
comparisons
If
that
is
descriptive
be improved
with
other
are to be correlated
differentiating
with
causes,
then
is necessary.
vegetation
If the sample
as
change.
be
determine
(1983);
especially
of the environment
criteria
be adequate
such
the
could
for both
statistical
depth
important,
of a study
factors
same
sampling?
factors,
require
aims
account
to Edwards
monitor
environment
of environmental
Is sample
not
the
However,
sampling
to
description
standardization
environmental
extremely
to the
A descriptive
studies.
are
cover;
according
re-sampled
environmental
terms.
by
be
in metres;
in degrees;
structure
items
longitude);
in degrees;
rock and outcrop
relating
required
(latitude,
(above sea level)
three
to
questions
co-ordinates
sampling
size for vegetation
then
using
how can
if
these
and
sampling
a sample
the same criteria?
lithostratigraphy
sampling
and environ-
soil
have
of,
for
Mappable
type
already
should
been
79
mapped.
However,
factors
sampled
by
are
equated
with
also
what
the
be taken
hill could
factors
correlations
The
data
plant
affect
scale
and
possible
influence
field
area
of
and
of
namely,
suitable
a) stand
L,
in
extent
categorization
those
Cognizance
should
factors.
is increased.
what
be
can
be
A
What
derived
community
and
environment
the value of a classification.
from
two
sources,
namely,
at the time of sampling
and geological
direct
and indirectly,
maps.
data
community
purposes,
when
are
cannot
environmental
sampled
from topographic
A) Field
the
be
depth
which
for vegetation.
determined
in the
disparity
study,
samples,
considering
should
and soil
applied
were
measurements
3.2.5.2
point
as a mountain
and hence
data
as derived
when
could
Methods
Habitat
used
slope
sources?
questions
3.2.5.2
scale
be regarded
from other
These
essentially
samples
of
environmental
such as aspect,
between
a
stand
particular
sampling
habitat
those
data
suitable
for sampling
and
sampling
variable's
two
for
stand
for
influence
considerations
into
unit
on
necessitated
groups
for
this
the
the
recording
characterization
and
unit characterization.
data
Lithostratigraphy,
being
formation
and
surface
rock
type
80
co
TABLE 3.2. - Geomorphology
classes
in terms of facets and segments
from Scheeper~
(unpubl.)
f-'
LANDFORM
I
Major
facet
LANDFACETS
(regular uud irregular)
d:lSSCS .
I
FA~ETS
I
Fla"
Flat summits'
or plateaux and structural
terraces
rial to rolling pin ills :1I1d
intcrf'luvcs
I
(0-2')
Slopes
~1J
FI:lt pfuteau
Flat suuuult and 11:11intcrf'luve
Structurul tcrruccs
[I!J
Flat intcrlluvcs
I
12
01
rises
Acute rugged crests
or peaks, l'O/WCX
crests or waxing
slopes. Ircc races
em
(Planes)
(>2-45')
Rolling pia ills
Iluves
)2
33
:lIItl
or
(>45')
rolling inter-
I Convex crests & lntcrnuvcs
Seepage slopes
W~~illb slupcx
Creep slopes
wnxing slopes
[W
(Curves)
SEG~J':NTS
I
Cliff,
)7
Free face or scarp or cliff
1 smooth
Crilg, l'fnc&y tree Incc. ('f:I&tY
scarp or cruggy emf
Acute rUlU~cdpcuk
Acute rugged crest or ridge. knife
)8
udgo or :Il~IC
I\C\ltC SIIHlOI!!
34
or
35
)(1
shaq)
Q3
Rounded ro Sleep, smooth
(10 rugged) rocky prominences
[d
Coucavo-convcx slopes or uppcrslupcs
uf cots or suddles
Convcxo-cuncavc slupes or IO\\'l'!
slopes
cols or saddtcs
Upllmds
Intcrfluvc,
water shed or rid!;:!.!
Debris
I Scree
slopes
[W
<12
Talus
4)
Detrital
slope
~
Zones of inflexion
stope
slope or upper pediment
Plane stope ur nrkldlc Ih:lIiUll'nt
stope or uudslupc
m
Lower pcdintcn t slopc~
or w;lllin~ slopes'
Low fhus
pcm:pl:dns. SW:dl'S
&. Ilal!l lluodpJains tcr ruccs
L".t
~
I
72
Wel bottumland
flals ur vlcls
Dry
Huts
7)
Ffuod plain
River !l'H;ICCS
74
~
I
H'
~8'1
Levees &
ll\lIlutlll:IIIU
cuuvexitics
bottomland
or
Lower pediment
Swnlc
slope or wallinl: slope
uonombuds
J Ilumm'ocks
Levee
Stranded levee und ~I1HJbanks
Fossil strcuutbcd, boulder, rock
nnd ~r:lvcl runs anti sandbanks
[ll]
M5
H6
rYil
~
Chaotlel hanks
Gullies, dongns & hatllillu.Is
Microlnccrs
1 J.e. Sch:!epers, PrivateBag X05, Lynn East, 0039
~ J
I
hl:tnus& l slc ts
Sandbnnks, slltbunks,
mudbnnks,
pcbblcbanks.
stoncbunks, rock banks,
and bouldcrbunk s
ClwlIlld banks
92 _ Gullies, dougus ami budlunds
M icroscgmcn
IS
a
FIGURE
82
a
3.3. - A flexible plastic cross for selecting random
pairs of plants in ground layer vegetation. a) intersecting arms, b) wing nut pivot - the first component of
the plant pair is that which is nearest to the pivot,
and c) marked - the second component of the plant pair is
that which first intercepts the arm or continuation
thereof.
FIGURE
83
3.4. - A tripod-mounted,
manual spinner for selecting
consecutive plants in pairs, at random, in vegetation
above the ground layer. a) spinner, b) sight, and c)
telescopic tripod.
TABLE
3.3. - A dual
level
system
Primary
symbol
Land use:
(level 1)
Vegetation utilization
None observable
Component mostly utilized:
Grass
Herbs
Shrubs
Trees
All
Vegetation replaced by:
Pasture
dry land
Pasture - irrigated
dry land
Crops
- irrigated
Crops
Plantation
Exotics (other)
-
-
Construction:
Dam
Road - unsurfaced
Road - surfaced
Footpath/track
Railway
Single storey
Multi-storey
Cleared
84
of land use categorization
A
B
C
D
E
F
I
J
K
L
M
N
Q
R
S
T
U
V
W
X
Land use:
(level 2)
Production:
Cattle
Game
Goats
Horses
ostriches
Sheep
Other stock
Bark
Hay
Flowers
Fruits
Leaves
Root
Seeds
Wood
Cereals
Vegetables
Alien
Waste:
solid
Liquid
Gas
Sectmdary
symbol
A
B
C
D
E
F
G
H
I
J
K
L
M
N
a
P
Q
R
W
X
Y
according
to
Jansen
(1982)
with
field
verification
of
out-
crops.
ii.
Altitude,
taken
being the altitude
from
nearest
the
half-contour
iii. Geomorphology
Table
iv.
3.2
class,
classes
by a cover
les were
meter
v.
as visually
SA topo
series
categorized
taken
and
being the total
suggested
(Westfall
in each
up to one metre,
mounted
000
in
metres,
map,
to
the
i.e. 10 m or 25 feet.
structure,
height
thrown
1:50
centre
according
to
.
Vegetation
the
relevant
of the stand
for the
spinner
height
height
cover
by Edwards· (1983)
& Panagos
a plastic
canopy
1984).
class.
For
cross
(Figure
classes
above
as determined
Five
the
assessment
random
height
3.3) was
according
samp-
classes
arbitrarily
one metre,
(Figure 3.4) was used to select
Land use, being a visual
for each of
a tripod-
random
pairs.
to the categor-
ies in Table 3.3 .
vi.
Insolation
angles
tal,
The
exposure,
of the
at
the
eastern
stand
directions
study
area,
difference
the
Tropic
taken
being
and western
centre
of the
were
of
as being
and
as
vii. Temperature
of bare ground,
tation
height,
five
stand
arbitrarily
centre.
due
Temperatures
western
east
from
with
and
stand
Maximum
solar
an
the
inclinometer.
horizons,
due
centre
horizon-
for
the
less
the
latitude
and
west
insolation
is
thus
solstice.
ground
being
selected
the sum of the
horizons
the
at the summer
at chest
and
between
Capricorn.
minus
measured
eastern
taken
in degrees
180 degrees
layer vegetation
the mean
sites
of the measurements
for each
were measured
and vege-
category
with
around
at
the
an Instatherm
85
infra-red
non-contact,
dard
grey
parative
scale
x.
according
for com-
to asssess
the eff-
to the following
layer grazed
short
1, ground
layer grazed
short and patchy;
2, ground
layer grazed
evenly;
Browsing,
four-
assessed
0, browse
line evident;
1, browse
line not evident.
Erosion,
(less than
100 mrn);
and
evident.
as visually
as visually
according
to a two-class
scale:
and
assessed
according
to the following
four-
scale:
0, donga
erosion
evident,
by donga or gully
1, sheet
erosion
evident,
by general
2, donga
direction
Fire,
soil accumulation
in a
plants;
evident;
assessed
according
of fire present;
1, no evidence
and
to a two-class
scale:
and
of fire present.
unit data
Slope
inclination,
mined
with
Aspect
formation;
evident.
as visually
0, evidence
b) Sampling
against
and sheet erosion
3, no erosion
ii.
assessed
0, ground
similar
i.
stand
scale:
class
xi.
of a stan-
cover on soil temperature.
as visually
3, no grazing
ix.
at each
These data are intended
ect of vegetation
class
The temperature
card was also measured
purposes.
viii.Grazing,
thermometer.
from the horizontal
in degrees
as
deter-
an inclinometer.
of inclination,
in degrees
as determined
with
a magne86
TABLE
3.4. - The "sausage" method of determining the clay percentage or texture of soil at field capacity (from F.S.S.A.
1974)
No sausage
Sand
C::""'~
Loamy
sand
10
-
15% clay
~
Sandy
loam
>15
-
20% clay
~
Sandy
clay loan
>20
-
35% clay
Sandy
clay
>35
-
55% clay
~
"
«»
"
'.,.
TABLE
"--
Clay
>55% clay
'
.
3.5. - The "finger test" method of determining the clay
percentage or texture of soil at field capacity (from
F.S.S.A. 1974)
Fingers
stay clean
Fingers
slightly
dirty
Fingers
slightly
sticky
Slight
glossy
smear
Highly
glossy
smear
87
<10% clay
and concretions
0
-
6% clay
>6
-
12% clay
>12
-
20% clay
>20
-
35% clay
>35% clay
and allowing
tic compass
iii. Percentage
surface
for magnetic
rock and outcrop
deviation.
cover,
as a visual
esti-
mate.
iv.
Soil surface
at the
sampling
trometer,
v.
centre
These
if the
corners,
recordings
Soil texture,
percentage
clay
"sausage
(Table
greater
spike measurements
if the
soil
depth
a pocket
than
pene-
300 rom or as
at the sampling
was
less
than
(Table
3.4)
unit
300
rom.
10 mm.
the mean of the class midpoints
in the A horizon,
method"
as estimated
and
the
of the
using
"finger
both
test"
the
method
3.5).
vii. Soil colour
city
was
are to the nearest
being
using
with a soil auger at the sampling
soil depth
of five depth
and
and corners,
in kg.cm-2.
as determined
centre,
the mean
being the mean of five measurements
unit centre
calibrated
Soil depth,
unit
vi.
compaction,
of the A and B horizons,
with
a
soil
colour
chart
determined
(Munsell
at field
Soil
Color
capaCharts
1954) .
viii.Soil
form,
according
ix.
Litter
as determined
to MacVicar
cover,
by diagnostic
et al.
horizon
combinations
(1977).
as an estimated
percentage
of soil surface
covered.
x.
Litter
the
depth,
nearest
being
10 mm
the mean of five depths measured
at
the
sampling
unit
centre
and
in rom to
the
four
corners.
xi.
Relative
biomass
of the ground
measurements
at the
disc pasture
meter
sampling
(Trollope
layer, being
unit
centre
& Potgieter
the mean of five
and corners,
using
a
1983), calibrated
88
in
cm.
Conversion
clipped,
made
material.
between
A program
the
dry
stands
HABIMEAN
sampling
generally
unit
based
applicable.
rence
of the
However,
relative
although
written
in BASIC
habitat
data
forms
is calculated
as the
each
unit.
case
of soil
mean
soil
values
for brightness,
Munsell
class.
This
program
because
program
data.
form,
conversion
the
of
percentage
slope
aspect
aspect
as the means
part
is
occur-
Mean
and
and hue,
form
in
be
to SI units,
is calculated.
saturation
difficulty
can
PC 1500 converts
The
is taken
not
with
less precise.
vectors
colour
does
comparisons
are transformed
stands
of the
Munsell
the
stand
which
in the
Mean
of
to
correlations
for the Sharp
v~lues
In the
soil
sampling
requires
directly,
on mean
where
kg.ha-2
to
of the
to the
of the
standardizing
for
nearest
PHYTOTAB-PC
environmental
variables.
3.2.5.2
B) Derived
Derived
data
case,
were
included
are those
derived
in
the
in extracting
will
make
this
not usually
drainage
approach
here
in the
as
maps.
to
because
show
The
the
maps.
stand
position
which,
current
type
of
is not
developments
systems
description
information
of this
information
of difficulties
relative
in this
SIDA
information
program
Some
field because
stand
The program
from geographic
obsolete.
from topographic
such
for each
package
information
recorded
determinations,
data
from topographic
included
be derived
habitat
PHYTOTAB-PC
such
nevertheless,
can
data
to
is,
that
is
in field
watershed
and
line.
89
The
co-ordinates,
the program
in degrees,
PHYTOLOC
(section
plotted
on the relevant
ference
of each
scribed
around
stand
was
stand
the
noted
minutes
3.1.6)
and
(5 mm radius
on the
centres
sheets.
from the topo
sheets
for the Sharp
PC 1500 computer:
generated
for stand centre
1:50 000 SA topo
stand
seconds
series
at 1:50
and
The
the
scale)
releve
following
were
The circumwas
number
was
for i-nput to the program
location
sheets.
000
with
then
circumfor
each
determined
SIDA written
in BASIC
a) Input
i.
Stand
or releve
ii.
Contour
interval,
interval
iii. Sheet
iv.
number.
in feet or metres,
used on the relevant
grid number,
Latitude,
topo
for reference
in degrees,
minutes
depending
series
on the contour
sheet.
purposes.
and seconds
of the stand
centre.
v.
Altitude
dent)
vi.
vii.
of the stand centre,
to the nearest
Aspect,
in degrees
being
the
point
on
from
lowest
of the stand inclination
the
contour,
stand
centre
most
adjacent
This forms the aspect
Lowest
altitude,
tour most
viii. Lowest
stand
adjacent
contour
nearest
ix.
from north
circumference.
contour
point
to
the
to
nearest
the
stand
line.
in feet or metres
of the lowest
con-
to the stand circumference.
distance,
on
(map depen-
half-contour.
direction
the
in feet or metres
the
in mm from the stand centre
lowest
contour,
most
to the
adj acent
to
the
circumference.
Highest
contour
contour
most
altitude,
adjacent
in feet or metres
to the stand
of the highest
circumference.
90
x.
Highest
contour
nearest
point
stand
xi.
the
distance
stand
highest
contour,
most
stand
along
crest
line
drainage
adj acent
contour
the
distance,
nearest
stand
is
the
This is the watershed
a crest
then
the
line
to
line.
If
line
is
watershed
opposite
the
so that the stand position
first
can be re-
lines.
in feet or metres,
of the watershed
in mm, to the nearest
drainage
situated
nearest
watershed
related
from
distance
or
half-contour,
from the stand centre along the lowest contour
to
to
line.
xiii. Drainage
the
to the
half-contour
line approximately
(see xiii)
altitude,
drainage
highest
or watershed.
lated to two drainage
xii. Watershed
the
is situate-d on
to the nearest
drainage
in mm, to the nearest
centre
the nearest
the
on
in mm from the stand centre
circumference.
Watershed
the
distance,
line
line.
This
is the drainage
in a depression
watershed
(see
xi)
then
the
approximately
so
distance
that
the
line.
drainage
opposite
stand
line to
If the
line
the
position
is
first
can
be
to two watersheds.
xiv. Drainage
altitude,
in feet or metres
of the drainage
line or
watershed.
xv.
Angle
and
horizon
western
erence
east or west. The directions
horizons
between
in
section
are
determined
contour
centre,
line
the
are
stand
3.2.5.2
by
a
A)
due
east
centre
( a)
and
west
latitude
vi) .
subroutine
altitudes
and
of the eastern
of
The
the
distances
in mm, along the directions
and
plus
the
23~o
south
horizon
program
from
of the eastern
SIDA
the
diff(as
angles
when
stand
and
91
western
horizons,
extrapolated
number
contour
input.
so
that
lines
to
The
the
be
subroutine
user
input.
displays
can
determine
Contour
lines
the
lower
stand are ignored.
xvi. Cumulative
distance,
drainage
less
altitudes
of
than the
are
line
than
to
the
the watershed
in mm, being the distance
watershed
stand- centre
or
if
then
to
from
watershed
the
altitude
highest
point
is
along
line.
b) Output
The output
is based mainly
input variables.
The following
i.
stand
or releve
ii.
grid number,
iii. altitude,
iv.
aspect,
v.
mean
number,
slope
mean
annual
the
combined
as inputi
in mm, based on a linear regression
and
viii.relative
and
watershed
and
drainage
exceeded
on a linear
and
the
stations
regression
recorded
in the
of
rainfall
study
area
20 yearsi
in degreesi
moisture,
slopes,
latitude,
latitude
rainfall
generally
available
altitude,
in mm, based
recognized
exposure,
rainfall,
between
of the
soil depthi
altitude
recordings
insolation
position
rainfall,
for officially
annual
for each stand:
in degreesi
line and measured
vii.
of the
as inputi
soil depth,
where
are printed
functions
in metresi
combined
vi.
on simple trigonometric
in mm, as a function
aspect,
insolation
position
exposure
of mean
in the landscape,
and mean
soil
depthi
92
Watershed
a)
Drainage
Drainage
FIGURE
93
line (e)
Depression
(D)
Flat (B)
Rise Knoll
line (G)
3.5. - Geomorphology
classes determined with the SIDA program with class codes in brackets. a) Landscape slope
<=2°
b) Landscape slope >2°. Slope categorization
is
restricted if the slope is less than 12x stand radius (R)
to: crestjkoppie
if slope <4R; midslope if slope <8R;
lower slope and upper slope if slope <12R.
(A)
ix.
topographic
stand
from watershed
profile,
diameter
proportionally
x.
topographic
profile
slope
xi.
stand
slope,
xii.
scale
of topographic
indicated
line with the
according
to position;
(mean), in degrees;
in degrees;
profile;
xiii. geomorphology,
in classes
xiv.
class· according
stand drainage
stand
to drainage
as listed
in Figure
3.5; and,
to the topographic
profile
and
slopes where:
A
Outflow:
B
Throughflow:
C
No flow: Run-off
D
Inflow:
> Run-in
Run-off
=
Run-off
=
Run-in
0; Run-in
0
< Run-in.
Run-off
3.3 CLASSIFICATION
Classification
in
this
sense
presence/absence
data
releves
stands)
values
(sampled
at the
quantity
intercepts
a matrix
and
rows
of columns
to
form
Background
Gauch
(1982)
defines
English
Language
(1985)
animals
and plants
because
of similarities
The
New
defines
in a series
ordering
where
species.
species
represent
The
matrix
presence
and
symbol.
as "grouping
Collins
Concise
classification
of increasingly
in structure,
of
columns
and rows indicate
classification
in clusters".
the
represent
by a cover or cover-abundance
3.3.1
together
in
refers
origin,
similar
Dictionary
entities
of the
as "the placing
specialized
etc. that
of
groups
indicate
94
common
relationships".
description
appears
tion,
of
more
scatter
as
data
hierarchically
The
procedure
proposed
identification
releves".
occur,
The
concatenation
to achieve
and
diagnostic
refers
fluenced
scale
by
ties
species
group
be
were
show
the
achieved
within
according
releve
distributions
the
study
practice,
(Westfall
be
releves.
for
however,
Programs
such
and
initial
plant
species
study
area
these
two
species
process.
a totally
releves
are not
not
groups
If a set of
speciesclassi-
often
grouped
as a first
step in
(Hill
(Westfall
classification
distri-
are
different
are
as TWINSPAN
21
at the
on a set of diagnos-
gradients
PHYTO
in-
communi-
the
to that based
environmental
e t: e I , 1982)
used
of
area,
but definite
of
sense,
are
from the non-diagnostic
concatenation,
compared
to noticeable
sequencing
also
for example,
emerge,
In
recognition
en-
species
In this
Non-diagnostic
within
(1974)
to the problem
recognition.
have limited
limits
relation-
selected
rise
after the classification
chosen,
could
species.
their
of
as a key to grouping
in which
whose
gradients
reliable
for initial
fication
can
species
therefore,
classifica-
Ellenberg
&
gives
species
in the hierarchy.
and
Clearly,
can only
20
level
influenced
clear.
tic
They,
to
definition
diagnosis
in the study area and can be used to delimit
at any
thus
those
environmental
concerned.
butions
to
and
groups
reLeve s
sequence,
non-diagnostic
second
to a
such units.
species
of
a releve
diagnostic
and
Mueller-Dornbois
of "common
suited
for a vegetation
units
within,
by
the
identification
vegetation
and variation
seems more
whereas,
to the needs
reduction,
grouped
between
tails
diagrams,
appropriate
such
ships
The first definition
but
&
these
1979),
De
Wet
PHYTO
1988)
programs
do
95
1
1 2 1 3 1 415
a
FIGURE
c
b
3.6. - communities
(numbered) sequenced according to the
main environmental
gradient responsible for their differentiation
(a); subsidiary gradients (b and c) in which
communities
are sequenced similarly but inserted after
the main environmental
gradient.
li===1=1
=2==3=1 =4=11 10 1 II
a
FIGURE
96
b
a
c
3.7. - Communities
(numbered) sequenced according to the
main environmental
gradient responsible for their differentiation
(a); subsidiary gradients (b and c) in which
communities
are sequenced similarly but inserted in the
positions to which they correspond on the main environmental gradient.
a
not meet
the
needs
ing is generally
In a matrix
This
can
responsible
be shown
by
a
further
subsidiary
of
sequenc-
These
relationships
of insertion,
than
are
are
they
are
linear.
gradients
are
diff-
at the end of the comor inserted
respective
within
the
positions
on
is preferred,
as
likely that the two rele-
of a community
more
similar
to the
illustrated
is
communities
possibility
it is often
gradient,
floristics
Should
at their
last-mentioned
However,
sequence
environmental
be placed
gradient
to the point
terms
multiple
by the main gradient
The
later.
yes adjacent
and
number
differentiation.
of the main
gradient?
releve
gradients
differentiated
the main
ted
where
by subsidiary
communities
will
problems
the
for community
erentiated
classification
required.
classification
pose
munities
for a final
to
releves
differentiaeach
to
schematically
other
be
in
in
inserted.
Figures
3.6
and 3.7.
In
sequencing
problem
which
was
releves
found,
the other
tested
in the
first
and
factorial
number
namely,
releves
first
of
the
required
is a value
sequence
of releves.
similarity,
selection
and each
releves,
number
even
how can the
are compared
to
of the
are to be compared.
total
is prohibitive
the
position
subsequent
Furthermore,
releves
according
then
of
"best"
sequence
number
to the
of
with
sequences
be determined
effectiveness
are
with
tests
magnitude
at a time and not the entire
relating
releve
is compared
r-eLeve s . The
if mirror-image
first
additional
If all the releves
releve
the
an
the
is
of
a
this
are excluded.
if only
sequence?
What
two
is
of a particular
97
The
questions
Blanquet
i.
that
arise
classification
given
that
ii.
can
iii. what
is the basis
before
v.
objective
sampling
two
regarded
matrix
data
Should
species
similarity
in
be
distribu-
(1974)?
should
workers
be
possible
hence
to
achieve
where
the
discon-
and,
process
be automated
by
repeatable,
For
the
thereby
reducing
process?
automated
objective.
and classification
classification
a
will
classification
processes
of
should not have observer
to
not
be
stratification,
bias.
applied
adopted
for classification
to obtain
formed,
to
independent
It
followed
classification
The approach
Braun-
(1956)?
in the classification
sampling
3.3.2 Methods
i.
by Werger
are obvious;
and
be
in the classified
according
can the classification
the
a
the same
classification
of the classification?
for
decision-making
make
by Poore
classification?
Subjective
to
in a data set with little or no observable
discontinuities
tinuities
approach
have exactly
particular
releves
is it possible
same
species
relationships
tion as suggested
iv.
a
as suggested
sorted
objective
and, if not, what are the alternatives?
are the community
linear,
an
are:
no two plant
distribution,
as unique
from
in this
is in accordance
and has the following
a releve
based
study,
sequence
on floristic
where
with
the needs
aims:
(a) releve-groups
similarity;
and
can be
(b) the releve98
groups
formed
thus
similarity.
This
ii.
to delimit
iii. to obtain
according
a species
Releve
matrix.
The
an
total
gaps
often,
can
outset
form
taken
would
be
into
the
of
and their
because
formation.
Judgements
of
value
can
gaps
to
of
were
can
have
on
to
and
a species
the
last
of
after
in the
sequence
considered
of
then
sequence.
and,
species.
The
more
Visual
from
the
impossible.
not
be
in
and
in two or more
Pattern
be based
before
the
often
absence
minimization
as to be virtually
judgements
the
outliers
plant
gap
gaps
first
of releve
many
of species
1982) .
gaps
omissions,
according
is the occurrence
quantified
all
of
by
so that
irrespective
so difficult
(Gauch
If
sequence
the
caused
respectively,
spacing
rei eves
between
matrix,
sampling
irregular
reduction
consideration,
same
by
and
releve
re leve s . The
total.
be the
caused
sequenced,
in the
occurrence,
affect
pattern
of the possi-
the releve-groups
blanks,
last
is, however,
Redundancy
included
relevant
not
the
sequencing
are
in the
and
can
where
assessment
a species
species
r e Leve s
the
gaps
of
are
because
sequen-
sequencing
intuitive
matrix
the same as simply
are emphasized.
on
first
floristic
gradients;
sequence
depends
the
to
and,
of a preliminarily
that
according
to similarity
The refinement
occurrence
sequenced
of subsidiary
releve-groups;
relationships
3.3.2.1
also
is not necessarily
cing the releves
ble insertions
are
be
to
adequately
taken
a combination
releves
of
on
pattern
number
of
99
species
and
the releves
Gauch
re Leve s
forming
(1982)
and
the
the pattern.
defines
pattern
"noise",
the unco-ordinated
regards
"noise"
reduced
in
(1982)
does
noise
The
last
releve
matrix,
not
sequence.
"total
to
total
each
adequacy.
separation
releve
be
possibilities
program
units".
was
are:
possibilities
small
matrix
2
000
sequences
in
to
each
a
gaps
and
million
effect
of
of included
between
all
the
first
a
given
for
the
units
and hence
This
time
B7900
x 33 species.
Even
related
to
presupposes
that
The
of
it would
The
inversely
where
number
mainframe
as
is that
position.
total
species
species
are
are
classification
n is the
years.
is
Gauch
for a single
for
minimum.
the
the
He
by addition.
possible
determine
of 25 releves
this
are
"noise"
Although
matrix,
separation
optimum
on a Burroughs
million,
A heuristic
The
in, where
written
the
in the matrix
units
tested
these
of
in
units"
total
if
in terms
absences
the included
The
that
in the matrix
and
in the matrix.
quantifiable,
species
"separation
to redundancy
classification
the
each
occurrence
enhanced.
are species
sequence
as
separation
proportional
to gaps
of
is
so
in
are subjective.
of species
pattern
as
occurrence
co-ordinated
opposites,
"noise"
species
judgements
This can be quantified
releve
referred
as
blanks
occurrence
These
occurrence
then
regard
of
as _the
pattern
included
For a given
are
and
can be attributed
blanks.
and
a
constancy
resulting
releves.
take
to
computer,
answer
excluding
was
A
test
for a
in excess
mirror-image
is impracticable.
approach,
where
not
all,
but
only
the
best
possibili100
ties
are
also
playa
the
tested,
role
initial
It was
(Hill
was
in excluding
releve
found,
1979)
sequence
the initial
change
when
sequence
The
and hence
arbitrary.
The
to obtain
same,
irrespective
re1eve
should
releve
sequence.
done
for
each
proved
releve
either
of
The
have
other
extreme
releve
the
the
This
generally
This
this
species
sequence
The
that each
is to
of any
count
in a releve.
sequence,
can
have
species
the
the
This
then
to the
can exist
called
occurring
be the
independent
according
is
would
releves
highly
is, there-
implies
aim,
of the sequence
occurring
the least generally
releve
attribute
this
altered
is often
sequence
first
some similarity
the one extreme
most
of
was
as possible.
or descending,
extremes
TWINSPAN
could be to standardize
sequence.
matrix.
sequence.
using
sequence
could
However,
final
sequence
releve
achieving
the
the
numbering,
the
of all the
although
releves.
because
in
ascending
releve.
which
initial
some sort of unique
to be unique
sequence
the
of the sampling
A method
each
intermediate
final
in which
in the matrix,
sequenced,
for
a sequence
have
affect
a manner
index
permutations.
the
aim of standardizing
fore,
occurrences
still
in as unique
sequence,
similarity
to this problem
3.3.2.1 A) Commonality
sampling
a
many possible
that
4). A solution
sequence
Clearly,
could
for example,
could
(see Chapter
adopted.
is
be
totals
in tests,
between
some
commonality
represents
releves
in common
and the
species
in common.
101
TABLE
3.6.- Illustration
of the possible permutations of species
presence and absence (rows) in three releves (columns).
Presence is indicated by a "+" and absence by a "0".
Columns
3
1 2
Rows
102
1
+
+
+
2
+
+
0
3
0
+
+
4
+
0
+
5
+
0
0
6
0
+
0
7
0
0
+
8
0
0
0
3.3.2.1
B) Similarity
A problem
of
encountered
Jaccard,
Ellenberg
with
For
1974)
example
different
two
This
Ellenberg
taken
could
is because
found
releves
and
Gleason
are compared
re Lev es
have
in
a
in three's
position
in pairs
as those
&
in isolation,
they
matrix,
an extremely
in similarity
such
(Mueller-Dombois
of which
of the weight
are considered
central
co-efficients
of the matrix
adjacent
generally
If releves
similarity
is that releves
communities
efficient.
the
with
Sorensen,
no cognizance
ciation,
sequence
form
a part.
representing
low
attached
two
similarity
to negative
coas so-
co-efficients.
(Table 3.6), then
in the matrix,
the
in terms
following
of
can
be
stated:
i.
positive
associations
exist
in the case of the first
three
rows;
ii.
a possible
resulted
fourth
positive
association
from sampling
association
not be shown because
spacings
absence
Therefore,
then
in all three
if
only
positive
occurrences),
5 and
absence
omission
or irregular
spacing,
in the
exists,
but negative
association
of sampling
omissions
in the fifth to seventh
but not a negative
3.6)
if single
row;
iii. no positive
iv.
could exist
the
first
could
positive
or irregular
rows; and,
could be considered
association,
associations
possible
6 (absences
columns
two
in the eighth
releves
exist
for
associations
be because
can
are
row.
considered
rows
exist
of sampling
a similarity
1
and
for
(Table
2.
rows
omissions),
(joint
3, 4,
and
no
103
negative
association
indicate
similarity).
Similarity
on
and
given,
releves.
in rows
7 and
for
pair
z e Lev es
weighting
positive
values
exists
possible
as
The weightings
association
possible
positive
similarity
matrix
absence
therefore,
with
cannot
could
be
no
based
negative
shown
for
two
are a~ follows:
2
association
co-efficient
1.
for
(C)
comparing
=
1/2 S + J x100
S + J
two
releves
in
a
where
S
number
of single
J
number
of joint occurrences.
3.3.2.1
C) Separation
standardization
of
according
the
is calculated.
matrix
sively,
one
tion units
repeated
position
total
with
calculations
of releves,
the
The
in the
in both releves
sampling
new
separation
all the
releve
and
units,
releves
and the
have
and
This
been
units
moved,
and
for
succes-
total- separa-
The releve
where
releves
separation
is retained
sequence.
of
is then
sequence
each move.
units
releve
sequence
the total
first
releve
after
separation
of total
after
the
to similarity,
are calculated
lowest
occurrences
unit sequence
sequencing
the
is,
associations
associations
given
(joint
is, therefore:
C%
After
of
positive
negative
Positive
The
a
8
sequence
the
results
process
with
is
in
n2-n+1
n is the total
number
moved
in this
manner.
104
The
releve
sequence
reversed
and
decrease
in total
releve
more
the
sequence
separation
til
the total
lowest
continues.
tends
represent
a
occurring
is
in which
Releve
the
releves
process
represent
plant
achieved
represent
the
a
increment
groups
is reversed
a
are
when
separation
and
where
minimum
of
are
iterated
the
iteration
and the
sequence
then
achievrelevant
should
blanks.
affect
in the releve
un-
with
have been
included
do not
new similarity
the
units
releve
releve
of the
releve
the sequence
are terminated
This
they occur,
have been
communities.
it could
releve-group
then
species
the position
of
sequence.
decrement
differentiated
last occurrence
the
optimum
be expected
be
because
of
the next
of
by different
releve
towards
a sine
curve
Furthermore,
species
distribution
then
classi-
of which
could
sequence
that the releves
situated
species.
of each species
step in the
rei eve-groups , some
If
would
releve-group,
and
sequenced
is to delimit
then
relevant
a
no
grouping
fication
been
units
with
of
processes
increase
iterations
until
exits.
units
retained.
matrix
position
and these
total
is then
The reversal
is then used to determine
in all or a single
the releves
After
the
units
successively
is achieved.
for that releve
separation
separation
is repeated
units
sequences
manner,
total
average. the
sequence
unit
sequence
3.3.2.2
to
When the lowest
in this
releve
process
separation
total
lowest
separation
than one position
and
the
entire
The last retained
ed
with
the
that
middle
effect
if
the
the
of
has
best
of
the
releve-
first
and
in the matrix .could be
105
significant,
curve,
terms
of
or approximation
obtained
ing
in
releve-group
thereof,
by the difference
first
Species
and
last
occurring
in
could
between
occurrences
all
the
delimitation.
be expected
the number
in
the
in
for
a
a
sine
from the values
of species
matrix,
ze Leve s or
Thus
represent-
each
single
releve.
releve
are
ignored.
The
number
of releve-groups
of difference
to the
be
permitted
scale
closest
of the
between
study,
is taken
into
programs
allow
be recognized
and
Thus,
flexibility
sampling
at the appropriate
so that
This
are
Species
step
the
in vegetation
process.
and redundancy
should
stratified,
stratification
process,
However,
of releve-groups
that
the
can
even if the stratifi-
according
into the
releve
classification
and their
by
increasing
with
similar
combinations
species-groups.
are not
units
relate
to scale.
sequence
Releve-
automatic-
positions.
releve-groups
species
groups. or
in the number
inserted
this will
sequencing
is achieved
whereby
on the
on the degree
of releve-groups
classification
processes
ally,
final
Clearly,
so that data can be classified,
delimiters
The
depends
of vegetation
based
in the
group
3.3.2.3
number
scale,
account
releves.
and the number
to the original
for objectivity.
cation
to be recognized
A
of
balance
or pattern.
relationships
pattern
through
distributions
releve-groups
must
is to sequence
also
Various
be
are
are
placed
achieved
combinations
emphasized.
species
according
species
sequencing
to
releve-
together
between
of outliers
in
outliers
to re106
dundancy
were tested without
of total
separation
in the matrix
tern.
or
similarly
the
Outliers
are
the
pattern
Furthermore,
improvement
does
not
each
species
are
in terms
sequence
or the
dependent
between
on minimum
included
relationships
of outliers
on
an
that
the
outliers
and
It must
is present
species
does
not
releve-groups.
can be allowed
appreciation
or
corresponds
blanks
of
This is obviously
grouping
occurren-
outliers.
between
exceed
to pattern.
and also
and that sequencing
number
releve
on tests
of
a
releve-
not
because
could contribute
The relationship
to be based
for
the
pattern
subjective,
relationships
in
but
between
as has been stated.
also
sequenced
species-group
Species-group
any
formed.
..
in the matrix,
is based
"noise"
the
two,
does
that can be effected.
affect
as
of outliers
fraction.
is
taken
number
the maximum
releve-groups,
Species
is
is
the total
grouping
matrix
no pattern
in
of species
releve
given
occurrence
to pat-
significant
occurrences
for outliers
that
considered
because
species
in a releve-group
single
can be considered
be emphasized
affect
releves
be
of
to
for that
Fibonacci
regardless
cannot
In terms
contribute
of a species
than one in a releve-group
limit
occurrences
of releve-groups
limited
that
total
ces of more
with
of
species
also
provided
38%
occurrence
or combination
38% of the
The
or groups
single
as they cannot
releve-groups
minimum
releve-group
group,
of
group
Therefore,
a
a single
ratios being evident.
however,
are not significant
combination
for
units,
definite
order
in descending
in the diagnostic
is according
species
to presence
order
of
portion
occurrence
in
of the matrix.
in:
107
Releve-groups
3
1
2
Species-groups
1
[!]
I
3
4
5
6
8
I
6
7
1[2]
8
8
10
11
108
.
5
9
FIGURE
5
[!]
2
3
4
9
10
I
I
I
I
11
3.8. - Simplified schematic diagram showing the sequence
according to the releve
species-groups
construction,
group sequence.
of
i.
first
releve-group;
ii.
second
iii.
first
and second
iv.
third
releve-group;
v.
second
and third
vi.
first,
second
releve-group;
illustrated
Species-group
releve-groups;
releve-groups;
and third
in Figure
delimiters
and,
releve-groups
and so forth as
3.8.
are
then
inserted
into
the
species
sequence.
Species-groups
determine
combine
with
whether
such
species
a
single
or not this
species-groups.
sequencing
can
consisting
spond
limits
C). This
the
simplifies
can
occur.
is significant
The
be
species-groups,
with
species
flexibility
carried
even
of two or more
of environmental
the matrix,
in terms
The
user
and either
that
should
retain
is possible
further
or
with
by
combining
releve-groups,
to corre-
gradients
(section
of species-groups,
3.5.2.1
without
loss of information.
3.4 VERIFICATION
A classification
more
than
testing
3.4.1
one
should
solution
procedure
be tested
to ensure
to a classification
can be referred
its validity,
could
because
be possible.
to as a verification
This
process.
Background
109
Werger
(1974)
stresses
between
the
pattern
habitat
conditions".
on
relationships,
confirm
a
any
to
factors
with
between
community
fication
of
adequacy
of
necessarily
not
a
The
floristic
it
is
accepted
produce
environmental
total
but
variable.
separation
possible
each
is
rating
units
that
correspon-
Correspondence
but
position.
to jointly verify
that
not necessarily
units
the
also
distribution
or other
a classification,
separation
can
it is feasible
some
is, therefore,
of
every
it, then
and
a
can
support
rating
obtained
because
other
veri-
the
means
a
of
the
is
not
releves
should,
a classification.
sections
describe
some methods
in which
a classifi-
can be verified.
3.4.2.1
Classification
A measure
of
ship between
rix and the
which
into
specific
applied
following
cation
and
"co-incidence"
if
limit
could
total
be sought
3.4.2 Methods
may
classification
in
"coincidence
table has its own particular
validity
the minimum
of
table
insight
of
However,
other
The
tested
therefore,
degree
and habitat
the
classification.
are
the
which
or
confirmation
phytosociological
of releves
one
for
from gaining
classification.
combination
dence
the
in the classified
according
need
Apart
habitat
species
the
the
efficiency
efficiency
all the blanks
total
separation
classifications
of a classification
(not just included
units.
This
can be compared,
is the
blanks)
is a relative
unlike
the total
relation-
in the matvalue,
with
separation
110
units,
which
absolute
is an
(E) is calculated
value.
TSU = total
A matrix
have
with
a
ciency
values
indicate
step
is
groups.
to
releve-groups
formed
indicate
A
the
quantified
the other
the
spatial
exhibit
where
therefore,
appears
that,
classification
lower
than
process
A
the
in
effi-
this
could
or data.
low
an overlay
is
of
the
high
grouped
the releves
degree
that
number
according
grouped
of
to
spatial
can
be
could
or data.
relationships
classification
two
In general
integrity
process
The degree
releve-
mosaics
spatial
spatial
adequacy.
the
between
technique.
such
degree
in the
and
of a classifica-
relationships
scale
comparing
releves
representing
adequacy
in the classification
process
by
the
a relatively
the
of classification
representing
have
in the matrix.
would,
It
markedly
in verifying
correspondence
stratification
be
matrices
can be done by using
shortcomings
indicative
can
suggested
should
of
100%.
in the classification
releve-groups.
measure
efficiency
relationships
except
by
classified
of
60%. Values
determine
This
integrity,
efficiency
of above
Spatial
next
tion
and correctly
classified
shortcomings
3.4.2.2
The
no noise
adequately
x 100)
AG
units and AG = all gaps
separation
classification
tests,
classification
as follows:
E%=100-(TSU
where
The
process
between
can
be
of correspondence
sequences;
the
according
the
one
stratification;
to the classi-
111
a)
Habitat
1 Altitude
(m1O)
2 Soil depth
ranges
40-45
35-42
25-42
31-471
55-68
53-70
52-65
60-83
(cm) 10-14
11-13
15-20
16-181 10-13
12-16
17-18
17-19
Sand
Clay
Clay
sa:d I
Clay
3 Soil texture
4
variable
Y
Sand
Communities:
I C1:
1
3
Sand
4
5
6
8
b)
Habitat
1 Altitude
(m1O)
3 Soil texture
4
ranges
15-20
11-13
16-18
10-13
17-18
12-16
17-19
40-45
25-42
35-42
31-471
55-68
52-65
53-70
60-83
Sand
Sand
Clay
Clay
Sand
Sand
Clay
Clay
(cm) 10-14
2 Soil depth
variable
Communities:
1
3
2
4
5
7
6
c)
Soil depth
10-13
FIGURE
112
10-14
listed
11-13
according
12-16
15-20
to ascending
16-18
range
17-18
17-19
3.9. - Hypothetical
model illustrating the sequence of three.
environmental
factors, namely, altitude, soil depth and
soil texture to hierarchically
differentiate eight communities (a). If soil depth is first in the sequence (b)
then differentiation
of only four groups of two communities each, occurs. This is because in (b) the ranges
overlap so that no groups can be formed, as can be seen
in (c), whereas, the division of soil depth by altitude
in (a) creates non-intersecting
sub-groups of soil depth.
8
ication.
of
A
program
grouped
number
sequences
degree
to which
package.
The
intersect
the
indicates
stratification
with
the
the
two
3.4.2.3
an
Floristic
understanding
of
factors
which
derived.
process
factors
correspondence
ordination
CANOCO
between
techniques
(Ter Braak
selected
scattergram
to assess
cess
The
environment
available,
the
method
is limited
tested
floristics
tested
between
whereas,
the
relationships
can
also
lead
to
differentiation.
to those
habitat
that have been recorded
cause
problems
relationships
and habitat
in
between
can be assessed
such as DECORANA
However,
the former
variables
correspondence,
is based
is
the
the
then be incomplete.
programs
floristics
but
community
could
would
so that the classification
following
of
environmental
and
sequences
correspondence
process
process
1987).
which
and habitat
such as those
using
PHYTOTAB-PC
Furthermore,
hypothesis,
of floristic
causes
as
sets
relationships
and floristics
overlaying
bines
not
of
two
for mapping.
classification
Factors
the
in both
hypothesis
the
the basis
are available,
in
correspondence.
degree
inval idate
the
between
releve-groups
of
low
correspondence
correspondence
habitat
the
and habitat
the
included
a basic
A
can an examination
the
is
form
can
in verifying
correspondence
amount
can also provide
Obviously,
and
the
can
processes
Not only
aid
determine
classification.
converse
or
to
and the
in a single
is not based
on a simple
(Hill 1989b)
program
over
with
the
latter
requires
floristic
program
classification
on floristics
hierarchical
compro-
alone.
model
which
113
associated
step
correspondence
direct
tests
habitat
factors
is to construct
hypothetical
factors,
model
because
available
is merely
where
communities
has changed.
to test
sity
to
model
could
construct
the
particular.
mean
annual
are
because
in
programmatically,
factors
should
example,
Figure
the
The
Such
in
3. 9b
sequence
limitations
hence
the
neces-
a
hypothetical
of relevancy
and economy,
to record
usually
soil
3.9a.
is necessary
because
model.
first
of habitat
of permutation
in terms
The
the
in the
field.
The
be from the general
texture
should
not
to
precede
rainfall.
the
with
for
the sequence
illustrated
hypothetical
factors
basis.
hierarchical,
is
and
as in Figure
differentiated
habitat
For
one
class
sYmbol.
other
sets
model,
for
each
or
If
more
all
then
correspondence
first,
the
determined
ranges
the
which
from
determined
are
sequences
of habitat
starting
This
also be of benefit,
in determining
sequence
being
It is not possible
all possible
model,
to determine
the sequence,
correspondence.
releve-groups
on a releve
a hypothetical
determining
fewer
between
a
for
the
general,
ranges
the
range
sets
class
habitat
of
releve-group.
releve-groups
single
that
most
that
Each
is
habitat
unique
allocated
have
sYmbol
factor,
a
factor
set
habitat
factor
and
is then
tested
in the
allocated
the
of
different
interceptions
is
as
with
and
no
rei eve-group
is
implied.
The
next
habitat
separately
previous
within
habitat
factor
each
factor.
of
the
unique
The process
sets,
same
manner,
determined
is repeated
with
but
the
for all the
114
habitat
at
factors.
various
levels
differentiation
conditions
Individual
in
of
the
releve-groups
habitat
releve-groups
can
factor
is
be
differentiated
hierarchy.
achieved
then
If
a
poor
the
following
factor do not correspond
precisely
could apply:
i.
the classification
is inadequate;
ii.
the habitat
selection
iii.
the model
iv.
the limits
with
factor
is inadequate;
releves,
values
or,
for a habitat
the limits
of a releve-group.
representing
which
is inadequate;
two
In other words,
communities
do not correspond
with
have
adjacent
habitat
the border
factor
between
the
two communities.
In
the
using
case
of
either
releve-group
3.4.2.4
The
classified
step
condition
releves
or
in
and field
the
releve-groups
This
of the
plant
the
process
omitting
can
be
releves
repeated
adjacent
to
relationships
verification
with
the
step is not to confirm
species
in the
process
units
to
is
be
comparing
mapped,
the occurrence,
relevant
on
the
the
on the ground,
releve-groups,
which
should
be the case, but the following:
to assess
the degree
ant rei eve-group
ii.
last
delimiters.
ground.
i.
nodal
Classification
final
clearly
the
to assess
levant
to which
the plant
are representative
the variation
rei eve-group
represented
and the degree
species
of the relev-
of the unit to be mapped;
by the releves
to which
this
of the recorresponds
115
with
the variation
of the unit to be mapped,
at the relevant
scale;
iii. to assess
plant
the value of plant
community
community
diagnosis
iv.
habitat
the reliability
mapped;
the validity
evant
the
60%
or
sponds
which
can be
floristic
and
of the units
to be
of the releve-groups
are mappable
at the rel-
scale.
more,
invalidate
well
does
with
stratification,
not
the
processes
habitat
tion,
but
does
rather
the
a
classification
habitat
satisfactory.
these
derived
so that the community
of the borders
has
relevancy
size and regular-
of the hypothesized
classification
be deemed
spacing,
the plant
and,
to assess
If
of identifying
for
relationships;
to assess
vi.
is important
can be used
recognized;
to assess
v.
which
as a means
in the field. Here plant
ity of occurrence
easily
species
the
data
However,
not
concerned
of the classification
field
of
and
and classification
invalidate
could
will ultimately
be
about
corre-
assessment,
lack of correspondence
necessarily
process
stratification,
and the
correspondence
efficiency
the
then
can all
in any of
classifica-
questioned.
The
depend
on the uses
information
that can be
therefrom.
3.5 DERIVATIVES
The term derivative
derived
directly
is taken
(primary
here, to mean
derivatives)
or
indirectly
(secondary
116
from
derivatives)
as
the
locality
mappable
maps,
the
of
factors
vegetation
cess,
which
can
being
derivatives
3.5.1
Background
be
of such
interpret
tion.
and
Those
can
uses
be
and a map
as much
demand
information
able
to interpret
be the
from
matrix
be visually
from
units,
as
pro-
rather
than
relationships
will
to workers
who
are
able
the
classifica-
information
thereof.
in this
field
from classifications,
a
that
determine
for such work.
to
from
Information
largely
and hence the need
as possible
The classified
the
It is,
to derive
so that
the
of coded variables
for
can be increased.
can include
factor,
habitat
compilers
importance
applied
workers
and derive
a classification
for such work
of
with
of a study then the value
additional
information
fication
spatial
derive
of great
habitat
illustrating
to those
3.5.2 Methods
each
vegetation
classification,
be limited
for classifications
therefore,
with
of the stratification
are the end-products
should
derived
co-incidence
can be illustrated
will
best
classification
the
such
of the classification.
units
a study
by
relationships
their
maps. Mapped
are the product
modified
If a classification
of vegetation
or
or soils
or integrated
in this study,
Spatial
units
such as geology
map overlays
proposed
classification.
above
each releve,
variables,
correlated
a summary
(Deall
to form a passive
&
Westfall
1989)
classi-
which
can
with the releve-groups.
117
There
is usually
formation
much
contained
ucting
a synoptic
duced
to a single
redundancy
therein
matrix
A five class
Ellenberg
1974).
tance
of single
i.
This
according
scale
tends
a "+" sYmbol
classified
is generally
which
used
habitat
data
dent on the classification
and
based
are re-
on
species
&
(Mueller-Dombois
can be overcome
the
impor-
by using:
or,
in section
3.2.3.
matrix,
although
synoptic
in-
by constr-
over-emphasize
scale described
the
releve-groups
for single occurrences;
number
and
reduced
to constancy
to visually
occurrences
ii. the plant
The
can be considerably
(Werger 1974) where
column
presence.
in a classification
are not considered
derivatives,
depenin this
study.
3.5.2.1
Primary
The primary
derivatives
derivatives
that
follow,
can be derived
from the
syn-
optic matrix.
3.5.2.1
A) Plant
Braun-Blanquet
terized
theses
less
by
of
(1928)
"its
Terms
species
ional
species
1974).
Clearly,
cular
own
matrices
clear.
character
communities
study
were
species".
took
such
as
place
as
well
its
a plant
As
and
more
community
data
community
exclusive,
a plant
to
clarify
community
composition,
as being
were
and
situation
is restricted
diagnosis
and
synbecame
preferential
territorial
the
charac-
gathered,
distinctions
selective
as differential,
introduced
unless
area,
defined
and
reg-
(Werger
to a parti-
description
for
118
the study
area,
ity of which
could
differ
it is a part.
considerably
This
community
variation
present
not known
where
community
area.
The
following
areas
where
i.
a rei eve-group
the
plant
A
is
community
area,
because
the species
formal
that
of
is only
which
is usually
included
in a study
for individual
areas
included
only where
in
the
adequately
partially
study
is involved:
the
study
area
represent
included
of the variation
plant
the
of plant
that
a
in a study
represented
by
in the
study
is
the
area,
and
species;
should
but only with
insufficient
to the
community,
diagnostic
communities
study areas,
is restricted
plant
community,
to as community
is usually
of the
is not known;
that
ranking
area,
community
cannot
representing
for individual
there
completely
of the species-group
are referred
Hi.
study
a plant
the proportion
releve-group
diagnosis
study
suggested
releve-group
plant
to the proportion
is not entirely
with other
the releve-group,
ii.
in the
represents
community
concerned.
relates
is, therefore,
no synthesis
from that of the commun-
information
not be applied
syntheses,
because
to elucidate
rela-
tionships;
iv.
v.
informal
ranking
of plant
based
on combined
apply
for the combined
releve-groups
resent
munity,
tent
releve-groups,
without
ecotones,
communities
where
a study
conditions
area
is
i. and ii.
group;
community
extremes
or communities
within
diagnostic
in variation
for which
species
for a particular
the particular
in the study area is inappropriate;
can rep-
scale
comor ex-
or,
119
-
a)
Releves
community
diagnostic
species
+
+
+
+
+
+
+
+
+ + + +
+ + + +
+ + +
+
+ +
c)
Releves
Community
diagnostic
species
+
+
+
+
FIGURE
120
+
+
+
+
+
+
+
b)
Releves
Community
diagnostic
species
+ + + +
+ + +
+ +
+
+
+ + + + + + +
+ + + + +
+ + +
+
Releves
d)
community
diagnostic
species
+ + + +
+ + +
+ +
+
+
+
+
+
+
+
+
+
+
+
+
+
+
3.10. - Species presence in community diagnostic speciesgroups: a) few blanks indicating a well-defined community
corresponding
to an abrupt environmental change at the
community limits; b) species presence approximates a
normal curve, indicating a well-defined community but
with a less abrupt change at the community limits than in
(a); c) a variation of (b) where the releves have been re
-arranged so that the strongest community expression, in
terms of species presence, is at the left of the group,
decreasing towards the right. This pattern does not adequately show spatial relationships because the releves on
the extreme right can often be adjacent to different
communities;
and d) a poorly-defined
community.
vi.
plant
community
species,
composition
both
diagnostic
the
releve-group/s
is,
however,
is represented
and
non-diagnostic,
representing
only
a
sample
by all the
the
of
plant
the
present
community.
species
present
in
This
in
the
community.
This
information
as well
as plant
directly
from the synoptic
3.5.2.1.
B) Community
Community
degree
is according
unity
Figure
3.l0a,
differentiated
limits
pattern
co~munity
to minimum
diagnostic
can be obtained
matrix.
species
the
names
definition
diagnostic
to which
community
total
species
should
indicate
and can, therefore,
be
is defined,
separation
with
by an abrupt
can
no or
used
if the
units.
with
releve
as
a well-defined
be mapped
infer
comm-
illustrated
community
change
the
sequence
For example,
few gaps,
environmental
to
which
in
is
at the
community
a high degree
of preci-
sion.
A
community
where
approximates
abrupt
a normal
is
according
to
skewness,
the
in the
re-arranged
without
width
central
within
units
diagnostic
(Figure
at
3.10b)
the
well-defined.
r e Leve s
affecting
separation
change
still
the
community
curve
environmental
community
those
the
of
best
part
the
community
ecotone.
expressing
then
increase
Depending
a less
but
on
in Figure
the
the
original
be
sample
composition
However,
the
will
The releves
as illustrated
because
indicate
limits
community
relationships.
pattern
precision
of the releve-group.
community
will
should
Mapping
the releve-group
species
are
can be
3.10c
total
releve
121
Releve-groups
1
2
3
4
5
6
7
8
9
species-groups
FIGURE
3.11. - Species-groups
representing two or more communities, sequenced to correspond to an environm~ntal
gradient of which both the upper and lower limits for differentiated communities can be ascertained. This also
applies to the horizontal mirror-image of the illustrated pattern.
Releve-groups
1
2
3
4
5
6
7
8
9
Species-groups
FIGURE
122
3.12. - Species-groups
representing two or more communities, sequenced to correspond to an environmental
gradient of which only the upper or lower limits for differentiated communities can be ascertained. This also
applies to the horizontal mirror-image of the illustrated pattern.
sequence
A
is based
community
irregular
on minimum
diagnostic
spacing
to any of the
community;
caused
sample
Mapping
precision
3.5.2.1
C)
be
environmental
a)
a
(Figure
3.11),
of
should
be
(section
simplification
releves
matrix
the
limits;
over the
would be low.
total
rather
both
those
evident
after
patterns
and
lower
with either
3.12).
illustrated
in
overlapping
releve-groups,
upper
(Figure
grouping
Two main
or b) a gradient
only
by
two or more
with
correlation
of
can emerge,
environmental
upper
Both these
Figures
to corre-
or lower
types
3.11
can
3.12.
&
communities
be
The
with
3.4.2.3).
of
the
species-groups,
separation
or releve-groups,
but
can be attributed
disturbance
simplified
gradients.
gradient
mirror-images
affect
and
size.
representing
limits
The
blanks
over the community
by widespread
further
environmental
habitat
change
of such a community
can
species-groups,
limits
3.10d)
many
Gradients
matrix
limits
(Figure
with
or,
iv. inadequate
namely,
pattern
units.
community;
environmental
iii. heterogeneity
with
species
of occurrences
defined
ii. a gradual
spond
separation
following:
i. a poorly
The
total
units
so that
additional
or
in this
the
will
relationships
information
pattern
way,
becomes
not
between
is not lost from the
evident.
The
matrix
123
1
species-groups
2
3
4
5
6
7
1
2
3
a
4
5
6
7
8
9
10
11
b
12
13
14
15
16
FIGURE
124
_____
I}c
3.13. - Schematic illustration of the simplification of a
phytosociological
table into 3 sections: a) community
diagnostic species-groups;
b) species-groups corresponding to environmental
gradients. Horizontal
mirror-images
of the illustrated pattern also apply;
and c) non-diagnostic
species.
will
now
unity
consist
diagnostic
environmental
3.5.2.2
The
of
gradients
secondary
data
parts,
(Figure
species-groups;
Secondary
recorded
three
b)
3.13)
namely,
species-groups
and c) non-diagnostic·
a)
comm-
indicating
species.
derivatives
derivatives
but
are
those
are dependent
results
on the
computed
classification
from
for
the
group-
ing.
3.5.2.2
The
A)
Structure
general
height
correspondence
(section
ture,
according
within
to
each growth
to be made.
into
3.2.4),
This
grasses
recorded
as total
3.5.2.2
B)
A plant
species
dual plant
permit
total
which
canopy
growth
forms
of vegetation
cover
differentiation
which
is not
cover per height
Community
recorded
for
all
and
strucspecies
(after Ito 1979) for each community,
allows
forbs
the
layer diagrams
recorded
form class
also
and
between
composition
utilizes
possible
of the
where
ground
layer
structure
is
class.
analysis
a large
resource-space
per
indivi-
is likely to:
i. have a correspondingly
high crown cover;
ii. have a high resource-space
requirement,in
the mature
phase;
and,
iii. have a high cover
in relation
to its frequency.
125
The
converse
with
a
low
requires
plant
could
cover
a
in
be
valid,
relation
correspondingly
in the
mature
cover-to-frequency
with trees
also
namely,
to
its
small
phase.
as the disparity
a plant
frequency
of
resource-space
It is obvious
ratios, grasses
that
that
species
occurrence,
per
individual
in comparisons
of
for example, cannot be compared
in resource-space
requirements
is too
great in the mature phase. These statements
are generalizations
as
individual
in their
to
external
plants
factors.
dynamics
section
within
the
elucidation
and
in
each
between
for the
regressions
decreasing
to
the
plant
growth
Cover-to-frequency
calculated
relation
a defined
according
3.2.4.
frequencies
cover
However,
considerably
of
reactions
intra-community
not that
plants.
species
categorized
then
differ
in this study is dependent on species reactions,
of individual
Plant
can
a
linear
growth
cover
actual
and
ratios
class
frequency
classes
for
each
for
is
is
are,
therefore,
defined
species,
the
assumed.
The
A
in
are
cover
determined.
of each species,
is also calculated.
order
form
regression
form
frequency
community
and
linear
expected
according
to the
The
species
are then arranged
of the differences
between
actual
in
and predicted
cover.
Species
outside
frequency
with
a
standard
strong
the standard error of the mean,
regressions,
higher
error
form
cover,
and
of
mean.
competitor
the
species
two
those
The
distinct
with
first
because
a
groups,
lower
group
of
for the cover-tonamely,
cover
than
is referred
their
those
to
the
as
resource-space
126
requirements
and
weak
competitor
range
between
referred
a
their
these
can
to the
of
3.5.2.2
After
at
groups
a third
15
obtained
for
analysis.
For
used
only
and
differences
at
Thus,
to
as
within
a
which
can be
species
within
classes
(i.e.
form classes)
which
and the effect
of
requirements.
calculations
is the determina-
growth
classes,
of the
cover within
form
each growth
within
a
form class.
analysis
which
the
community
practical
in
group
in the community
composition
the
fall
different
analysis,
it is then
the cover of any stand of vegetation
scale
referred
species
range.
into
programmatic
C) Stand phase
the
are
in each of five growth
proportions,
community
most
competition
present
of these
group
forming
categorized
and the total
the
compare
be
second
Generally
resource-space
cover
community
the
two groups
species
differing
An extension
tion
species.
competitor
relate
in
to as the normal
community
three
those
community
with
purposes
cover
differences
cover
between
was
the
the
stand
with
community
into
and
that
composition
frequency
account.
the
to
the community,
sampled,
community
taken
the
within
possible
can
Thus
community
be
the
are
emphasized.
Five
distinct
community,
Phase
phases,
in
can be identified,
0: all the species
i.e.
there
the
comparison
between
stand
and
namely:
fall into the normal
are no strong
or weak
competition
competitor
range.
classes
and,
127
therefore,
Phase
1: weak
no class
competitors
indicating
Phase
2: strong
3: weak
The
competitors
competitors
class
could
of
classes
these
strong
Phase
4: Strong
These
the
main
stand
stages.
indicate
the
growth
form class,
form classes.
biome
or vegetation
a progression
to trees
juvenile
through
is possible.
stage
are
Some
potential
and
occur
in the woody
growth
as
in Phase
3,
A
tendency
towards
form
could
be
biome
or
unidominance
could exist.
the
stage
to the community
thus
Reversal
southern
3.5.2.2
phases
is
in
shrubs
classes,
moribundity
growth
on the
of savanna
dependent.
relative
Dynamics
case
competitors;
formation
in the grass
in the woody
from dwarf
The
form class,
utilization;
be dependent
competitors
growth
grass utilization;
occur
species
classes.
and/or
occur
In the
or disadvantaged;
in the grass
selective
excessive
formation.
the
occur
possible
indicating
Phase
is advantaged
inferred,
of vegetation
of which
assuming
of the sequence
development
the stand
progression
is probably
of
is a part.
through
the
a rare occurrence
in
Africa.
D) community
The
ideal
vegetation
the
type
of
composition
the
ideal
the
type
composition
utilization
ideally
for goats.
of
cover assessment
animal
suited
for utilization
applied.
to
For
example,
cattle
grazing
It would
appear
far more
utilization
and
degree
is dependent
the
is not
on
vegetation
necessarily
efficient
of utilization
to
adapt
to
the
128
specific
vegetation
resource
than to try to manage
the resource
to
suit the form of utilization.
condition
Vegetation
perspectives,
namely,
utilization
and
condition.
ecological
condition
vegetation
to
with
Tree
should
cover
analysis,
and
relationship
and error
the
mean
between
c
tree
This
is
merely
required
ground
adequacy
relate
particular
and
as
accordingly,
an
of
resilience
or
(section
1.3)
adequate
slope
two
type
to the ability
implies
penetration
from
of the
ground
increases,
describes
for
an elementary
adequacy.
in
community
slope
tree
and
cover
the
are
slope
community
used
was
as
taken
composition
input.
The
after
trial
minimum
tree
to be :
where
level
a
justification
This
determined
c
cover
of the
soil.
root
assessed
production
tree cover
values,
for
primarily
following,
for assessing
be
suitability
In terms
increasing
The
therefore,
sustained
protect
soil binding.
method
a)
b)
ecological
cover
can,
cover
a
for
= tan s x 100
(%) and s
provisional
root
soil
indication
binding
to 100% on a 45° slope.
of ground
of
which
the
varies
No assessment
from
is made
0%
on
of the
cover.
3.6 PHYTOTAB-PC
PHYTOTAB-PC
is a program
package
to
facilitate
the
methods
des129
cribed
in this
logical
data.
files
written
AT
286
or
5~"
in TURBO
drive,
C:\PT
b) sub-directory
Classification
tory,
Menu
a
by
items
single
menu
"PT".
according
items
are
can
be
executed
the
program
of BASIC,
disk
and
The
for an
a 360 KB,
package
is
namely,
and processing,
for a floristic
from the C:\PT
and
data bank.
main
menu
by entering
preceding
the
to the numeric
used
package
necessary.
are
file,
selection
the
symbol,
item.
then
which
invoked.
is bound
by
sequence
is
items. Alphabetic
functions
described,
from any direc-
is
and utility
The
or if ";C:\PT"
Application
main menu
for documentation
when
directory,
of the AUTOEXEC.BAT
The
are executed
bracket,
normally
and text
and processing
in the path
typing
version
configuration.
for classification
of phytosocio-
of 147 program
20 MB hard
both menu driven,
are executed
is included
a compiled
minimum
C:\PT\PD
handling
consists
640 KB RAM,
into two parts,
The programs
package
BASIC,
with
disk
a) directory
3.6.1
and also facilitate
The program
IBM-PC
floppy
divided
chapter
according
programs
main
which
and procedures
to
the
menus,
for
as
follows:
D) Documentation
This
details
initiation
on-line
as well
Data
initiates
format:
of
directories,
as program
program
usage
transfer
by displaying
and
program
or printing
the
manual.
1) Matrix
This
setup
releve
Input
keyboard
input
of
the
matrix
(species;
cover-abundance)
in either
or TABIN
format:
PHYTOTAB
species
130
releve;
cover-abundance).
continuation
or eight
of
input.
character
Provision
Species
is made
for a new data
are abbreviated
set or
to an eight
digit
code.
2) File Handling
This
option
view
and edit
files
can be used if PC-WRITE
sequential
to diskette
files,
Matrix
to QMAT
QMAT
random
access
process
done.
must
Outputs
codename
file
file
files
and
be
not
(if PHYTOTAB
be
transformed
species
sequence
split.
before
Eleven
option,
time,
000.
another
first
Matrix
is
the
This
are
It should
to
TABIN
and QMAT
file.
be
that
noted
computer
TABIN
need
not,
releve
2 cannot
Corformat
the
allocated
The
releve
numbers
therefore,
data
occur
should
in part
number.
is
in two
used
limitation
product
in the order
code numbers,
or
numbers
be
file,
of computation.
1 to n. However,
releve
can
sequence
PHYTOTAB
re Lev e number
This
Sequencing
option
option
the
Releve
from
available
size
for speed
contain
example,
and species
the
releves.
186
For
options
menu,
files
processing
input)
the
processing.
species
computer
either
again.
or range
and after
4) Releve
to
further
is the
file
and releves.
in sequence
be
format
further
file,
and releve
access
done
and
species
is resident,
as to back up and restore
for
any
sequence
for species
should
for the
before
releve
is a random
rections
file used
completed
are:
file
format
last
be
as well
3.00)
or, for· file transfer.
3) Transform
is the
(Version
of
can
of hours
for
On the
programmatically
for releve
species
also
submenus.
and
utilize
using
greater
considerable
to days, depending
sub-
sequencing
sequencing,
reLev es not
first
on matrix
this
than
processing
size,
131
noise
The
(see section
second
releve
the
previous
third
and
option,
sequence
to group
for
where
accordingly
full
inserting
characters
ed using
or
this
option
user
sequencing
_option
is
(see sections
and
the
downloading.
numbers
in
and releve-group
of subsequent
a
checking
or
to
option
is
Separation
can be insert-
Deleting
by one.
a number
It is important
that the numbers
on the releve
file are computer
which
be
the
numbers,
then
converted
On the
of
converted
a
second
submenu
or
from
species
1 to
sequence),
delimiters;
and a random
fifth
and
conversions
the previous
mutation
sixth
and
reversal
which
four options
and
an
is not
sequence
versa.
of
are
These
an
existing
generation
of
a number
affected
options
sequence
by the
number
must
releve
(for
presence
a
of
replicates.
to-sequence
be used
The final option
the optimum
entry
to
existing
sequence
and
submenu.
keyboard
from 1 to n, without
are executed.
to obtain
permit
codes
processed
the second
corrections
be used),
options
vice
of releves
sequence,
can
ni
mirror-image
The
invokes
the first four options
(or an editor
sequence
original
back. The fifth option
releve
sequence
to
The
be
to remember
must
used
can
fourth
delimiters
numbers
also
file
sequence.
can be used.
or using
& 3.3.3).
codename
The
the
alphabetically
abbreviation
or an editor
the value
3.3.2
abbreviations
facilitate
deleting
after
by the
species
when
is executed
either
numerically
names
for species
decreases
the
to
species
species
required
numbers
species,
determined,
The
sequences
releve
updated
match
has been
releves
and processor.
for sequencing
option.
option
the
3.3.1).
code
if any
of
is a per-
sequence,
exclud132
ing mirror
option
images,
should
extremely
on minimum
not be used
total
separation
for more than about
units.
20 releves,
This
as it is
time-consuming.
5) Peripheral
This
based
option
data
input
invokes
ity names,
table
classif ied
codes.
a submenu
title,
for
community
codes
and midpoint
of environmental
habitat
dat.a
Other
codes
for input
data descriptions
which
composition
values
of species
can
be
analysis
used in the matrix
variables
which
commun-
and habitat
data
are
form
input
and
names,
check
growth
lists;
cover
and the actual
can be either
descriptive
values
or num-
erical.
6) Print
options
files
1 to 7 are used to print
checking
or
sequence
of species
frequency
reference
as
alphabetic
species
7) Print
option
full
species
occurrences
for each
names
option
and other
optionally,
This
option
growth
created,
an
alphabetic
forms,
in PC-WRITE
can
also
be
for
species
format
used
if
for
an
is required.
Table
printing
have
separation
total
classification
Final
8 prepares
units per species,
listing
names
and
Option
with,
checklist.
permits
relev8;
8) Print
This
a
Working
This
centage
purposes.
and separation
annotating
files that have been
of a-matrix
been
added.
units
before
The
for each
habitat
output
species;
separation
units
efficiency
(see section
data
includes
total
for all species;
and
total
species
and per-
3.4.1).
Table
permits
peripheral
printing
of a matrix
with
full
species
names
data.
133
9) Process
This
Data
option
invokes
Synoptic
1.
be
table
according
with
number
4. Community
cover
5. Community
and habitat
indicated
u) utility
do
first
three
as
not
which
can
class
"+"
a
3.5).
can be applied
or
scale
the
All
33
options
here.
includes
community
& 3.5.2.2).
(see section
3.5.2.3).
(see section
correlation
3.5.2.4)
with relationships
of a dendrogram
part
options
errors
species
names
being
internal
form
format
(see section
(see section
that
fourth
of
can
the
3.4.3).
3.2.3.7).
invoked
occur
after
the
main
External
program
submenu
are
downloading
menu,
utility
are
pro-
package.
for
The
correcting
matrix
or data bank or to change
data
or
the para-
name file.
(see
for key or other
purposes.
The
converts
option
PHYTOTAB-PC
is used to generate
keys
from
programs.
the
from a mainframe
option
identification
executable
utility
with
for the species
fifth
five
with
section
3.5.2.1
relationships
programs,
to
grams
meters
cover
a
values
programs
utility
referred
(see
assessment
by means
scale;
analysis
analysis
the matrix
indicated
(see sections
phase
6. Species
class
scale
composition
structure
The
five
to the full matrix
2. Community
3. Stand
a
in which
occurrences
plant
available
for the following:
production
to
single
class
The
a submenu
section
species
3.2.2.8)
the matrix
and
sequences
to
split
for plant
matrices
data to the CONDENSED
format
134
for the Cornell
1979b)
or
required
to
the
programs
CANOCO
for
the
(Ter Braak
CANOCO
necessary
available
are those
sixth
option
paired
or
means,
range,
Graphics.
values
The
using
The ninth
output
can
other
used
options
& Istock
include
as well
as
can also be sorted
according
1987)
applications
applications
be
be
et al. 1982).
statistical
variance
data
format
(Scheiner
(Hill
as
input
various
linear
for
according
on
re-
Harvard
to x or y
option
fixed
raster
mapping
according
to
large sequential
files,
such as
for raster mapping.
is used
for inputting
information
point
system.
co-ordinates
This
is
for the
required
for
localities.
option
This
for
is used to merge
geographic
releve
tenth
allows
the printer.
option
ARC/INFO-PC
scale.
The
habitat
loaded,
(Westfall
basic
deviation,
be
editor.
format
for
Should
can
Analysis
variables.
variables
option
could be required
The
an
and DECORANA
ascending.
The eighth
mapping
used
Optional
seventh
grids,
is
1987).
these
using
mainframe
standard
Paired
1986,
for Affinity
unpaired
gressions.
(Hill 1979a)
programs
format,
and for the PHYTOTAB
The
TWINSPAN
does
(C)
not
converts
improve
cover
codes
the precision
to
the
of the
plant
original
number
esti-
mates.
The eleventh
option
(L) is used for generating
and locating
random
135
TABLE 3.7.
- A model of the
primary
options
with
keystroke
sequences
IIHaincutplt
PHYTOTAB-PCprogram package,
square brackets
indicating
II
A. Instructions
on-Line rranual
=11~tion
[D]II
B. Classification
Data i.npJt
Print data files
:l:trh data
=> Cover axles [5-4]
Matrix data [1]
[
~re
<r
fil~
Edit files
i:1Matrhdata
1<
[6-11
[2-HI
leve sequence [6-2-2] ~
Releve sequence [2-1-2]
Sp3cies SEq.lence [6-2-3] <---> Sp3cies SEq.lence [2-1-3]
Nu'rb:r axles
[6-3]
Nurrber axles
[2-1-4]
> Releve sequence [3] 11----'> '---------------'
Sp3cies sequence
Matrix file
Reduce data set
L
,-
Nurrber axles
> ISplit rratr ix [U-4-C]
I[
Classily
I
-> Releves
[4-1]1
<-
.
:]Sp3cies
[4-2]
L::
-
sort; SfECi~------'
dich::Jtcm::usly [U-4-S]
<
--'
,--
Print v..orkingtable
r
136
> User SEq.lence [4-5-2/3-1]
Ascen:li.ngsequence [4-5-2/3-2]
Reverse SEq.lence [4-5-2/3-3]Rarrl:::m
sequence [4-5-2/3-4]
Permutation SEq.lence [4-5-2-7]
Plant '""'"
I
~>IIPrint v..orkingtable
Optional sequences
for speci.es & releves
1
Convert releve nurrbers
ISe:Juencenurt:er to SEq.lencecede [4-5-2-5]
Syn:Jptic rratrix
1<-
[7]1<-
<
;-------c>ISyn:JptiC
files
[9-1]1
1<
- A model of the PHYTOTAB-PC
program,
TABLE 3.7. (continued)
package primary options with square brackets
indicating
keystroke
sequences
C. Final table
Print final table
>I'Final
table
1
PeriIberal data inp.It
------------------------------------------------,
Print checklist
Title, taxa, syntaxa, habitat descriptors, foot.n:Jtes [5-1)
II
II
Habitat data classified cedes or plant characters [5-2)
-> S}::ecieschecklist [6-8)
G=<th form axles
[>I=~cdata
[5-31
.nli..I.~
baM
[Q-PD-l)
D. Envirorrrental aJrrespondence
~
Envirorrrental data inp.It
~I
aJrresporrence
Envirorrrental variables [5-5) 1--------------------------:>
E. G::mfOSition
analyses
II
II
> stan::ifhase analysis [9-3)
======;]
)1 ::t:rorrurli'
ty ccrrpos
i tion
II
Cbmunity &
habitat [9-5 )
[9-2 )
=====================dJ---------------->rr=======================~
Cbmunity oover assesSTel1t9-4]
~II
:37
S}::eciesoover relationshirs
[9-6) II
TABLE 3.7. (continued)
- A model of the PHYTOTAB-PCprogram,
package primary
options
with square brackets
indicating
keystroke
sequences
F. Internal
utility
prc:grcmc;
blanks fran matrix file
(usually wiEn transferring data) (lJ-1]
Convert mY'J:1:BlI.S SPFNAM
to \Pr\DESaU
(for transfer fran data bank)
(lJ-2]
Reclirrensionor rarove nares or blanks
fran file \Pr\DESClU
(lJ-3]
M,
Rarove
IIstatistics
1"=Je
Convert to other forrrats (c:7I,NX)J,
mY'IOTI\B-ma.inrarre
)
(lJ-S]
/ Linear re:;p:-essions (lJ-6] II
I""",_W
Cb--Drdinateconversions or generate
ran::k:mpoints
(U-L]
files
I
!!Sarrple/stand dirrensions
IINurTber
set cxnp:rri9Jns
G. External utility
Cover cx:de conversions
138
(U-q
(U-N]II
prc:grarrs
13-0 onlination
Plotter
Digital ll\3fPing
artp.>t is requized
[C>N3DJ
I
species ccrrpartsons with PREdata
set, search or ad:! autbors
(SP]
I
points
grid.
according
to
Provision
or from decimal
The twelfth
is also made
degrees
option
sequences
same data
set.
stand
ii.
sampling
number
Two
function
number
keys
estimation
are
invokes
a problem
operative
section
solving
files generated
copies
minutes
a 4 mrn
degrees
to
and seconds.
of two sets of group-
two releve
sequences
the following:
stand area and stand
from species
number
for a given
of the
radius;
and area;
and,
area from species
with
for main
of the on-line
PT are saved
tags
are
of primary
items.
manual
to separate
with multiple
are made both data integrity
identification
menu
The
Fl
and the
key
F2 key
guide.
file maintenance
A flow chart
fractional
for determining
from scale,
and
per unit area.
the relevant
simplifies
(S) is used
unit dimensions
invokes
All
and degrees,
dimensions
iii. species
co-ordinates
for converting
such as comparing
option
i.
longitude
(N) is for the comparison
ed number
The thirteenth
latitude,
printed
program
users
and security
automatically
options
This
and if backup
file
are ensured.
File
for
is shown
diskettes.
each
in Table
diskette.
3.7.
3.6.2 Data bank
This
is a separate
functioning
set of programs,
as a floristic
data
on a subdirectory
bank.
The data
bank
C:\PT\PD,
progr.ams are
139
the
C:\PT
directory,
executed
from
the
of the AUTOEXEC. BAT
path
"PO". Four
species
diskettes
menu
file,
files are resident
names,
community
generated
selections
or
if ";C:\PT"
from
any .directory,
on a hard disk,
names
and an index.
the
classification
with
is included
namely,
Data
by typing
matrix
are
in
data,
loaded
from
The
main
programs.
are the following:
1. Load data
This
is from
final
table
grams
are to
used
in the
has been printed
be used
for analysing
2. List
The
diskette
and deemed
for a data
a single
data
bank.
format,
only
satisfactory,
can
splitting
the
if the pro-
The programs
set and
after
also
be
matrices.
data
following
data
can be
listed:
to code or alphabetically;
identifiers,
species
PHYTOTAB-PC
being
sequences
for a particular
names,
an index of data
computer
being
species
codes
the
and
order
according
sets resident;
original
in which
data set and community
both
releve
species
names,
releve
numbers;
should
occur
both numerically
or
alphabetically.
3. Search
Searches
bers,
routines
can be made
data
starting
sets or multiple
subdivision
and merging
ssification
are generated
Searches
names
when
can
also
or truncated
more
species
than
should
one
be
with releve
groups
identifiers,
of releve
numbers.
of data sets. The necessary
name
be present
num-
This permits
files
for cla-
codes,
species
on diskette.
made
species
releve
starting
names.
is
The
entered
in a releve
from
species
Boolean
for
a
"and"
search,
for the releve
logic
applies
i.e.
all
the
to be select140
ed. Apart
from
programs,
a list of species
releve
basis,
4. Correct
This
random
to
and floristic
Species
The
Search
is limited
loading
with selected
community
species,
on a
For
name
files.
The
index
with an editor.
for table
large
is
data
production,
can be
compared
for errors.
by the
time
capacity
of the hard
proportional
sets,
an 80386
to
the
or 80486
disk
amount
drive.
of
processor
data
is an
advantage.
utility
Apart
ASCII
from
an
the
higher),
vegetation
programs
file
editor
following
sb ch
as
programs
PC-WRITE
can
(version
further
3.00
facilitate
analyses:
DECORANA:
The
for ordination
of sample
AA:
the classification
names
3.6.3 External
or
and
on the data base,
and
obvious
species
on diskette,
bank
present.
associated
data can be corrected
names
data
for use with
files
the
species
with those
diskettes
can be obtained.
applies
5. Check
creating
CANOCO
because
and species
Affinity
variation
analysis
of
communities
degrees
of a data
CAN3D:
This
program
DECORANA
output.
(Ter Braak
1986;
of the options
numbers
version
available
in the scatter
(Scheiner
of
1987)
& Istock
compositional
is preferred
and the
inclusion
diagrams.
1987)
for analysing
relatedness
among
the
the
set.
is
A plotter
used
for
three-dimensional
plotting
of
is required.
141
SPECOM:
(Westfall
species
names
in
prep.).
against
list and adding
the
author
This
program
computerized
names
is
used
National
for
checking
Herbarium
synonym
for checklists.
3.7 REFERENCES
ACOCKS,
J.P.H.
Botanical
BAKER,
H.A.
cards.
BIANCHI,
1975. Veld Types of South Africa.
Survey
of South Africa
A.T.J.
of South African
1931. Een nieuwe
of the
40: 1-128.
1970. A key for the genus Erica
Journal
Memoirs
Botany
using
edge-punched
36: 151-156.
determinatie-methode.
Tectona
24:
884-893.
BOLUS,
H.M.L.
Notes
BOUGHEY,
1958. Key to the genera
on Mesembryanthemum
A.S.,
BRIDGES,
of Mesembryanthemaceae.
and allied
Genera
K.W. & IKEDA, A.G.
biological
identification
key. Museum
University
of California,
Irvine,
2: 392-403.
1968. An automated
of Systematic
Research
Series,
Biology,
2: i-xix,
1-35.
BRAUN-BLANQUET,
J. 1928. pflanzensoziologie.
1 Aufl.
springer,
Wien.
COETZEE,
B.J.
Rustenburg
DEALL,
G.B.
1975. A phytosociological
Nature
& WESTFALL,
floristic/habitat
tables.
DUKE,
J.A.
Reserve.
Bothalia
1969.
Bothalia
R.H. 1989.
pattern
classification
of the
11: 561-580.
Improving
correlations
the resolution
of
on phytosociological
19: 263-266.
On tropical
systems
and systematics.
Gardens
56: 125-161.
tree seedlings.
Annals
I. Seeds,
of the Missouri
seedling
Botanical
142
A broad-scale
=
.eqe~a~~~
~=o
,J.M.
samples
F.S.S.A.,
of benthic
1974. Manual
H.G.
Society
G.E.,
List of species
GIBBS
1987. List
of southern
ecology.
of species
Edition
2.
Part 1. Cryptogams,
Memoirs
W.G.,
of the Botanical
RETIEF,
and synonyms.
Survey
E., IMMELMAN,
K.L.,
B.J., VAN WYK, M. & NICHOLAS,
of southern
African
revision
M.Sc.
Identification
plants.
2.
Memoirs
59: 1-270
of the subtribe
thesis,
A.
Edition
Part 2. Dicotyledons.
of South Africa
1974. A taxonomic
1968.
plants.
51: 1-152.
G. PIENAAR,
literature
D.W.
The
in community
African
and synonyms.
(Mesembryanthemaceae).
GOODALL,
of
C., VAN ROOY, J. & SMOOK, L. 1985.
G.E., WELMAN,
of the Botanical
H.F.
37: 1-65.
analysis
analysis
Wilson.
of South Africa
Monocotyledons.
GERMISHUIZEN,
GLEN,
Kendall,
methods.
of South Africa
RUSSELL,
Recent
14: 705-712.
of soil analysis
REID,
literature
Gymnosperms,
Survey
Bothalia
of
New York.
RUSSELL,
Recent
classification
for the statistical
invertebrates.
1982. Multivariate
Cambridge,
GIBBS
purposes.
1983. Some methods
Fertilizer
GAUCH,
r practical
structural
University
by computer.
Gibbaeinae
of Cape Town.
BioScience
18:
485-488.
GYLLENBERG.
H.G.
1965. A model
micro-organisms.
HANSEN,
Journal
for the computer
of General
Microbiology
B. & RAHN, K. 1969. Determination
by means
of a punched-card
1-46 & 172 punched
system.
identification
39: 401-405.
of Angiosperm
Dansk botanisk
of
families
Archiv
26:
cards.
143
J.S.
HENKEL,
1934. A field book of the woody plants
Zululand.
HERRE,
Robinson,
A.G.J.
of Natal
and
Durban.
1971. Genera
of
Mesembryanthemaceae.
Tafelberg,
Cape Town.
HILL,
M.O.
1979a.
TWINSPAN
classification
Unpublished
M.O.
report.
Cornell
DECORANA
correspondence
A
A. & Okuda,
pp.357-368.
program
two-way
for arranging
table by
and attributes.
University,
FORTRAN
Ithaca,
program
and reciprocal
University,
ITO, K. 1979. A tentative
Japans.
-
analysis
Cornell
Miyawaki,
FORTRAN
of the individuals
1979b.
report.
A
data in an ordered
multivariate
HILL,
-
Ithaca,
for detrended
averaging.
Unpublished
New York.
study of stratification
S. (eds). Vegetation
The Yokohama
New York.
diagrams.
In
und Landschaft
Phytosociological
Society,
Yokohama.
JANSEN,
H. 1982. The geology
Transvaal,
Republic
Geological
Survey
LAMARCK,
J.B.A.P.M.
Royale,
LAPAGE,
Computer
S. WILCOX,
identification
in the
of the
71: 1-98.
vols.
1-22. Academic
BASCOMB,
Identification
perspectives.
M.,
of bacteria
MALAN,
Press,
Imprimerie
O.G.
&
& R.J.
in
London.
by computer:
TURNER,
1970.
and Data Handling
W.R. & CURTIS,
of General
M.A.
In A. Baillie
Mechanisation
S. WILCOX,
Journal
W.R. & CURTIS,
of bacteria.
(eds), Automation,
S.P.,
LASSERRE,
of South Africa
Memoirs
de. 1778. Flore Francaise3
BASCOMB,
Microbiology
LAPAGE,
of South Africa.
Basins
Paris.
S.P.,
Gilbert
of the Waterberg
M.A.
general
Microbiology
B.
1983.The
1973.
aspects
and
77: 273-290.
application
of
144
component
principal
of seminar
and earth
LEENHOUTS,
analysis
on Principal
sciences,
1966. Keys
a new kind.
Proceedings
Akademie
MACVICAR,
J.J.N.,
T.H.
Agricultural
MALAN,
quantitative
African
symposium
METCALFE,
C.R.
P.G.
S.W.A.
of digital
image
of the third
image processing,
Oxford
University
1969. Eintuehrung
Wissenschaft1iches
L.E.
L.E.
1971.
Specimen
data;
South
Durban
von Suedwest
22-23
Afrika.
matrices.
University"
Biological
1975.
Recent
Press,
Gesel1schaft,
identification
Taxon
programs
and description
Publications
of the Dicotyledons.
advances
Suedwestafrikas.
Windhoek.
and key construction
20: 269-282.
for specimen
printing
of the Museum,
Series
Vol.
Oxford.
in die Pflanzenwelt
computers.
1974. Computer
key construction
L.E.
einer Flora
& CHALK, L. 1979. Anatomy
with time-sharing
MORSE,
of
Lehre.
1, 2nd edition.
MORSE,
Department
Pretoria.
Proceedings
H. 1972. Prodromus
Cramer,
MORSE,
J. VAN
1985.
MERXMUELLER,
MEYER,
E.,
M. 1977. Soil classifica-
Services,
on Digital
C. 69: 571-596.
R.F., VERSTER,
for South Africa.
and optimal.
of
Nederlandsche
F.R., LE ROUX,
& LAMB, A.D. 1985. Display
O.G.
July
H.J.VON
and a proposal
Series
J.M., LOXTON,
system
Technical
1983.
a survey
Amsterdam,
MERRYWEATHER,
& HARMSE,
tion: A binomial
7-8 February
Proceedings
in the atmosphere
of the Koninklijke
DE VILLIERS,
LAMBRECHTS,
MSS data.
analysis
in biology:
van wetenschappe
C.N.,
ROOYEN,
component
Pretoria
P.W.
to Landsat
identification,
using
Michigan
taxonomic
data
State
5: 1-128.
in the
theory
and
practice
of
145
specimen
identification
vegetation
with computers:
D. & ELLENBERG,
MUELLER-DOMBOIS,
ecology.
MUNSELL
SOIL COLOR
NEWELL,
I.M.
Wiley,
CHARTS,
Press,
London.
H. 1974. Aims and methods
1954. Munsell
Color
of
co., Baltimore.
and use of tabular
1963. Some aspects
keys. New Phytologist
R.J.
Watsonia
PANKHURST,
11-52. Academic
keys. Pacific
12: 25-37.
D.V.
PANKHURST,
(Ed.), Biological
New York.
1970. Construction
Insects
OSBORNE,
In R.J. Pankhurst
identification.
of the theory
of dichotomous
62: 144-160.
1971. Botanical
keys generated
by computer.
8: 357-368.
R.J.
1978. Biological
identification.
Edward
Arnold,
London.
PARTON,
& INNIS, G.S. 1972. Some graphs and their functional
W.J.
forms.
U.S.
Project
Technical
RUTHERFORD,
M.C.
Province
RUTHERFORD,
B.W.,
for the
CLAPPER,
J.C.
methodologies
Biome
153: 1-41.
Bothalia
R.H. 1986. Biomes
of South Africa
BOWEN,
of the Transvaal
15: 294-295.
categorization.
W.E.,
Grassland
of southern
Memoirs
of the
54: 1-98.
I.G. & BABB, R. 1967. A model
of bacteria.
Journal
of General
46: 407-424.
(Unpubl.)
Centre,
S.M.,
Program,
R.H. 1984. Sectors
& WESTFALL,
identification
Research
SCHEINER,
& WESTFALL,
Survey
Microbiology
SCHEEPERS,
Report
- An objective
Botanical
RYPKA,
Biological
of South Africa.
M.C.
Africa
International
Geomorphology
classes.
Grassland
Pretoria.
& ISTOCK, C.A. 1987. Affinity
and statistical
inference.
analysis:
Vegetatio
72: 89-93.
146
SNEATH,
& SaKAL, R.R. 1973. Numerical
P.H.A.
Taxonomy.
Freeman,
San Francisco.
TAINTON,
N.M.,
BRANS BY , 0.1. & BOOYSEN,
and Pasture
Grasses
of Natal~
P. de V. 1976. Common
Shuter
Veld
& Shooter,
Pietermaritzburg.
TER BRAAK,
C.J.F.
1986. Canonical
new Eigenvector
analysis.
TER BRAAK,
Ecology
C.J.F.
relationships
Vegetatio
TROLLOPE,
Department
W.A.
& POTGIETER,
A.L.F.
analysis.
1983. Characteristics
National
University
Park.
of
Report.
of Fort Hare, Alice.
of keys and phylogenetic
of the Science
Laboratory,
trees
in
Denison
43: 1-25.
& NELSON,
M.J.A.
with
gradient
of vegetation-environment
correspondence
in the Kruger
Journal
of Colorado
WERGER,
1987. The analysis
of Agronomy,
University
WEBER,
direct
A
67: 1167-1179.
1952. The history
biology.
analysis:
69: 69-77.
W.S.W.
L.G.
for multivariate
by canonical
fire behaviour
VOSS,
technique
correspondence
P.P. 1972. Random
mosses.
University
1972. Species-area
some examples
access
of Colorado
key to the genera
Museum,
relationships
from South African
Boulder.
and plot
vegetation.
size:
Bothalia
10: 583-594.
WERGER,
M.J.A.
1974. On concepts
Zllrich-Montpellier
method
and techniques
of vegetation
applied
survey.
in the
Bothalia
11: 309-323.
WESTFALL,
R.H.
Thabazimbi
1981. The plant
ecology
of the farm Groothoek,
District.
thesis,
University
M.SC.
of Pretoria.
147
WESTFALL,
R.H.
comparing
WESTFALL,
the PRECIS
R.H.,
PHYTOTAB
derived
G.K.
and
sets.
1982.
tables.
& MALAN,
derived
recognition
O.G.
satellite
District.
in remote
2-3 July
Bothalia
M.D.
classification
Proceedings
sensing
of the
and
for vegetation
16: 263-268.
1984. A cover meter
Bothalia
relations
and determination
in the Transvaal
for canopy
15: 241-244.
J.M. & PANAGOS,
of Botany
with
vegetation-enhanced
VAN STADEN,
sampling
imagery
1984.
1986. A method
cover estimations.
species-area
data
of vegetation
the floristic
scale-related,
& PANAGOS,
R.H.,
from
Thabazimbi
using
imagery.
R.H.
vegetation
units
Groothoek,
stratification
and basal
in phytosociological
from vegetation-enhanced
Johannesburg,
R.H.
satellite
species
for preliminary
18: 122-123.
on Pattern
geophysics,
Journal
name data
for Braun-Blanquet
& MALAN, O.G. 1984. A comparison
R.H.
symposium
WESTFALL,
package
of relevesand
of the farm
WESTFALL,
species
& DE WET, B.C. 1988. New programs
R.H.
the vegetation
WESTFALL,
list with
for interrogating
G., VAN ROOYEN,. N., & THERON,
DEDNAM,
sets. Bothalia
units
A program
49: 35-37.
sequencing
WESTFALL,
SPECOM:
synonym
- A program
Vegetatio
WESTFALL,
(in prep.).
M.D.
1987. Predictive
of subsample
Waterberg.
South
size
for
African
53: 241-244.
148
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement