Content uploaded by Vivienne L. Williams
Author content
All content in this area was uploaded by Vivienne L. Williams
Content may be subject to copyright.
Application of diversity indices to appraise plant
availability in the traditional medicinal markets
of Johannesburg, South Africa
VIVIENNE L. WILLIAMS
*
, EDWARD T.F. WITKOWSKI and
Kevin Balkwill
School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, PO Wits,
2050, South Africa; *Author for correspondence (e-mail: vivwill@planetac.co.za; fax: +27-11-403-
1429)
Received 4 June 2003; accepted in revised form 19 April 2004
Key words: Abundance, Diversity, Evenness, Quantitative ethnobotany, Species accumulation
curve, Species richness, Traditional medicinal trade
Abstract. The lack of scientific rigour in analysing ethnobotanical surveys has prompted
researchers to investigate ways of quantitatively describing their data, including the use of eco-
logical diversity indices. There are numerous indices and measures available to describe sample
diversity. Twenty-two measures of species richness, diversity and evenness were reviewed using six
sets of ethnomedicinal data derived from 50 formal muti shop traders (of different ethnicities) and
100 informal street traders of traditional medicine in Johannesburg, South Africa, and a seventh
data set from traders on the western boundary of the Kruger National Park, South Africa. The
diversity measures were coupled with species accumulation curves to construct cumulative diversity
curves used to determine the minimum viable sample size on which a diversity index should be
based, and to better understand the differences in the relative diversities of the samples. Distinct
differences in the relative abundance and diversity of plants sold by street traders and shop traders
were evident. Species diversity and evenness was found to be higher in shops, thus resulting in a
lower dominance in the sale of certain plant species compared to the street traders. A survey of an
informal market should include no less than 35 research participants compared to no less than 20
for the muti shops. The use of selected indices of species richness (Margalef’s), diversity (Shannon,
Simpson’s, Fisher’s alpha, Hill’s numbers) and evenness are recommended as a means of describing
patterns exhibited within ethnobotanical data.
Introduction
An emergent trend in ethnobotanical studies has been the use of quantitative
methods and models to describe patterns of plant use and availability in sur-
veys or assessments of natural resources (e.g., Prance et al. 1987; Phillips and
Gentry 1993a, b; Johns et al. 1994; Begossi 1996; Ho
¨ft et al. 1999; Hanazaki
et al. 2000; Luoga et al. 2000; Cunningham 2001; Wong et al. 2001), thereby
allowing for a more rigorous statistical approach to the discipline. While a
quantitative approach to the discipline is not always possible or even necessary,
the benefits include: a greater depth to the understanding of the subject under
investigation; a conscious attempt at reporting and refining methods employed
to collect and evaluate the data; attention to sampling effort; economy of
Biodiversity and Conservation 14:2971–3001 ÓSpringer 2005
DOI 10.1007/s10531-004-0256-4
description and examination of patterns in the data; hypothesis testing; and the
ability to question and describe more precisely the results of surveys.
Begossi (1996) first demonstrated how diversity indices were useful quanti-
tative tools that could assist researchers analysing ethnobotanical data by
allowing comparisons of diversity among different communities in different or
similar environments. In ecology there are numerous indices available for
exploring species diversity between different communities. Begossi (1996)
demonstrated how the Shannon–Wiener index of diversity and evenness, and
the rarefaction curve, might be used to compare differences in diversity, uni-
formity of species use and sampling effort in eight samples from South
America, Thailand and Tonga. This paper broadens the spectrum of diversity
indices to include 22 measures of richness, evenness and diversity. The goals are
to: evaluate the performance of the indices in relation to samples of different
sizes, and trader profiles and to examine the kind of information they provide;
make recommendations on measures appropriate for quantifying ethnobo-
tanical data; assess whether the survey sites were adequately sampled, and
determine the minimum viable sample size on which a diversity measure should
be based for the type of survey data collected; and lastly, to compare the
species diversity of sites within the formal and informal sector, and thereby
appraise plant availability within the ethnobotanical trade in the region.
Species diversity
Diversity indices are used to characterise species abundance relationships in
communities (Ludwig and Reynolds 1988). The literature reveals a multifari-
ous array of ecological indices, usually expressing species diversity as a single
number (Magurran 1988). How can the appropriate measure be determined?
The answer largely depends on the question the index is being used to answer,
the component of diversity being measured, and whether the index is simple to
use and understand.
Diversity measures take into account two factors: species richness (i.e., the
number of species, S) and evenness/equitability (i.e., how uniformly abundant
species are in a sample) (Magurran 1988). An ‘index’ of species diversity (also
called an index of heterogeneity) incorporates both richness and evenness into
a single value. Species diversity measures are broadly divided into four main
categories: indices of species richness; indices of evenness; indices of species
diversity/heterogeneity; and species abundance/distribution models (Ludwig
and Reynolds 1988; Magurran 1988). These concepts may be translated into
ethnobotanical terms to answer the following questions: (1) what is the species
richness of plants used/sold within a sampled market or group of resource
users?; (2) how does the species diversity of plants sold differ between different
groups of traders?; (3) are the same plants sold by most traders?; and (4) is
sampling effort adequate, i.e., have sufficient numbers of research/survey
participants been interviewed?
2972
Recent research by Colwell and Coddington (1995) and Gotelli and Colwell
(2001) recommend the measurement and comparison of species richness by the
use of taxon sampling or accumulation curves. The curves may be computed
from EstimateS (Colwell 2001) software that computes randomised species
accumulation curves and also several indices of diversity and the parameters to
calculate others. Plotting the performance of the indices on a diversity curve
demonstrates the performance of the indices and differences in relative abun-
dance as sample size increases.
Study area
Johannesburg is located within a region of Gauteng Province called the
‘Witwatersrand’, the name given to an extensively urbanised axis of approxi-
mately 100 40 km and part of a geological super-group consisting of gold-
bearing conglomerates (Lowrey and Wright 1987). The Witwatersrand
emerged from a small mining town in the 1880s and labour for the mines were
provided by mainly rural people in the migrant labour system. The ensuing
rural–urban oscillation of Black labour from around the country enhanced the
introduction of activities related to Black ‘rural’ culture (Dauskardt 1990,
1991). Traditional herbalism was incorporated into the developing urban mine
culture to meet the needs of both the Black migrant labourers and the rapidly
expanding, permanent urban population for traditional medicine (Dauskardt
1991). Various historical processes shaped the preponderance of different
ethnic and language groups within sectors of the emerging South African
capitalist economy and the traditional medicine trade. The Witwatersrand is
South Africa’s second largest market for medicinal plants after the markets in
KwaZulu-Natal, and the ethnic diversity of the region’s traders, healers,
gatherers and consumers is influential in determining the traded floristic
diversity and sources of the plants harvested for the multinational trade
(Williams et al. 2000).
The trade is differentiated into two sectors, namely formal business and
informal street markets (Williams et al. 1997). Traders, including traditional
healers, selling traditional medicines from premises called ‘muti’ shops, repre-
sent the formal sector. In 1994, there were estimated to be 244 shops in the
region selling traditional medicines (Williams et al. 1997), the majority of which
were owned by Black traditional healers (52%) and Indian merchants (25%).
Commercial gatherers and traders selling plants from the pavements and
street markets, on the other hand, represent the informal sector. Located in
Johannesburg, the ‘Faraday Street’ market is the Witwatersrand’s only infor-
mal wholesale and retail street market for traditional medicine. Ninety-seven
percent of the approximately 166 traders are migrants to the Witwatersrand, of
whom 90% regard the province of KwaZulu-Natal as ‘‘home’’ (Williams 2003).
Customers to the market are primarily traditional healers from Gauteng
townships, owners of muti shops, and sometimes patients seeking treatment
2973
from the traders that are traditional healers. Most of the traders earn less than
R100 a week (US$1 R6.51, June 2004) (Williams 2003).
Methods
Market surveys
Between February 1994 and January 2001, two semi-quantitative surveys of
plants traded for traditional medicine were conducted within the Witwaters-
rand. The first survey in 1994, based on a stratified random sample of 50
research participants from 50 muti shops, appraised the nature of the plant
trade in the formal sector. The second survey in 2001, a stratified random
survey of 100 street traders in the Faraday Street informal market, was con-
ducted on behalf of the provincial Directorate for Nature Conservation for
Gauteng. The surveys were based on questionnaires and structured interviews,
and an inventory of all common names of plants sold by each trader was
compiled. The characteristics of the trade and the species sold within the shops
and at the market have already been quantitatively and qualitatively described
(Williams et al. 2000, 2001; Williams 2003).
Synthesis of plant inventories
Identification of the species traded was mainly achieved by matching vernac-
ular names to botanical names from previously published studies that, for the
most part, are reliable because of the credible body of literature existing for
ethnobotanical names in South Africa. In some cases, species were visually
identifiable or were identified later from purchased specimens. Species identi-
fication through published records is problematic, and errors in identification
are likely to have occurred. However, this was considered the most expedient
mechanism for identifying the large numbers of inventoried species sold by each
of the traders surveyed. In order to make a distinction between vernacular
names synonymous with >1 species, analysis by ‘ethnospecies’ instead of by
botanical species was used. ‘Ethnospecies’ (Hanazaki et al. 2000) takes into
account the folk or common name given to one or several species quoted during
interviews. The Zulu name ‘iNgwavuma’, for example, is the ethnospecies name
designating Elaeodendron transvaalense (Burtt Davy) R.H. Archer, whilst the
ethnospecies ‘iMphepho’ applies to at least six species of Helichrysum.iMphe-
pho was cited 17 times during the Faraday market survey, however, only one of
the six potential Helichrysum species would have been sold at each stall and the
most prevalent species is not known. Therefore, wherever appropriate, the data
were quantified based on the number and frequency of occurrence of ethno-
species to avoid repetitions and hence any bias/inaccuracies in reporting the
results. In this paper, all citations of species are citations for plant ethnospecies.
2974
Six data sets derived from the Witwatersrand medicinal plant trade were
subject to analysis with the 22 diversity measures. This was to evaluate the effect
of different sample sizes, trader ethnicities and formal/informal market sectors
in the appraisal of patterns in the utilisation and trade of traditional plant
medicines. The sample of 50 muti shops (abbreviated as MS: All 1994 in the
graphs) was subdivided into three smaller subsamples based on the ethnicity of
the shop owner, namely Black (n= 28 shops; MS: Black 1994), Indian (n=20
shops; MS: Indian 1994) and White (n= 2 shops; MS: White 1994). The
Faraday market data (ST: Faraday) were not subdivided for the initial appraisal
of the indices. An earlier survey conducted by the author in 1992 of seven muti
shops (MS: 1992) was included to compare the effect of sample size (Williams
1992). Whenever appropriate, the performance of the indices was compared
with a seventh data set – a sample of 17 informal vendors derived from an
inventory compiled for medicinal plants traded on the western boundary of the
Kruger National Park, South Africa (Botha 2001; Botha et al. 2001). The
dataset is abbreviated as ‘ST: WBKP’ in the graphs. Later, by way of an
independent example comparing intra-sample diversity for selected indices, the
Faraday data matrix was subdivided into ‘healer’ and ‘non-healer’ traders.
Calculating indices
The calculation of an index to evaluate ecological diversity relies on infor-
mation regarding the number and frequency of occurrence of species in a
sampled community. In order to calculate diversity indices for ethnobotanical
purposes it is necessary to have data on the number of individual informants
(e.g., resource users/traders) who cited the plant species (Begossi 1996). Since
the inventory of plants sold by each trader in this study recorded the presence
of an ethnospecies, each ethnospecies was recorded/cited once per trader and,
therefore, incidence/occurrence equals abundance of the ethnospecies per tra-
der sample/inventory.
References for the formulae and software used to calculate the indices are
listed in Table 1. The statistical distributions used to fit species abundance
observations may be used for fitting species occurrences (Hayek and Buzas
1997). Noccurrences may be substituted for Nindividuals (Hayek and Buzas
1997). In the calculations, n= number of samples (e.g., number of muti shops
or street traders surveyed) and N= number of citations or occurrences of
ethnospecies.
The species accumulation curves and cumulative diversity curves were con-
structed from variables and diversity statistics computed by EstimateS (Colwell
2001). In cases where EstimateS did not directly compute the diversity measure
(e.g., Margalef and Menhinick’s indices, Hill’s diversity numbers N1 and N2,
evenness measure E1–E5), then the formulae cited in Magurran (1988) and the
appropriate variables computed by EstimateS were used to calculate the indices
and construct the graphs. It is important to note that EstimateS computes the
2975
reciprocal of Simpson’s k. In the formulae for diversity indices, any logarithmic
base may be used (Zar 1984). As a way of standardising the results, the natural
log (ln) was used throughout.
Results and discussion
Species richness
Numerical species richness (S), or the number of species in a sample of a
specified size, is an instantly comprehensible expression of species diversity
Table 1. References for methods used to compute the indices and measures applied to the data.
Index/Measure Reference
Species richness indices
# Species (Sor N0) Discerned from observation of the data set
Margalef (DMg) Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
Menhinick (DMn) Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
Diversity indices
Shannon–Wiener (H0) Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Colwell (2001)
*
Shannon–Wiener (Hmax) Magurran (1988)
Comparing Shannon indices Zar (1984); Magurran (1988); Murali et al. (1996)
or use standard deviation for H0from Colwell (2001)
*
Brillouin measure (HB) Zar (1984); Magurran (1988); Krebs (1989)
Simpson (k) Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Colwell (2001)
*
Simpson (ln k) Pielou (1975); Colwell (2001)
*
Berger–Parker (d) Magurran (1988)
McIntosh’s dominance (D) Magurran (1988)
Fisher’s alpha (a) Magurran (1988); Krebs (1989); Colwell (2001)
*
Hill’s Diversity Number N1 Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
Hill’s Diversity Number N2 Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
Hill’s Diversity Number N1Magurran (1988)
Evenness measures
Shannon (J0or E1) (or, Pielou’s J) Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
E2 (or, Buzas0& Gibson’s E) Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
E3, E4 and E5 Ludwig and Reynolds (1989) – program ‘SPDIVERS.BAS’
or Magurran (1988) and Colwell (2001)
*
Brillouin (J) Krebs (1989)
McIntosh’s (E) Magurran (1988)
*
Indicates that EstimateS (Colwell 2001) either computes the index and/or the parameters that can
be inserted into the equations obtained from the other references listed.
2976
(Magurran 1988). Sis related to the total number of individuals (N) summed
over all Sspecies recorded. As sampling effort increases, more individuals are
encountered and more species are likely to be recorded (Hayek and Buzas
1997). The relative abundance of species, however, is important and a number
of simple indices have been derived using some combination of Sand N. These
indices include Margalef’s and Menhinick’s index of species richness. While
these indices are easy to calculate, they are (like S) sensitive to sample size.
The relationship between Sand Nmay be viewed by plotting a species
accumulation curve (Figure 1a, b), also termed a species effort curve or collector’s
curve – so called because the cumulative number of species is plotted against
some measure of the effort it took to obtain that sample of species (Hayek and
Buzas 1997). Compared to interpreting the single numerical value of species
(205) MS: Survey 1992
(310) MS: Black 1994
(312) MS: Indian 1994
(371) MS: All 1994
(349) ST: Faraday
0
50
100
150
200
250
300
350
400
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Number of species (S)
(205) MS: Survey 1992
(310) MS: Black 1994
(312) MS: Indian 1994
(371) MS: All 1994
(349) ST: Faraday
0
50
100
150
200
250
300
350
400
1000 1500 2000 2500 3000 3500 4000 45000 500
Number of individuals (N)
Number of species (S)
(a)
(b)
Figure 1. Species accumulation curves (or collector’s curves) for plant ethnospecies traded as
traditional medicine in muti shops and in a street market in the Witwatersrand. The curves rep-
resent successively pooled and randomly ordered samples (a) and individuals (b). The curves were
computed using EstimateS (Colwell 2001). The total number of ethnospecies per sample (S)is
labelled in brackets at the end of each curve. (MS: Muti shops; ST: Street traders).
2977
richness for the randomly pooled samples (Table 2), plotting the curves facili-
tates improved interpretation of species richness results for different samples of
varying size. Comparing raw taxon counts (and index values) for two or more
assemblages/samples will quite generally produce misleading results (Gotelli
and Colwell 2001). Differences in measured species richness between commu-
nities may be the result of differences in underlying species richness, differences
in the shape of the relative abundance distribution, or because of differences in
the number of individuals counted (Gotelli and Colwell 2001).
Whereas fewer muti shop traders were sampled compared to street traders
(n) (Figure 1a), the number of individual plants (N) recorded in the shops was
greater (Figure 1b), and hence the numerical richness per trader is greater for
the muti shops. There is also a similarity in shape and clustering of curves for
the shop data (‘MS’), even the 1992 survey of seven shops, making them
distinct from the curve of the street trader (‘ST ’) sample (Figure 1a). The initial
steep gradients of the curves for the shop data show that more ethnospecies per
shop are sold (mean = 83) and consequently the accumulation of ethnospecies
is more rapid, even for smaller sample sizes. Street traders, by contrast, sell
fewer ethnospecies per trader (mean = 24) and consequently the accumulation
of ethnospecies is slower. When samples are highly variable in terms of plant
diversity amongst traders, then more samples are needed to fully represent the
trade in medicinal plants, while if the samples show little variation then fewer
traders need to be sampled.
The Margalef and Menhinick indices have been cited as being inadequate by
several authors (e.g., Brower and Zar 1977; Magurran 1988; Hayek and Buzas
1997) because the indices lack the ability to differentiate the species richness of
samples having similar Sand N. Looking at Table 2, the performance of these
indices as a single numerical value for pooled samples is difficult to adequately
judge. However, plotting the performance of an index as samples are succes-
sively pooled and individuals are accumulated is a useful procedure for aiding
the interpretation of plant availability within the different trader groups
(Figures 2 and 3).
The graphs of Margalef’s index show how species richness increases until
eventually the curve levels off with increasing sample size and the number of
individuals inventoried (Figure 2a, b). The point at which the curve flattens
indicates a minimum viable sample size on which a diversity or richness index
should be based (Magurran 1988). The curve of S(Figure 1) can also be
constructed for this purpose i.e., to determine the minimum requisite sampling
effort. The sample of species sold by White shop traders (n= 2) is numerically
inadequate and only two points of richness could be plotted on the graphs.
However, the sampling strategy for the 50 traders selected for the shop survey
(including the 2 White traders) was stratified random, and therefore trader
ethnicities were proportionately representative within the sample to minimize
participation biases (Williams et al. 1997). Sampling more White traders would
only have been necessary if it had been an important aim to compare the shops
of different trader classes, but it would have biased the overall results of the
2978
Table 2. Species richness indices calculated for six data sets sampled from Witwatersrand traders of traditional medicine. n= the number of samples (traders
and/or shops surveyed); N= number of individual ethnospecies observed; S= number of ethnospecies counted.
Index/measure 1992 Muti Shop
Survey (n=7)
N= 809
1994 Muti Shop
Survey (N= 50)
2001 Street trader survey
(n= 100) N= 2402
‘White-owned’
(n=2)N= 193
‘Indian-owned’
(n= 20) N= 2168
‘Black-owned’
(n= 28) N= 1769
Total (‘All’) shops
(n= 50) N= 4129
# Species (S or N0 or e
H
max
) 205 144 313 310 371 349
Margalef (DMg) 30.5 27.2 40.5 41.3 44.4 44.7
Menhinick (DMn) 7.2 10.4 6.7 7.4 5.8 7.1
2979
study. The minimum viable sample size (i.e., the number of research partici-
pants) necessary for assessing species richness is larger for informal street
traders than for shop traders. The evidence for this is reflected in the species
accumulation and diversity graphs for street traders, which show that a larger
sampling effort is necessary before the curves begin to reach an asymptote
(Figures 1 and 2).
Evidence for the distinctive trading patterns in species richness in formal and
informal trading sectors are supported by the sample of 17 informal vendors
from the western boundary of the Kruger National Park (ST: WBKP,
Figure 2a). Despite the small sample size (which was conducted with 73% of
the vendors at the site), the curve adopts a similar shape to the Faraday Street
trader sample (ST: Faraday) and exhibits low ethnospecies richness per trader.
(44.7) ST: Faraday
(44.4) MS: All 1994
(41.3) MS: Black 1994
(40.5) MS: Indian 1994
(30.5) MS: Survey 1992
(25.6) ST: WBKP
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Margalef's index
(44.7) ST: Faraday (44.4) MS:All 1994
(41.3) MS: Black 1994
(40.5) MS: Indian 1994
(30.5) MS: Survey 1992
(25.6) ST: WBKP
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
Number of individuals (N)
Magalef's Index
0 500 1000 1500 2000 2500 3000 3500 4000 4500
(a)
(b)
Figure 2. The cumulative species richness curves of Margalef’s index for ethnospecies traded on
the Witwatersrand for both samples (a) and individuals (b). ‘WBKP’ is a sample of 17 informal
vendors trading medicinal plants on the western boundary of the Kruger National Park, South
Africa (Botha 2001). The overall value of the index for the randomly pooled samples is labelled in
brackets at the end of each curve. The formula for the index is DMg =(S1)/ln N. (MS: Muti
shops; ST: Street traders).
2980
Gotelli and Colwell (2001) recommend that when comparing species richness
between samples, the number of taxa should be plotted as a function of the
accumulated number of individuals (not samples) because datasets may differ
systematically in the mean number of individuals per sample. Figure 2b,
therefore, shows that the relative abundance and richness of ethnospecies sold
at the Faraday street market is higher than that of the shops. Margalef’s index
for Faraday is 44.7 compared to 44.4 for ‘All shops’ (MS: All). The relative
abundance and richness of ethnospecies sold by Black traders is slightly higher
than that of Indian traders.
The graph of Menhinick’s index (Figure 3) corroborates earlier evidence
derived from Margalef’s index that street traders keep a lower species richness
per trader, and therefore increased sampling effort is required for the curve to
reach a peak before declining as nand Nincrease. The ST: WBKP sample
shows a similar pattern to the Faraday Street trader data, which is different to
the pattern shown by the muti shops. The numerical richness values for the
index similarly indicate the Faraday sample to be relatively more species rich
than the ‘All shops’, and the Black trader sample to be more species rich than
the Indian trader sample.
Species diversity or heterogeneity
Indices of diversity or heterogeneity incorporate both richness and evenness
into a single value and are based on the proportional abundance of species in a
sample (Ludwig and Reynolds 1988; Magurran 1988). These measures are
(7.1) ST: Faraday
(5.8) MS: All 1994
(7.4) MS: Black 1994
(6.7) MS: Indian 1994
(7.3) ST: WBKP (7.2) MS: Survey 1992
2.0
4.0
6.0
8.0
10.0
12.0
Menhinick's index
Number of individuals (N)
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Figure 3. The cumulative species richness curves of Menhinick’s index for ethnospecies traded on
the Witwatersrand. The overall value of the index for the randomly pooled samples is labelled at
the end of each curve. The formula for the index is DMn =S/ ffiffiffiffi
N
p. (MS: Muti shops; ST: Street
traders; WBKP: Western Boundary Kruger National Park).
2981
attractive to researchers because, unlike the species abundance models, no
assumptions are made about the underlying distributions of the data (Hayek
and Buzas 1997). There are four categories of indices. First are measures de-
rived from information theory (e.g., Shannon–Wiener and Brillouin), based on
the rationale that diversity or information in a natural system may be measured
in a similar way to information contained in a code or message (Magurran
1988). The second category of indices are the dominance indices (e.g., Simpson,
McIntosh and Berger–Parker), so called because they are weighted towards the
abundances of the commonest species (Magurran 1988). A third category of
diversity index is ‘Hill’s diversity numbers’. The numbers, developed by Hill
(1973), show how diversity indices are mathematically related and may be
arranged in a sequence depending on whether they measure species richness
(weighted towards uncommon species) or dominance (weighted towards
abundant species) (Magurran 1988). Interpreting the single statistic for each
index can be problematic (Table 3). A fourth category of diversity index is
derived from the logarithmic series abundance model, namely Fisher’s alpha.
Information theory indices
The Shannon index (H0) measures the average degree of ‘‘uncertainty’’ in
predicting to what species individuals chosen at random will belong (Ludwig
and Reynolds 1988). Uncertainty may be visualised as being synonymous with
diversity (Krebs 1989), therefore, the higher the degree of uncertainty, the
higher the diversity and the greater the degree of uncertainty in correctly
predicting the identity of the next species chosen at random. The average
uncertainty (H0) increases as Sincreases, as seen in Figure 4 when compared
with Figure 1. Figure 4a shows that there is a distinction in the curves between
formal and informal traders (Figure 4a), with the overall degree of uncertainty
and diversity being higher in the formal sector (Figure 4b). The higher pre-
dictability of a species’ identity in the street markets is because of the lower
mean species richness per trader, as discussed in the previous section. Informal
traders sell fewer species and one may more comfortably predict what popular
species most traders are likely to sell. In terms of determining the minimum
sample size necessary to assess the Shannon index (as indicated by the point at
which the curve levels off), the curve for shop traders begins to reach an
asymptote at around 15–25 samples, compared to 25–30 for the street traders.
Maximum uncertainty (Hmax) will occur when each species in a sample is
equally represented (Table 3) (Hayek and Buzas 1997). The more species there
are, the more evenly the individuals are spread across the species, the higher
will be the value of H0because there will be greater uncertainty as to which
species will most likely be observed next time they are chosen at random. It
appears that a characteristic of ethnobotanical samples (especially those of
large sample sizes) is that H0is high. In examples described in Magurran (1988)
for ‘‘natural’’ communities (e.g., diversity of birds in woodlands; species
diversity in plantations etc.,) H0ranged between 1.38 and 3.54. By contrast,
Begossi (1996) estimated H0to be between 2.99 and 5.95 (average = 4.6) for
2982
Table 3. Species diversity indices calculated for six data sets sampled from Witwatersrand traders of traditional plant medicine.
Index/Measure 1992 Muti Shop
Survey (n=7)
N= 809
1994 Muti Shop Survey (N= 50) 2001 Street trader survey
(n= 100) N= 2402
‘White-owned’
(n=2)
N= 193
‘Indian-owned’
(n= 20)
N= 2168
‘Black-owned’
(n= 28)
N= 1769
Total (‘All’)
shops (n= 50)
N= 4129
Shannon-Wiener (H0) 5.16 4.91 5.43 5.38 5.46 5.33
Shannon-Wiener (Hmax) 5.32 4.97 5.75 5.74 5.92 5.86
Brillouin measure (HB) 2.08 1.78 2.25 2.22 2.30 2.21
Simpson (k) 0.0050 0.0026 0.0047 0.0051 0.0049 0.0066
Simpson (ln k) 5.29 5.95 5.36 5.28 5.32 5.02
Berger-Parker (d) 0.0087 0.0104 0.0088 0.0113 0.0094 0.0249
McIntosh’s dominance (D) 0.954 0.982 0.949 0.947 0.942 0.935
Fisher’s alpha (a) 88.49 257.28 99.95 108.85 98.75 112.26
Hill’s Diversity Number N1(eH0)
*
174.9 135.7 227.5 216.8 233.9 206.5
Hill’s Diversity Number N2(1/k)
*
198.9 378.1 213.2 195.8 200.1 150.9
Hill’s Diversity Number N1
*
115.6 96.5 114.1 88.4 105.9 40.0
*
The result is a count of the number of ethnospecies.
2983
eight ethnobotanical samples from mainly South America. In the South Afri-
can study, H0ranged between 4.91 and 5.46 (Table 3).
When the Shannon index is obtained for two or more samples it is possible
to test the null hypothesis that the diversities of the samples are equal (Zar
1984). The differences in the index between the samples is mostly significant at
p< 0.000001, except for three comparisons which are approaching an equal
diversity. The diversity of plants sold by Black traders when compared with
both the diversities of the Faraday traders and Indian-owned shops is
p= 0.0021. The least significantly different samples are those of Indian traders
and All shops, where p= 0.0017 – this would suggest that the characteristics
and diversity of All Shops (n=50) is largely due to the influence of the sample
of Indian traders (n=20) within it.
(5.33) ST: Faraday
(5.46) MS: All 1994
(5.38) MS: Black 1994
(5.42) MS: Indian 1994
(5.16) MS: Survey 1992
(4.75) ST: WBKP
3
3.5
4
4.5
5
5.5
6
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Shannon diversity index (H')
(5.33) ST: Faraday (5.46) MS: All 1994
(5.38) MS: Black 1994
(5.42) MS: Indian 1994
(5.16) MS: Survey 1992
(4.75) ST: WBKP
3
3.5
4
4.5
5
5.5
6
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Number of individuals (N)
Shannon diversity index (H')
(a)
(b)
Figure 4. The cumulative diversity curves of Shannon’s diversity index (H0) for ethnospecies
traded on the Witwatersrand for samples (a) and individuals (b). The overall value of the index for
the randomly pooled samples is labelled in brackets at the end of each curve. (MS: Muti shops; ST:
Street traders; WBKP: Western Boundary Kruger National Park).
2984
The Brillouin index (HB) is similar to Shannon, and the use of this index
instead of Shannon is recommended when randomness of a sample cannot be
guaranteed (Magurran 1988). The values for HB are lower than H0(Table 3)
and show similar numerical rankings for species diversity between samples.
The obstacle to using this index is the calculation of very large factorials;
additionally, it was not possible to derive the accumulated diversity curve.
While many authors recommend the use of HB over H0(e.g., Magurran 1988),
the simplicity of calculating Shannon is to its advantage and has led to its
widespread acceptability as an index – however, Shannon is sensitive to sample
size, thus indices such as Simpson’s and Fisher’s are sometimes preferred by
researchers. Magurran (1988) says that ‘‘ideally (Fisher’s) alpha should replace
the Shannon index as the preferred measure’’.
Dominance indices
Simpson’s index (k) proposes that diversity is inversely related to the proba-
bility that two individuals picked at random from a sample belong to the same
species. Simply stated, if the probability (k) is high (kapproaches 1) that both
individuals belong to the same species, then the diversity of the sample is low
and vice versa (Pielou 1975; Ludwig and Reynolds 1988; Krebs 1989) (e.g.,
Figure 5a). However, because kdecreases as diversity increases, Simpson’s
index is usually expressed as 1 k(the probability that two individuals chosen
at random are different species)or1/k(also known as Hill’s number N2, which
is a measure of the number of very abundant species in a sample). A rarely
cited function of kis ln k(Pielou 1975) (Figure 5b, c), and is preferred by the
authors as a diversity indicator. The function does not represent a probability,
but a single diversity statistic that increases as diversity increases.
Indian shop traders, followed by the total sample of shops (MS: All) and
Black traders, have the largest diversity of plants for sale, especially compared
to the street traders (Figure 5c). As with the other indices discussed so far, there
is a distinction between the diversity of plants for sale by the muti shops and
street traders. There is, therefore, a higher probability that two plants selected
at random from different street trader stalls belong to the same species than for
two plants selected from different muti shops (Figure 5a). As a result, there is a
greater dominance of certain species sold by the street traders.
When k= 1, most individuals from a sample are concentrated in a single
species, and therefore dominance of species within the sample is high. Values
for kfor the ethnobotanical samples investigated in this paper are relatively
low (k< 0.008), and therefore the overall dominance of species is relatively
low and diversity is relatively high. By comparison, Hanazaki et al. (2000)
calculated kto be between 0.015 and 0.06 for plants used by native inhabitants
from the Atlantic Forest coast in south-eastern Brazil. The overall diversity of
plants used by the community investigated in Brazil is, therefore, lower and
dominance of plants is higher when compared with the South African study.
The minimum sample size necessary for evaluating Simpson’s index is be-
tween 15–20 for street traders and 20+ for muti shops (Figure 5b). The graph
2985
shows that the sample ‘MS: Survey 1992’ is too small to assess the index, and
therefore comparisons of diversity with the other samples. Additionally, curves
for the street traders show that despite the smaller number of individuals sold
(0.0066) ST: Faraday
(0.0050): MS: All 1994
(0.0051) MS: Black 1994
(0.0047) MS: Indian 1994
(0.0050) MS: Survey 1992
(0.0090)ST: WBKP
0.000
0.001
0.002
0.003
0.004
0.005
0.006
0.007
0.008
0.009
0.010
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Simpson's diversity index (λ)
(5.02) ST: Faraday
(5.30) MS: All 1994
(5.28) MS: Black 1994
(5.36) MS: Indian 1994
(5.29) MS: Survey 1992
(4.71) ST: WBKP
4.5
4.7
4.9
5.1
5.3
5.5
5.7
5.9
6.1
6.3
6.5
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Simpson's diversity index (-ln λ)
(5.02) ST: Faraday
(5.30) MS: All 1994
(5.28) MS: Black 1994
(5.36) MS: Indian 1994
(5.29) MS: Survey 1992
(4.71) ST: WBKP
4.5
4.7
4.9
5.1
5.3
5.5
5.7
5.9
6.1
6.3
6.5
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Number of individuals (N)
Simpson's diversity index (-ln λ)
(a)
(b)
(c)
2986
per street trader, that dominance of individuals related to their relative
abundance is established sooner (indicated by the point at which the curve
levels off). This confirms the Shannon index results, i.e., that the probability of
encountering the same species is higher between street traders. Additionally,
there are fewer abundant species sold by the street traders compared to the
muti shops (as demonstrated by the curve for N2 [or 1/k] in Figure 6).
The Berger–Parker index is a dominance measure that expresses the pro-
portional abundance of the most abundant species (d=N
max/N) (Table 3).
The index is independent of S, but is subject to bias caused by fluctuations in
the abundance of the commonest species (Magurran 1988). Like Simpson’s
index (k), diversity increases and dominance decreases as ddecreases. The
results are concordant with Shannon and Simpson’s index, except for the ‘MS:
Survey 1992’ sample which shows a higher diversity than for the Indian traders.
However, it was established in Figure 5a that the sample size was too small in
this sample to assess species dominance and hence the result is disregarded in
Figure 5. The cumulative diversity curves of Simpson’s diversity index (k) for ethnospecies traded
on the Witwatersrand for samples (a, b) and individuals (c). (a) plots the standard form of the
index, namely k, the probability that two individuals will belong to the same species. b and c plot
ln k, a rarely cited form of the index recommended by Pielou (1975) that expresses kas a diversity
statistic that increases as diversity increases. The overall value of the index for the randomly pooled
samples is labelled in brackets at the end of each curve. (MS: Muti shops; ST: Street traders;
WBKP: Western Boundary Kruger National Park).
b
N0
N1
N2
N•
0
50
100
150
200
250
300
350
400
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Hill's number (S)
number of species
exp Shannon
reciprocal Simpson
reciprocal Berger-Parker
Figure 6. The cumulative diversity curves for Hill’s numbers (N0, N1, N2 and N1) for ethno-
species traded in the Faraday Street market. The curves for N0, N1, and N2 were based on
successively pooled and randomly ordered samples, whilst N1has been estimated from non-
randomised values directly from the dataset because the requisite parameters for a randomised and
cumulative curve were not available on the software. The final value for N1is shown as because
it does not depend on the order of the randomised data matrix. The unit on the y-axis is number of
plant ethnospecies.
2987
this instance. It was not possible to plot the curves for the performance of this
index as EstimateS does not calculate the requisite parameters.
McIntosh’s index (D) is a measure of diversity independent of N. As dom-
inance increases (related to the increase in abundance of species in the sample),
the value of Dappears to decrease. However, the discriminant ability of this
index in samples of different sizes is poor and, coupled with the inability to
graph the index, the performance and usefulness of this index was difficult to
evaluate.
Hill’s diversity numbers
Hill’s numbers are the easiest diversity statistic to interpret. The diversity
numbers are in units of number of species and measure what Hill calls the
effective number of species present in a sample (Hill 1973; Ludwig and Reynolds
1988). The numbers are mathematically related to the Shannon, Simpson and
Berger–Parker indices (Table 3). As the number of species (N0 or S) increases,
less weight is placed on ‘rare’ species, and lower values are obtained for N1, N2
and N1since they measure the number of abundant,very abundant and most
abundant species in a sample respectively (Table 3). ‘Rare’ species in an eth-
nobotanical context are those species with low incidences/abundances in the
sample (i.e., low values of n).
Hill’s numbers may be plotted, for example the Faraday Street trader sample
in Figure 6. Like the other indices previously described, the point at which the
curve levels off gives an indication of the minimum viable sample size needed to
assess the index. The minimum sample size required for the street traders is
around 30 (Figure 6), compared to 20 for shop traders (not shown graphi-
cally). The distinctive high values for the first part of the N2 curve (derived
from Simpson’s index) is because N2 is weighted in favour of the commonest
species. Because incidence equals abundance in these samples, adding new or
‘rare’ species to the sample (as nor Nincreases), decreases the relative abun-
dance of the commonest species until the value stabilises when fewer new
species are added. N1, on the other hand, shows a steady increase until the
curve reaches an asymptote – this is because the function is derived from the
Shannon index, and therefore weighted in favour of species richness. As the
number of new species increases, the value of the curve increases until it
eventually levels off when very few new species are added to the sample as n
increases.
It was not possible to plot N1for the randomised and successively pooled
samples because the requisite parameters were not available on the EstimateS
software. Except for the final value of N1(shown as rin Figure 6 for the
total pooled sample), the projection of the curve has been estimated from non-
randomised values from the data matrix for Faraday. The curves for the muti
shop samples are similar in shape to the street trader data, but higher in value.
The curves are not plotted because Table 3 adequately expresses the results.
Values for N1, N2 and N1are higher for the muti shops than for the
Faraday Street market (Table 3), indicating that if species abundance equates
2988
with plant popularity, then there is greater dominance of a few popular
(abundant) species in the street market. In other words, the street traders sell a
smaller number of ethnospecies that have very high occurrences within the
market. Muti shops, on the other hand, sell a larger range of species with
equally high abundances.
This pattern is related to factors already discussed in this paper, i.e., that
street traders have neither the space nor the financial capacity to sell large
numbers of species. They, therefore, sell a smaller range of ethnospecies that
have assured commercial value and are likely to have a higher restocking
potential. Additionally, if one trader does not have a certain plant that the
customer is looking for, then there are at least 160+ other traders in the
market that might sell the plant. Muti shop traders, on the other hand, have
larger trading spaces and financial freedom and can afford to stock a large
range of species in their shop – i.e., they are ‘one-stop-shops’. The number of
species represented by N2 and N1are indicators of the number of ethno-
species within the markets that are candidates for more immediate conserva-
tion action, assuming that high incidence is an indicator of potential risk.
Fisher’s alpha
Generated from a species abundance model, Fisher’s alpha (a) is a constant
used to fit the logarithmic series distribution model once the parameter xhas
been solved for iteratively. The index is also known as the log series alpha
(Magurran 1988), and has been adjudicated as a good, if not one of the best,
measures of species diversity by several authors (e.g., Magurran 1988; Krebs
1989; Hayek and Buzas 1997) even when the underlying species abundances do
not follow a log series distribution.
Alpha is low when the number of species is low (Table 3, Figure 7), and
therefore the smaller samples with fewer ethnospecies have smaller values of a
(e.g., WBKP and MS: Survey 1992). The index is less affected by the abun-
dance of the rarest or commonest species than either H0or k, respectively
(Magurran 1988), and depends more on the number of species of intermediate
abundance. According to Hayek and Buzas (1997), Fisher’s ais a number close
to the number of species we expect to be represented by one individual – this
would account for the high values of ain the initial part of the curves in Figure
7. Because the incidence of species sold at trader’s shops/stalls equals abun-
dance, the samples all initially have very high numbers of ethnospecies repre-
sented by one individual due to the nature of the sampling methods.
Hayek and Buzas (1997) recommend that ais used as a diversity measure
when the parameter xof the log series model is 1 x0.61 because when
x0.61 then a>Sand the statistic becomes unacceptable and biologically
meaningless (where x=N/[N+a]). Greater confidence in the true value of a
occurs when xis close to 1 and Nis large. The point on the curves at which
x0.61 is marked with *, and the final value of xis also shown (Figure 7).
Eventhough ais a constant, it would appear from the data that aincreases with
Nand S. Hayek and Buzas (1997) explain this apparent paradox in the
2989
following way: if the data fits the log series model exactly, then ais a constant
independent of N. However, data rarely fits the models exactly in ‘‘Mother
nature’’, and therefore an increase in ais observed as Sand Naccumulate. The
results in Figure 7 show the Faraday Street traders to have greater species
diversity than the other samples.
Evenness
Measures of evenness (or equitability) attempt to quantify the unequal repre-
sentation of species against a hypothetical sample in which all species are
equally abundant (Krebs 1989), i.e., the ratio of observed diversity to maxi-
mum possible diversity. Hence, evenness may be referred to as relative diversity
or homogeneity (Brower and Zar 1977; Zar 1984). A low evenness means a high
dominance in the use (or presence) of a few species (Begossi 1996). When all
species are equally abundant, an evenness index would be at a maximum (of
1.0) and decrease towards zero as the relative abundances of the species diverge
away from evenness (Ludwig and Reynolds 1988). Different measures of
evenness have been proposed, most of which are expressions of Hill’s numbers
(Table 4).
All the indices gave different values but consistent rankings for the samples
and subsamples (Table 4). Which index should, therefore, be chosen as a
representative measure of how evenly distributed are the species sold by the
Witwatersrand traders? Magurran (1988) recommends the use of the Brillouin
(α=112; x=0.96) ST: Faraday
(α=99; x=0.98) MS: All 1994
(α=109; x=0.94) MS: Black 1994
(α=100; x=0.96) MS: Indian 1994
(α=89; x=0.90) MS: Survey 1992
(α=84; x=0.88) ST: WBKP
50
100
150
200
250
300
0 500 1000 1500 2000 2500 3000 3500 4000 4500
Number of individuals (N)
Fisher's alpha (α)
Figure 7. The cumulative diversity curves of Fisher’s a for ethnospecies traded in the Witwa-
tersrand. The overall value of aand x(a parameter required to fit the log series model) for the
randomly pooled samples is labelled in brackets at the end of each curve. The point marked on
the curves is where x0.61. (MS: Muti shops; ST: Street traders; WBKP: Western Boundary
Kruger National Park).
2990
evenness index, but the computation of very large factorials made it impossible
to calculate. E1, also called the Shannon J0or Pielou’s J, is probably the most
common evenness index in use but is strongly affected by species richness, and
the addition of rare species (or singletons) can greatly change the value of E1
(Ludwig and Reynolds 1988). Hayek and Buzas (1997) recommend the use of
E1 and E2 (also known as the Buzas and Gibson E). Ludwig and Reynolds
(1988) further describe E3–E5, but consider E1–E3 to be of limited value be-
cause they are highly sensitive to the number of species in the sample. A general
problem with all measures of evenness, however, is that they assume that the
total number of species that could possibly be sampled is known (Krebs 1989).
Since observed species numbers must always be less than true species richness,
the evenness ratios are always overestimated, with the possible exception of E4
and E5.
E4 and E5 remain relatively constant with sampling variations and hence
tend to be independent of sample size (Ludwig and Reynolds 1988). This is
because E4 and E5 are computed as ratios where Sis in both the numerator
and the denominator, thus effectively cancelling the impact of the number of
species in the sample (Ludwig and Reynolds 1988). However, E4 and E5 are
not totally unaffected by the large number of singletons found in small samples,
including the samples collected in the initial stages of research at a site before
an adequate sample size is accumulated. Figure 8a shows how E5 > 1.0 until
about sample 17 in the muti shops and about sample 22 in the street traders. An
explanation for this feature of the index is as follows: initially N2 > N1
(Figure 6) because ‘rare’ species and singletons are to begin with very abundant
in the ethnobotanical samples, thereafter declining in numbers as samples
accumulate and the more dominant species become evident in the sample. This
feature of the index is useful for determining the minimum viable sample size
required for assessing evenness. E5 for samples MS: White and MS: Survey
1992 is never less than 1.0 (Table 4), and therefore their evenness cannot be
compared with the other samples. The results demonstrate that evenness is
higher in the sample of Indian shop traders, followed by Black, All and Far-
aday Street, and therefore there is greater dominance in the sale of few species
within the street market (Table 4). This result is consistent with the observa-
tions described earlier, namely that a high sample diversity means that it is
more difficult to correctly predict a species chosen at random from a sample,
and therefore the dominance of species is lower and evenness higher.
The values of E2 and E3, as well as E4 and E5, are similar, and therefore
either may be used. However, Ludwig and Reynolds (1988) recommend the use
of E5 as a measure of evenness because it is the least ambiguous. The authors
also suggest calculating E1 because it is more widely used as a comparative
index (e.g., Begossi 1996). The performance of E1 as an index is shown in
Figure 8b for the sample of Faraday Street traders and shops (MS: All). The
sensitivity of E1 to the addition of ‘rare’ species (singletons) is evident in the
first part of the curve. When the Witwatersrand results are compared with six
ethnobotanical samples from South America (Begossi 1996), it is evident that
2991
Table 4. Species evenness indices calculated for six data sets sampled from Witwatersrand traders of traditional medicine.
Index/Measure 1992 Muti Shop
Survey (n=7)
N= 809
1994 Muti Shop Survey (N= 50) 2001 Street Trader
Survey (n= 100)
N= 2402
‘White-owned’
(n=2)
N= 193
‘Indian-owned’
(n= 20)
N= 2168
‘Black-owned’
(n= 28)
N= 1769
Total (‘All’)
shops (n= 50)
N= 4129
E1 (Shannon J0or Pielou’s J) (H0=Hmax) 0.970 0.988 0.945 0.938 0.922 0.910
E2 (N1/N0, or Buzas & Gibson E) 0.853 0.943 0.727 0.699 0.631 0.592
E3 (N11/N01) 0.853 0.942 0.726 0.698 0.629 0.590
E4 (N2/N1) 1.137 2.786 0.937 0.903 0.855 0.731
E5 (N21/N11) 1.138 2.799 0.937 0.903 0.855 0.730
McIntosh’s (E) 0.990 0.994 0.984 0.980 0.979 0.968
Brillouin (J) Can’t compute Can’t compute Can’t compute Can’t compute Can’t compute Can’t compute
2992
evenness is high overall for ethnobotanical samples (E1 > 0.90 on average),
with very little overall dominance of certain species for use/sale. Values for E1
in the South American study range between 0.78 and 0.97 (average 0.91)
compared to an average of 0.92 for the Witwatersrand traders.
Assessment of rare, intermediate and common ethnospecies
Indicator species are a useful adjunct to investigations of diversity (Magurran
1988). In ecology, they can provide an additional clue to how community
structure is changing (Magurran 1988). In ethnobotany, indicator species are
usually those in high demand by resource users and are at risk of over-
exploitation and population decline. A question that frequently arises is: how
does one objectively select criteria and categories for delimiting high risk
species from those that are lower risk? Cunningham (2001) describes some of
ST: Faraday
MS: All 1994
MS: Black 1994
MS: Indian1994
MS: 1992 Survey
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
4.5
5.0
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Evenness (E5)
(a)
ST: Faraday
MS: All 1994
0.900
0.910
0.920
0.930
0.940
0.950
0.960
0.970
0.980
0.990
0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100 105 110
Number of samples (n)
Evenness (E1)
(b)
Figure 8. The cumulative evenness curves for (a) E5 and (b) E1 for ethnospecies traded in the
Witwatersrand. (MS: Muti shops; ST: Street traders).
2993
the categories used for choosing priority species for monitoring as ‘‘filters’’
which help to sift out species that are likely to be more vulnerable to over-
harvesting. While complex and comprehensive models and methods exist for
‘filtering’ species (the authors are currently addressing this question in more
detail in forthcoming papers), a simple (albeit crude) method obtained from
Hill’s numbers can be used as a first step in the process of prioritising species
for monitoring.
Hill’s numbers N1, N2 and N1measure the number of abundant, very
abundant and most abundant species in a sample, respectively. In addition, N0
equals S– the total number of species. The number of species of rare, inter-
mediate and common abundances within a sample were defined in the following
way: common N1; intermediate N1–N1; and rare N0–N1. ‘Rare’
species, as previously mentioned, are those species with low incidences/abun-
36.5%
51.7%
11.8%
37.3%
34.9%
27.8%
37.4%
34.2%
28.4%
32.9%
29.1%
38.0%
0%
10%
20%
30%
40%
50%
60%
Rare Intermediate Common
Percentage of ethnospecies
Faraday All shops Black Indian
127
180
41
138 129
103
116 106
88
103 91
119
0
20
40
60
80
100
120
140
160
180
200
Rare intermediate Common
Number of ethnospecies (S)
Faraday All shops Black Indian
(a)
(b)
Figure 9. The percentage (a) and number (b) of species of rare, intermediate and common
abundances within the Faraday street trader sample and the muti shop survey of 1994. The cate-
gories are derived from Hill’s numbers N0, N1 and N1.
2994
dances in the sample (i.e., low values of n) and are not necessarily endangered.
Rare species are, therefore, the remaining species in a sample that are not
abundant (N1). Because N1 and N2 represent species present in abundance in a
sample (but not of ‘most’ abundance like N1), they were combined to produce
the category ‘intermediate’.
Figure 9 shows the number and percentage of ethnospecies categorised as
rare, intermediate and common in abundance from the Faraday street market
survey and the muti shop survey of 1994. The purpose of this paper, is not to
discuss what those species actually are, but to derive a method to more easily
delimit the species based on their abundance within a data matrix. The his-
tograms in Figure 9 corroborate the results of the evenness and diversity
measures, i.e., there is a higher evenness in the representation of species in the
muti shops. Additionally, there are fewer common or abundant species sold by
the street traders compared to the muti shops and, therefore greater dominance
of a few species in samples of the former. Further, street traders sell a smaller
range of species that are more prevalent in the market than other species. It is
known from research conducted in the market (Williams 2003) that many of
these species are currently threatened or have the potential to be so if current
harvesting and utilisation trends continue.
An example of intra-sample diversity: healers and non-healer traders
from the faraday market
The main use of diversity measures within this paper has been to compare
inter- and intra-sample diversity and sampling effort with respect to traders of
different ethnic groups selling plants within formal and informal markets. A
(295) Non-healers
(281) Healers
00 250 500 750 1000 1250 1500
50
100
150
200
250
300
350
Number of individuals (N)
Number of ethnospecies (S)
Figure 10. Species accumulation curves for ethnospecies sold by healer and non-healer traders in
the Faraday street market.
2995
further example of the application of diversity indices to intra-sample differ-
ences is the comparison of plants sold by traders that are traditional healers
and traders that are not traditional healers within the Faraday market. The
data matrix for Faraday was subdivided into ‘healers’ and ‘non-healers’ and
analysed accordingly. The results are shown in Table 5. The results make it
clear as to why it is necessary to describe a sample in terms of more than one
index of species richness, evenness and diversity as well as the necessity for a
species accumulation curve (Figure 10).
Numerical species richness of the plants sold by healers is lower than non-
healers (Table 5) – however, this is a function of the smaller subsample size. The
curve for healers lies above the curve for non-healers (Figure 10), and therefore
the subsample is comparatively more species rich. Additionally, Margalef’s
index for species richness indicates that species richness is almost the same for
both subsamples, with the fractionally higher value for non-healers being the
result of a larger subsample. Shannon’s index of diversity is exactly the same for
both subsamples, thereby underlining the importance of using additional
indices to discriminate between sample diversity where sample sizes differ.
Simpson’s index and Fisher’s ashow plant diversity sold by the healers to be
higher, and therefore there is lower dominance of species sold by the healers.
Additionally, the probability of encountering the same species amongst the non-
healer traders is higher – the low evenness values for the non-healer subsample
corroborate this evidence. The greater dominance of a few species in the non-
healer subsample can be accounted for in the following way: unlike healers,
non-healers do not supplement their trading incomes with paid consultations by
patients. With the high level of competition in the market, the non-healer
traders cannot afford to keep too many species that have intermediate demand
and commercial value (wholesale), and they therefore, tend to keep more of the
species known to be in demand by customers. The plant knowledge traditional
healers have, by contrast, allows more flexibility in the range of plants sold –
some of which are added to mixtures and preparations sold to patients.
Conclusions
The true value of diversity measures will be determined by whether or not they
are empirically useful (Magurran 1988). In non-ethnobotanical studies there
are two main areas in which diversity measures have potential application,
namely: conservation management, which concentrates almost exclusively on
measures of species richness (underpinned by the idea that species-rich com-
munities are better than species-poor ones); and, environmental monitoring,
which makes extensive use of diversity indices and species abundance distri-
butions (where, for example, the adverse affects of pollution will be reflected in
a reduction in diversity or a change in the shape of the species abundance
distribution) (Magurran 1988).
2996
According to Begossi (1996), diversity measures can be used to evaluate the
intensity of resources used by human populations, to allow comparisons
among different populations in different environments, and to allow evalua-
tions of sampling effort. Further, Begossi (1996) used diversity indices to help
answer the following questions: does the diversity of plant use in an area
represent the floristic abundance available?; are the same plants used by most
individuals?; and, are there differences in the diversity of plant uses per cate-
gory of user (e.g., gender/age)?
Diversity indices, in the broad sense of the term, therefore, have a useful role
to play in the quantitative analysis of ethnobotanical data. Which measures are
‘best’ cannot be decided without first knowing why diversity should be of
interest (Pielou 1975) or how it can help appraise the availability of plants used
and/or traded commercially. The primary goal of this paper was to evaluate the
performance of a large variety of indices in relation to samples of different sizes
and trader profiles, and to examine the kind of useful information they pro-
vided. The primary criterion used for recommending certain measures is the
value and economy that can be added to the description of plant availability/
use, and therefore the degree of insight that can be acquired into interpreting
relative abundances. Second, indices are recommended based on ease of cal-
culation and the extent of use by other researchers so that comparisons may be
made with other data sets similarly analysed (e.g., Begossi 1996; Hanazaki et
al. 2000). The relative merits and shortcomings of some diversity measures
have been previously described in Magurran (1988), and an awareness of their
limitations is necessary (Ludwig and Reynolds 1988). Ultimately, the choice of
index depends on the requirements of the researcher and the value that the
index adds to the quantitative description and understanding of the resources
under investigation.
A single index of diversity will most often not be sufficient to describe inter-
and intra-sample diversity (Hayek and Buzas 1997). Additionally, to describe a
sample only in terms of its diversity/heterogeneity index is to confound the two
factors of species richness and evenness (Pielou 1975). It is, therefore, judicious
to describe a sample in terms of richness, evenness and diversity. To this end,
we recommend the use of the following measures: (1) species richness (Sor
N0); (2) Margalef’s index; (3) Shannon index (H0); (4) Simpson’s index (both k
and ln k); (5) Hill’s diversity numbers N1, N2 and N1; (6) Fisher’s a; and (7)
evenness indices E1 and E5. It is additionally essential that these indices are
graphed as diversity accumulation curves so that the performance of an index
may be comprehensively evaluated, and the minimum viable sample size can
be determined. According to Gotelli and Colwell (2001), comparing richness
without reference to a taxon sampling curve is problematic and graphing the
results is necessary to detect differences in measured species richness (and
diversity) related to the relative abundance shown in the species accumulation
or species diversity curves. We recommend the use of EstimateS (Colwell 2001)
as a basis for calculating the accumulation curves and the input values nec-
essary for computing most of the other indices. Hill’s numbers have an
2997
additional beneficial use in the delimitation of species that are rare, interme-
diate and common in abundance within a sample, and this is a crucial first step
in prioritising species for monitoring and/or remedial conservation action.
The second objective of the paper was to assess whether the survey sites were
adequately sampled, and determine the minimum viable sample size on which a
diversity measure should be based for the type of data collected. Rarefaction is
a commonly used method for estimating species richness, and can be applied to
evaluating sampling effort (Magurran 1988; Begossi 1996; Williams et al. 2000;
Gotelli and Colwell 2001). However, the size of the sampling unit should be
chosen according to factors other than richness because relative abundance
affects the performance of the indices. When evaluating species diversity
measures (including richness and evenness) a sample size of at least 20 muti
shops and 35–40 street traders is necessary for the formal and informal sectors
in the Witwatersrand, respectively, (Table 6). The actual number of traders
surveyed depends on what aspect of diversity is being measured (Table 6). One
reason for the necessity to sample more street traders than muti shops is be-
cause of the lower mean number of species sold per street trader, therefore,
requiring additional sampling effort to increase the number of individuals
Table 5. Comparisons of selected measures of diversity between healer and non-healer traders in
the Faraday street market.
Index/Measure Healer traders (n= 39)
N= 1023
Non-healer traders
(n= 60) N= 1351
Species richness (S/N0) 281 295
Margalef 40.4 40.8
Shannon (H0) 5.25 5.25
Simpson (k) 0.0062 0.0067
Simpson (ln k) 5.08 5.00
Fisher’s a127.9 116.4
Hill’s N1 190.6 190.6
Hills N2 161.1 148.6
Hill’s N140.9 39.7
Evenness E1 0.931 0.923
Evenness E2 0.845 0.779
Table 6. The minimum viable sample size on which a species diversity measure should be based.
Index/Measure Muti shops Street traders
Species richness At least 20 At least 40
Diversity indices
Information theory (e.g., Shannon) 15–25 25–30
Dominance indices (e.g., Simpson) 20+ 15–20
Hill’s numbers 20+ 30+
Fisher’s alpha 30 35
Evenness 17 22
Summary: minimum viable sample size 20–30 35–40
2998
recorded for the sampling curves to reach a horizontal asymptote. Overall,
however, sampling effort was found to be more than adequate.
The third objective was to compare the species diversity of sites within the
formal and informal sector, and thereby appraise plant availability within the
ethnomedicinal trade in the region. As a result, inter- and intra-sample simi-
larities and differences in the sale of plants were identified. Numerical species
richness was found to be higher for muti shops than street traders, despite the
smaller sample size (n). This is related to the large number of individuals (N)
and ethnospecies sold per shop trader. Most of the diversity accumulation
curves for the indices showed a distinction in the availability and relative
abundance of plants sold by street traders compared to shop traders. Different
trading factors, therefore, operate within the trading sectors to determine the
plant diversity for sale. The graphs of the diversity curves are, therefore,
essential for interpreting the different mechanisms operating within the dif-
ferent markets. The average degree of uncertainty in predicting the identity of
species sold by the traders is higher in the formal sector, therefore, diversity is
higher and dominance of a few species is lower. The higher dominance of
certain plants sold by the street traders is confirmed by the lower evenness
values of the samples. Intra-sample differences in the muti shops showed Indian
traders to sell a larger diversity of plants compared to the Black traders, and
therefore dominance of plants in the latter was higher. In general, all the
indices gave different values but consistent rankings for the different sites.
The high diversity of plants sold within both the formal and informal sector
in the study area is likely to be related to a number of factors. Cities (like
Johannesburg) are more likely to have more culturally diverse populations,
drawn in from many rural communities (Cunningham 2001). Diversity of
species sold increases with increasing size of the marketing area, and therefore
more species are sold in regional markets (such as the Witwatersrand), fewer in
central markets and still fewer in minor or local markets (Cunningham 2001).
Begossi (1996) suggests that local resistance to Western medicine may result in
a greater demand for traditional medicines, thereby increasing the diversity of
plants used. At least 12–15 million people are estimated to consult traditional
healers in South Africa annually, and urbanisation has not precluded the use of
traditional medicine. In one ‘township’ southwest of Johannesburg (Soweto),
there were estimated to be at least 18,000 traditional healers. The Faraday
Street market functions primarily as a wholesale market to the traditional
healers in townships in the region (Williams 2003). The high diversity of tra-
ditional medicines sold in the region is, therefore, indicative of the high de-
mand and the acceptability of traditional healing practices – which to some
extent is related to the affordability of primary health care.
Ecologists have long known of species richness, diversity and evenness, and
it is only recently that these measures have been applied to the quantitative
analysis of ethnobotanical data (e.g., Begossi 1996; Williams et al. 2000). The
methods add greater depth to the exploration and understanding of mecha-
nisms and patterns operating within the field of indigenous plant use and trade.
2999
While a quantitative approach to analysing ethnobotanical data might not
always be possible, the approach is highly recommended.
References
Begossi A. 1996. Use of ecological methods in ethnobotany. Econ. Bot. 50: 280–289.
Botha J. 2001. Perceptions of Availability and Values of Medicinal Plants Traded on the Western
Boundary of the Kruger National Park, South Africa. MSc dissertation, University of the
Witwatersrand, Johannesburg.
Botha J., Witkowski E.T.F. and Shackleton C.M. 2001. An inventory of medicinal plants traded on
the western boundary of the Kruger National Park, South Africa. Koedoe 44(2): 7–46.
Brower J.E. and Zar J.H. 1977. Field and Laboratory Methods for General Ecology. Wm. C.
Brown Publishers, Dubuque, Iowa.
Colwell R.K. and Coddington J.A. 1995. Estimating terrestrial biodiversity through extrapolation.
In: Hawksworth D.L. (ed.), Biodiversity Measurement and Estimation. Chapman and Hall, pp.
101–118.
Colwell R.K. 2001. EstimateS: Statistical estimation of species richness and shared species from
samples. Version 6, User’s guide and application published at: http://viceroy.eeb.uconn.edu/
estimates.
Cunningham A.B. 2001. Applied ethnobotany: people, wild plant use and conservation. People and
Plants Conservation Manual, Earthscan.
Dauskardt R.P.A. 1990. The changing geography of traditional medicine: urban herbalism on the
Witwatersrand, Johannesburg. GeoJournal 22(3): 275–283.
Dauskardt R.P.A. 1991. ‘Urban herbalism’: the restructuring of informal survival in Johannesburg.
In: Preston-Whyte E. and Rogerson C. (eds), South Africa’s Informal Economy. Oxford Uni-
versity Press, Cape Town, pp. 87–100.
Gotelli N.J. and Colwell R.K. 2001. Quantifying biodiversity: procedures and pitfalls in the
measurement and comparison of species richness. Ecol. Lett. 4: 379–391.
Hanazaki N., Tamashiro J.Y., Leita
˜o-Filho H.F. and Begossi A. 2000. Diversity of plant use in two
Caic¸ ara communities from the Atlantic Forest coast, Brazil. Biodivers. Conserv. 9: 597–615.
Hayek L.C. and Buzas M.A. 1997. Surveying Natural Populations. Columbia University Press,
New York.
Ho
¨ft M., Barik S.K. and Lykke A.M. 1999. Quantitative ethnobotany: applications of multivariate
and statistical analysis in ethnobotany. People and Plants working paper 6. UNESCO, Paris.
Hill M.O. 1973. Diversity and evenness: a unifying notation and its consequences. Ecology 54(2):
427–432.
Johns T., Mhoro E.B., Sanaya P. and Kimanani E.K. 1994. Herbal remedies of the Batemi of
Ngorongoro District, Tanzania: a quantitative appraisal. Econ. Bot. 48: 90–95.
Krebs C.J. 1989. Ecological Methodology. Harper and Row Publications, New York.
Lowrey T.K. and Wright S., eds. 1987. The Flora of the Witwatersrand, Volume I, The Mono-
cotyledonae. Wits University Press, Johannesburg.
Ludwig J.A. and Reynolds J.F. 1988. Statistical Ecology – a Primer on Methods and Computing.
John Wiley and Sons, Toronto.
Luoga E.J., Witkowski E.T.F. and Balkwill K. 2000. Subsistence use of wood products and shifting
cultivation within miombo woodland of eastern Tanzania, with some notes on commercial uses.
S. Afr. J. Bot. 66(1): 72–85.
Magurran A. 1988. Ecological Diversity and its Measurement. Princeton University Press,
Princeton.
Murali K.S., Shankar U., Shaanker R.U., Ganeshaiah K.N. and Bawa K.S. 1996. Extraction of
non-timber forest products in the forests of Biligiri Rangan Hills, India. 2. Impact of NTFP
extraction on regeneration, population structure, and species composition. Econ. Bot. 50(3):
252–269.
3000
Phillips O. and Gentry A.H. 1993a. The useful plants of Tambopata, Peru: I. Statistical hypotheses
tests with a new quantitative technique. Econ. Bot. 47: 15–32.
Phillips O. and Gentry A.H. 1993b. The useful plants of Tambopata, Peru: II. Additional
hypothesis testing in quantitative ethnobotany. Econ. Bot. 47: 33–43.
Pielou E.C. 1975. Ecological Diversity. John Wiley and Sons, New York.
Prance G.T., Bale
´e W., Boom B.M. and Carneiro R.L. 1987. Quantitative ethnobotany and the
case for conservation in Amazonia. Conserv. Biol. 1: 296–310.
Williams V.L., Balkwill K. and Witkowski E.T.F. 1997. Muthi traders on the Witwatersrand,
South Africa – an urban mosaic. S. Afr. J. Bot. 63(6): 378–381.
Williams V.L., Balkwill K. and Witkowski E.T.F. 2000. Unravelling the commercial market for
medicinal plants and plant parts on the Witwatersrand, South Africa. Econ. Bot. 54(3): 310–327.
Williams V.L., Balkwill K. and Witkowski E.T.F. 2001. A lexicon of plants traded in the Wit-
watersrand umuthi shops, South Africa. Bothalia 31(1): 71–98.
Williams V.L. 1992. An investigation of the herbal medicine or ‘muti’ trade on Witwatersrand. BSc
(Hons) Dissertation, Department of Geography and Environmental Sciences, University of the
Witwatersrand, Johannesburg.
Williams V.L. 2003. Hawkers of health: an investigation of the Faraday Street traditional medicine
market in Johannesburg, Gauteng. Report to the Gauteng Directorate of Nature Conservation,
DACEL, Johannesburg.
Wong J., Thornber K. and Baker N. 2001. Resource assessment of non-wood forest products:
experience and biometric principles. NWFP Series 13, FAO, Rome.
Zar J.H. 1984. Biostatistical Analysis. Prentice-Hall, Inc., Englewood Cliffs.
3001
