Professional Documents
Culture Documents
Comparative Methods For Abundance
Comparative Methods For Abundance
Meester
Department of Biology, University of Natal, King George V Avenue, Durban, 4001 Republic of South Africa
Monthly population estimates generated by five standard methods (numbers caught; direct enumeration;
Bailey's Lincoln index; Hayne's removal method; Jolly-Seber stochastic model) were compared using vari-
ation coefficients and cluster analysis. Assessments from the first three methods were the most similar.
Hayne's method apparently underestimated numbers when sample sizes were low, and overestimated when
numbers caught on different trapping occasions were dissimilar. The Jolly-Seber model apparently over-
estimated population size consistently. The inaccuracies of the three statistical models resulted from inherent
statistical and biological pitfalls, which render them less heuristic than the simpler enumeration methods. The
direct enumeration method is the best available when trapping data indicate unequal catchability and high
population gains and losses.
Vyf standaardmetodes (aantal gevang; direkte telling; Bailey se Lincoln-indeks; Hayne se uitvangmetode;
Jolly-Seber se stochastiese model) is gebruik om maandelikse bevolkingskattings te verkry wat met behulp
van variasie-koeffisiente en groepanalise vergelyk is. Ramings gebaseer op die eerste drie metodes het die
meeste ooreenkomste getoon. Hayne se metode het getalle onderskat wanneer steekproewe klein was, en
dikwels oorskat wanneer die getalle wat met verskillende vangpogings versamel is, verskillend was. Die Jolly-
Seber metode het bevolkingsgrootte konsekwent oorskat. Die onakkuraatheid van die drie statistiese mode lie
spruit voort uit inherente statistiese en biologiese tekortkominge wat hulle minder heuristies as die meer
eenvoudige telmetodes maak. Die direkte telmetode is die beste metode beskikbaar wanneer vangdata dui op
ongelyke vangbaarheid en hoe vlakke van bevolkingsaanwins en verlies.
'To whom correspondence should be addressed at: Department of Mammals, Transvaal Museum, P.O. Box 413, Pretoria,
0001
njZjRj ZR· mj
Nj = nj+ _ OJ = mj + __ 1 _1_ aj = __
12
rj nj
10
OJ-mj + Rj 1 1 1-aj
SE j = V N;(N j - nJ----{---}+--
1Ii 8
OJ rj Rj mj
I
6
where Nj, nj and mj are as above
§ OJ = number of animals in the population imme-
4
'0 diately preceding the i th occasion
.;
z aj = proportion of marked animals in the i th sam-
ple
0
Rj = number of individuals released at time i
0
Zj = number of individuals marked before the i th
occasion which were not in the i th sample,
Figure 1 Regression line calculated by Hayne's method for but which are subsequently recaptured.
data on trapping of M. natalensis. The intersection of this line rj = number of animals released on the i th occa-
with the x-axis (18,5) represents total population size. sion that are subsequently recaptured.
construct phenograms (Figure 2) showing the degree of
Jolly-Seber estimates are usually accurate when 9% or
more of a population is sampled, and if the survival rate overall similarity I dissimilarity between population
is not less than 0,5 between trapping occasions (Bishop assessments yielded by the various estimation methods.
& Sheppard 1973). However, its reliability strictly de-
pends on the condition that survival rates are not age re- Results
lated (A 1;2;3;4b;5). Evidence used to test which of the implicit assumptions
More information, including formulae for the calcu- of the different methods were satisfied is shown in Table
lation of recruitment rates, is given by Caughley (1977), 3. Only assumptions 1 and 2 were not violated during the
Seber (1973) and Southwood (1978). study.
With methods (i), (iii) and (iv), data for days one to The overall trapability rate, expressed as the number
three of each trapping session were combined to es- of captures / number of individuals, was high for both
timate numbers on day four. Owing to low daily sample the grids (3,42) and the traplines (3,99). The overall re-
sizes, Jolly-Seber estimates were calculated by com- capture rates were 70,7% and 75% for the grids and lines
bining data for each four-day session, and treating these respectively. Recaptures outnumbered intial captures by
data as if they applied to a single trapping occasion, so a factor of three, indicating a high degree of trap-
that the effective sampling period was 12 months (d. proneness.
four days per month). Grid population estimates for January-March 1986
The mean absolute differences (ie. 'Manhattan dis- (Table 2) were not significantly different (X2 =
tances') between monthly estimates provided by each 9,18 x 10-2; p = 0,99). Numbers caught and MNA were
particular combination of techniques (Table 2) were identical in each of the three months, and agreed most
used with Sneath & Sokal's (1973) unweighted closely with Hayne's removal estimates (Figure 2). How-
pair-group arithmetic averaging clustering (UPGMA) to ever, the latter method underestimated population size
in January 1986. Bailey's modified Lincoln indices were,
on average, 14% higher than estimates given by the first
Table 2 Grid population estimates yielded by the var-
ious methods used to assess numbers of M. nata- three methods. The Jolly-Seber estimate for February
lensis. Estimates are as follows: NC - numbers
1986 was markedly higher (23%) than MNA and num-
caught; MNA - minimum number alive; BLI -
bers caught.
Bailey's Lincoln index; HTO - Hayne's removal (trap- Although monthly line estimates provided by the five
out) estimator; J-S - Jolly-Seber estimate methods for March 1985-March 1986 (Table 4) varied
concordantly, differences between them were statistic-
ally significant (X2 = 231,74; P < 0,001). In general,
greater disparity between monthly estimates occurred
between April and December 1985 when population
gains and losses were high. Numbers caught, MNA and
January 1986 19 19 21,O±3,6 17,8* / Bailey's Lincoln indices were comparatively the most
February 16 16 18,6±2,7 16,4 19,8±1,2 similar (Figure 2), and did not differ significantly
March 18 18 20,6±3,4 18,5 / (X2 = 8,18; p = 0,997). However, the last method
yielded considerably higher (over)estimates in May and
*Based on a statistically non-significant linear regression. December 1985.
/ Estimation of population size not possible.