RANDOM SAMPLING OVERESTIMATES SPECIES RICHNESS OF SHRUBLAND VEGETATION Lara Galea & Sandro Lanfranco Department of Biology, University of Malta, Msida MSD2080, Malta
[email protected]
INTRODUCTION
OBJECTIVE
• Measurement of alpha diversity of plants is sensitive to the field method used for
• To determine whether the alpha diversity of plants, measured at different spatial
assessment, with different methods potentially returning very different results.
scales, varies according to the sampling design used. Two quadrat sampling designs, one using ‘random’ quadrats and the other with ‘nested’ quadrats were compared.
• We propose to compare the values of plant alpha diversity returned by two
contrasting sampling designs: a multiscale ‘nested’ quadrat design, and a ‘random’ quadrat design. • The sampled quadrats in a nested design are not independent of each other as
the constraint of contiguity implies that neighbouring quadrats (with similar abiotic conditions and, presumably, species) will always be sampled. • Conversely, the same sampling effort carried out randomly will not be as
sensitive to the effects of these interactions and can therefore detect many more species, as different parts of the same habitat would be sampled.
MATERIALS & METHODS • The study was carried out using a 20cm x 20cm frame quadrat as the basic
sampling unit in six comparable shrubland sites in the Central Mediterranean islands of Malta and Gozo.
DATA COLLECTED • Binary presence/absence data for each species present was collected from each
quadrat. This was used to calculate the indices of diversity that were used to describe the characteristics of the sample. These were the Observed Species Richness, S(est), and the Chao-1 estimator of species richness.
MULTISCALE NESTED QUADRAT DESIGN • For each study site, the ‘basic unit quadrat’ (20cm x 20cm) was initially replicated
symmetrically five times in an area measuring 1m x 1m, with four unit quadrats 2 being placed at the vertices of the area, and one in the centre (Figure 1). This 1m block was, in turn, nested within an area measuring 5m x 5m and symmetrically replicated five times. This was repeated within a 25m x 25m area and then a 125m x 125m area, containing 625 basic unit quadrats.
Figure 1:
Showing the hierarchically nested sampling design at four spatial scales superimposed on one of the six sites of study. Each unit comprised five subunits of the next lower level with the ‘basic unit quadrat’ being 0.2m x 0.2m.
RANDOM QUADRAT DESIGN • The basic unit quadrat was randomly positioned 125 times within each site within
the boundaries of the 125m x 125m scale of the nested multiscale approach.
RESULTS
CONCLUSIONS
40
S(est) & Chao-1 from random quadrats design
Chao-1 from random quadrat design
S(est) from random quadrat design
GL-125 GL-125
30 IS-125 PB-25 RT-125
20 GL-25 IS-25 XH-25
10
XH-125
MF-125
RT-25
MF-25
0
PB-125 IS-125
40 GL-25 PB-25
30 RT-125 IS-25 RT-25
20
MF-125 XH-125
XH-25
MF-25
10
0 0
10
20
30
40
S(est) from nested quadrat design
Figure 2: Comparison of S(est) in ‘nested’ and ‘random’ designs at the 25quadrat and 125-quadrat scales. The straight line (y=x) represents the ‘equality’ trend across the two designs. Triangle symbols represent data from the 125-quadrat scale, and circles represent data from the 25-quadrat scale
• The random design would be suitable for
50
50 PB-125
0
10
20
30
40
40
30
20
10
0 5
10
15
20
25
30
35
S(est) & Chao-1 from nested quadrats design
Chao-1 from nested quadrat design
Figure 3: Comparison of Chao-1 in ‘nested’ and ‘random’ designs at the 25quadrat and 125-quadrat scales. The straight line (y=x) represents the ‘equality’ trend across the two designs. Triangle symbols represent data from the 125-quadrat scale, and circles represent data from the 25-quadrat scale
Figure 4: Comparison of the mean magnitude of residuals in ‘nested’ and ‘random’ designs at the 25-quadrat and 125-quadrat scales across all sites. The data points shaded in yellow represent the S(est) value and the points shaded in orange represent the Chao-1 value. The mean magnitude of residuals is indicated by the diameter of the data point
• The species diversity, both S(est) and the Chao-1 estimator, given by random quadrat sampling was generally higher
than that given by the nested quadrat design at comparable spatial scales (Figure 2, Figure 3). These differences in the species richness given by the two designs at the 25-quadrat and 125-quadrat spatial scales were statistically significant for S(est) (t= -3.931; P=0.002) and Chao-1 (t= -4.734; P< 0.001). • The difference in diversity between ‘nested’ and ‘random’ designs was relatively higher in species-rich sites and was
closer to ‘equality’ in sites with fewer species (Figure 4).
The research work disclosed in this publication is partially funded by the Endeavour Scholarship Scheme (Malta). Scholarships are part-financed by the European Union - European Social Fund (ESF) – Operational Programme II – Cohesion Policy 2014-2020 “Investing in human capital to create more opportunities and promote the well-being of society”
a species inventory, but may misrepresent local diversity as it would detect species populations that, despite sharing the same habitat, would be distant from one another and probably not interact significantly. As such, it may be less useful for understanding ecosystem processes, since most such processes work most intensely at a short range. • We argue that in cases where diversity is
linked with ecological function then using a nested design is preferable, as this would census the immediate environs of the individual plants, the parts of the habitat that are most relevant from the ecological point of view.