Tropical Forest Community Ecology

16 Jérôme Chave

rely heavily on complex statistical concepts and it
is very easy to get lost along the way. Here I try to
avoid technicalities, while pointin gthe interested
reader towards the relevant literature.
The most intuitive approach in perform-
in gspecies association analyses is where the
abundance or occurrence of one species is cor-
related with environmental descriptors, indepen-
dently of all other species in the community. This
approach has been used largely for temperate
plants. Because tropical tree species are usually
rare and infrequent, however, analyses for tropical
tree assemblages often lack statistical power and
can be applied only to abundant species (Newbery
and Proctor 1984, Baillieet al. 1987, Swaine
1996, Pitmanet al. 1999, Svennin g1999, Clark
et al. 1999a, Webb and Peart 2000, Pykeet al.
2001, Phillipset al. 2003, Svenninget al. 2004).

This omission for rare species, which compose

the majority of species in most tropical forests,

may bias species association analyses because

rare species may behave very differently from

abundant ones (Conditet al. 2000).

A measure of environmental dissimilarity

may be defined by the absolute value of the dif-

ference in the environmental variable between

samplin gunits (e. g., rainfall, nutrient concen-

tration). Before applyin gmany of these statisti-

cal methods, it is important to make sure that

environmental dissimilarity data are normally

distributed, so a non-linear transform may be

advisable (Phillipset al. 2003, Tuomistoet al.

2003a, cf. Figure 2.2). The environmental vari-

able may also be qualitative: soil types may be

classified as fertile or infertile, dry or wet (Swaine

1996), well drained or poorly drained, sandy or

12345

Log(AI)

Range AI: 5–765 Range Ca: 3–11 069 Range Cu: 0.1–28

0

67 02468

Log(Ca)

0

10 − 3 − 112

Log(Cu)

0

34

12345

Log(Fe)

Range Fe: 3–759 Range Mg: 2–1620 Range K: 8–1284

0

0.4

67 0 2 4 6

Log(Mg)

0

867

Log(K)

0

2345 8

−20 2

Log(P)

Range P: 0.1–146

Range N: 196–15 090

Range Na: 1–356 Range Si: 3–478

0

0.4

4

59876

Log(N)

0

0.4

10

023145

Log(Na)

0

667

Log(Si)

0

0.4

12345

Figure 2.2 Histograms of the log-transformed soil chemical concentrations for 69 plots reported in
Gentry (1988) (see Clinebellet al. 1995). All variables are measured in ppm (or m gk g−^1 ).