12 Jérôme Chave
at various spatial scales (Levin 1992, Huston
1999, Mouquet and Loreau 2003, Ricklefs 2004).
Historically, however, tropical plant ecology has
focused almost exclusively on the mechanisms of
local species coexistence, placin gless emphasis on
intermediate-scale patterns of diversity.
Our limited knowledge of the scaling proper-
ties of biodiversity is partly due to the scarcity
of datasets available to quantify these patterns.
This, of course, is a result of the difficulty
of gathering large-scale and consistent diver-
sity data in species-rich ecological communities
(Ashton 1964, Gentry 1982). A second cause
for the limited interest in documentin gpatterns
of spatial species turnover is the complexity of
statistical measurement procedures, and also the
lack of a consistent theoretical framework for
testin gecolo gical hypotheses. While phytosoci-
ologists predicted that dispersal, together with
biotic and abiotic factors, played a prominent
role in spatial plant turnover (Braun-Blanquet
1932), they lacked statistical approaches to test
appropriate biological hypotheses. A number of
studies have contributed to the recent revival of
interest in tree beta-diversity patterns (Tuomisto
and Ruokolainen 1994, Duivenvoorden 1995,
Tuomistoet al. 1995, Terborgh et al. 1996,
Hubbell 1997, Ruokolainenet al. 1997, Pitman
et al. 1999, Conditet al. 2002), and beyond
(McKnightet al. 2007, Woodcocket al. 2007).
By usin gnetworks of plots established in sev-
eral neotropical forests, these studies tested the
theoretical expectation that among-plot species
similarity might be predicted by abiotic environ-
ment or that it should decrease predictably with
between-site geographical distance. Their effort
benefited from improvements in methods to anal-
yse spatial turnover in diversity. For example,
rapid and normalized techniques for chemical soil
analysis are now available, with modern instru-
ments such as inductively coupled plasma-mass
spectrometers (Lucaset al. 1993, Clinebellet al.
1995) now bein groutinely used to measure the
concentration of macro- and micronutrients in
the soil. Major advances have also been made in
the long-term, cross-scale, prediction of climatic
variables (Newet al. 2002, Hijmanset al. 2005,
Flikkemaet al. 2006), in topography (worldwide
mappin gat a 90 m resolution from the Shuttle
Radar Topography Mission), and in mapping spa-
tial envelopes by remote sensin g(Tuomisto 1998,
Clarket al. 2004).
The conjunction of these conceptual advances,
technological progress, new empirical work, and
social demand make this field a very excitin gone,
and recent achievements are evidence for this
claim. My goal here is to convey the message that
even more remains ahead of us. I will review
available tools for measurin gspatial variation in
floristicdiversity,andstatisticalanddynamicmod-
els. I will then discuss evidence for and against the
role of environmental variation in predictin gbeta-
diversity. New methods are available to partition
beta-diversity into deterministic and stochastic
processes, and these approaches should be used
more consistently across a broad array of tropical
forest landscapes.DOCUMENTING PATTERNS OF
SPATIAL VARIATION IN SPECIES
DIVERSITY
There has been a tremendous wealth of research
on the statistical measurement of biodiversity.
General discussions of these techniques can be
found in Pielou (1975), Engen (1978), Gaston
(1994), Colwell and Coddington (1994), Krebs
(1999), and Magurran (2004). I restrict the
present section to a selective introduction of com-
mon measurements of diversity across scales in
the context of tropical tree communities.
Whittaker (1960, 1967, 1972) pioneered the
study of spatial diversity. He offered a conceptual
spatial diversity partitionin gscheme by distin-
guishing four scales at which diversity could
be measured: samples (point diversity), habitats
(alpha-diversity), landscapes (gamma-diversity),
and biogeographic provinces (epsilon-diversity).
Although intuitively appealing, this classifica-
tion has been interpreted differently amon gdif-
ferent authors. For instance, point diversity and
alpha-diversity are often confused, given the diffi-
culty of delimitin gobjectively habitats for plants.
Further, many measures of diversity depend on
samplin geffort (they are “biased”). This makes
it difficult to compare samplin gunits of unequal
size. To resolve this problem, one may choose to