C H A PTER 6

Distance Measures

Background • O ne can calculate distances am ong either the

T he first step o f m ost m ultivariate analyses is to
calculate a m atrix o f distances or sim ilarities am ong a
set o f item s in a m ultidim ensional space. T his is
analogous to constructing the triangular "m ileage
chart" provided w ith m any road maps. But in our case,
we need to build a m atrix o f distances in hyperspace,
rath er th an the tw o-dim ensional m ap space. Fortu­
nately, it is ju st as easy to calculate distances in a
m u ltidim ensional space as it is in a tw o-dim ensional
space.
T h is first step is extrem ely im portant. If
inform ation is ignored in th is step, then it cannot be
expressed in the results. Likewise, if noise or outliers
are exaggerated by the distance m easure, then these
unw anted features o f our data w ill have undue
influence on the results, perhaps obscuring m eaningful
patterns.
Distance concepts
D istance m easures are flexible:
• Resem blance can be m easured either as a
distance (dissim ilarity) or a sim ilarity.
• M ost distance m easures can readily be con­
verted into sim ilarities and vice-versa.
rows o f your data m atrix or the colum ns of
your data m atrix. W ith com m unity data this
m eans you can calculate distances am ong your
sam ple units (SUs) in species space or am ong
your species in sam ple space.
Figure 6.1 show s two species as points in sample
space, corresponding to the tiny data set below (Table
6.1). We can also represent sam ple units as points in
species space, as on the right side o f Figure 6.1. using
the sam e data set.
T here are m any distance m easures. A selection of
the m ost com m only used and m ost effective m easures
are described below. It is im portant to know the
dom ain of acceptable data values for each distance
m easure (Table 6.2). M any distance m easures are not
com patible w ith negative num bers. O ther distance
m easures assum e th at the data are proportions ranging
betw een zero and one, inclusive
Table 6.1. Exam ple data set A bundance o f two
species in tw o sam ple units.
Species
Sam ple unit 1 2

• All o f the distance m easures described below A 1 4

can be applied to either binary (presence- B 5 2
absence) or quantitative data.

5 j S am ple sp a ce S p ecies sp a ce
ш q
♦SUA
3 3
<L> Sp 2 SUB
ex 2 tU
CL
C /3
C /3

- I ------------------ 1------------------1------------------ L_

» 1 2 3 4 5 1 2 3
Sample U nit A S p ecies 1

F igure 6.1. G raphical representation o f the data set in T able 6.1. T h e left-hand graph
show s species as points in sam ple space. T he rieh t-h an d nranh shnwc eamnip unite ·>-
 Chapter 6

Table 6.2 Reasonable and acceptable dom ains o f input data. л\ and ranges o f distance m easures, d - fix).

D om ain
Name (synonym s) of X Range o f d = f i x ) C om m ents
Sorensen x >0 0 <d < 1 p roportion coefficient in city-
(Brav & Curtis; ( o r ( ) < x < 100%) block space, sem im etric
Czekanovvski)
Relative Sorensen x >0 0 <d< 1 proportion coefficient in city -
(Kulczyński; Q uantitative (or 0 < x < 100%) block space; sam e as Sorensen
Symmetric) but data points relativized by
sam ple unit totals; sem im etric
Jaccard x>0 ()< £ /< 1 proportion coefficient in city-
(orO < d < 100%) block space; m etric

Euclidean (Pythagorean) all non-negative m etric

Relative Euclidean all E uclidean distance between
0 < d < 4 2 for quarter
(Chord distance, points on unit hy persphere:
hypersphere; 0 < d < 2
standardized E uclidean) m etric
for full hvpersphere
C orrelation distance all 0 <d < I converted from correlation to
distance; proportional to arc
distance betw een points on unit
hypersphere; cosine of angle
from centroid to points; m etric
Chi-square x>0 d>0 E uclidean but doubly w eighted
by variable an d sam ple unit
totals; m etric
Squared Euclidean all d> 0 m etric
M ahalanobis all d> 0 distance between groups
weighted by w ithin-group
dispersion: m etric

D istance m easures can be categorized as metric,
scm im etric. or nonm etric A m e tric distance m easure
must satisfy the follow ing rules:
1 The m inim um value is zero w hen two item s are
identical.
2 W hen two item s differ, the distance is positive
(negative distances are not allowed).
3 Symmetry: the distance from objects A to object
B is the sam e as the distance from B to A.
4 T riangle inequality axiom: W ith three objects.
the distance between two o f these objects
cannot be larger than the sum of the two other
distances.
Kulczyński distances Seim m ctrics are extremely use­
ful in com m unity ecology but obey a non-Euclidean
geometry N o n m etrics violate one or m ore o f the other
rules and are seldom used in ecology
Distance measures
T h e equations use the follow ing conventions: Our
data m atrix A has q rows, w hich are sam ple units and
p colum ns, w hich are species. E ach elem ent of the
m atrix, a, ,, is the abundance of species j in sam ple unit
i. Most of the following distance m easures can also be
used on binary data (1 o r 0 for presence or absence).
In each o f the follow ing equations, we are calculating
the distance between sam ple units / and h.
 D istance Λle asu res

Euclidean distance

oo EUCLIDEAN
E E ) ,,и 'У ' ( α , , ~ ай,]} Щ DISTANCE
D
ω
Cu
oo
T his form ula is sim ply the Pythagorean theorem
applied to p dim ensions rather th an the usual two
dim ensions (Fig. 6.2).
S P E C IE S 1

City-block distance (= M a n h attan distance) 00

ω
0 f CITY-BLOCK
ω [ d is t a n c e
CBo ъ a¡ , " Ok,, α,
oo
7=1

ln city-block space you can only move along one

dim ension o f the space at a tim e (Fig. 6.2). By S P E C IE S 1
analogy, in a city o f rectangular blocks, you caimot cut
diagonally through a block, but m ust w alk along either
of the two dim ensions of th e block. In the m athem a­
tical space, size o f the blocks does not affect distances cos α

in the space Note also that m any equal-length paths 00

exist betw een two points in city-block space.

E uclidean distance and city-block distance are
c e n t r o id S P E C IE S 1
special cases (different values of k) o f the M inkow ski
metric. In two dim ensions:

Figure 6.2. G eom etric representations o f basic dis­

Distance \[xk + y k
tance m easures betw een two sam ple units (A and B) in
species space In the upper two graphs the axes meet
where x and v are distances in each o f two dim ensions.
at the origin; in the lowest graph, at the centroid
G eneralizing this to p dim ensions, and using the form
of the equation for ED:
cos(18()“). Tw o sam ple units lying on the sam e radius
from the centroid have r = 1 = cos(0°). If two sam ple
units form a rig h t angle from the centroid then r = 0 =
Distance,h = at] - a h j t
cos(90°).
The correlation coefficient can be rescaled to a
Note that k = 1 gives city-block distance, k = 2 gives distance m easure o f range 0-1 by
Euclidean distance. As k increases, increasing
em phasis is given to large differences in individual t d ista n c e ( Î ~ ~

dim ensions

Correlation Proportion coefficients

The correlation coefficient (r) is cosine a (third
panel in Fig. 6.2) w here the origin o f the coordinate
system is the m ean species com position o f a sam ple
tuut in species space (the "centroid"; see Fig. 6.2). If
the data have not been transform ed and the origin is at
(0,0). then this is a noncentered correlation coefficient
For exam ple, if two sam ple units lie at 180“ from
each other relative to the centroid, then r - -1 =
Proportion coefficients are city-block distance
m easures expressed as proportions o f the m axim um
distance possible. T he Sorensen. Jaccard. and QSK
coefficients described below are all proportion coef­
ficients. O ne can represent proportion coefficients as
the overlap betw een the area under curves. T his is
easiest to visualize w ith two curves o f species abun­
dance on an environm ental g rad ien t (Fig. 6 3). If A is
the area under one curve, B is the area under the other.
 ( 'hapt ur 6

and n' is the overlap (intersection) of the two areas,

(hen the Sorensen coefficient is 2и>/(. I · lí).
Jaccard coefficient is w i l l - И-w).
T he
Σ tin - Cthj

IX,
W ritten in set notation: Σ». +Σ a.
2( A Г) B)
Sorensen similarity A nother way o f w riting this, w here MÍN is the
( A ' u B) - (A rsB) sm aller o f two values is:
АглВ
Jaccard similarity
A sj B Clh,)

Proportion coefficients as distance m easures are D,h

foreign to classical statistics, w hich are based on
squared Euclidean distances. Ecologists latched onto
+Σ
proportion coefficients for their sim ple, intuitive appeal
despite their falling outside of m ainstream statistics.
Nevertheless, Roberts (1986) showed how proportion O ne can convert th is dissim ilarity' (or any o f the
coefficients can be derived from the m athem atics o f follow ing proportion coefficients) to a percentage dissi­
fuzzy sets an increasingly important branch of m ilarity ( r ø ) :
m athem atics For exam ple, when applied to q u an tita­
tive data. Sorensen sim ilarity is the intersection
between two fuzzy sets. Sorensen distance ■=- BC, PD,i 100 D„
Sorensen sim ilarity (also know n as "BC" for
Bray-Curtis coefficient) is thus shared abundance
Jaccard dissimilarity is the proportion of the
combined abundance that is not shared, or u (A ■ B w)
(Jaccard 1901):

- Σ a ¡j “ ci

JD ,

E n v iro n m e n ta l G ra d ie n t
Σ ач + Σ ^ ”^ Σ ^ο ~ Qhj

Figure 6.3. O verlap between two species abundances

along an environm ental gradient. T he abundance
shared between species A and B is shown by w. Q uantitative sym m etric dissim ilarity (also
know n as the K ulczyński or Q SK coefficient: see
Faith et al. 1987):
divided by total abundance. It is frequently know n as
"2u/(A + B )" for short. T his logical-seem ing m easure
was also proposed by Czekanow ski (1913).
Originii Ily used for binary (0/1) data, it w orks
£ MIN(¿/y, a hj ) X M1N(û„ , üHj)
equally well for quantitative data It is often called
the "Bray-Curtis coefficient" (Faith et al. 1987) Q S K ih= 1- v
w hen applied to quantitative data, as in Bray and
Σ a '¡ Σ α >ν
Curtis (1957). R ew riting the equation as a
dissimilarity (or distance) m easure, dissim ilarity
A lthough Faith et a f (1987) stated that this measure
between item s / and h is.
has a "built-in stan d ard izatio n " it is not a standardized
city-block distance in the sam e way that relative
Euclidean distance is a standardized m easure of
Euclidean distance. In contrast with the "relative
Sorensen distance" (below), the Q SK coefficient gives
 D istance M easures

different results w ith raw data versus data standardized

by SU totals (Ch. 9). A fter such relativization. how­
ever. Q SK gives the sam e results as Sorensen, relative
Sorensen, and city-block distance
R elative S ø ren sen (also know n as relativized
M anhattan coefficient in Faith et al. 1987) is m athem a­ RED ih
tically equivalent to the Bray-C urtis coefficient on data
relativized by SU total. T his distance m easure builds
VV;
Σ
in a standardization by sam ple unit totals, each sam ple
unit contributing equally to the distance m easure.
U sing this relativization shifts the em phasis o f the
analysis to proportions o f species in a sam ple unit,
1.0
rather than absolute abundances.

cn
\ A rc d istan ce
/ y w
u RED \
Ш (ch o rd )
es­
co
X j a k
D„ Σ M I N
J- 1 P

Σ«. =1 7 1.0
Vj-i J S P E C IE S 1
Figure 6.4. Relative E uclidean distance is the
chord distance betw een two points on the surface
An alternate version, using an absolute value o f a unit hypersphere
instead of the MIN function, is also m athem atically
equivalent to Bray-C urtis coefficient on data relativized
RED builds in a standardization It puts differ­
bv SU total:
ently scaled variables on the sam e footing, elim inating
any signal other than relative abundance. Note that the
correlation coefficient also accom plishes this standard­
a ization, but arccos(r) gives the arc distance on the
IX, qu arter hypersphere. not the chord distance Also,

Σ α >< Σa. w ith the correlation coefficient the surface is a full

hypersphere. not a quarter hypersphere. and the center
o f the hypersphere is the centroid of the cloud of
points, not the origin

Note that w ith standardization or relativization of

the data before application o f the distance m easure,
many o f the city-block distance m easures become
m athem atically equivalent to each other. After relati-
vization by sam ple unit totals. PD (Bray-Curtis) = CB
= QSK = Relative Sorensen
Relative Euclidean distance (RED)
RED is a variant o f ED that elim inates differences
m total abundance (actually totals o f squared abun­
dances) am ong sam ple units. RED ranges from 0 to
the square root of 2 w hen all abundance values are
nonnegative (i.e., a quarter hypersphere). Visualize
the SUs as being placed on the surface of a quarter
hypersphere w ith a radius o f one (Fig. 6.4). RED is
the chord distance between two points on this surface
Chi-square distance
T he chi-square distance m easure is used in corre­
spondence analysis and related ordination techniques
(Chardy et al 1976. M inchin 1987a). A lthough Faith
et al. (1987) found that this distance m easure
perform ed poorly, it is im portant to be aw are o f it.
since it is the im plicit distance m etric in some o f the
m ore popular ordination techniques (detrended
correspondence analysis and canonical correspondence
analysis)
Let
p
ah, = total for sam ple unit h (i.e.. ' У \ а . )
 ( ’hapler 6

a,. = tolal for sam ple unit / (i.e.. f ' tf„ ) C ase
/I o o·
o oo A

a . , = total for species j (i.e.. ^ a ¡} )

th en the chi-square distance (C hardv et al. 1976) is

C ase
B
Cl hj Clij
x:
a h- ci,~ G ro u p ƒ G ro u p h

Note (hat this distance m easure is sim ilar to
E uclidean distance, but it is w eighted by the inverse o f
the species totals. If the data are prerelativized by
sam ple unit totals (i.e.. b„ a,., ), then the equation
sim plifies to:
X' Σ (fibL-AÌ
The num erator is the squared difference in relative
abundance It is expressed as a proportion o f the
species total (the denom inator) and sum m ed over all
species.
M inchin (1987a) offered the following critique o f
this distance measure:
The appropriateness o f C hi-squared distance as a
m easure o f com positional dissim ilarity in ecology
may be questioned (F a ith et al 1987). T he m ea­
sure accords high w eight to species w hose total
abundance in the data is low. [Conversely, it de-
em phasizes abundant species ] It thus tends to
exaggerate the d istin c tiv e n ess o f sam ples contain­
ing several rare species. U nlike the B ray-C urtis
coefficient and re la te d m easures. C hi-squared
distance does not reach a constant, m axim al value
for sam ple pairs w ith no species in com m on, but
fluctuates according to variations in the rep resen ­
tation o f species w ith high or low total abundances.
These properties o f C hi-squared distance may
account for som e o f the d istortions observed in
DCA ordinations.
M ahalanobis distance (Ό2)
M ahalanobis distance O f is used as a distance
measure between two groups (/ and h). It is com m only
used in discrim inant analysts and in testing for
outliers. If a„ is the m ean for the ith variable in group
J. and vt'y is an elem ent from the inverse of the pooled
vvithin-groups covariance m atrix. representing
' >n ihtpc / and /'. then
Figure 6.5. Illustration o f the influence o f vvithin-
group variance on M ahalanobis distance
Dfh (w - g ) ΣΣ
r i J- 1
W 'J ' K * Cl.h ) ( « л - a jh )
w here n is the num ber o f sam ple units, g is the num ber
o f groups, an d i * j. Note that differences are w eighted
m ore heavily by w„ w hen v ariables / and j are uncorre­
lated Thus. M ahalanobis distance corrects for the
correlation structure o f the original variables (the
dim ensions o f the space). T h e built-in standardization
m eans that it is independent o f the m easurem ent units
o f the original variables.
In w hich case in Figure 6.5 are groups ƒ and h
more distant? Because the M ahalanobis distance
inversely w eights the distance between centroids by the
variance, the two groups are m ore distant in Case B.
even though the centroids are equidistant in the two
cases.
Note the conceptual sim ilarity to an T -ratio of
between- to w itlnn-group variance. Indeed, the M aha­
lanobis distance can be used to calculate an F-test for
m ultivariate differences between groups. Sim ilarly, it
can be used to test for outliers by calculating the
distance between each point and the cloud of rem ain­
ing points.
Performance of distance
measures
Loss o f sensitivity with heterogeneity
Perform ance o f distance m easures can be
evaluated by com paring the relationship between
environm ental distance (distance along an environ­
m ental gradient, such as elevation) vs. sociological
distance (the difference in com m unities as reflected by
the distance in species space) T his method oi
 D istance ΛJe asures
evaluating distance m easures w as used by Beals
(1984). Faith et al. (1987), De ath (1999a). an d Boyce
and Ellison (2001). 11' species respond noiselessly to
environm ental gradients and the environm ental
gradients are know n, then we seek a perfect linear
relationship betw een distances in species space and
distances in environm ental space. Any departure from
that relationship represents a partial failure o f our
distance measures.
Two exam ples help clarify' the variability in the
relationship betw een distance in species space and
environm ental space. T hese exam ples are based on
synthetic data sets w ith a know n underlying structure
and noiseless responses o f species to two environm en­
tal gradients.
The first exam ple is an "easy" data set. consisting
o f 25 sam ple units and 16 species. It is easy because
the beta diversity is fairly low (average Sorensen
distance am ong SUs = 0.59: 1.3 h alf changes), the
sam ple unit totals are fairly even (coefficient of
variation (CV) o f SU totals = 17%). and the species are
all sim ilarly abundant (CV o f species totals = 37%).
Despite this being an easy data set. all o f the
distance m easures show a curvilinear relationship with
environm ental distance (Fig. 6.6). Specifically, we see
the loss in sensitivity o f our distance m easures at large
environm ental distances. T he problem is least
apparent in the Sorensen, chi-square, and Jaccard
distances. T he problem is w orst with the correlation
distance, w here the curve not only flattens at high
environm ental distances, but starts to decline at the
highest distances. T he drop in the curve for the corre­
lation coefficient (actually (1 - r)l2 w hich converts r
into a distance rather than a sim ilarity m easure) is due
to interpreting shared zeros (0.0) as positive associ­
ation.
T he second exam ple is a m ore difficult data set.
consisting o f 100 sam ple units and 25 species. The
difficulty has nothing to do w ith the size o f the data
set Rather its beta diversity is higher (average Soren­
sen distance am ong SUs = 0.79: 2.3 h alf changes), the
sam ple unit totals vary m ore widely than in the
previous exam ple (CV o f SU totals = 40% ). and the
species vary realistically in abundance (CV of species
totals = 183°»).
A gain, all o f the distance m easures lose sensitiv ity
with increasing environm ental distance (Fig. 6,7),
T his loss is greatest for distance based on the co rrela­
tion coefficient. E uclidean distance not only loses
sensitivity at high distances, but introduces consid­
erable error, even at m oderate distances. N ote also that
E uclidean distance shows no fixed upper bound for
sam ple units that have nothing in com m on
Sorensen distance loses sensitiv ity ov er a distance
about h alf the length o f the env ironm ental gradients.
T he flat top on the Sorensen scatterplot results because
it has a fixed m axim um for SUs having no species in
com m on. M any ecologists consider this a desirable,
intuitive property for species data. C hi-square distance
perform s reasonably well at sm all environm ental
distances but m isinterprets many distant SUs as being
close in species space.
T ransform ation of the data to binary (presence-
absence) in both exam ples results in a more linear
relationship w ith most distance measures. T his was
show n for E uclidean distance and Sorensen distance
w ith real data from an elevation gradient (Beals 1984).
G iven the apparently poor perform ance o f all of
the distance m easures, it is rem arkable that m u ltiv ari­
ate analysis is able to extract clear, sensible patterns
(Fig. 6 8) We are rescued by the redundancy m the
data — all ordination and classification techniques
benefit front this redundancy in the data W here two
species fail to be inform ative of a difference, another
two species are inform ative. Some ordination tech­
niques. nonm etric m ultidim ensional scaling (NM S) in
particular, are able to linearize the relationship
between distance in species space an d distance in a
reduced ordination space. NM S has an advantage over
other ordination techniques: it is based on ranked
distances, w hich im proves its ability to extract
inform ation from the nonlinear relationships illustrated
in the two exam ples.
 C hapter 6

« 0 0
u
o 20
c 0.8 8 a
ce o
0 .6 «0
9
0
S 10 0.4

o 0.2
сл
0
2 4 6 2 4

E nvironm ental D istance E nvironm ental D istance

0.08 0.9
a
3 0.06
co
D 0.05
υ.07 0.8
0.7
06
V ? "
05
<υ
u 0.04
03 j 0.4
¡3 0.03 аз
cr ТЗ 0.3
.2 0 02 fc 0.2
O
O 0.01 CJ 0.1
0
2 4 6 2 4 6

E nvironm ental D istance E nvironm ental D istance

700 1 o o
8 *8
a 600 o
d) o 0.8
Г2 500 o 0
73 0
Z3 400 0.6
Q O
ш 0
-а 300 T3
w
P 03 0.4
o
g 200
cr 0.2
C/3 100
0
2 4 6 2 4 6

E nvironm ental D istance E nvironm ental D istance

Figure 6 6. R elationship between distance in species space for an "easy” data set, using various distance
m easures and environm ental distance. T he graphs above are based on a synthetic data set w ith noiseless
species responses to two know n underlying environm ental gradients T he gradients w ere sam pled w ith a
5 x 5 grid T his is an "easy” data set because the average distance is reasonably sm all (Sorensen distance
= 0.59: 1.3 h a lf changes), all species are sim ilar in abundance (CV o f species totals = 37% ), and sam ple
units have sim ilar totals (CV o f SU totals = 17%).
 D istance Λteasures

0 1 2 3 4 5 6 7 8 9 10 11 12 13 0 1 2 3 4 5 6 7 X 9 10 1 1 1 2 13
Environm ental D istance Environm ental D istance

& a 0.6

o 0. 4

ω 0.3 &

0 1 2 3 4 5 6 7 8 9 10 1 1 1 2 13 0 1 2 3 4 5 6 7 X 9 10 11 12 i:
Environm ental D istance Environm ental Distance

Figure 6.7. R elationship between distance in species space for a "m ore difficult” data set, using various
distance m easures, and environm ental distance. T he graphs above are based on a synthetic data set with
noiseless species responses to two know n underlying environm ental gradients. T he gradients were
sam pled w ith a 10 x 10 grid. T his is a "m ore difficult ' data set because the average distance is rather
large (Sorensen distance = 0.79: 2.3 h a lf changes), species vary in abundance (C V of species totals =
183%). and sam ple units have moderately variable totals (CV o f SU totals = 40% ).
 ------- V 'J J

C hapter 6

O 1 2 3 4 5 6 7 8 9 10 11 12 13

Environmental Distance

Figure 6 8. D istance in a 2-D nonm etric m ultidim ensional scaling

ordination (NM S) in relation to environm ental distances, using the sam e
data set as in Figure 6.7. Note how the ordination overcam e the lim ita­
tion o f th e Sorensen coefficient at expressing large distances.

for selecting one distance m easure over another The

Availability
T he Sorensen (B ray-C urtis) index has repeatedly
been shown to be one o f the m ost effective m easures o f
sam ple or species sim ilarity, yet it is not w idely avail­
able in general statistical software.
Compatibility
T he prim ary disadvantage o f city-block distance
m easures is that they are not com patible w ith many
standard m ultivariate analyses (e.g.. discrim inant
analysis, canonical correlation, and canonical corre­
spondence analysis). T his becom es less an d less
im portant, however, as non-E uclidean alternatives
becom e increasingly available. C ertain m ethods of
classification and ordination are effective w ith
ecological data (e.g.. B ray-Curtis ordination an d non-
m etnc m ultidim ensional scaling), in part because they
are am enable to distance m easures that perform well
w ith ecological data.
Theoretical basis
Beyond the choice of a proportion coefficient or
not (Box 6.1) and the choice o f a relativized distance
m easure or not, there is little basis in ecological theory
rationale for our choice is prim arily em pirical: we
should select m easures that have show n superior
perform ance, based on the other criteria listed here
O ne im portant theoretical difference betw een E uclid­
ean an d city-block distance is, however, apparent.
Long E uclidean distances in species space are
m easured through an uninhabitable portion o f species
space — in other w ords the straig h t-lin e segm ents tend
to pass through areas o f im possibly species-rich and
overly full com m unities. In contrast, city-block
distances are m easured alo n g the edges o f species
space — exactly w here sam ple units lie in the dust
bunny distribution.
Intuitive criteria
Does E uclidean or city-block distance better match
our intuition on how com m unity distances should be
m easured? C onsider the follow ing exam ples.
In city-block space, the im portance o f a gradient is
proportional to the num ber o f species responding to it.
For exam ple, assum e that 20 species respond only to
gradient X an d 4 species only to gradient Y. In city-
block space, g radient X is 5 tim es as im portant as
gradient Y. In E uclidean space, gradient X is square-
 D istance M easures
root-of-5 tim es as im portant as gradient Y W hich
space m atches your intuition?
W ith Euclidean distance, large differences are
w eighted more heavily than several sm all differences
(Box 6.2). T his results in greater sensitivity to outliers
w ith E uclidean distance than w ith city-block distance
m easures For exam ple, assum e we have four species
and three sam ple umts. A, B. and C T he data and
differences in abundance o f each species for each pair
o f sam ple units are listed in Box 6.2.
Geodesic distance
P erform ance o f all o f the traditional distance
m easures declines as distances in species space
increase (Figs. 6.7 and 6 8). An innovative solution to
the problem o f m easuring long distances in nonlinear
structures is the geodesic distance (Tenenbaunr et al.
2000) T his concept is sim ilar to the "shortest path"
adjustm ents to a distance m atrix (W illiam son 1978.
1983: Clvm o 1980. Bradfield & K enkei 1987. De ath
1999a). W illiam son sum m ed distances between
sam ple unit pairs representing the shortest path
betw een two distant SUs. but only applied this to SUs
w ith no species in com m on Bradfield and Kenkei
(1987) added flexibility by varying the threshold for
the num ber of species in com m on. Bradfield and
Kenkei found better results w ith a low er threshold; i.e..
adjusting a larger proportion o f the distance m atrix.
De ath (1999a) further extended the m ethod by using
city-block distance m easures, changing the threshold to
a quantitative dissim ilarity value, and allow ing
inultiple-step paths betw een very distant SUs.
A geodesic distance betw een two points is m eas­
ured by accum ulating distances betw een nearby points.
T enenbaum et al. (2000) used E uclidean distances, but
geodesic distances can be built from other distance
m easures "N earby" can be defined as a fixed radius or
as the n nearest neighbors G eodesic distances should
be able to find effectively the curvature of com position­
al gradients in species space Geodesic distances arc
one o f the m ost prom ising new m ethodological deve-
lopments. A key issue w ill be objectively defining
"nearby" to optim ize the recovery o f patterns in
ecological com m unities.
T he difference betw een T enenbaum s geodesic
distance an d the ecologists' shortest path (SP) m ethods
can be visualized w ith an analogy to crossing a stream
dotted w ith stepping stones We w ant to find the
shortest route from a p articular point on one bank to a
p articular point on the opposite bank. T he SP m ethod
m ust find a single stepping stone that gets us across the
stream in the two shortest possible leaps (one to the
stepping stone and one to the far bank). T he geodesic
m ethod, however, defines a com fortably sm all step,
then seeks the shortest series o f steps w ithout ever
exceeding that sm all step length. T he geodesic m ethod
thus considers the w hole array o f stepping stones,
w hile the SP m ethod can consider only one stone at a
tim e
A problem w ith the SP m ethod is that if the stream
is broader th an two leaps, th en no single stepping stone
will work. T his corresponds to two SUs so different
th at there is no th ird SU that shares species with both
o f them. De ath (1999a) solved this problem by allow ­
ing m ultiple passes o f the SP method, in essence
allow ing m ultiple stepping stones.
D espite the excellent o rdinations in Bradfield and
K enkei (1987), Boyce an d E llison (2001), and De ath
(1999a). the geodesic distances and related m ethods
have not been widely adopted, probably because they
have not been included in p opular softw are packages.
W hether T enenbaum et a l .’s (2000) geodesic distances
offer further im provem ents over the SP m ethods used
by ecologists rem ains to be seen
 C'hapter 6

Box 6.1. C om parison o f E uclidean distance w ith a proportion coefficient (Sorensen distance). Relative proportions
o f species 1 and 2 are the sam e between Plots 1 and 2 and Plots 3 and 4

Data m atrix containing abundances

E xam ple calculations o f distance m easures
for Plots 3 an d 4. ED = E uclidean distance.
S pi Sp2 PD = Sorensen d istance as percentage
Plot 1 1 0
ED, . V ( 1 0 - It))2 + ( 1 0 - o r = 10
Plot 2 1 1
Plot 3 10 0 io o ( |io - io | + |i o - o |)
Plot 4 10 10 PD , 4 = = 33.3%
10 + 20

Plot 1 Plot 2 Plot 3 Plot 4

Plot 1 0
Plot 2 33.3 0
Plot 3 81.8 83.3 0
Plot 4 90.5 83.3 33.3 0

E uclidean distance m atrix

Plot 1 Plot 2 Plot 3 Plot 4
Plot 1 0
Plot 2 10 0
Plot 3 .9.0 9.1 0
Plot 4 13.4 12.7 10.0 0

T he Sorensen distance betw een Plots 1 and 2 is 0.333 (33.3% ). as is the Sorensen
distance between Plots 3 and 4. as illustrated below In both cases the shared abundance is one
third o f the total abundance. In contrast, the E uclidean distance betw een Plots 1 an d 2 is I,
w hile the E uclidean distance betw een Plots 3 and 4 is 10. T hus the Sorensen coefficient
expresses the shared abundance as a proportion of the total abundance, w hile E uclidean distance
is unconcerned w ith proportions.

10 Plot 4 ♦

8
СЧ
00 6
<υ
Õ 33.3
<D
Q* 4
СП Plot 2
2 I 33.3
♦ / Plot 1 Plot 3
0

2 4 6 8 10
Species 1
 D istance M easures

Box 6.2. E xam ple data set com paring E uclidean and citx -block distances, contrasting the effect
o f squaring differences versus not.

H ypothetical data: abundance o f four species in three sam ple um ts (SU).

Sp
SU 1 2 3 4
A 4 2 0 1
B 5 1 1 10
C 7 5 3 4

Sam ple units A.B: species differences d ~ 1, 1, 1, 9 for each o f the four species.
Sam ple units A.C: species differences </= 3. 3, 3, 3

D istance M easure

P air o f SUs E uclidean City-block

AB 9.165 12

AC 6 12

T he sim ple sum o f differences (city-block distance) is the sam e for AB and AC
E uclidean distance sum s the squared differences, so that difference o f 9 is given more
em phasis w ith E uclidean distance than w ith city-block distance. Thus, the Euclidean
distance betw een A and B is larger th an the distance betw een A and C The city-
block distance between A and B is the sam e as that betw een A and C. W hich dis­
tance m easure m atches your intuition ’