Limnol. Ocem7ogr.. 42(5), 1997, 938-949
0 1997. by the American Society 01 Limnology and Oceanography, IK.
The intermediate disturbance hypothesis, refugia, and biodiversity in streams
Colin R. Townsend and Mike R. Scarsbrook'
Zoology Department, University of Otago, PO. Box 56, Dunedin, New Zealand
Sylvain Dolgdec
ESA CNRS 5023, Ecologic des Eaux Deuces et des Grands Fleuves, Universite Claude Bernard, Lyon 1, 69622
Villeurbanne Cedex, France
Abstract
The intermediate disturbance hypothesis has been influential in the development of ecological theory and has
important practical implications for the maintenanceof biodiversity but has received few rigorous tests.WC tested
the hypothesisthat maximum taxon richnessof macroinvertebrateswill occur in communities subjectto intermediate
levels of disturbance at 54 streamsites that differed in the frequency and intensity of flood-related episodesof bed
movement. Our results support the intermediate disturbance hypothesis, with both highly mobile and relatively
sedentarytaxa conforming to the predicted bell-shaped curve. Taxon richness was not related to habitat area(stream
width), distancefrom the headwater,or the diversity of microhabitats (particle size categories)but was significantly
and negatively related to the proportion of the substratum made up of small particles. Of all the factors measured,
however, bed disturbance was by far the best at accounting for variation in taxonomic richness. We also quantified
several kinds of potential refugia for invertebrates and found a positive relationship between richness and a refugia
axis that combines amount of dead space with proportion of large substratum particles.
The intermediate disturbance hypothesis, whose genesis
can be traced back to Hutchinson (1953) and Horn (1975),
was proposed by Connell (1978) to account for patterns of
diversity in tropical rainforests and coral reefs. It has occupied
a central place in the development of ecological theory
because all communities are subject to disturbances that exhibit
different frequencies and intensities. It also has important
practical implications for the maintenance of biodiversity,
of which species richness (the number of species
present) is the most basic component. The hypothesis is
based on the argument that ecological communities seldom
reach an equilibrium state, in which the competitively superior
species exclude others, because disturbances that kill
or damage individuals will continually set back the process
of competitive elimination by opening space for colonization
by less competitive individuals. For this idea to work, there
must be a trade-off between species' colonizing ability and
competitive ability, and the habitat must be patchy and dynamic
(Wilson 1994). At one extreme, patches that are frequently
and/or intensely disturbed are expected to exhibit
low species richness because few species are able to colonize
during the brief periods between disturbances or tolerate the
high intensities of their impact. At the other end of the scale,
patches in which disturbances are infrequent and/or of low
intensity are expected also to be poor in species because they
I Present address: National Institute
Research, Hamilton, New Zealand.
of Water and Atmospheric
Acknowledgments
We thank Alex Huryn for his critique of an earlier version of this
manuscript, Chris Arbuckle for his management of the Taicri database,
and Luc Corstens, Femke Jansma, and Richard Montgomery
for help in the field.
This research was supported by a grant from the Foundation for
Science, Research and Technology of New Zealand.
become dominated by competitively superior taxa. Richness
should be highest at intermediate levels of disturbance because
both rapid colonizers and more competitive species
co-occur.
Despite its importance, the intermediate disturbance hypothesis
has received few rigorous tests. An appropriate test
requires, first, a measure of disturbance that is relevant to
the organisms and, second, a sufficient number of replicate
communities that are similar, in general, but which differ in
their disturbance regimes. Streams offer an excellent opportunity
for testing the hypothesis (Townsend 1996). Disturbances
in streams often take the form of bed movements
during periods of high discharge; because of differences in
flow regimes and in the substrata of stream beds, some
stream communities are disturbed more frequently and to a
greater extent than others. We quantified the bed disturbance
regime at a number of stream sites and tested whether the
pattern of richness in benthic macroinvertebrate taxa conforms
to the intermediate disturbance hypothesis.
Refugia, which may act to lessen the effects of disturbance,
have been postulated to exist at various spatial scales
in streams (Sedell et al. 1990; Townsend 1989; Townsend
and Hildrew 1994), but their availability and use by invertebrates
have received little study (Scarsbrook and Townsend
1994). Local refugia that may be exploited by benthic invertebrates
include large stable substratum particles (Townsend
1989), dead zones (where shear stresses on the bed are
always low, even at high discharge) (Lancaster and Hildrew
1993, 1994), and the hyporheic zone (interstitial habitat deep
in the stream bed) (Palmer et al. 1992). On a larger scale,
areas upstream may act as a refugium for colonizers that
drift downstream to disturbed locations (Townsend 1989),
nearby tributaries may provide colonists arriving as flying
adults (Townsend 1989), and floodplains may provide refugia
into which animals move during a flood and then return
938
Disturbance in streams 939
Fig. 1. Location of 54 stream sites in seven subcatchmcnts of the Taieri River. Each number
represents a single tributary, within which upper and lower sites (black circles) were sampled.
as discharge subsides (Badri et al. 1987). We attempted to
quantify several categories of potential refugia and to determine
relationships between their availability and taxon rich-
ness.
Methods
Study sites-Our study involved two sites on each of 27
tributaries of similar order scattered throughout a 2991 -km2
area in the Taieri River catchment on the South Island of
New Zealand (Fig. 1). Each site was approximately 30 m in
length and comprised a single pool-riffle sequence. Distance
between the upper and lower sites on each tributary ranged
from 200 m to 15 km. Sites varied in mean width (1.07-
7.25 m-measured as wetted width at baseflow in spring for
six cross sections), mean discharge (9-1,760 I s-l; discharge
was continuously monitored on the main stem of each catchment
from July 1993 to June 1994 by the National Institute
940 Townsend et al.
Table 1. Invertebrate taxa identified in this study.
Higher taxonomic
group Family' Taxon
Ephemeroptera Ephemcridae
Leptophlebiidae
Oligoneuriidae
Siphlonuridae
Plccoptera
Tiichoptera
Austropcrlidae
Eusteniidae
Gripopterygidae
Conocsucidae
Helicopsychidae
Hydrobiosidae
Coleoptera
Diptera
Hydropsychidae
Hydroptilidae
Leptoceridae
Philopotamidae
Polycentropodidae
Dytiscidae
Ehnidac
Hydraenidae
Hydrophilidae
Ptilodactylidae
Scirtidae
Tipulidae
Ceratopogonidae
Chironomidae
Hemiptera
Mecoptera
Megaloptera
Odonata
Empididac
Muscidac
Psychodidae
Simuliidae
Tabanidae
Tanyderidae
Corixidae
Nannochoristidae
Corydalidae
Anisoptera
Zygoptera
Ichthybotus bicolor Tillyard
Austroclima spp.
Deleatidium spp.
Neozephlebia scita (Walker)
Zephlebia spectabilis
Coloburiscus humerulis (Walker)
Ameletopsis perscitus (Eaton)
Nesameletus spp.
Oniscigaster wakejeldi McLachlan
Austroperla cyrene (Newman)
Stenoperla prasina (Newman)
Megaleptoperla spp.
Zelandobius spp.
Zelundoperla spp.
Beraeoptera roria Mosely
Olinga feredayi (McLachlan)
Pycnocentrodes sp.
Helicopsyche albescens Tillyard
Costachorema spp.
Hydrobiosis spp.
Neurochorema spp.
Psilochorema spp.
Tiphobiosis sp.
Aoteupsyche spp.*
Oxyethira albiceps McLachlan
Paroxyethira sp.
Hudsonema amabilis (McLachlan)
Triplecides obsoleata (McLachlan)
Hydrobiosella stenocerca Tillyard*
Polyplectropus sp.*
Liodessus plicatus (Sharp)
Antiporus sp.
Hydora spp.
Hydraenidae
Berosus sp.
Byrrhocryptus sp.
Scirtidae
Aphrophila neozelandica (Edwards)*
Eriopterini
Paralimnophila skusei Hutton
Limonia nigrescens (Hutton)
Zelandotipulu sp.
Tanypodinae
Non-tanypod Chironomidae
Limnophoru sp.
Austrosimulium australense gp."
Mischoderus sp.
Sigara sp.
Nannochoristu philpotti (Tillyard)
Archichauliodes diversus (Walker)
Anisoptera
Zygoptera
Disturbance in streams 941
Table 1. Continued.
Higher taxonomic
group
Molluscs
Crnstacca
Platyhelminthes
Acarina
Oligochaeta
Family
Planorbidae
Lymnaeidae
Physidae
Hydrobiidae
Sphaeriidae
Amphipoda
Isopoda
Ostracoda
Decapoda
Taxon
Gyraulus corinna (Gray)
Lymnaea sp.
Physa acuta (Draparnaud)
Potamopyrgus antipodarum (Gray)
Sphaerium novaezelandiae (Deshayes)"
Amphipoda
Isopoda
Ostracoda
Paranephrops zealandicus (White)
Platyhelminthes
Acarina
Oligochaeta
* Taxa that are relatively sedentary.
of Water and Atmospheric Research Ltd. while each tributary
had maximum flow indicators and was regularly gauged
to allow the correlation of flows in the tributaries with those
of the mainstem), stream slope (0.11% to 9.92%-measured
over a 30-m reach with an Abney level), distance from furthest
headwaters (0.65-45.15 km-estimated from topographical
maps) and median particle size (medium gravel,
16 mm diameter, to large cobble, 256 mm diameter-assessed
from 100 randomly chosen particles). Diversity of
particle sizes at a site was measured as the evenness of representation
of 12 substratum size classes (E = H'/ln S with
S as the number of size classes); all were present at every
site so that evenness is a measure of the extent to which
many microhabitats are available in substantial quantities.
Assessment of disturbance-To assess bed disturbance
during high discharge events, particles taken from the immediate
area of each study reach and corresponding to the
50th, 75th, and 90th percentiles of the substratum size distribution
(excluding bedrock) were painted and arranged on
the surface of the stream bed in regular arrays (five rows of
three particles in each of the three size classes randomly
assigned to transects placed 1 m apart). The movement of
these particles was monitored on five occasions from September
93 to June 1994, a period when some unusually large
discharge events occurred. On each occasion, particles that
had moved were noted and replaced. Mean intensity of bed
disturbance at a site was calculated as the average of the
percentage of the painted tracer particles of all size classes
combined that moved between consecutive sampling occasions
(log-transformed). Frequency of disturbance was taken
to be the proportion of periods (of the five) during which at
least an arbitrarily chosen 40% of particles moved (all size
classes combined).
Assessment of refugia-Our assessment of refugia is tentative
because it has not generally been determined whether
invertebrates actually use the refugia proposed by theorists.
We have measured six variables that we consider to be potential
indices of potential refugia. (1) The proportion of 100
randomly selected substratum particles from the streambed
that were large (diameter > 128 mm). (2) The amount of
dead space on the stream bed, assessed as the percentage of
100 random shear stress measurements, taken at base flow
in summer, which were less than 0.771 X 10 PN cmP2.This
value corresponds to the critical shear stress required to
move a hemisphere (3.85 cm radius) with density 1.015 g
cm-", the lightest in a set of 24 hemispheres differing only
in density which are used to estimate near-bed shear stress
(Statzner and Mtiller 1989; Statzner et al. 1991). It was not
possible to estimate shear stress under high discharge conditions.
(3) An index of the potential depth of the hyporheos
(interstitial space in the bed inhabited by invertebrates), obtained
by averaging the depths to which a metal sounding
rod (16 mm diameter, 1.5 m length) could be hammered into
the substratum at 10 randomly chosen points. Invertebrates
occur in loose bed material to the depths measured (pers.
obs.), whereas none can occur at a depth where bedrock
exists close to or at the bed surface. (4) An index of the drift
refugium, measured as the length in km of all stream channels
upstream of the site. (5) An index of the aerial refugium,
measured as the length in km of stream channels within an
arbitrary 1 km of the site. (6) An index of floodplain availability,
classified into four classes according to the angle of
the upper streambank; l-highly constrained channel with
minimal floodplain (bank gradient >50"), 4-unconstrained
channel with maximal floodplain (bank gradient <30"), 2
and 3 are intermediate.
Sampling of invertebrates-While the physical measurements
at each site were taken during 1993-1994, invertebrates
were sampled in two periods. On tributaries 1-21,
four Surber samples (area 0.06 m2, 300 pm mesh, two each
in a pool and a riffle) were taken from random locations in
the 30-m study reaches in the austral summer between 2
January and 7 February 1990 as part of a large-scale, longterm
study of sites within the Taieri River system (Townsend
et al. 1997). Size-class composition of the bed material and
other physical conditions at these sites did not change between
1990 and 1994 (pers. obs. on frequent site visits).
Thus, the indices of disturbance and refugia availability derived
in 199311994 can provide an acceptable comparative
942 Townsend et al.
representation of conditions in 1990. Macroinvertebrates at
sites in tributaries 22-27 were sampled, using the same techniques,
between 17 January and 1 March 1994, a few weeks
after some uncharacteristically large .floods in the Taieri
catchment, the largest of which (1,467.2 rn? s-l measured
below the confluence of tributaries 5-27; see Fig. 1) had a
return period of about 30 yr (National Institute of Water and
Atmospheric Research, Dunedin, New Zealand, pers.
comm.). In contrast, the 1990 sampling dates followed a period
of low flow without major flood disturbances (mean
discharge for the period 30 November 1989 to 31 March
1990 = 6.2 rn" s I, maximum discharge = 110.3 m3 s-l).
This presented an opportunity to assess the influence of the
timing of the most recent major disturbance on invertebrate
richness.
Invertebrate samples were preserved in the field in 70%
ethanol. In the laboratory, animals were removed from the
samples and identified to species wherever possible, using
the keys of Winterbourn and Gregson (1989) for Insecta,
Winterbourn (1973) for Molluscs, and Chapman and Lewis
(1976) for Crustacea and Acarina. When specific names
could not be reliably assigned, taxa were identified to the
genus level. Likewise where species have not been formally
described, family or subfamily names are given. A list of
taxa is provided in Table 1. Taxon richness was calculated
as the number of invertebrate taxa recorded at each site.
Treatment of data-The relationships among physical
variables were explored by correlation analysis (Pearson's
r), applying Bonferroni-adjusted probabilities to provide
protection for multiple tests.
The relationships between invertebrate taxon richness at
the 54 stream sites and our a priori indices of disturbance
were explored using regression analysis. Simple and stepwise
multiple linear regressions were used to discover the
models that produced the highest coefficients of determination
(R2). Moreover, because the intermediate disturbance
hypothesis predicts a bell-shaped curve we repeated the exercise
using polynomial regression models.
The relationships between taxon richness and the availability
of each of our postulated refugia were also investigated
by regression analysis. In addition, we constructed
combined refugia axes as follows. Each variable was scaled
between 0 and 1 using: xv = (zi, - min)/(max - min) where
z,~is the value of the jth variable for the ith site, min is the
smallest value of the jth variable, and max is the highest
value of the jth variable. A noncentered principal components
analysis then ordered the sites according to their reFugia
availability. This analysis takes into account the potential
synergistic effects of the refugium variables. All
combinations of refugia variables were investigated for their
relationships with taxon richness.
The significance of regression coefficients (slopes and intercepts)
against 0 was tested using analysis of variance (null
hypothesis rejected if Foba1 F,-,,,,,). The use of linear regression
models requires data to be distributed normally and
homoscedastic. Tests of normality (Lilliefors 1967) and the
examination of residuals showed that log transformations
were needed to meet these assumptions for intensity of bed
disturbance, distance to the furthest headwater, index of hyporheos,
percentage of large particles, in drift refuge, aerial
refuge and combined abundance of invertebrates.
Results
The relationships among physical variables are shown in
Table 2. The intensity and frequency of disturbance were
positively correlated. The percentage of particles that were
small (8-32 mm in diameter) was positively correlated with
the intensity and frequency of disturbance and hyporheos
depth. Drift refugium, aerial refugium, and distance to the
furthest headwater were strongly positively correlated with
each other. Deadspace was negatively correlated with frequency
of disturbance, drift refugium, and distance to the
furthest headwater.
Taxon richness was highest at intermediate intensities and
frequencies of disturbance (Fig. 2), conforming to the intermediate
disturbance hypothesis. The overall relationship between
richness and the intensity of disturbance (IoD) was
well described by a second-order quadratic regression [richness
= 177.0 (563.2, 95% CL) log IoD - 61.2 (k21.1) log
loD2 - 102.3 (246.8); regression line shown in Fig. 2; 95%
confidence limits are significantly different from zero at P
< O.OOOl].The coefficient of determination for the quadratic
regression model (R2 = 0.42, P < 0.0001) was much greater
than that obtained from a simple linear regression model (R"
= 0.04, P = 0.16). The possibility that site 15 was an outlier
biasing the overall result (see Fig. 2) was tested by reanalyzing
the data after excluding the datum for this site; the
quadratic relationship remained highly significant (R2 =
0.38, P < 0.0001). The relationship between richness and
frequency of disturbance also conformed to the predicted
bell-shaped curve (ANOVA of richness in six frequency categories,
F-value = 10.08, P < 0.0001). (A Tukey HSD multiple
comparison test demonstrated significantly lower taxon
richness for FoD = 0.0 than 0.4, and also for FoD = 0.2
than either 0.4 and 0.6; taxon richness was significantly
higher for FoD = 0.4 than any of 0.6, 0.8, and 1.0.) The
relationships with richness for IoD and FoD are essentially
the same because intensity and frequency of disturbance
were strongly correlated in our study. Results presented in
the rest of this paper are restricted to intensity of disturbance.
Relationships between taxon richness and the intensity of
disturbance were similar in the upper and lower sets of sites
from the 27 tributaries but the percentage of variance explained
was somewhat higher in the upper set of sites [upper
sites: richness = 200.1 (58 1.O, 95% CL) log IoD - 70.5
(528.4) log IoD2 - 115.2 (k57.3); R2 = 0.52, P < 0.0001;
95% CL are significantly different from zero at P < 0.0005,
P < 0.0001, and P < 0.0001, respectively; lower sites: richness
= 196.0 (5143.1, 95% CL) log ToD - 66.1 (k46.0)
log IoD2 - 120.1 (L 109.7); R2 = 0.36, P = 0.005; 95%
CL at P < 0.01, P < 0.007, and P < 0.04, respectively]. In
both cases the coefficients of determination for quadratic regression
models (see above) were greater than those obtained
for simple regression models (upper: R2 = 0.00, P > 0.95;
lower: R2 = 0.12, P = 0.08). Perusal of Fig. 2 also shows
that the 1994 (postflood) subset of sites (tributaries 22-27)
conformed to the overall relationship just as well as the 1990
subset (tributaries 1-21).
Disturbance in streams 943
n
l.E
n
s
The number of sampled organisms differed from site to
site (Fig. 3a) and samples containing more individuals are
expected to contain more taxa. To check the possibility that
the pattern of richness might be an artifact arising through
a relationship between abundance and disturbance, we investigated
the ratio of taxon richness to abundance (logtransformed)
and found a second-order quadratic relationship
between this ratio and intensity of disturbance similar
to that between richness and disturbance [Fig. 3b; richness/
abundance = 48.9 (220.4, 95% CL) log IoD - 17.0 (26.9)
log IoD2 - 26.9 (215.2); R2 = 0.37, P < 0.0001; 95% CL
all at P < O.OOOl].The R2 value resulting from the quadratic
regression model was again higher than that obtained from
a simple regression model (R2 = 0.05, P = 0.09).
The Shannon diversity index (H' = -x:f..,pi In pi, where
pi is the proportional abundance of the ith taxon) also
showed a quadratic relationship with intensity of disturbance
[Shannon diversity = 10.1 (k7.7, 95% CL) log IoD - 3.8
(k2.9) log IoD2 - 3.9 (25.7); R2 = 0.27, P < 0.001; 95%
CL at P < 0.015, P < 0.005, and P = 0.27, respectively]
(Fig. 3~). This index has two components-richness and the
evenness of the distribution of individuals among taxa.
Evenness (E = H'/ln S with S as the richness of invertebrate
taxa) was negatively, linearly correlated with the intensity of
disturbance [evenness = 0.9 (20.2, 95% CL) - 0.2 (50.1)
log IoD, R2 = 0.13, P < 0.008; 95% CL at P < 0.0001 and
P < 0.008, respectively] (Fig. 3d). A quadratic regression
model was not appropriate in this case (quadratic term not
significantly different from zero).
Invertebrate taxa may be divided into those that are mobile
and those that are more or less sedentary (Table 1).
Figure 4 shows that both mobile and sedentary taxa conform
to the pattern predicted by the intermediate disturbance hypothesis
[mobile taxon richness = 156.2 (L55.9, 95% CL)
log IoD - 54.4 (+ 18.7) log IoD2 - 90.0 (241.5); R2 =
0.44, P < 0.0001; 95% CL all at P < 0.0001; sedentary
taxon richness = 20.8 (5 15.2, 95% CL) log IoD - 7.1
(25.1) log IoD2 - 12.0 (211.3); R2 = 0.14, P = 0.021;
95% CL at P < 0.01, P < 0.01, and P < 0.04, respectively].
The R2 values resulting from quadratic regression models
were again higher than those obtained from simple regression
models (mobile: R2 = 0.06, P = 0.07; sedentary: R2 =
0.01, P = 0.55).
Taxon richness was not significantly related to stream
width (Fig. 5a), distance to the furthest headwater (Fig. 5b),
or the diversity of particle sizes at a site (Fig. 5~). When
individual particle size classes were considered, however,
richness decreased with an increase in the proportion of the
substratum composed of small particles [richness = 26.7 -
0.14X, where X = proportion of particles 8-32 mm in diameter;
R2 = 0.13, P < 0.009; 95% confidence limits 2.8
(P < 0.0001) and 0.1 (P = 0.009), respectively].
The relationships between taxon richness at the 54 stream
sites and the availability of each of the six postulated refugia
are shown in Fig. 6. Taking the variables separately, only
the amount of dead space showed a significant relationship
with richness [richness = 15.8 (24.8, 95% CL) + 0.5 deadspace
(L-0.3) - 0.005 (+0.004) deadspace2, R2 = 0.19, P
< 0.005; 95% CL at P < 0.0001, P < 0.008, and P = 0.03,
respectively]. A refugia axis that combines the amount of
944 Townsend et al.
30
cn 25
E
5
0
[1: 20
5
F
15
IO
(a) 4m 164
01 la O3
16C)5lIl-)132R 8M
010
27(]19W 25()(29
7a
15
w-----r- ! I
--I-T- .-r- ,- - ,- ] --. ,--. , -.7- .I.-T--,
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
LOG INTENSITY OF DlSTlJRBANCE
(W 4161
03
72@1
582627@1316
0" --. ------ -_ -___
014
__
20018
23m
(18 024
15m 023
24W
r ---- -------7 I--- - I ------- --- --7 -1
0 0.2 0.4 0.6 0.8 1
FREQUENCY OF DISTURBANCE
Fig. 2. Richness of invertebrate taxa in relation to (a) intensity of disturbance and (b) frequency
of disturbance at 54 stream sites in the Taieri River. The invertebrate communities of tributaries l-
21 and 22-27 were sampled in the summers of 1990 and 1994, respectively. In each tributary two
sites were selected; lower sites are shown as open circles; upper sites as black squares (numbers
are in italic). A second-order quadratic regression line is fitted in a; lines arc drawn between means
for five frequency categories in b.
dead space with the proportion of large substratum particles
was positively and linearly related to taxon richness [richness
= 19.1 (24.3) + 6.9 (25.7) refuge; R2 = 0.13; P <
0.008; 95% CL at P < 0.0001 and P = 0.016, respectively].
A stepwise multiple regression involving all of the physical
variables failed to detect a combination of variables that
could explain more of the variation in taxon richness than
was accounted for by the intensity of disturbance alone.
advantages. The first, used here, concentrates on patterns in
natural communities and seeks to establish whether taxon
richness peaks, as predicted, at intermediate levels of disturbance.
Because it deals with natural situations this appreach
has the advantages of realism and, perhaps more important,
utility. Support for the hypothesis may provide the
knowledge to engineer the maintenance or enhancement of
biodiversity by, for example, manipulating discharge re-
Discussion
gimes. The disadvantages are that the mechanisms underlying
the intermediate disturbance hypothesis are not directly
There are two approaches to testing the intermediate disturbance
hypothesis, each with inherent advantages and disaddressed
and, furthermore, that a relationship between richness
and disturbance may result from the effect of another
Disturbance in streams 945
41
( )a
+j 3.6 z
-
2 3.2 -
5 - 0
$ 2.0 - 0
8 02.4 -
.A
I
2JIIIIIIIIIIII(IIIIII(I
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
1.0'1.1 1.2'1.3 1:4'1:5'1:6'1.7'1:8 1:9'
4 c
1
( )
3.5 z
3- 0
z
k 2.5 - Y!!!
A--'
I a
P 2-
1.5 - n +4
1 - III-T--T- T-T- T-IIT--TI
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
0.8 l (d)
o-7-
8 0.6 -
%
E 0.5 zi
0.4 -
0.3 LOG
INTENSITY OF DISTURBANCE
Fig. 3. Relationships of intensity of disturbance with (a) invertebrate
abundance(m 2), (b) richness/log abundance,(c) Shannon
diversity index, and (d) evenness.Sites sampledin 1990areshown
ascircles and in 1994astriangles; lower sitesasopen symbols and
upper sites as closed symbols.
30
25
lj 20
g 15 -
z
10' -
'5-
O-
5
r4
z
!z3
E
ii2.
1
( )a
IrI 1 r--1- 1 I l-r7 7-r-r-1 1 I I I T I I
1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9
LOG INTENSITY OF DISTURBANCE
' 1:O'lil 'I:2 11!3'1~4'1:5' 1!6' 1!7' 1:8'1:9'
LOG INTENSITY OF DISTURBANCE
Fig. 4. Relationships of intensity of disturbancewith (a) number
of mobile taxa and (b) number of sedentary taxa. A second-order
quadratic regressionis fitted.
factor correlated with disturbance; correlation does not prove
causation. We have tried to deal with the potential confounding
effects of intercorrelated factors in our analytical approach;
however, of all the factors measured, bed disturbance
accounted for the most variation in taxonomic richness. The
second approach seeks to check the veracity of the mechanisms
underlying the intermediate disturbance hypothesis;
that is, the trade-off between competitive and colonizing
ability under contrasting disturbance regimes (e.g. Sommer
1995). Such an approach can be informative but is unsatisfying
unless it has been established that the predicted pattern
in richness actually exists and that the proposed mechanism
is actually at work in the natural environment. The ideal
study, which has not yet been performed, will combine both
approaches. One reason for this lack of a combined approach
is the enormous amount of work required either to determine
the pattern or to establish the mechanism. To provide evidence
of competitive exclusion among just a pair of stream
invertebrate species requires a substantial effort (e.g. Hemphill
and Cooper 1983). The extension of pairwise competition
studies to large species assemblages, and the incorporation
of investigations of colonizing as well as competitive
abilities of taxa, provides a major challenge for the future.
In our study, it is likely that low taxon richness at high
frequencies and intensities of disturbance reflects the poor
946 Townsend et al.
( )a
i
0
30
l 0
4A ub l OU
l o an
i
25-a~09eAoqp~~ . 0
a *
au
8
p 20e3b
0
Y 0
0 A a
cc 15- m A
l n,
On
10 +-l----r--r- -I-r --- AT ----T-- - .,----.~
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
LOG STREAM WIDTH (m)
25
20
15
110----.---r
0,
-r-----`r __- .,-- ___.I
-- --
_ --I--- _- , __---__(_,
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
(Cl
LOG DISTANCE DOWNSTREAM (km)
-
30 0
0
o o A 4A)A
25 -
20 -
15 -
0
0
0
ma
A o
,(?a
10 o*___ ..__7-d--1-
0.7 0.75 0.8 0.85 0.9
SUBSTRATE DIVERSITY
Fig. 5. Relationships between richness of invertebrate taxa and
(a) streamwidth, (b) distance downstream (distance from furthest
headwaterto sampling site), and (c) substratumdiversity (measured
asthe evennessof rcprcsentation of 12 substratumsize classes).
ability of some stream invertebrates to colonize or persist in
such situations. Whether low richness at low frequencies and
intensities of disturbance is due to competitive exclusion, as
proposed in the intermediate disturbance hypothesis, is uncertain.
Competitive exclusion in streams is expected to be
more likely between species that are relatively sedentary and
compete for space (Hemphill and Cooper 1983; McAuliffe
1984; Frid and Townsend 1989). However, in the River
Taieri both sedentary and highly mobile taxa conform to the
intermediate disturbance pattern, indicating that competition
between mobile taxa (e.g. Kohler 1992) may be more influential
than previously thought. The finding that evenness was
higher in less disturbed sites is consistent with the hypothesis
that competitive forces structure these communities by eliminating
rare species, but the pattern is also consistent with
some rare taxa gaining more members. Further work will be
needed to resolve these conflicting hypotheses.
The upper and lower sites in each tributary often differed
substantially in their disturbance regimes because of increased
discharge between them and/or differences in the
size class composition of their substrata. The predicted quadratic
relationships were similar in the two sets of sites, but
with the percentage of variance explained somewhat higher
in the upper set of sites. It is possible that the tendency of
invertebrates to drift downstream (Brittain and Eikeland
1988) confounds the relationship between taxon richness and
disturbance regime at downstream sites.
The 1990 samples were taken after a period of low flow
without major floods, whereas the 1994 samples were taken
a few weeks after some large floods. We might have expected
the 1994 samples to contain fewer species, being
dominated by a few, rapidly colonizing species. It is of interest,
therefore, that the relationship for the 1994 subset
conforms to the overall relationship just as well as the 1990
subset, indicating that the timing of the most recent major
disturbance had little impact on the pattern of taxon richness.
Other proposed patterns of species richness in streams
were investigated but not supported. Thus, there were no
significant relationships between species richness and habitat
area (assessed as stream width) (Townsend and Hildrew
1994) or distance downstream (Vannote et al. 1980), nor was
there a relationship between species richness and the diversity
of substrata (microhabitats) available at a site (Townsend
1991). Richness, however, was observed to decrease with an
increase in the proportion of the substratum made up of
small particles (8-32 mm diameter). Because sites with
smaller particles will experience a higher level of disturbance
for a given discharge event, the decrease in richness
observed from intermediate to high disturbance levels might
conceivably be due to a substratum size effect and not to
disturbance per se. However, while intensity of disturbance
explains more than 42% of the variance in richness, the variance
explained by small substratum was only 12.5%. In addition,
the relationship with substratum particle size cannot
explain a decrease in richness from intermediate to low disturbance
levels.
Whether frequency or intensity of disturbance is the more
influential in determining species richness in our streams is
a moot point. A greater frequency of disturbance may reduce
richness by excluding taxa that cannot quickly recolonize in
the intervals between disturbances and a lesser frequency
may permit competitive exclusion of species that are capable
colonists but poor competitors. An intermediate frequency
allows both sets of species to co-occur. On the other hand,
a greater intensity of disturbance may exclude species because
local refugia become inoperative (even large particles,
dead space and the hyporheos are disturbed) whereas largescale
refugia (drift, aerial, floodplain) predominate and may
provide sources of long-distance colonists. A lesser intensity
of disturbance may lead to low richness because most of the
bed is relatively stable over the long term and competitive
exclusion occurs. Finally, an intermediate intensity of disturbance
permits coexistence of a wider variety of species
because of the presence of a range of local and large-scale
refugia. In operational terms, it may be impossible to dis-
Disturbance in streams 947
0
LOG :/tRGE PARTIdES (%)
15
0 10 20 30 40 50 60 70
DEAD SPACE
0 0.05 0 1 015 02 025 03
LOG HYPORHEOS DEPTH (m)
2 25
W
$ 20
0
a 15
IO
LOG DRIFT REFUGIUM (km)
02 03 04 05 0.G 0.7 08 0.9 1
LOG AERIAL REFUGIUM (km)
IO I / 1 I
1 2 3 4
FLOODPLAIN AVAILABILITY
`i\ A (,'
Acj 'j
Cl
A` e
I I I I I
0 02 04 06 08 1 12 14
COMBINED REFUGIA INDEX
Fig. 6. Relationships between richness of invertebrate taxa and various possible refugium categories:
(a) proportion of streambcdsubstratumparticles that were large (diameter > 128mm), (b)
potential amount of dead space(percentageof the streambcdwhcrc shear stresswas negligible,
measuredat baseflow in summer), (c) potential depth of hyporhcos, (d) potential extent of drift
refugium, (e) potential extent of aerial refugium, (f) potential availability of floodplain refugium
(l-highly constrained channel with minimal floodplain, &-unconstrained channel with maximal
floodplain), and (g) a refugia axis that combines dead spaceand proportion of large substratum
particles. The relationships were statistically significant in b (second-order quadratic regression
fitted) and g (simple linear regressionfitted).
entangle frequency and intensity when intensity is measured
as the percentage of space disturbed and frequency as the
number of occasions when more than an arbitrarily defined
percentage of area is disturbed. More frequently disturbed
stream sites are also more intensely disturbed.
The central importance of disturbances in the development
of ecological theory (Begon et al. 1996) places a high premium
on finding appropriate ways to measure disturbance.
Our measures of disturbance are undoubtedly imperfect, particularly
because the tracer particles were placed on the
streambed when, in reality, much of the substratum consists
of embedded particles that are less likely to move at a given
discharge than surface particles. Moreover, other potentially
influential aspects of disturbance for benthic invertebrates,
such as abrasion and high shear stress, are ignored. Nevertheless,
our a priori choices of disturbance measures were
related to taxon richness as predicted by the intermediate
disturbance hypothesis. The method provides a comparative
index that may be of general value in studies of disturbance
in streams. Tests of disturbance theory require that disturbance
be defined relative to the organisms of interest. Thus,
for example, the measurement in streams of the frequency
948 Townsend et al.
with which certain sizes of discharge event occur (Resh et
al. 1988) is not likely to be appropriate for organisms that
inhabit the streambed because the impact of these events also
will vary with substratum composition. Disturbance needs
to be assessed in terms of bed movement itself.
Floods with their associated bed movements can have a
severe impact on the community of organisms in streams
and rivers (Siegfried and Knight 1977; Scrimgeour and Winterbourn
1989; Giller et al. 1991; Flecker and Feifarek
1994). Despite the sometimes catastrophic effects of floods,
community recovery can be very rapid, resulting in a community
structure similar to that preceding the disturbance
(Fisher et al. 1982; Molles 1985; Townsend et al. 1987;
Scrimgeour et al. 1988). The presence of a range of refugia,
each likely to be used by different sets of species, must be
largely responsible for this resilience. Our analysis of the
potential importance of refugia in maintaining biodiversity
is very preliminary but indicates that the availability of large
substratum particles and dead space may be influential. Doledec
and Statzner (1994) have also recorded a positive relationship
between species richness in habitats of the Rhiine
river and a spatial variability index defined partly by grain
size.
The preservation (or restoration) of natural and diverse
communities in our streams and rivers is likely to depend
on identifying and protecting (and enhancing) refugia. Moreover,
an intermediate level of disturbance may often itself
be responsible for generating some of the richness of communities,
as indicated here. Patterns of abstraction or impoundment
of river water for human benefit will need to
consider these patterns if we wish to achieve the objective
of maintaining high biodiversity.
References
BAIIRI, A., J. GIUDICELLI, AND G. PRBv~T. 1987. Effects of flood
on the benthic macroinvertebrate community in a mediterranean
river, the Rdat (Morocco). Acta Ecol. Ecol. Gen. 8: 481-
500.
BEGON, M., J. L. HARPER, AND C. R. TOWNSEND. 1996. Ecology:
Individuals, populations and communities, 3rd ed. Blackwell
Science.
BRITTAIN, J. E., AND T. I. EIKELAND. 1988. Invertebrate drift-a
review. Hydrobiologia 166: 77-93.
CHAPMAN, A., AND M. LEWIS. 1976. An introduction to the freshwater
Crustacea of New Zealand. Collins.
CONNELL, J. H. 1978. Diversity in tropical rain forests and coral
reefs. Science 199: 1302-1310.
DOLBDEC, S., AND B. STATZNER. 1994. Theoretical habitat templets,
species traits, and species richness: 548 plant and animal
species in the Upper RhGne River and its floodplain. Freshwater
Biol. 31: 523-538.
FISHER, S. G., L. J. GRAY, N. B. GRIMM, AND D. E. BUSCX 1982.
Temporal succession in a desert stream following flooding.
Ecol. Monogr. 52: 93-110.
FIAXXER, A. S., AND B. FEIFAREK. 1994. Disturbance and the temporal
variability of invertebrate assemblages in two Andean
streams. Freshwater Biol. 31: 131-142.
FRID, C. L. J., AND C. R. TOWNSEND. 1989. An appraisal of the
patch dynamics concept in stream and marine benthic communities
whose members are highly mobile. Oikos 56: 137-
141.
GILLER, P S., N. SANCJPRADUB, AND H. TWOMEY. 1991. Catastrophic
flooding and macroinvertebrate community structure.
Verh. Int. Ver. Limnol. 24: 1724-1729.
HEMPMILL, N., AND S. D. COOPER. 1983. The effect of physical
disturbance on the relative abundances of two filter-feeding insects
in a small stream. Oecologia 58: 378-382.
HORN, H. S. 1975. Marckovian processes of forest succession, p.
196213. Zn M. L. Cody and J. M. Diamond [eds.], Ecology
and evolution of communities. Belknap.
HUTCHINSON, G. E. 1953. The concept of pattern in ecology. Proc.
Acad. Nat. Sci. Phila. 105: 1-12.
KOHLER, S. I,. 1992. Competition and the structure of a benthic
stream community. Ecol. Monogr. 62: 165-188.
LANCASTER, J., AND A. G. HILDR~W. 1993. Characterizing instream
flow refugia. Can. J. Fish. Aquat. Sci. 50: 1663-1675.
-, AND -. 1994. Flow refugia and the microdistribution
of lotic macroinvertebrates. J. N. Am. Benthol. Sot. 12: 385-
393.
LILUHE`ORS, H. W. 1967. On the Kolmogorov-Smirnov test for normality
with mean and variance unknown. 5. Am. Stat. Assoc.
64: 387-389.
MCAULI~;FE, J. R. 1984. Competition for space, disturbance and
the structure of a benthic stream community Ecology 65: 894-
908.
MOLLES, M. C., JR. 1985. Recovery of a stream invertebrate community
from a flash flood in Tesuque Creek, New Mexico.
Southwest. Nat. 30: 279-287.
PAI~MER, M. A., A. E. BELY, ANL) K. E. BERG. 1992. Response of
invertebrates to lotic disturbance: A test of the hyporheic refuge
hypothesis. Oecologia 89: 182-194.
RESH, V. H., A. V. BROWN, A. l? COVICH, M. E. GURTZ, H. W. LI,
G. W. MINSHALL, S. R. REICE, A. L. SHELDON, J. B. WALLACE,
AND R. C. WISSMAR. 1988. The role of disturbance in stream
ecoology. J. N. Am. Benthol. Sot. 7: 433-455.
SCARSBR~OK, M. R., AND C. R. TOWNSEND. 1994. Stream community
structure in relation to spatial and temporal variation:
A habitat templet study of two contrasting New Zealand
streams. Freshwater Biol. 29: 395-410.
SCRIMGEOUR, G. J., R. J. DAVIDSON, AND J. M. DAVIDSON. 1988.
Recovery of benthic macroinvertebrate and epilithic communities
following a large flood, in an unstable, braided, New
Zealand river. N. Z. J. Mar. Freshwater Res. 22: 337-344.
AND M. J. WINTERBOURN. 1989. Effects of floods on epilithon
and benthic macroinvertebrate populations in an unstable
New Zealand river. Hydrobiologia 171: 33-44.
&DELL, J. R., G. H. REEVES, E R. HAUER, J. A. STANFORD, AND
C. l? HAWKINS. 1990. Role of refugia in recovery from disturbances:
Modern fragmented and disconnected river systems.
Environ. Manage. 14: 71l-724.
SIEGFRIED, C. A., AND A. W. KNIC;HT. 1977. The effects of washout
in a Sierra foothill stream. Am. Midl. Nat. 98: 200-207.
SOMMER, U. 1995. An experimental test of the intermediate disturbance
hypothesis using cultures of marine phytoplankton.
Limnol. Oceanogr. 40: 1271-1277.
STATZNER, B., E KOHMANN, AND A. G. HILDREW. 1991. Calibration
of FST-hemispheres against bottom shear stress in a laboratory
flume. Freshwater Biol. 26: 227-231.
, AND R. MULLER. 1989. Standard hemispheres as indicators
of flow characteristics in lotic benthos research. Freshwater
Biol. 21: 445-459.
TOWNSEND, C. R. 1989. The patch dynamics concept of stream
community ecology. J. N. Am. Benthol. Sot. 8: 36-50.
-. 1991. Community organization in marine and freshwater
environments, p. 125-144. In R. S. K. Barnes and K. H. Mann
[eds.], Fundamentals of aquatic ecology. Blackwell Sci. Publ.
-. 1996. Concepts in river ecology: Pattern and process in
Disturbance in streams 949
the catchment hierarchy. Arch. Hydrobiol. 113(suppl. 10): 3-
21.
-, T A. CROWL, C. J. ARBUCKLE, AND M. R. SCARSBROOK.
1997. The relationship between land USCand physicochemistry,
food resources and macroinvertebrate communities in tributaries
of the Taicri River, New Zealand: A hierarchically
scaled approach. Freshwater Biol. 37: 177-192.
AND A. G. HILDREW. 1994. Species traits in relation to a
habitat templet for river systems. Freshwater Biol. 31: 265-
275.
3-, AND K. SCHOFIEI.D. 1987. Persistence of stream
invertebrate communities in relation to environmental variability.
J. Anim. Ecol. 56: 597-613.
VANNOTE, R. L., G. W. MINSHALL, K. W. CUMMINS, J. R. SEDELL,
AND C. E. GUSHING. 1980. The river continuum concept. Can.
J. Fish. Aquat. Sci. 37: 130-137.
WILSON, J. B. 1994. The `intermediate disturbance hypothesis' of
species coexistence is based on patch dynamics. N.Z. J. Ecol.
18: 176-181.
WINTERBOURN, M. J. 1973. A guide to the freshwater molluscs of
New Zealand. Tuatara 20: 141-159.
- AND K. L. D. GREGSON. 1989. Guide to the aquatic insects
of `New Zealand. Bull. N.Z. Entomol. Sot. 9. 95 p.
Submitted: 1 November 1995
Accepted: 18 December 1996