Shifting dominance of riparian Populus and Tamarix along gradients of flow alteration in western North American rivers.

Abstract

Tamarix ramosissima is a naturalized, nonnative plant species which has become widespread along riparian corridors throughout the western United States. We test the hypothesis that the distribution and success of Tamarix result from human modification of river-flow regimes. We conducted a natural experiment in eight ecoregions in arid and semiarid portions of the western United States, measuring Tamarix and native Populus recruitment and abundance at 64 sites along 13 perennial rivers spanning a range of altered flow regimes. We quantified biologically relevant attributes of flow alteration as an integrated measure (the index of flow modification, IFM), which was then used to explain between-site variation in abundance and recruitment of native and nonnative riparian plant species. We found the likelihood of successful recruitment of Tamarix to be highest along unregulated river reaches and to remain high across a gradient of regulated flows. Recruitment probability for Populus, in contrast, was highest under free-flowing conditions and declined abruptly under even slight flow modification (IFM > 0.1). Adult Tamarix was most abundant at intermediate levels of IFM. Populus abundance declined sharply with modest flow regulation (IFM > 0.2) and was not present at the most flow-regulated sites. Dominance of Tamarix was highest along rivers with the most altered flow regimes. At the 16 least regulated sites, Tamarix and Populus were equally abundant. Given observed patterns of Tamarix recruitment and abundance, we infer that Tamarix would likely have naturalized, spread, and established widely in riparian communities in the absence of dam construction, diversions, and flow regulation in western North America. However, Tamarix dominance over native species would likely be less extensive in the absence of human alteration of river-flow regimes. Restoration that combines active mechanical removal of established stands of Tamarix with a program of flow releases conducive to native species establishment and persistence is hypothesized to facilitate the codominance of Populus in reaches where it has become rare. Our findings have implications for planning flow-related stream restoration, for developing realistic expectations for yield on investment in prescribed flow releases, and for planning flow-related interventions that might be possible if control and management of invasive plant species along rivers is a goal.

Full text

Page 1

Ecological Applications, 20(1), 2010, pp. 135–152
? 2010 by the Ecological Society of America
Shifting dominance of riparian Populus and Tamarix along gradients
of flow alteration in western North American rivers
DAVID M. MERRITT1,2,5AND N. LEROY POFF3,4
1National Watershed, Fish, and Wildlife Program, USDA Forest Service, 2150A Centre Avenue, Suite 368,
Fort Collins, Colorado 80526 USA
2Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado 80523 USA
3Department of Biology, Colorado State University, Fort Collins, Colorado 80523 USA
4Graduate Degree Program in Ecology, Colorado State University, Fort Collins, Colorado 80523 USA
Abstract.
become widespread along riparian corridors throughout the western United States. We test the
hypothesis that the distribution and success of Tamarix result from human modification of
river-flow regimes. We conducted a natural experiment in eight ecoregions in arid and
semiarid portions of the western United States, measuring Tamarix and native Populus
recruitment and abundance at 64 sites along 13 perennial rivers spanning a range of altered
flow regimes. We quantified biologically relevant attributes of flow alteration as an integrated
measure (the index of flow modification, IFM), which was then used to explain between-site
variation in abundance and recruitment of native and nonnative riparian plant species.
We found the likelihood of successful recruitment of Tamarix to be highest along
unregulated river reaches and to remain high across a gradient of regulated flows. Recruitment
probability for Populus, in contrast, was highest under free-flowing conditions and declined
abruptly under even slight flow modification (IFM . 0.1). Adult Tamarix was most abundant
at intermediate levels of IFM. Populus abundance declined sharply with modest flow
regulation (IFM . 0.2) and was not present at the most flow-regulated sites. Dominance of
Tamarix was highest along rivers with the most altered flow regimes. At the 16 least regulated
sites, Tamarix and Populus were equally abundant.
Given observed patterns of Tamarix recruitment and abundance, we infer that Tamarix
would likely have naturalized, spread, and established widely in riparian communities in the
absence of dam construction, diversions, and flow regulation in western North America.
However, Tamarix dominance over native species would likely be less extensive in the absence
of human alteration of river-flow regimes. Restoration that combines active mechanical
removal of established stands of Tamarix with a program of flow releases conducive to native
species establishment and persistence is hypothesized to facilitate the codominance of Populus
in reaches where it has become rare. Our findings have implications for planning flow-related
stream restoration, for developing realistic expectations for yield on investment in prescribed
flow releases, and for planning flow-related interventions that might be possible if control and
management of invasive plant species along rivers is a goal.
Tamarix ramosissima is a naturalized, nonnative plant species which has
Key words:
river management; river restoration; tamarisk; Tamarix ramosissima; vegetation change; water development.
dams; flow management; flow regime; index of flow modification; invasive species; Populus;
INTRODUCTION
River bottomlands and riparian habitats serve as key
corridors for the establishment and spread of introduced
species (Planty-Tabacchi et al. 1996, Stohlgren et al.
1999, Brown and Peet 2003) for many of the same
reasons that they harbor disproportionately diverse
assemblages of plants relative to uplands worldwide
(Naiman et al. 1993, Naiman and De ´ camps 1997, NRC
2002). Fluvial disturbance, habitat heterogeneity, varied
flow regime and hydrologic gradients, and connectivity
through landscapes maintain high species diversity and
also provide opportunities for the spread of nonnative,
ruderal species (Pysˇek and Prach 1993, 1994, Hood and
Naiman 2000, Aguiar et al. 2001). Although debate over
the relative importance of species traits vs. site charac-
teristics in explaining the causes and consequences of
plant invasion is far from resolved (Levine and D’An-
tonio 1999, Levine et al. 2003, 2004), it is clear that
disruption or modification of natural disturbance pro-
cesses (e.g., fire and flow regimes) can render some
ecosystems more vulnerable to invasion by nonnative
species (Vitousek et al. 1997, Lonsdale 1999, Thuiller et
al. 2006). Once established, species that regulate distur-
bance processes (e.g., increase fire frequency or intensity)
tend to cause the greatest degree of change within an
ecosystem into which they become established (Busch
Manuscript received 8 December 2008; revised 22 April 2009;
accepted 27 April 2009. Corresponding Editor: C. Nilsson.
5E-mail: dmmerritt@fs.fed.us
135

Page 2

and Smith 1993, Levine et al. 2003, Brooks et al. 2004).
Because patterns of fluvial disturbance, stream flow, and
groundwater regime structure the habitat template upon
which riparian plant communities form, changes in these
hydrologic and channel processes can have significant
negative effects on native, flow-regime-adapted taxa
(Braatne et al. 1996, Stromberg 2001) and may increase
the susceptibility of these systems to invasion by
generalist, invasive, and ruderal plant species.
Worldwide development of rivers for water resources
and flood control represents one of the most deliberate
and pervasive human alterations of natural processes.
Over half of the world’s largest river systems are
regulated and fragmented by dams and storage reser-
voirs which cumulatively detain ;15% of the total
global river runoff annually (Nilsson et al. 2005) and
significantly alter natural hydrologic regimes. In the
western United States the 15000 dams .2 m high and
tens of thousands more smaller dams and diversions
were constructed over the course of the 20th century to
provide municipal and agricultural water supplies,
recreation, enhanced navigation, hydroelectric power,
and flood control (Graf 1999, Lytle and Merritt 2004,
NID 2004). Changes to hydrologic and sediment
regimes on regulated rivers around the world have
induced shifts in the distributions of plant populations
(Busch and Smith 1995, Scott et al. 1997, Lonsdale
1999), altered plant community composition (Nilsson et
al. 1997, Jansson et al. 2000, Merritt and Cooper 2000,
Merritt and Wohl 2006), diminished plant species
diversity (Nilsson et al. 1997, Jansson et al. 2000), and
modified ecosystem processes (Molles et al. 1998, Valett
et al. 2005). Altering the timing, magnitude, duration,
and interannual variability in flow regimes can be
detrimental to species that have phenologic and
physiological adaptations to the characteristics of
natural flow regimes (Lytle and Poff 2004) and may
advantage generalist and ruderal species.
Over the last 50 years along regulated rivers through-
out western North America, some native riparian tree
species, such as cottonwood (Populus deltoides Bartram
ex Marsh. and subspecies), have declined dramatically,
in concurrence with the rapid invasion of Eurasian
saltcedar, or Tamarix (Friedman et al. 2005). Because
Populus (and some other native species) has reproduc-
tive and physiological adaptations synchronized with the
regional timing, magnitude, duration, and interannual
variability of historic stream flow (Karrenberg et al.
2002, Lytle and Poff 2004, Stella et al. 2006), flow
alteration has contributed directly to declines in the
regeneration, persistence, and population structure of
these taxa (Poff et al. 1997, Rood et al. 2003, Lytle and
Merritt 2004). Thus, a widely held view is that the
dramatic decline in Populus over the last 50 years reflects
extensive regional flow alteration (Rood and Mahoney
1990).
However, flow alteration can also influence patterns,
rate, and extent of colonization of exotic species along
rivers by creating new habitat, displacing native species,
and providing conditions that suit the nonnative species’
life-history requirements (The ´ baud and Debussche 1991,
Pysˇek and Prach 1994). Therefore, some have argued
that regional flow-regime modification has facilitated the
spread of Tamarix (Shafroth et al. 1998, Stromberg et al.
2007). As many as a dozen species of Tamarix have been
introduced from southern Europe and Asia into western
North America over the past century and a half. Widely
introduced for horticulture, as windbreaks, and to
stabilize river banks, Tamarix (primarily the hybrid T.
ramosissima Ledeb. 3 T. chinensis Lour.) rapidly
naturalized and spread, now occupying an estimated
500000 (Robinson 1965, Everitt 1980, Gaskin and
Schaal 2002) to 1000000 ha (Brock 1994, Zavaleta
2000) in an area of North America ranging from Kansas
to coastal California and from northern Mexico to
northern Montana (Friedman et al. 2005).
The presence and spread of Tamarix has been
associated with a variety of contentious scientific and
social issues, ranging from the compromised health and
collapse of native riparian ecosystems to economic
implications about water use and water availability
along rivers in the western United States (Di Tomaso
1998). Perceptions and concerns have led to large-scale
efforts to salvage water and manage invasive species
through Tamarix removal (Graf 1981, Culler et al. 1982,
Zavaleta 2000). In the popular and peer-reviewed
literature, Tamarix has been implicated as a major
driver of environmental change: increasing salinity of
floodplain soil, depleting soil and groundwater and
reducing stream flow (attributed to increased evapo-
transpiration by Tamarix), increasing fuel loading and
fire intensity, displacing native species, reducing habitat
quality, and suppressing species diversity (Cleverly et al.
1997, Di Tomaso 1998, Pataki et al. 2005). Despite these
concerns, empirical data directly linking Tamarix to
such ecosystem changes are surprisingly sparse or
absent, and conclusions about cause and effect remain
largely speculative. The presumed ecological and eco-
nomic costs associated with the presence of Tamarix
(Zavaleta 2000) have resulted in an expenditure of tens
of millions of U.S. dollars at local, state, and federal
levels annually for Tamarix control and ‘‘restoration,’’
aimed primarily at removing Tamarix to salvage water,
restoring native plant assemblages, and supporting the
fundamental goal of invasive species management (Hart
et al. 2005, Shafroth et al. 2005, U.S. House of
Representatives 2005–2006).
In this study, our goal was to disentangle the poorly
understood relationships between flow regulation, Tam-
arix spread, and Populus decline. We tested the
hypotheses that the dominance of Tamarix relative to
Populus (and by implication other native riparian
species) is a function of the degree of hydrologic
alteration caused by dams and that successful recruit-
ment of Tamarix is a function of flow alteration (Fig. 1).
Despite the widely held belief that altered hydrologic
DAVID M. MERRITT AND LEROY POFF136
Ecological Applications
Vol. 20, No. 1

Page 3

regimes are a principal mechanism facilitating domi-
nance shifts toward Tamarix, supporting evidence has
been based either on small-scale experiments (Shafroth
et al. 1995, Sher et al. 2000, Beauchamp and Stromberg
2007) or on examination of dammed and undammed
river segments in a particular area (Stromberg et al.
2007). Therefore, to test this hypothesis we conducted an
extensive, regional analysis along 13 rivers in five states
with different degrees of flow alteration by dams and
other water infrastructure. Previous regional analyses
have been either small in extent and limited to
intermittent vs. perennial classes of streams (Stromberg
et al. 2007) or geographically extensive but designed to
examine the relationship between Tamarix dominance
with flow-regime alteration (Friedman et al. 2005). Our
approach enabled us to conduct a ‘‘natural experiment’’
across a wide range of conditions, capturing both
catchment and local-scale environmental variables along
a gradient of flow regulation. We expected evidence in
support of our hypothesis to manifest as increasing
recruitment success and abundance of Tamarix along a
gradient of intensifying flow regulation. Similarly, we
predicted Populus would have greater success under the
least altered conditions, as has been suggested elsewhere
(Stromberg et al. 2007). One goal of this work is to gain
insight into the possible efficacy of active stream flow
management to control the abundance of Tamarix and
to restore native Populus on dam-modified rivers in the
arid and semiarid western United States.
METHODS
Sites and geographical extent
Major perennial rivers from each of eight arid and
semiarid ecoregions west of the 100th meridian and east
of the Sierra Nevada Mountains were selected for
measurement of Tamarix and Populus recruitment and
abundance (Fig. 2). Perennial rivers of varying degrees
of flow regulation were selected in the following
ecoregions (sensu Bailey 1995): American semidesert
and desert (includes the Mojave, Colorado, and
Sonoran Deserts); Arizona–New Mexico semidesert
open woodland–coniferous forest–alpine meadow; Chi-
huahuan semidesert; Colorado Plateau semidesert;
intermountain semidesert and desert; intermountain
semidesert; southern Rocky Mountain steppe?open
woodland?coniferous forest?alpine meadow; and
southwest plateau and plains dry steppe and shrub.
The region surveyed includes the most arid portion of
North America. The sites that we surveyed spanned an
elevation gradient of 79–1989 m above mean sea level
and ranged in annual precipitation from 102 to 424 mm.
As in many other arid and semiarid regions of the world,
the perennial desert streams sampled derive flows
primarily from snowmelt in mountainous headwaters,
although in the southernmost ecoregions summer
monsoons also influence late-season flows. Most of
these rivers flow from narrow, confined mountain
channels out onto broad alluvial valleys. Segments of
each river were selected to represent varying levels of
flow modification ranging from nearly free-flowing (the
upper Colorado, upper Gila, San Miguel, and Yampa
Rivers), to moderately regulated by dams (upper
Humboldt and upper Pecos Rivers), to heavily regulated
(lower Colorado, middle Rio Grande, Little Colorado,
and lower Pecos Rivers; Table 1). Once a sampling
segment was identified (based upon degree of regulation,
accessibility, and permission for access), each was
systematically sampled by randomly choosing an initial
point and evenly spacing sampling reaches along the
segment. In this study, a river segment is a length of
FIG. 1.
allowing Tamarix to occupy new space which can lead to Tamarix dominance of habitat formerly dominated by native species.
Alternatively, hydrologic alteration may facilitate the establishment of Tamarix by creating conditions to which Tamarix is better
suited than native species, leading either directly to dominance by Tamarix or to the initiation of autogenic processes (changes in
soil properties, litter accumulation, soil moisture or groundwater levels, light availability, and so forth) that inhibit native species
and lead to dominance. A third alternative pathway to Tamarix dominance is that inherent traits of Tamarix, rather than some
connection to hydrologic alterations, facilitate its spread and initiate autogenic processes that inhibit native species and lead to
Tamarix dominance.
Possible pathways leading to Tamarix dominance. Hydrologic modification may lead to the decline of native species,
January 2010137RIVER ALTERATION AND RIPARIAN DOMINANCE

Page 4

stream separated from another by a dam or a major
tributary.
Sampling of recruitment and abundance
of Tamarix and Populus
We inventoried recruitment and abundance of Tam-
arix and Populus along 64 river reaches on 13 perennial
rivers in the western United States (Fig. 2). Inventories
were conducted during the growing seasons in
1999?2003. At each sampling reach, a 200-m transect
was established parallel to the river channel at the
margin of the active channel. Locations of candidate
200-m transects were systematically selected by dividing
each river segment into 0.5-km intervals on 1:300000
scale topographic maps. Subsampling of candidate sites
(every other site) was necessary in some cases due to
time constraints. In the field, the beginning point of the
200-m reaches was chosen to be representative of the
larger stream segment. Along each reach, we collected
recruitment and abundance data for Tamarix and
Populus.
Recruitment was determined by census of 2–5 year-
old saplings of Tamarix and Populus along both river
margins along the 200-m reach; a single presence
constituted positive recruitment. Ages were verified by
excavating and taking a cross section of the stem of the
oldest individual in the plot and counting annual growth
rings. In cases when Populus was present but not
abundant, rather than excavating individuals, we made
the assumption that saplings 5 mm or more in diameter
at their base were within the 2?5 year range.
Abundance of the two species at the segment scale was
based on estimates derived from a census of the
presence?absence of adults (.5 years old) at 1-m
intervals along the 200-m transect. We recorded the
presence of Tamarix or Populus within a 1 m wide belt
(positioned perpendicular to the 200-m transect) visu-
ally extended across the entire floodplain. Lengths of the
belts varied from 100 to 400 m depending on the width
of the valley bottom along the reach. Abundances of
Tamarix and Populus were calculated as the proportion
of occurrences in the 200 1 m wide belts (i.e., the
occurrence of at least one individual in each belt would
FIG. 2.
USGS stream flow gages are indicated with black and white symbols, dams are indicated by black lines normal to streams, and gray
dots indicate Tamarix sampling sites.
Study-site map showing the western United States. Only rivers included in the analysis are included in the diagram.
DAVID M. MERRITT AND LEROY POFF 138
Ecological Applications
Vol. 20, No. 1

Page 5

yield an abundance of 1.00). This did not provide an
aerial estimate of abundance, but rather an indication of
occurrence along river bottomlands (presence–absence
across the entire floodplain). Relative abundance of the
two species was calculated as the abundance of the
species plus one divided by the abundance of the other
species plus one. Relative abundance is a measure of
dominance.
Several local environmental variables were also
recorded at sampled sites. Soil grab samples (upper 15
cm) were collected in two randomly located points along
the 200-m transects; river morphology at the scale of the
reach was recorded and dominant land-use patterns
were noted (presence of recent fire, grazing, and
agriculture). In the laboratory, soil samples were
analyzed for pH by the soil paste method (Richards
1954). Salinity (electrical conductivity) and concentra-
tions of B, Naþ, Ca2þ, K, and Mg2þwere determined
using inductively coupled plasma (ICP) analysis on an
extraction of the soil paste. Sodium absorption ratio was
calculated using the formula Naþ/[0.5(Ca2þþ Mg2þ)]0.5
(Harron et al. 1983). Salinity of the catchment above the
sample site was estimated by examining the mean
salinity of grid cells within a 100-m buffer of all USGS
National Hydrography dataset for stream channels in
the watershed upstream of the sample reach (USGS
1999). The USGS National Hydrography data set
(NHD, 1:100000 scale; USGS 1999) was used for
stream coverage and the U.S. General Soil Map
(STATSGO) salinity data set was used to calculate
mean salinity in the upstream watershed (data available
online).6Estimates of mean annual precipitation for the
period 1970?2000 at each site was derived from PRISM
climatic data (Daly et al. 2001). Valley width through
the reach was measured from 1:300000 scale topograph-
ical maps and latitude and elevation were included in the
data set for each of the 64 reaches sampled.
Index of flow modification (IFM)
We obtained stream flow data from the U.S.
Geological Survey for the period of record available
from 26 gages on the 13 rivers nearest the sampling sites
(USGS 2002; Table 1). Dam locations, date of
construction, and purpose for construction were ob-
tained from the National Inventory of Dams database
(NID 2004). For each sampled river segment, we used
existing stream flow records to develop a composite
metric of the degree of flow modification compared to a
relatively unregulated (pre-dam) condition based on
biologically relevant flow metrics that we could extract
from the hydrologic records. We first summarized the
daily mean flow data for the period before and after dam
construction. The mean number of pre-dam flow records
was 31 years (range 12–111 years) and the mean number
of post-dam years of flow record was 49 years (range 21–
70 years). Dams were built between 1905 and 1984. We
fitted flood-frequency curves to instantaneous peak-flow
series for each river segment studied. Log Pearson Type
III distributions were fitted to series of annual instan-
taneous peak flow using methods outlined by the U.S.
Geological Survey (USGS 1981).
For each sampling site, we developed a profile of eight
biologically relevant hydrologic variables to characterize
alterations associated with river regulation. For regu-
lated systems, the date of upstream dam installation was
obtained and the pre-dam record compared to the post-
dam record. For relatively free-flowing river segments,
the entire hydrologic record was divided into two equal
time periods, and the absolute or percentage change
from the first half to the second half of the record was
calculated. The ‘‘modification’’ or change represented by
the difference in the two equal time periods represents
natural variation in the unregulated rivers and incorpo-
rates climatically and human-caused changes in flow
regime. Dams have a much larger effect on modifying
flow regimes over the 20th century than do decadal-scale
changes attributable to climate variability (Poff et al.
2007).
The biologically relevant hydrologic variables includ-
ed in this analysis represented key flow-regime compo-
nents. Magnitude and frequency of high and low flows
comprised six metrics based on percentage change in
spring flow (mean of April through June), summer flow
(mean of July through September), low flow (mean of
October through February flows), and the 2-, 10-, and
25-year recurrence-interval peak flows. Timing variables
included change in the number of days of minimum flow
(number of days shifted) and change in maximum flow.
FloodFreq software (version 4.1) was used to
calculate flood frequency (USGS 2002). All other flow-
related summary statistics were calculated using Indica-
tors of Hydrologic Alteration software (Richter et al.
1996). To see if there was an association between recent
flooding and Tamarix and Populus recruitment, we
determined the number of times that the 5-year
recurrence-interval flood had been exceeded in the 5-
year period 1998?2002 (Table 1).
Principal components analysis (PCA) was conducted
using the eight flow variables from the 26 gages
representing all of the study sites (Table 1). To better
represent the full range of altered flow regimes in rivers
in the western United States, we also included gage
records from 21 additional regulated rivers (not field
sampled) for which long-term stream flow data were
available (Table 2). The broken-stick rule was used to
determine the statistical significance of the principal
component axes (Jackson 1993): In this rule, eigenvalues
from the PCA are compared to eigenvalues from
randomly generated data. Only significant axes (P ,
0.05) were included in the calculation of index of
hydrologic modification.
We developed an index of flow modification (IFM) by
calculating the Euclidean distance of each observation
6hhttp://soildatamart.nrcs.usda.govi
January 2010 139RIVER ALTERATION AND RIPARIAN DOMINANCE

Page 6

(study reach) from the centroid of the significant PCA
axis scores of relatively unregulated rivers (no major
dams) for the eight hydrologic metrics (Fig. 3). Because
hydrologic changes on unregulated systems were rela-
tively small between the earlier and later periods, the
unregulated streams (upper Colorado, upper Gila, San
Miguel, and Yampa Rivers) clustered together in the
ordination. The more significant the change in hydro-
logic alteration of a particular reach, the further the
observations lie from the centroid of unregulated sites.
IFM was rescaled from 0 to 1.0 using the most flow-
modified segment (Bill Williams River) as the maximum.
IFM was determined for each of the 64 reaches sampled
for Tamarix and Populus in the field. Multiple reaches
located along segments between dams were assigned
equal IFM values.
Statistical analysis
We developed models to test Tamarix and Populus
response to flow alteration (IFM) in terms of three
response variables: recruitment, abundance, and domi-
nance. For recruitment we used logistic regression to test
the hypotheses that Tamarix recruitment increases as a
function of flow modification and Populus recruitment
decreases as a function of flow modification. After fitting
fixed-effects models and determining that models were
significant, we accounted for the lack of independence
between sites within a river or within a river segment in
order to more accurately estimate variance components
(SAS Institute 2004). We added the random effects
RIVER and SEGMENT nested within RIVER in the
logistic regression models and compared the fixed-effects
and mixed models using corrected Akaike’s information
criterion (AICc; Burnham and Anderson 2002).
For the abundance and dominance response variable,
we first used locally weighted regression (LOESS) as an
exploratory approach (Cleveland et al. 1988, Cohen
1999) to determine the form of the relationship between
IFM and the abundance of Populus and Tamarix. We
then used linear or nonlinear regression as appropriate
to test the hypothesis that Tamarix and Populus
abundance at the sites is related to flow alteration. For
each analysis of abundance, we first fitted fixed-effects
models with IFM (adding quadratic terms as appropri-
ate) and then added the random effect RIVER followed
by the SEGMENT nested within RIVER term if it
improved the model, using AICcas the criterion (SAS
Institute 2004).
We were also interested in determining which
combinations of the measured environmental variables
at the site and segment scale accounted for variability in
recruitment and abundance of Tamarix and Populus.
TABLE 1.
American rivers.
Attributes of U.S. Geological Survey stream flow gages that represent each of the sampled segments in western North
River (gage no.)
Catchment
area (km2)
Elevation
(m)
Sites at
gage no. Ecoregion?
Dam
constructed
(year)
Five-year
peak
exceeded?
IFM
Gunnison (9128000)
Rio Grande (8332010)
Rio Grande (8358400)
Rio Grande (8362500)
Pecos (8383500)
Pecos (8384500)
Colorado (9095500)
Colorado (9163500)
Colorado (9429100)
Dolores (9169500)
San Miguel (9177000)
Dolores (9180000)
Yampa (9251000)
Little Colorado (9384000)
Little Colorado (9388000)
Little Colorado (9394500)
Little Colorado (9402000)
Gila (9430500)
Verde (9504000)
Verde (9506000)
Verde (9511300)
Humboldt (10327500)
Humboldt (10335000)
Truckee (10351700)
14576
49806
71899
79514
10282
11370
20849
46212
471880
5242
3882
11862
19839
1823
2608
20903
68526
4828
8091
12030
17132
31649
41697
4341
1564
1440
1358
1262
1314
1262
1467
1091
79
1506
1524
1269
1783
1832
1694
1563
1213
1482
1067
876
402
1329
1240
1255
2
3
1
1
2
3
2
4
ISDD
CPSD
CSD
CSD
SPPS
SPPS
ISDD
ISDD
ASDD
ISDD
ISDD
ISDD
SRMS
CPSD
CPSD
CPSD
CPSD
CSD
CPSD
CPSD
ASDD
ISDD
ISDD
ISDD
1966
1973
1938
1938
1979
1937
NA
1966
1938
1984
NA
1984
NA
NA
1912
1939
1939
NA
NA
NA
1939
NA
1936
1905
0
0
0
0
0
0
0
0
1
0
0
0
1
0
0
0
0
0
0
0
1
1
1
0
0.18
0.43
0.28
0.25
0.18
0.53
0.07
0.05
0.18
0.21
0.07
0.12
0.04
0.19
0.59
0.53
0.15
0.12
0.22
0.09
0.16
0.12
0.08
0.31
14
2
1
1
3
2
3
3
2
1
2
1
2
5
2
2
Notes: Index of flow modification (IFM) is calculated by examining pre- and post-impact change in eight biologically relevant
flow attributes. Elevation is above mean sea level. Values reported for Tamarix and Populus recruitment, abundance, and
dominance are the means for sites along each segment. Abundance is a proportion; recruitment is a probability; and dominance is a
ratio of abundance of one species relative to abundance of the other.
? Ecoregion abbreviations are ISDD, intermountain semidesert and desert; CPSD, Colorado plateau semidesert; ASDD,
American semidesert and desert; CSD, Chihuahuan semidesert; SRMS, southern Rocky Mountain steppe?open wood-
land?coniferous forest?alpine meadow; and SPPS, southwest plateau and plains dry steppe and shrub.
? Number of times the five-year recurrence-interval flow occurred within the five years prior to sampling.
DAVID M. MERRITT AND LEROY POFF 140
Ecological Applications
Vol. 20, No. 1

Page 7

First, we constructed fixed-effects models using stepwise
selection in both logistic regression (for recruitment) and
linear regression (for abundance). Variables included the
following site attributes: latitude, longitude, drainage
area, elevation, valley width, mean annual precipitation,
estimated salinity from STATSGO maps. The following
soil anions and cations were included in model selection:
CO3, HCO3, NO3, NO3N, and SO4. Electrical conduc-
tivity (EC) was highly correlated with B, Ca, K, Mg, Na,
Cl; so only EC was included in model selection. Sodium
absorption ratio and pH were also included. We also
included the percentage change in pre- to post-flow
regulation percentage for April to June flow, July to
September flow, and October to February flow, as well
as the 2-, 10-, and 25-year recurrence interval flows and
the change in the number of days in minimum and
maximum flows. After selecting the most parsimonious
model, we entered the random effects terms and
removed fixed effects that were insignificant after
accounting for variation due to random effects.
SAS/STAT version 9.1 was used to perform all
statistical analyses (SAS Institute 2004). Adjusted R2
model selection in REG procedure was used to select
variables to be included in mixed linear models; LOESS
procedure was used for locally weighted regression;
NLIN procedure was used to fit nonlinear models; and
stepwise selection in LOGISTIC procedure was used to
select variables for inclusion in logistic regression
models. MIXED procedure was used for mixed linear
modeling, and GLIMMIX procedure was used for
mixed logistic regression modeling (Littell et al. 1996).
Independent variables were transformed to natural
logarithm or square root when necessary to more closely
meet the assumptions of normality and homogeneity of
variance.
RESULTS
Index of flow modification
The first three axes of the PCA were significant (P ,
0.05, broken-stick model) and cumulatively accounted
for 74% of the variation in the hydrologic data. Axis 1
represents increases in flood magnitudes (2-, 10-, and 25-
year recurrence interval flows) and earlier minimum
flow. Axis 2 represents an increase in summer flow, a
reduction in 10- and 25-year flows, and an earlier
minimum and maximum flow. Axis 3 represents elevated
summer flow and later maximum and minimum flows
(Fig. 3).
Relationships between IFM and Populus and Tamarix
recruitment and abundance
Tamarix recruitment was negatively associated with
flow modification but not significantly so (rescaled R2¼
0.07, likelihood ratio (LR) v2¼ 3.1, P ¼ 0.08). The
probability of recruitment was highest (up to 0.8) along
unaltered stream segments and declined as a function of
IFM (Fig. 4). Inclusion of a quadratic term did not
improve model fit, nor did inclusion of random effects
(RIVER and SEGMENT nested within RIVER).
Tamarix abundance (proportion of mature individu-
als occurring in 200-m transects) was significantly higher
on average (0.75) than Populus (0.26) over the 64 sites.
TABLE 1. Extended.
RecruitmentAbundance Dominance
Tamarix PopulusTamarix PopulusTamarixPopulus
1.00
1.00
1.00
0.98
0.95
0.92
0.77
0.75
1.00
0.85
0.60
0.98
0.09
0.00
0.37
0.94
0.90
0.30
0.69
0.01
0.84
0.61
0.76
0.36
0.28
0.25
0.00
0.40
0.30
0.03
0.74
0.68
0.00
0.30
0.75
0.00
0.75
0.21
0.00
0.12
0.00
0.85
0.31
0.10
1.00
0.00
0.04
0.00
0.50
1.00
0.00
1.00
1.00
0.67
1.00
1.00
1.00
1.00
1.00
1.00
1.00
0.00
0.33
0.67
0.50
1.00
0.00
1.00
0.50
0.60
1.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
1.00
0.25
0.00
0.00
1.00
0.00
1.00
0.00
0.00
0.00
0.00
1.00
0.00
1.00
0.50
0.00
0.00
0.50
1.58
1.60
2.00
1.41
1.52
1.86
1.04
1.05
2.00
1.42
0.91
1.98
0.65
0.85
1.37
1.78
1.90
0.70
1.29
0.92
0.92
1.61
1.70
1.36
0.64
0.63
0.50
0.71
0.67
0.54
1.00
0.96
0.50
0.70
1.09
0.51
1.62
1.21
0.80
0.58
0.53
1.42
0.77
1.09
1.09
0.66
0.60
0.76
TABLE 2.
flow gages used to get full representation of modified river
flow regimes throughout the study region (no sites sampled
along these segments).
Attributes of U.S. Geological Survey (USGS) stream
River (USGS gage no.)
Catchment
area (km2)
Elevation
(m) IFM
Strawberry (9285000)
Truckee (10337500)
San Pedro (9471000)
Carson (10312000)
Sevier (10191500)
Sevier (10217000)
Gila (9431500)
Gila (9519500)
South Platte (6710000)
San Juan (9355500)
White River (9306500)
Bill Williams (9426000)
Arkansas (7099400)
Arkansas (7109500)
Arkansas (7123000)
Arkansas (7130500)
Arkansas (7138000)
Green (9217000)
Green (9315000)
Pecos (8396500)
Colorado (9380000)
111
1313
3196
3372
6322
12745
7327
128592
7948
8443
10411
11999
12328
16386
31623
48989
61916
25226
116161
39626
289560
2243
1867
1205
1274
1804
1532
1247
230
1617
1723
1508
295
1445
1374
1232
1139
978
1847
1231
1003
947
0.50
0.42
0.17
0.43
0.07
0.44
0.14
0.67
0.31
0.29
0.15
0.99
0.17
0.18
0.26
0.71
0.35
0.05
0.10
0.29
0.37
Notes: Index of flow modification (IFM) is calculated by
examining pre- and post-impact (dams) change in eight
biologically relevant flow attributes. Elevation is above mean
sea level.
January 2010141RIVER ALTERATION AND RIPARIAN DOMINANCE

Page 8

FIG. 3.
western United States. Axis 1 represents increases in flood magnitudes (2-, 10-, and 25-year recurrence interval flows) and earlier
minimum flow. Axis 2 represents an increase in summer flow, a reduction in 10- and 25-year flows, and an earlier minimum and
maximum flow. The size of the circle in the main figure indicates the degree of hydrologic alteration, measured as the Euclidean
distance (in three-dimensional flow space) from the centroid of unregulated rivers (generally just right of center). Letters a?f within
or beside circles correspond to the hydrograph panels (a)–(f) from selected streams in the PCA. Vertical dashed lines on each
hydrograph demark the pre-dam and post-dam hydrographs or the first and second half of the flow record for which hydrologic
attributes were compared. For the two tracings in the hydrograph panels, the dark lines are a representative hydrographs in the
80th percentile, and the light lines are representative hydrographs for a 10th-percentile daily average flow.
First two principal component axes (PCA) of biologically relevant attributes of flow regime for 64 river segments in the
DAVID M. MERRITT AND LEROY POFF 142
Ecological Applications
Vol. 20, No. 1

Page 9

Tamarix abundance exhibited a curvilinear form in
locally weighted regression, peaking at IFM of ;0.3 and
dropping off at extreme values of IFM (Fig. 5). Using
linear regression with a polynomial term, the index of
flow modification was significantly but weakly related to
Tamarix abundance (R2¼ 0.12, P ¼ 0.02). Both linear
and quadratic terms were significant (P , 0.007).
Including random effects improved the model fit (AICc
reduced from 41 to 13) and the fixed-effects IFM and
IFM2remained significant (Type III test P , 0.05; Fig.
5).
Loess regression indicated that a power function was
the appropriate form to fit to the Populus abundance
data. This function had a significant fit (P , 0.0001) and
was chosen over the polynomial regression (Fig. 5).
Populus recruitment was significantly inversely related to
degree of flow modification (rescaled R2¼0.32, LR v2¼
13.7, P , 0.0002). Populus recruitment probability was
highest (.0.7) along the least altered stream segments
and dropped precipitously to ,0.1 at IFM of 0.2 (Fig.
4).
The quadratic relationship between Populus abun-
dance and IFM was also significant (R2¼ 0.30, P ,
0.0001; Fig. 5). Populus abundance was highest at very
low values of IFM and quickly declined as a function of
IFM (Fig. 5). Including the random RIVER effect in the
model improved the model (AICcreduced from 32.8 to
9.2), and the slope parameter was significantly different
than 0 (P ¼ 0.003). Adding the SEGMENT nested
within RIVER random effect further improved the
model (AICc¼?3.3), but the fixed-effect IFM was not
significant after accounting for the variation due to this
nested effect (Type III test P . 0.05).
Tamarix adults were present at 97% of the sites
sampled and showed only a 10% difference in abun-
dance between the lower and upper quantiles of IFM
(0.64 and 0.74, respectively). In contrast, Populus adults
declined in abundance from a mean of 0.45 to 0.26 in the
upper and lower quantiles. Tamarix was present and
abundant in the 16 least regulated segments (median
abundance of 0.71). Populus abundance was 0.72 along
these same segments. Along the nine most regulated
segments, Populus was absent and Tamarix had a
median abundance of 0.91. Tamarix abundance did
not account for a significant proportion of the variation
in Populus abundance at the sites (R2¼ 0.02, P ¼ 0.32).
Relationships between site, segment and catchment-scale
factors, and Populus and Tamarix recruitment
and abundance
Eleven percent of the variation in Tamarix recruit-
ment was explained by elevation of the site above mean
sea level (P ¼ 0.02), and the model was significant (LR
v2¼ 7.6, P ¼ 0.006). After adding the random effects to
the model, elevation remained significant. Probability of
Tamarix recruitment declined from a probability of 0.9
to ,0.4 over the range of elevations sampled (79–1989 m
above mean sea level). Mean Tamarix recruitment was
not associated with the occurrence of a 5-year recur-
rence-interval flood during the 5 years prior to sampling
FIG. 5.
function of index of flow modification (IFM). Loess regression
was used to determine the form of the functions (gray dotted
curves, smoothing parameter ¼ 1). The nonlinear power
function fitted to Populus was significant (solid black line, P
, 0.0001), indicating that Populus abundance declines as a
function of flow modification. The polynomial regression fitted
to Tamarix abundance was also significant (R2¼ 0.12, P ¼
0.02), indicating that Tamarix abundance is highest at
intermediate levels of flow modification. Solid circles represent
Tamarix, and the open circles show Populus.
Models of Populus and Tamarix abundance as a
FIG. 4.
recruitment as a function of index of flow modification (IFM;
shaded area indicates upper and lower 95% confidence
intervals). Likelihood ratio tests indicated that Populus
regeneration is indeed a function of flow modification (P ¼
0.0002), declining rapidly with increasingly altered flows.
Tamarix recruitment also declined as a function of flow
modification, but this relationship was not statistically signif-
icant (P ¼ 0.08). Index of flow modification is the Euclidean
distance (re-scaled from 0 to 1.0) of a stream segment from the
centroid of free-flowing (unregulated) streams in ordination
space. The variables included in the ordination were changes in
eight biologically relevant attributes of flow regime in pre-
regulation and post-regulation time periods. Please refer to
Methods: Index of flow modification (IFM) for details.
Logistic regression models of Populus and Tamarix
January 2010143 RIVER ALTERATION AND RIPARIAN DOMINANCE

Page 10

(t¼0.2, P¼0.9). Recruitment was present along 69% of
river segments that had not experienced a flood during
this time and 66% of segments that had flooded (5 of 24
gages experienced a 5-year flood in last 5 years; Table 1).
Forty-nine percent of the variability in adult Tamarix
abundance was accounted for by five variables and the
model was significant (R2¼ 0.49, P , 0.0001). Tamarix
was more abundant along higher order streams (larger
watershed area, P¼0.03), sites with higher soil pH (P¼
0.02), sites receiving less precipitation (P ¼ 0.003), and
along segments with elevated late summer flows (P ¼
0.004). Tamarix was less abundant along segments that
had elevated 10-year recurrence-interval floods in post-
dam time periods (P¼0.0002). For Populus, 66% of the
variability in recruitment was accounted for by four
variables, and the model was significant (LR v2¼28.4, P
, 0.0001). Populus recruitment was higher along
narrower streams (P ¼ 0.006), at lower elevations (P ¼
0.05), on stream segments with an earlier minimum flow
(P ¼ 0.02), and at sites with earlier peak flow relative to
pre-dam periods (P¼0.008). Mean Populus recruitment
was not associated with the occurrence of a 5-year
recurrence-interval flood during the 5 years prior to
sampling (t¼?0.1, P ¼ 0.9). Recruitment averaged 0.34
on segments that had not experienced a flood during this
time and 0.36 on segments that had flooded.
Forty-two percent of the variability in adult Populus
abundance was accounted for by four variables and the
model was significant (R2¼ 0.42, P , 0.0001). Populus
abundance was highest along streams in areas receiving
more precipitation (P ¼ 0.03), and areas with earlier
peak flow (P ¼ 0.04) and earlier minimum flow (P ¼
0.0007). Sites in which the numbers of flow reversals
were fewer also had higher Populus abundance (P ¼
0.04).
Dominance of Populus and Tamarix
The IFM explained 9% of the variation in Tamarix
dominance (relative abundance) across sites (R2¼ 0.09,
P ¼ 0.02; Fig. 6). Stepwise model selection of site and
hydrologic factors resulted in a model that included a
negative association with precipitation (P , 0.001).
Tamarix dominance was positively associated with
delayed date of minimum flow (P ¼ 0.001) and delayed
date of maximum (peak) flow (P ¼ 0.0004; R2¼ 0.55, P
, 0.0001). Dominance of Populus declined significantly
as a function of IFM (R2¼ 0.11, P , 0.008; Fig. 6).
Variables selected in stepwise analysis included precip-
itation (P , 0.0008), date of peak flow (P ¼0.001), and
date of low flow (P ¼ 0.01). Populus dominance scores
increased across sites as annual precipitation increased,
as peak flows occurred earlier, and as the date of
minimum flow occurred earlier.
DISCUSSION
An unresolved question is whether the decline in
populations of native species and degradation of many
Populus-dominated riparian ecosystems can be attribut-
ed in part to the presence of nonnative Tamarix or
whether these effects are confounded with, or funda-
mentally driven by, altered stream flow regimes (Fig. 1).
Our approach was to use an index of flow modification
(IFM), which integrates biologically meaningful attri-
butes of flow, to quantitatively examine the relationships
between recruitment, abundance, and dominance of two
riparian plant species. Our findings are consistent with
the literature indicating that flow alteration is associated
with the decline of native species via reduced recruitment
and degradation of river bottomland habitat (Rood and
Mahoney 1990, Rood et al. 2003). Certainly, a number
of factors other than flow regime influence Populus
distributions along rivers; however, our data suggest
that even slight modifications in flow regime (IFM .
0.2) dramatically reduce the probability of successful
Populus recruitment and dominance in riparian wood-
lands across the region (Figs. 4 and 5). Our findings also
suggest that native Populus forest decline would have
occurred independently of Tamarix naturalization and
spread along rivers throughout the western United
States. Furthermore, our findings shed light on and
challenge many of the widely held views about the role
of flow alteration in Tamarix establishment on western
river systems by highlighting the inextricable link
between flow alteration, native species decline, and the
response of species better suited to the new bottomland
conditions.
Stream flow alteration and the decline of native species
We found that Populus recruitment, abundance, and
dominance were all inversely related to flow modifica-
tion at the regional scale, corroborating other studies
that have shown flow alteration (and associated changes
in fluvial processes) to be a principal driving factor in
Populus decline along rivers in the western North
FIG. 6.
(relative abundance) as a function of index of flow modification
(IFM). Solid circles represent Tamarix and the open circles
show Populus.
Models of Populus and Tamarix dominance
DAVID M. MERRITT AND LEROY POFF 144
Ecological Applications
Vol. 20, No. 1

Page 11

America (Rood and Mahoney 1990, Braatne et al.
1996). Because Populus releases its short-lived seeds in
synchrony with the falling limb of the natural hydro-
graph, it is important that open sites with suitable
conditions at appropriate locations above the channel
are available during the relatively short period of
Populus seed dispersal (Fenner et al. 1984; reviewed in
Lytle and Merritt 2004). Such sites are created by peak
flows of sufficient magnitude to create overbank
deposits, build depositional bars, and/or to scour
existing vegetation creating bare, moist habitat for
seedlings to establish (Scott et al. 1996, Karrenberg et
al. 2002). If hydrologic conditions are suitable and
Populus seedlings establish, they are able to grow rapidly
and have been shown to compete well with other species
for water, light, and nutrients (Sher et al. 2000, 2002).
In addition to providing the regenerative habitat and
establishment conditions for Populus, flow regime and
associated fluvial processes govern soil moisture condi-
tions (seasonal patterns of soil water availability and
depth to groundwater) as well as biogeochemical cycling
(e.g., floodplain decomposition, nutrient processing),
which influence the extent and distribution of Populus-
dominated forests (Braatne et al. 1996, Molles et al.
1998, Merritt and Cooper 2000, Valett et al. 2005). Our
findings reinforce our understanding of the well-
documented relationships between flow regime and
Populus recruitment and abundance in western North
America. Understanding the degree of stream flow
naturalness (IFM . 0.8) and the key components in
the flow regime necessary to maintain Populus-dominat-
ed forests has implications for both river management
and restoration. These findings suggest that Populus-
dominated forests can be managed while allowing for
only a minimal level of flow alteration (Lytle and
Merritt 2004).
Stream flow alteration and nonnative species
Our analysis of this extensive regional data set does
not support the widely held notion that river regulation
of perennial streams has facilitated or promoted the
recruitment of Tamarix. We found Tamarix recruitment
along free-flowing streams to equal or exceed that
observed on heavily flow-regulated streams. Tamarix
recruitment tended to be highest on average along the
least regulated segments, but this trend was not
statistically significant; mean probability of Tamarix
recruitment across the entire gradient of flow alteration
sampled (IFM of ;0–0.6) was 0.70 (Fig. 4). By
contrast, the abundance of mature Tamarix was a
function of degree of flow alteration (Figs. 4 and 5). In
addition, Tamarix became increasingly dominant over
Populus along a gradient of flow regulation (Fig. 6).
Examining the life-history traits of Tamarix as they
relate to components of the flow regime helps to explain
this pattern in recruitment and dominance of Tamarix.
Recruitment.—At the scale of the western landscape,
Tamarix can be considered a habitat specialist that
recruits in moist areas such as river floodplains, lake and
reservoir margins, irrigation ditches, springs, and vernal
pools. Within these areas, however, Tamarix may be
more appropriately described as a habitat generalist.
Within the riverine landscape, Tamarix has a wide
regeneration niche in that its seedlings may become
established over a broader range of soil salinities,
textures, and soil moistures than Populus and other
riparian specialists (Shafroth et al. 1995, Cooper et al.
1999). In the field sites sampled in the current work,
Tamarix recruitment occurred on soils with pH ranging
from 7.2 to 8.9, electrical conductivities ranging from 0.4
to 135 mmhos/cm, sodium levels ranging from 14 to
22038 mg/L, and soil textures ranging from clay to
coarse sand?cobble.
Because Tamarix flowers and releases seed later and
for a longer duration than native disturbance-adapted
species such as Populus, it has been suggested that
delayed peak flows associated with the storage of runoff
in reservoirs may put Tamarix at a reproductive
advantage over native species (Shafroth et al. 1998,
Merritt and Wohl 2002, Stromberg et al. 2007). In
warmer climates, Tamarix individuals may flower
multiple times during the growing season and disperse
seeds over a period of several months, extending the
‘‘window of opportunity’’ for establishment and the
probability that some seeds will encounter suitable
germination and establishment sites at some time during
the growing season. Delayed peak flow might reason-
ably be expected to enhance Tamarix recruitment.
Nonetheless, the sufficient interannual variability in
the timing of Tamarix seed release and in the timing of
peak flows along relatively unregulated rivers appears to
allow for successful Tamarix recruitment even under
free-flowing conditions. Whereas the timing of peak flow
is critical for reproductive specialists like Populus, our
data suggest that Tamarix is less sensitive to the timing
of peak flows and is not greatly advantaged by delayed
or advanced timing of peaks in its naturalized range.
Because shade inhibits Tamarix establishment, the
loss of Populus overstory due to the flow-related
disturbance could open up space for Tamarix to invade
if subsequent fluvial processes and hydrology are
conducive to Tamarix establishment (Taylor et al.
1999). The presence of Populus forests may inhibit
Tamarix establishment through shading and competi-
tion for water, although mixed stands are not uncom-
mon (Stromberg 1998b; D. M. Merritt, personal
observations). The loss of Populus forest overstory may
also increase Tamarix productivity in established stands
due to increased sunlight and elevated transpiration
rates. In addition, Tamarix rarely recruits into the
understory of stands already dominated by Tamarix.
Therefore, once formerly active channel margins and
bottomland habitat are occupied by Tamarix or other
colonizing species, the rate of establishment in these
areas is likely to decline under regulated flow conditions
given the absence or reduction of periodic fluvial
January 2010145 RIVER ALTERATION AND RIPARIAN DOMINANCE

Page 12

disturbance. Recruitment in these areas is then restricted
to shade-tolerant, secondary successional species (Dew-
ine and Cooper 2008). For Tamarix to continue to
successfully recruit in such areas, occasional fluvial
disturbance is necessary to form open sites. On free-
flowing and semi-naturally flowing rivers, such habitats
are likely to be regularly created, providing periodic
opportunities for both native and nonnative species to
establish and maintain populations. Indeed, the higher
probability of Tamarix recruitment associated with
lower values of IFM may be explained by the more
frequent disturbance events and recruitment site crea-
tion on free-flowing rivers (Fig. 4).
Abundance.—In contrast to the recruitment results,
we found that there is a strong and significant
relationship between the degree of flow alteration and
abundance of mature Tamarix along the rivers sampled,
but this relationship is complex (Fig. 5). Tamarix is
abundant across a wide gradient of flow alteration, even
along rivers in which natural patterns of flooding are
maintained, but it is highest along reaches with
intermediate levels of hydrologic alteration (between
IFM of 0.2 and 0.4). Stromberg et al. (2007) found that
Tamarix was more abundant along intermittent com-
pared to perennial streams in the American southwest,
suggesting that extreme flow regulation (e.g., depleting
or dewatering of channels), drying climate, and ground-
water depletion and decline may facilitate Tamarix
abundance and/or lead to decline of native riparian
species (Baird et al. 2005, Stromberg et al. 2007).
Along perennial streams like the Colorado River, sites
experiencing intermediate fluvial disturbance have been
shown to be associated with higher frequency and higher
cover of Tamarix (Tiegs et al. 2005), a finding consistent
with our observation that abundance is highest at
intermediate values of IFM (Fig. 5). We suspect that
the lower Tamarix abundance along relatively free-
flowing rivers in our study is explained by biological
processes (e.g., competition with native species for light),
whereas the decline under severely regulated conditions
is explained by the interruption of physical processes
necessary for recruitment (fluvial disturbance and new
site creation).
Dominance.—There is no question that Tamarix has
been more successful on floodplains in the arid and
semiarid American southwest because of the comple-
mentary effects of its generalist traits: high fecundity,
long dispersal duration, tolerance of salinity, herbivory,
fire, flooding, and drought. It is well-established that
Tamarix is more tolerant of high soil salinity than many
native riparian species and is able to establish and persist
in sites with such characteristics (Busch and Smith 1995,
Shafroth et al. 1995, Arndt et al. 2004). Tamarix foliage
is not as palatable as many native species to livestock,
wildlife, and insects (Andersen and Cooper 2000, Sexton
et al. 2006, Mortenson et al. 2008). Release from
Eurasian pathogens and insects in its naturalized range
may play some role in Tamarix abundance relative to
native species in some circumstances (Lonsdale 1999,
Shafroth et al. 2005). Tamarix suffers lower mortality
than native species following wildfire and may quickly
resprout from its root system (Busch 1995, McDaniel
and Taylor 2003). Despite these confounding factors, we
found a strong relationship between flow regulation and
Tamarix dominance.
Tamarix is recognized as more drought tolerant than
native Populus, due both to the ability of Tamarix to
function well at lower soil water potential than Populus
(Tyree et al. 1994, Cleverly et al. 1997, Pockman and
Sperry 2000, Naumburg et al. 2005) and to the fact that
Tamarix can develop deep taproots (Gries et al. 2003).
Drought tolerance relative to native Populus enables
Tamarix to persist at sites though extended periods of
drought, periods of dewatering, conversion of streams
from perennial to intermittent (Stromberg et al. 2007),
or other water-related circumstances that do not permit
survival of native phreatophytic species. The implication
is that increasing aridity in the southwestern United
States due to climate change (Seager et al. 2007) could
disadvantage reproductive specialists and favor gener-
alists such as Tamarix along rivers and lead to the
formation and persistence of Tamarix-dominated ripar-
ian woodlands.
Because Tamarix recruitment and frequency were
high even along the least regulated rivers in this study,
we hypothesize that, all other factors held constant (e.g.,
patterns of human introduction), Tamarix would likely
have naturalized, spread, and established an important
role in riparian communities (coexisting with Populus
and other native riparian woodland species) in the
absence of dam construction, diversions, groundwater
pumping, and flow regulation along rivers in arid and
semiarid regions of western North America. However,
the combination of the decline of native species resulting
from altered flows and the pre-adaptation (and gener-
alist traits) of Tamarix to the habitat associated with
rivers and their floodplains in western North America
has contributed to Tamarix dominance in many western
riparian ecosystems. We conclude that the rapid spread
and persistent presence of Tamarix along western rivers
is not simply a consequence of flow regulation by dam
building in the 20th century.
Autogenic changes due to Tamarix presence
The potential for Tamarix to alter ecosystem structure
and function is broadly recognized (Vitousek et al.
1997), though not well demonstrated (but see Whitcraft
et al. 2008). Increased soil salinity, inhibited decom-
position/reduced organic material processing, altered
nutrient status, changes in fire-related disturbance
frequency and magnitude, and stream channel form
have all been associated with the establishment and
presence of Tamarix (Graf 1978, Molles et al. 1998,
Valett et al. 2005, Follstad Shah and Dahm 2008,
Whitcraft et al. 2008).
DAVID M. MERRITT AND LEROY POFF 146
Ecological Applications
Vol. 20, No. 1

Page 13

Salinity.—Because Tamarix is better adapted to high
soil salinity than Populus, its presence in sites with these
characteristics has resulted in poorly supported suppo-
sitions that Tamarix may be the cause of or a
contributor to elevated soil salinity (Busch and Smith
1995, Pataki et al. 2005). There is little empirical
evidence that Tamarix produces allelopathic chemicals
or that Tamarix actively concentrates salts in the soil at
levels that could inhibit establishment of native species
(but see Di Tomaso 1998, Laudenberger et al. 2006).
Populus and other glycophytic riparian species are
intolerant of highly saline soils and may suffer from
leaf necrosis, reduced transpiration, and inhibited water
and nutrient uptake in highly saline soils (Busch and
Smith 1993, Chen et al. 2003).
One factor associated with some forms of river
regulation is isolation of floodplains from rivers which
can result in chronic drying of soils and elevated soil
salinity. Soil salinization is a function of reduced
flooding, evaporation, agricultural practices, and elevat-
ed water table associated with river regulation (Jolly et
al. 1993). Indeed, sodium absorption ratio increased as a
function of IFM at our sites (R2¼ 0.11, F ¼ 3.3, P ¼
0.04). Soil salinity and other factors associated with flow
regulation may put Tamarix at a competitive advantage
over some native riparian dominants (e.g., Pluchea
sericea, Baccharis salicifolia, Populus fremontii, and
Salix gooddingii) under conditions of elevated soil
salinity and drought stress (Stromberg 1998a, Vander-
sande et al. 2001). Altered soil properties are likely to
contribute to Tamarix establishing dominance over
native species along regulated rivers (Busch and Smith
1995, Glenn et al. 1998, Lite and Stromberg 2005).
Fire.—It is recognized that invasive nonnative species
that modify disturbance regimes can have the greatest
impact to ecosystems (D’Antonio 2000). One means by
which Tamarix and other invasive plant species can
affect species composition and ecosystem properties is
through its effect on fire regimes in riparian areas.
Tamarix produces flammable biomass that can influence
the frequency and intensity of fires, causing the
mortality of fire intolerant native species, such as
Populus and several species of Salix (D’Antonio and
Vitousek 1992, Busch 1995). Indirect effects of river
damming such as increased fire frequency and intensity
due to fuel accumulation in drier riparian habitats may
explain the increase in Tamarix abundance observed at
several sites at the highest levels of flow modification in
this regional study (Fig. 5).
Channel narrowing.—There is a perceived linkage
between Tamarix spread and channel narrowing. This
has likely come, in part, from isolated periods of
opportunistic establishment along formerly active chan-
nel margins immediately following dam closure, as has
been reported along regulated rivers throughout the
western United States (Graf 1978, 2006, Everitt 1998,
Merritt and Cooper 2000). This vegetation may stabilize
banks and enhance sediment deposition when flooding
does occur (Allred and Schmidt 1999). Although
Tamarix may initially take advantage of exposed river
margins in the years following dam construction, the
opportunities for Tamarix establishment diminish
through time on regulated rivers (Graf 2006). Our data
strongly suggest that Tamarix is able to periodically and
opportunistically recruit in regulated rivers due to
occasional flood-related habitat creation rather than a
direct coupling of life-history attributes to regulated
conditions under the altered flow regime per se.
The generalist life-history traits of Tamarix and its
tolerance of high soil salinity, drought, and fire suggest
that it could be an ‘‘interactive’’ rather than ‘‘compet-
itive’’ species (sensu Macdougall and Turkington 2005).
Tamarix may be a passenger, merely establishing
dominance due to its wide ecological amplitude and
ability to tolerate conditions of both regulated and
unregulated systems. There are several examples of
Populus decline along regulated streams in the absence
of Tamarix, further suggesting independence in a direct
interspecific pattern (Rood et al. 1995, Scott et al. 1997).
There is little evidence suggesting that Populus and
Tamarix directly interact for limiting resources; howev-
er, is it likely that when water is limiting, Tamarix may
exacerbate stress to native species through lowering
groundwater levels and/or desiccating soils.
Flow regime and implications for Tamarix control
and river restoration
There is discussion throughout the western United
States and worldwide about implementing prescribed
floods to hinder nonnative species establishment or to
prevent spread along flow-regulated rivers (Merritt and
Wohl 2002, Shafroth et al. 2005, Tiegs et al. 2005,
Stromberg et al. 2007). It has been suggested that
restoring appropriately timed flooding might inhibit
Tamarix through interrupting dispersal and establish-
ment (Merritt and Wohl 2002, Harms and Hiebert 2006,
Beauchamp and Stromberg 2007). We found no
significant repression of Tamarix recruitment under
even the most ‘‘natural’’ flow regimes examined (Fig.
5). Instead we found the highest recruitment of Tamarix
at the lowest levels of flow alteration. We attribute this
pattern to the characteristics of the regeneration niche of
Tamarix. Occasional flow-related disturbances create
bare patches that are more ‘‘invasible’’ than older
patches, providing more opportunities for well-dispersed
colonizing species like Tamarix to become established
along free-flowing river margins compared to more
static, stable habitats associated with regulated flows
(Planty-Tabacchi et al. 1996).
Flood-driven recruitment of Tamarix has been
reported for rivers on the Colorado Plateau (Cooper et
al. 2003, Birken and Cooper 2006) and in the American
semidesert (Whiteman 2006). During our surveys, we
aged (using growth ring counts) several Tamarix stems
along the relatively free-flowing San Miguel River in the
upper Colorado River basin and found that a quite
January 2010 147RIVER ALTERATION AND RIPARIAN DOMINANCE

Page 14

extensive cohort had become established following a
year with exceptionally high peak flows and a relatively
wet summer (1983). Along the free-flowing Yampa
River in the upper Colorado River basin, Cooper et al.
(2003) found that Tamarix establishment had occurred
regularly over the past half-century despite relatively
natural flow regimes. Cooper et al. (2003) also noted
that Tamarix recruitment occurred regularly along the
free-flowing pre-dam Green River and occasionally
(associated with floods) following dam construction.
Another study along the Green River found that 90% of
Tamarix recruitment occurred in years with floods
exceeding the 2.5-year recurrence-interval flow in both
pre-dam and post-dam periods (Birken and Cooper
2006). At our sites, Tamarix recruitment exceeded that
of Populus across a broad range of flow alteration,
including relatively free-flowing conditions (IFM from
0.04 to 0.59). Tamarix has many adaptations to arid
riverine environments and natural flow regimes in its
Eurasian home-range (Gaskin and Schaal 2002). Indeed,
declines of Tamarix populations on rivers in northern
China have been attributed to flow alterations that
include reductions in peak flow (Si et al. 2007).
Given the fact that floods create regenerative habitat
for Tamarix (as well as for native species) and that
successful Tamarix recruitment was no lower along the
least regulated streams surveyed in this work, we believe
care should be taken to develop realistic expectations for
the outcome of managed floods downstream from dams.
Managed flooding, unless catastrophic or extremely
restrictive, is unlikely to significantly hinder Tamarix
establishment. This has led to some cautionary advice
about controlled releases of water under circumstances
that could facilitate Tamarix establishment (Birken and
Cooper 2006). However, we argue that the benefits of
using flows to restore native species arguably outweigh
the cost of accommodating nonnative species in the
process because of the many other ecosystem benefits
associated with flooding. For example, flooding is vital
to the maintenance and restoration of riparian vegeta-
tion that is governed by patterns of natural disturbance,
microbial activity, decomposition and nutrient cycling,
seed delivery, groundwater recharge, and a host of other
flow-driven processes (Junk et al. 1989, Molles et al.
1998, Valett et al. 2005, Follstad Shah and Dahm 2008).
Furthermore, based on our extensive survey, overbank
flooding benefits Populus establishment and abundance
and is therefore likely to reduce the likelihood of
Tamarix becoming dominant over native species in
riparian plant communities.
Our findings strongly suggest that restored flooding
alone will not serve to remove or eliminate Tamarix
from western riparian ecosystems; however, high flows
are essential for the recruitment and persistence of
Populus in these ecosystems. Given that mixed Tamar-
ix?Populus stands have high ecological value (Strom-
berg 1998b), a reasonable restoration goal for Tamarix-
dominated ecosystems would be to proactively manage
for Populus recruitment and maintenance using pre-
scribed flows. Mechanical removal of dense Tamarix
stands also may be necessary to open up space for native
species, as older stands are relatively resistant to
scouring (Friedman and Auble 1999), but control of
Tamarix alone is unlikely to result in recovery flood-
dependent Populus in the absence of altered flow
management or natural flooding (Nagler et al. 2005).
If the management objective is to recover Populus and
other native riparian species, the question of what
specific attributes of the flow regime (e.g., timing,
magnitude, seasonal variation, rates of change) should
be restored arises. Some have suggested that high flows
precisely timed to the peak of Populus seed release could
maximize Populus recruitment success and inhibit later
dispersing Tamarix (Levine and Stromberg 2001).
However, the seed-release period of Tamarix and
Populus coincide for a portion of the growing season
in most regions (Shafroth et al. 1998, Cooper et al.
1999); thus managing flow by rigidly restricting the
hydrograph for the sole purpose of inhibiting Tamarix is
problematic and potentially harmful to Populus and
other native species. Restricting the hydrograph to one
single peak or short duration to advantage Populus
would likely prove detrimental to many other native
species that require gradual drawdown and shallow
groundwater for root elongation during early establish-
ment (Mahoney and Rood 1998, Amlin and Rood 2002,
Karrenburg et al. 2002). Further, many native riparian
species that benefit from or require occasionally late
peak flows or drought years would likely be adversely
affected by restricted flows designed to accommodate
only Populus and to inhibit Tamarix. We would argue
for a more holistic approach to riparian flow manage-
ment in which managed flows are designed within the
natural range of variability in magnitude, timing, and
duration of historic peaks, to the benefit of the broader
riparian community while ensuring some Populus
recruitment.
The debate continues as to the relative ecological
‘‘value’’ and function of Tamarix-dominated, mixed
native and nonnative, and pure native-dominated
riparian stands. Some studies suggest that mixed stands
of Tamarix and native vegetation rival completely
native-dominated stands in richness of birds, insects,
and rodents (Ellis et al. 1997, Andersen and Nelson
1999, Sogge et al. 2008, van Riper et al. 2008).
Stromberg (1998b) suggests that nonnative Tamarix
stands may be functionally equivalent to native Populus-
dominated stands. Managing flows to restore functional
attributes of riparian areas may prove a preferred
alternative to stream flow management aimed at one
or a few species.
CONCLUSION
The region we sampled in this study is among the
most arid in North America, yet it is experiencing very
rapid human population growth rates and an associated
DAVID M. MERRITT AND LEROY POFF 148
Ecological Applications
Vol. 20, No. 1

Page 15

continuing demand for development of water infrastruc-
ture (Barnett et al. 2008). Already heavy regional levels
of river regulation are expected to increase over the next
several decades with potentially serious impacts on
riparian ecosystems (Graf 1999, Lytle and Merritt 2004).
Depleted streamflow can result in decreased alluvial
groundwater levels capable of inducing drought stress
on riparian vegetation; reduced frequency of high-
magnitude flood events can prevent the creation of
regenerative habitat; and, altered timing of flow can
decouple flows during key phenological times in plant
development and growth (Nilsson and Svedmark 2002).
All streams inventoried have some degree of human
water development in the form of groundwater pump-
ing, diversions, and dams, and most are subjected to
each at some point throughout their drainage networks
and demands for water will continue to intensify
(Barnett et al. 2008).
Tamarix was present along all rivers sampled in this
study. We did not find that Tamarix was more abundant
along perennial rivers with lower total mean annual
discharge; however, we did find significant relationships
between the degree of stream flow alteration (IFM) and
the relative success of native Populus and nonnative
Tamarix. Understanding patterns and relationships
between flow alteration and the recruitment and
abundance of nonnative species is critical to informing
decision making in river management, especially in this
era of environmental flow prescription and experimental
flooding to attain ecological goals. Our findings have
implications for how we set objectives for the restoration
of flow-altered streams, for what sorts of yields we
should realistically expect on investments in conducting
prescribed flow releases, and for what kinds of
interventions may be necessary to control and manage
nonnative plant species along western North American
rivers where it is desirable. Because plant establishment,
vegetation change, and shifts in dominance may be a
multi-staged process, there are several mechanisms that
explain the invasion success of Tamarix into riparian
ecosystems of the American southwest. Our findings
suggest that implicating Tamarix as the sole driver in the
decline of western riparian ecosystems is too simplistic.
Flow alteration has likely been the principle cause of the
decline of native-dominated riparian ecosystems in
western North America, and it may have facilitated
dominance of some riparian communities by Tamarix
(Fig. 1). The simultaneous collapse of Populus-dominat-
ed ecosystems both within and outside of the naturalized
range of Tamarix suggests that such collapse would
likely have occurred in the western United States in the
absence of Tamarix. Furthermore, the success of
Tamarix along the entire gradient of flow alteration
suggests that Tamarix would be an important compo-
nent of western riparian ecosystems even in the absence
of water development. Our findings clearly suggest that
active management for natural hydrograph characteris-
tics downstream from dams is likely to decrease the
likelihood of Tamarix establishing dominance over
native species in riparian ecosystems. Such active
management will also provide a range of ecological
benefits that removal of invasive species alone will not
accomplish.
ACKNOWLEDGMENTS
The authors thank Alexander Van Alen, John Randall, and
David Graf for field assistance. Thanks to the land owners and
managers who allowed access for field sampling. Thanks to
Daniel Auerbach and Christopher Peltz for constructive
comments on drafts of this manuscript. We appreciate David
Theobald and John Norman’s assistance with analysis of the
STATSGO data. Thanks to Christopher Peltz for assistance
with the figures. We appreciate constructive comments from
Juliet Stromberg and an anonymous reviewer, which improved
this paper. This work was funded by the David H. Smith
Conservation Postdoctorate Fellowship administered through
The Nature Conservancy. This work was also partially funded
by the U.S. Bureau of Reclamation. Thank you is extended to
David H. Smith (deceased) and his family for generous support
of conservation research.
LITERATURE CITED
Aguiar, F. C., M. T. Ferreira, and I. Moreira. 2001. Exotic and
native vegetation establishment following channelization of a
western Iberian river. Regulated Rivers: Research and
Management 17:509–526.
Allred, T. M., and J. C. Schmidt. 1999. Channel narrowing by
vertical accretion along the Green River near Green River,
Utah. Geological Society of American Bulletin 111:1757–
1772.
Amlin, N. M., and S. B. Rood. 2002. Comparative tolerances of
riparian willows and cottonwoods to water-table decline.
Wetlands 22:338–346.
Andersen, D. C., and D. J. Cooper. 2000. Plant?herbi-
vore?hydroperiod interactions: effects of native mammals
on floodplain tree recruitment. Ecological Applications 10:
1384–1399.
Andersen, D. C., and S. M. Nelson. 1999. Rodent use of
anthropogenic and ‘‘natural’’ desert riparian habitat, lower
Colorado River, Arizona. Regulated Rivers: Research and
Management 15:377–393.
Arndt, S. K., C. Arampatsis, A. Foetzki, X. Y. Li, F. J. Zeng,
and X. M. Zhang. 2004. Contrasting patterns of leaf solute
accumulation and salt adaptation in four phreatophytic
desert plants in a hyperarid desert with saline groundwater.
Journal of Arid Environments 59:259–270.
Bailey, R. G. 1995. Description of the ecoregions of the United
States. Second edition. USDA Forest Service Miscellaneous
Publication 1391. USDA Forest Service, Washington D.C.,
USA.
Baird, K. J., J. C. Stromberg, and T. Maddock. 2005. Linking
riparian dynamics and groundwater: an ecohydrologic
approach to modeling groundwater and riparian vegetation.
Environmental Management 36:551–564.
Barnett, T. P., D. W. Pierce, H. G. Hidalgo, C. Bonfils, B. D.
Santer, T. Das, G. Bala, A. W. Wood, T. Nozawa, A. A.
Mirin, D. R. Cayan, and M. D. Dettinger. 2008. Human-
induced changes in the hydrology of the western United
States. Science 319:1080–1083.
Beauchamp, V. B., and J. C. Stromberg. 2007. Flow regulation
of the Verde River, Arizona encourages Tamarix recruitment
but has minimal effect on Populus and Salix stand density.
Wetlands 27:381–389.
Birken, A. S., and D. J. Cooper. 2006. Processes of Tamarix
invasion and floodplain development along the lower Green
River, Utah. Ecological Applications 16:1103–1120.
January 2010149 RIVER ALTERATION AND RIPARIAN DOMINANCE