Water Resources Research

Channel response in a semiarid stream to removal of tamarisk and Russian olive



[1] We report observed short-term (3 years) channel adjustment in an incised, semiarid stream to the removal of invasive plants, tamarisk (Tamarix spp.) and Russian olive (Elaeagnus angustifolia) by (1) removing the above-ground portion of the plant (cut-stump method) and (2) removing the entire plant (whole-plant method). The stream flows through Canyon de Chelly National Monument in Arizona, USA., draining an ∼1500 km2 catchment. Average channel width is 13 m; average thalweg depth is 2–3 m, although channel banks exceed 8 m locally. Channels adjusted primarily through widening, with significantly larger changes occurring in whole-plant removal reaches; however, neither plant removal method elicited large-scale bank destabilization, and the channels remained entrenched. Particular site conditions limiting large-scale destabilization include the absence of sufficient streamflow magnitudes, the presence of clay layers at the bank toe, the remaining presence of native vegetation, and the entrenched morphology. Our findings serve as a cautionary note regarding the temporal and spatial variability in channel response to invasive plant removal and underscore the importance of considering site-specific conditions in future restoration projects that include invasive plant removal.

1. Introduction

[2] Invasive species management is of global concern [Vitousek et al., 1997; Zavaleta, 2000], and riparian areas are particularly vulnerable [Planty-Tabacchi et al., 1996; Levine, 2000]. River restoration through invasive plant removal has gained increasing popularity, as illustrated by HR 2720, legislation in which the U.S. Department of the Interior targets expenditures of $80 million for management of tamarisk (Tamarix ramosissima Ledebour, T. chinensis Loureiro, and hybrids), the third most prevalent riparian woody species in the United States [Gaskin and Schaal, 2002; Friedman et al., 2005]. Objectives of invasive riparian plant removal can include increased esthetic value of the landscape, water salvage, and restoration of ecological and geomorphic functions. Meeting these objectives, however, requires considering the plant removal methods that are most appropriate to the specific site conditions, particularly in stream channels characterized by a dynamic morphology. Streams in the southwestern United States, for example, exhibit dramatic vertical instability and width changes that can exceed an order of magnitude within a few decades [Schumm and Hadley, 1957; Cooke and Reeves, 1976; Graf, 1983]. We report short-term (3 years) geomorphic channel adjustment to two invasive vegetation removal methods in Canyon de Chelly National Monument, Arizona, United States, an incised, intermittent stream. The width and stand density of the riparian vegetation corridor in Canyon de Chelly have increased substantially in the past few decades [Cadol et al., 2011], and this vegetation is largely characterized by the exotic plant species tamarisk and Russian olive (Elaeagnus angustifolia L.). A primary objective of invasive plant removal was to increase visibility along the stream bed, although there was concern that excessive channel widening from plant removal might damage archeological sites. Extensive channel erosion was documented in the Rio Puerco, New Mexico, during a summer flood following vegetation removal along 12 km of the stream [Vincent et al., 2009]. It remains unclear whether this magnitude of response can be reasonably expected elsewhere following other invasive vegetation removal projects, given the variability of site-specific conditions.

2. Study Area

[3] Field work was conducted in the Canyon de Chelly side (1500 km2) of the national monument at four ∼1.2 km geomorphically similar stream reaches, numbered sequentially in the downstream direction (Figure 1). The channel at all study sites is a single-thread meandering morphology 5–15 m wide and from 1–2 m to > 8 m deep (Table 1). The stream banks consist of sequences of sand, silt, and clay. Clay, in layers of varying thickness that locally exceed 3 m, extends along several kilometers of the canyon bottom. Bed material throughout the canyon is mainly sand with a veneer of small gravel in patches and along channel margins. Streamflow is bimodal; spring snowmelt produces a runoff season of smaller magnitude and longer duration relative to the flashy hydrograph of late-summer flows resulting from convective thunderstorms. The streambed is dry between the spring runoff and summer flows.

Figure 1.

Map of Canyon de Chelly National Monument, Arizona, USA, and 4 invasive plant removal study sites.

Table 1. Study Site Summary Statistics Based on 2005 Pretreatment Cross Section Surveys
SiteDrainage Area (km2)ReachCross-Sectional Area (m2) (SD)Width (SD)Depth (SD)Width/Depth Ratio (sSD)Slope (SD)Bed MaterialVegetation
11091Control13.4 (2.96)13.0 (3.01)1.63 (0.13)8.1 (2.23)0.0073 (0.0022)boulder/cobbleRussian olive/tamarisk
11091Cut Stump11.6 (3.52)11.1 (4.04)1.60 (0.29)7.1 (3.15)0.0053 (0.0031)cobble/gravelRussian olive/tamarisk
11091Whole Plant11.4 (1.00)10.1 (0.96)1.55 (0.12)6.5 (0.47)0.0031 (0.002)cobble/gravelRussian olive/tamarisk
21448Control22.7 (10.15)11.2 (3.32)2.60 (0.43)4.4 (1.48)0.0037 (0.0008)sand/clayMixed cottonwood, willow, Russian olive, tamarisk
21448Cut Stump14.7 (3.64)8.5 (0.91)2.42 (0.57)3.6 (0.69)0.0034 (0.0027)sand/clayMixed cottonwood, willow, Russian olive, tamarisk
21448Whole Plant22.6 (5.01)11.3 (2.15)2.89 (0.39)4.0 (1.10)0.0052 (0.0031)sand/clayMixed cottonwood, willow, Russian olive, tamarisk
31455Control27.7 (11.02)11.9 (2.56)2.57 (0.67)4.8 (0.81)0.0034 (0.0017)sand/clayRussian olive/tamarisk
31455Cut Stump28.9 (12.46)13.9 (4.70)2.92 (0.69)4.8 (1.02)0.0029 (0.0008)sandRussian olive/tamarisk
31455Whole Plant34.7 (15.73)17.3 (6.22)2.86 (0.63)6.1 (1.61)0.0029 (0.0011)sandRussian olive/tamarisk
41461Control11.6 (2.87)10.0 (1.61)1.61 (0.27)6.4 (1.60)0.0026 (0.0013)sandRussian olive/tamarisk
41461Cut Stump20.7 (3.03)13.8 (2.27)2.14 (0.17)6.5 (1.31)0.0036 (0.0021)sandRussian olive/tamarisk
41461Whole Plant30.3 (15.35)17.1 (4.95)2.59 (0.85)6.8 (1.19)0.0021 (0.0008)sandRussian olive/tamarisk
Stage gauge1459 22.713. cottonwood, willow, Russian olive, tamarisk

3. Approach

3.1. Study Design

[4] Each study site included one 300 m control reach at the upstream extent of the site, followed by a 300 m cut-stump treatment reach, a 200 m untreated reach, and finally a 300 m whole-plant treatment reach. The cut-stump treatment reach included cutting the above-ground portion of the plant flush to the ground surface, leaving the subsurface root structure intact, and applying an herbicide to the cut portion. The whole-plant treatment reach involved removing the entire plant, including roots, using a backhoe. Each of the three reaches was subsampled with cross sections longitudinally spaced at ∼50 m. Native riparian vegetation along portions of sites 2 and 4 was left intact.

[5] The cut-stump removal method was carried out at all study sites in Fall 2005; whole-plant removal occurred in the Spring and Summer 2006 for sites 1, 2, and 3 and Fall 2006 for site 4. In sites 3 and 4, limited access for backhoes along both banks prevented whole-plant removal along the right bank and the left bank, respectively, necessitating cut-stump treatment application to these banks, which had occurred in Fall 2005. Channel change was measured by repeat annual cross-section surveys using a laser theodolite beginning in Summer 2005 prior to plant removal treatment.

3.2. Streamflow Discharge Estimates and Antecedent Moisture Conditions

[6] A stream gauge (ultrasonic depth sensor) was installed between sites 3 and 4 at a location representative of the morphologic conditions at the other study sites (Table 1). The hydrograph record is limited to the time period between February 2006 and August 2007, and is most complete for Summer 2006 and 2007 as a result of the difficulty of maintaining gauge operation at this location. From the gauge record, discharge magnitude estimates were calculated based on a stage-discharge relationship developed using one-dimensional hydraulic modeling [Jaeger, 2009]. Individual streamflow magnitudes were compared to discharge magnitudes of known recurrence intervals calculated using regional regression equations [Thomas et al., 1997] to provide context for the streamflows experienced during the study period.

[7] Available precipitation records from two nearby National Climatic Data Center (NCDC) stations (Lukachukai (NCDC 21634) and Canyon de Chelly (NCDC 21248)) and United States Geological Survey (USGS) drought surveys provide supplemental context to the climate and antecedent moisture conditions. Climate station monthly precipitation totals were compared to the stations' 1971–2000 average values for that month to qualitatively evaluate whether the particular month was climatically wetter or drier than normal.

3.3. Statistical Analysis

[8] To determine whether differences exist in mean channel change metrics between reach types (e.g., control, cut-stump, whole-plant), we used an ANOVA (α = 0.1) followed by a Tukey HSD post hoc test to identify group differences. Channel change metrics area, perimeter, width, and depth calculated for each cross section were averaged by reach, and tests for significant differences were conducted on these values (n = 12). To determine if potential differences in channel change correlate with different plant removal methods, we used both fixed and mixed (fixed and random) effects models (PROC MIXED, Statistical Analysis Software, Cary, North Carolina, USA, 2003). Explanatory variables included reach type, the particular year that the survey took place (2005, 2006, 2007, and 2008), and an interaction term reach*year. Model selection was based on the smallest Akaike Information Criterion value corrected for small sample sizes (AICc) [Akaike, 1974; Burnham and Anderson, 2004].

4. Results and Discussion

4.1. Flow Regime in Canyon de Chelly During Study Period

[9] In general, monthly precipitation totals at the two climate stations and the stream gauge record indicate drier-than-normal conditions for Water Year (WY) 2006 (1 October 2005 to 30 September 2006), normal to wetter-than-normal conditions in WY 2007 and generally normal conditions in WY 2008 (Figure 2). Following plant removal treatment application in 2005, Winter 2005–2006 was extremely dry based on Lukachukai precipitation values, drought surveys, and observed earlier stream drying. Summer 2006 was normal; the stream gauge recorded six short duration (> 24 h) flows that resulted in generally 1 m to 1.5 m of flow depth in the three downstream study sites and overbank flow at site 1. Recurrence intervals for these flows were calculated from 0.5 to 2 years.

Figure 2.

Monthly and 1971–2000 average precipitation values from (left) Lukachukai and (right) Canyon de Chelly climate stations for Water Years 2005–2008. Precipitation from July through September represent the late-summer season. November through February of the following year represent winter precipitation.

[10] WY 2007 was wetter compared to WY 2006 (Figure 2). Streamflow persisted longer in Spring 2007 and 3 short duration (> 24 h) summer flows produced 1 m to 1.5 m of flow depth at all study sites with two August flows producing bankfull and near bankfull depths in sites 2 and 3 (∼3 m) and overbank flow in sites 1 and 4. The stream gauge was destroyed during the first August flow event (estimated peak discharge of 79 m3/s).

[11] The Lukachukai precipitation gauge reported a moderately dry 2007–2008 winter and spring although the lower portion of the basin (Canyon de Chelly climate station) received above average amounts of precipitation (Figure 2). During the May 2008 annual surveys, more than 0.5 m of flow depth existed in sites 3 and 4, indicating at least a moderate spring runoff season.

4.2. Channel Response to Plant Removal Treatments

[12] Channel adjustment was primarily through increases in width, with whole-plant removal reaches on average widening significantly more (1.93 m, ANOVA: F(2,9) = 3.88, p = 0.061) compared to control (0.25 m) and cut-stump (0.54 m) reaches (Figure 3). Changes in width, however, were insufficient to cause significant differences in channel area or perimeter between control and treatment reaches, although whole-plant reaches generally experienced larger increases in cross-sectional geometry, with less increase in the cut-stump reaches, and the least change in the control reaches (Figures 3 and 4). Nevertheless, the channels maintained an entrenched morphology, including cross section locations in which widening rates were twice that of control reaches documented by Pollen-Bankhead et al. [2009]. The measured channel change translates to ≤ 5% of the pretreatment channel in control and cut-stump reaches for all cross-sectional parameters, and up to 12–14% of the pretreatment channel in whole-plant removal reaches for cross-sectional area and channel width.

Figure 3.

Channel change by reach type (control, cut-stump, whole-plant). (top) Initial (2005) and final (2008) cross section geometry of representative cross sections for each reach type. (bottom) Boxplots of channel change metrics: (a) area, (b) perimeter, (c) width, and (d) depth. Dashed horizontal line demarcates 0. Number values are mean channel change and percent change from 2005 value. Positive values indicate erosion and incision. Negative values indicate deposition and aggradation. Letters in Figure 3c represent the Turkey post hoc test for differences between groups. Changes in average width are significantly larger (F(2,9) = 3.88, p = 0.061) in whole-plant removal reaches (group b) compared to control and cut-stump reaches (group a).

Figure 4.

Boxplots of channel change metrics by study site and reach type: (a) area, (b) perimeter, (c) width, (d) depth. Study sites and treatments are sequenced from left to right in the downstream direction. Reach type is abbreviated: control (c), cut-stump (cs), whole-plant (w). Dashed horizontal line demarcates 0. Positive values indicate erosion and incision. Negative values indicate deposition and aggradation.

[13] In general, results from the mixed and fixed effects models were inconclusive, with no clear indication that observed channel change during the study period can be attributed to invasive plant removal, which would be demonstrated by the inclusion of the variable, reach, within a selected model. Reach is not consistently present in selected models for channel change metrics with the exception of width; year is just as likely to produce a statistically similar mixed effects model (Table S1 in the auxiliary material).

[14] The small scale (14% increase in mean channel width) of observed short-term channel change contrasts with the more dramatic channel change (84% increase) after invasive plant removal at the Rio Puerco, New Mexico [Vincent et al., 2009]. At the Rio Puerco site, 100% of the riparian vegetation was treated with herbicide along a ∼12 km reach and the above-ground portion of the vegetation was removed by subsequent streamflows. Dramatic channel widening was observed following a late-summer flow with a discharge magnitude that was the largest in the preceding 30 years [Vincent et al., 2009].

[15] We attribute the disparity in geomorphic response to invasive plant removal to differences between the two study locations in (1) plant removal methods, (2) stream bank composition, (3) channel morphology, and (4) streamflows. The combined effect of a considerably longer reach length in the Rio Puerco (12 km versus 300 m in Canyon de Chelly), the complete removal of all riparian vegetation, and the absence of clay along the base of stream banks (J. Friedman, personal communication, 2009) likely facilitated the substantial channel change observed at the Rio Puerco site. In addition, the less entrenched Rio Puerco channel allowed the extensive overbank flooding that removed the aboveground portions of riparian vegetation and enhanced channel widening (J. Friedman, personal communication, 2009). Finally, dramatic channel change in the Rio Puerco was in direct response to a streamflow that was the largest in 30 years. Our estimate of the discharge magnitude for the largest flow in Canyon de Chelly (August 2007) has a recurrence interval of 7 years. This is a first-order approximation, but it is unlikely that the true discharge approached that of a 30 year recurrence interval.

[16] Our findings serve as a cautionary note that channel response to invasive vegetation removal can be highly temporally and spatially variable as a result of these influencing factors. Initial fears of excessive channel widening in response to removal of exotic vegetation were unfounded, in part because of local controls such as remaining native vegetation, the presence of erosion-resistant bank layers (e.g., clay), and the entrenched morphology at the study sites. Because the stream channel in Canyon de Chelly is representative of other streams of its size in the southwestern United States, the variable relationships between invasive plant removal and channel adjustment observed here likely hold for streams elsewhere in the region.

5. Conclusion

[17] Invasive, exotic riparian plants are an ongoing management problem in fluvial landscapes. Canyon de Chelly National Monument was selected as a case study to implement removal of exotic plant species by two methods and to quantify channel adjustment following plant removal. Over a 3 year period of generally climatically normal conditions, channels widened significantly in whole-plant removal reaches, but the general morphology remained entrenched. This finding contrasts with observations from other regions in which substantial channel change has been measured and demonstrates the temporal and spatial variability in channel adjustment to plant removal. Factors influencing channel response include the method and extent of plant removal, streamflow conditions, variability in bank strength properties, and the degree of channel entrenchment. Not surprisingly, it therefore becomes critical in planning invasive plant removal projects to consider local factors likely to influence the rate and magnitude of channel response.


[18] The National Park Service provided funding for this project. We thank the people at Canyon de Chelly National Monument, including the Navajo canyon residents. This work was part of doctoral research at Colorado State University. The manuscript was substantially improved by the comments of three anonymous reviewers.