Restoration of riparian ecosystems in the Western United States increasingly includes removal of Elaeagnus angustifolia (Russian olive), an invasive, nitrogen-fixing, non-native tree, however there is currently little research that investigates the resulting ecosystem impact. Invasive species are those that expand dramatically beyond their native range and thereby effect changes to historic ecosystem structure or function [1]. As a nitrogen-fixer, E. angustifolia has the potential to significantly alter soil nitrogen dynamics not only when alive [2], but perhaps especially when it is killed and the unused nitrogen is released into the soil [3]. This pulse in soil nitrogen has the potential to change vegetative communities in sites where this tree is being removed. Because many invasive plants are associated with elevated nitrogen levels, there is particular concern that E. angustifolia removal sites may be prone to high exotic cover [4]. E. angustifolia is the fourth most dominant riparian tree species in the western United States [5,6]. Introduced in the early 1900’s from Eurasia, it is a shade-tolerant, climax species, spread primarily through bird dispersal of seeds [7].

In contrast, native Populus spp. are pioneer species requiring disturbance for germination success [8,9]. Due to changes in stream flow and geomorphology, some riparian corridors lose their dynamic disturbance patterns and shift from an early to late successional overstory. They eventually become more like the adjacent upland habitats, which are predominantly agricultural wastelands and/or xeric steppe where exotic species coverage is more abundant than native species coverage [10,11,12].

The majority of research in the western US in the field of riparian restoration focuses on invasive Tamarix spp. restoration, especially in terms of water storage and removal from the watershed [13,14,15]. Tamarix spp. are early successional invaders that are now more dominant than native cottonwood (Populus spp.) [5,16]. Two studies report on a quantitative synthesis of vegetation response across sites and over time after Tamarix spp. removal [17,18]. Still others report on the restoration of mixed native and invasive stands [12,19,20], but none have synthesized vegetation response across sites after removal of E. angustifolia alone.

Ecological statistical models can be used to synthesize the results of riparian restoration projects by evaluating the relative importance of various site characteristics (e.g., soil type, elevation, rainfall) for predicting restoration success (e.g., relative native plant cover and richness). Previous research to identify patterns in riparian revegetation where Tamarix spp. was removed has used ordination analysis [17] and regression trees [18]. While no significant pattern was found among abiotic and biotic variables in the first study, the later study was able to identify several, easy to measure environmental variables associated with higher native cover across the 28 Tamarix removal sites surveyed, including moisture availability and soil characteristics [18]. In light of the complexity of riparian ecosystems, regression tree analysis appears to be a useful way to deal with many variables across highly variable sites.

To determine if similar factors were important for determining restoration success following E. angustifolia removal, we sampled vegetation and soils at 25 riparian sites in Colorado, Montana, and Wyoming where E. angustifolia was invasive and was removed. Given that maximizing native species cover and overall species diversity are common goals in restoration [21], for the purposes of this research, success was evaluated by several different measures including total cover, total exotic (i.e., non-native) cover, total native cover, relative exotic cover, relative native cover, richness, and diversity. It is important to note that “exotic” is not synonymous with “invasive”, as not all introduced (i.e., non-native) species cause harm, even if maximizing native cover (at the expense of exotics) is usually considered desirable. Our interest was in determining which, if any, site characteristics such as those associated with the hydrology, soil, or management of a site were associated with success factors in sites where E. angustifolia had been removed. Given previous research on Tamarix removal sites, we expected that greater moisture availability (as measured by hydrological and weather patterns) would correspond significantly with success of native species over exotics.

Exotic species were present in all 25 study sites with a high degree of variability among sites (Figure 2), the dominance of which was explained at least in part by abiotic conditions. Russian olive accounted for only 5.6% ± 2.6 (SE) of total cover across sites. Moisture availability (e.g., precipitation and distance from water) and/or temperature (e.g., mean annual minimum temperature) were most often selected by the regression tree analysis as the most important descriptors of vegetative response after E. angustifolia removal (Table 2). Neither management approach (e.g., removal method) nor geography (e.g., elevation) was important in the models. The strongest models (i.e., with the highest PRE values) were those that predicted total exotic cover, total native cover, and relative native cover.

On average, when precipitation was more than 35.6 centimeters per year, total cover was 30% higher (Figure 3). Cover significantly increased with precipitation, although the amount of variability explained by this one-to-one relationship was small (Figure 4). Both soil texture and temperature helped explain variability in the regression tree models; in sites with less than 35.6 centimeters per year, total cover typically was higher in loamy soils than in sandy soil types and higher in sites with lower mean annual minimum temperatures. All sites had sandy loam, loamy sand, or loam soils.

On average, when the mean annual minimum temperature was less than 0 °C, the total exotic cover was only 40% (Figure 5) and relative native cover was higher. A regression of these two variables reveals a significantly positive, linear relationship between exotic cover and mean annual minimum temperature (Figure 6). When the mean annual minimum temperature was greater than 0 °C, relative exotic cover was also higher.

On average, when the distance down to the stream channel from the site was less than 1 meter, native plants cover an average of 70% of a site (Figure 7). On average, when the distance down from the stream channel was greater than 1 meter, total native plant coverage was only 32%. The opposite results were found when total exotic cover was used as the response variable. A distance greater than 1 meter down from the stream channel, roughly doubled total exotic cover (from 48% to 93%). In sites where distance down to the stream channel was less than 1 meter, and when the permanent water was more than 5.6 meters away from the site native coverage was 92%. In other words, in sites without an incised channel native plants thrived, especially beyond the zone of frequent disturbance, i.e., right at the water’s edge.

When using Shannon-Wiener Index values as the response data, the relative cover of each species was more disparate when organic matter was less than 5% of soil mass. In the group where soil organic matter was less than 5% soil mass, species abundance is more even when sandy soils were present. In the loamy soil sites, species were more even when mean annual precipitation was less than 34.3 centimeters. When electroconductivity was less than 2.35 mmhos, richness was 16 as compared to 13 when electroconductivity was higher than 2.35 mmhos. Among sites with less than 2.35 mmhos of electroconductivity, and when organic matter was less than 5.4%, richness was even higher [18].

Our survey of E. angustifolia removal sites suggests that although there was little evidence of regrowth by the target tree, many sites were still dominated by other exotic (i.e., non-native) species, with evidence that exotic plants are favored over natives in restoration sites with hotter, drier conditions. It is not uncommon for non-native species to dominate when disturbance regimes have been altered and create newly stressful conditions for natives [30,31,32], as they have been in these riparian systems [33]. Non-native Tamarix is the most well-understood example of this, dominating in areas that are more arid [6], generally as a consequence of dam building and channelization of rivers, which dis-favor native trees [9,33]. Previous surveys of riparian restoration sites have similarly found associations between increased exotic cover and low water availability [18,19].

Although not unexpected, our observed relationship between exotic species and aridity is particularly notable given that we only had indirect measures of moisture availability; precipitation, distance down to stream channel, and distance from permanent water were proxies for moisture availability in the absence of wells or direct soil moisture measurements which were logistically unfeasible in this study. With few exceptions, temperature and moisture predictor variables were most important for understanding vegetation response.

Understanding the mechanism of the relationship between moisture and exotic cover requires that we consider native species distribution. In all models, opposite trends are reported for native and exotic cover. In other words, conditions that promoted native species were associated with a decrease in the coverage of exotic species. This is consistent with similar surveys of Tamarix removal sites [18,19,34], and may indicate competitive interactions between native and exotics [35]. This theory offers the explanation that although both exotic and native species are expected to respond positively to increased water availability, exotics will dominate only where native species cannot—suggesting that exotics in this system are competitively excluded where conditions are generally favorable [34]. This mechanism has been repeatedly demonstrated for Tamarix trees [34,35,36], however further research is necessary to determine whether it holds true for other exotic species, particularly those that have been considered invasive [37].

Of the few native species we found in these sites, most were either early succession species adapted to disturbance or salt tolerant grasses. The latter are native species usually found in the more xeric, upland habitats adjacent to mesic riparian habitats. Future research should address the efficacy of restoring to mesic, riparian, native vegetation communities versus xeric, native, upland vegetation communities [12]. With reduced flooding to promote disturbance-adapted early succession species, the mechanisms to maintain riparian vegetation communities are absent. Therefore adjacent native xeric communities may be more desirable than exotic riparian species in the riparian zone.

It should be noted that many of our sites were unusually dry during a year of drought. Only five of our sites in Colorado actually had precipitation greater than 32.3 centimeters despite the statewide average of 38.4 centimeters that year. Native plants in the riparian corridor are accustomed to a mesic habitat, thus drought conditions may explain the abundance of exotic species in these sites. There were many plants that did not flower for the duration of our field collection period. All of the sites were in areas categorized as experiencing moderate to extreme drought [38]. Thus, the role of moisture for determining vegetative response may have been especially heightened the year of our study [39].

There are several site descriptor variables that would be expected to be important for explaining vegetative response but could not be used in the final models because of missing information; most managers simply did not have baseline data or good records of site histories. Even if including these data would be unlikely to change the patterns observed, there is no doubt that our models would have been more powerful and informative with the inclusion of site history data. It is our intent that our work and other research on the outcomes of restoration projects will encourage monitoring practices in general, including the collection of baseline information.

Removal of any dominant tree will result in a pulse of resources which are likely to affect vegetative response; these include an increase in light availability at the very least, and likely soil resources as well because of competitive release. It is critical to understand what factors are associated with greater or lesser exotic cover among restoration sites. We found soil factors to be more important for understanding species richness and diversity than for cover estimates, and to be less predictive overall, however we did not have good nutrient measurements and therefore can only speculate about their potential role. Nor was our experimental design intended to study the effects of E. angustifolia removal (e.g., relative to sites where it was not removed). As these are important questions, we would recommend their pursuit for future work.

This quantitative study incorporated site data from Colorado, Wyoming, and Montana where E. angustifolia was removed. The models suggest that moisture and temperature site attributes were the best predictors of plant cover and richness, the response variables measured to quantify restoration success. As was found in Tamarix removal sites in an earlier study [18], increased availability of water, as measured by precipitation or by distance to stream channel, was associated with higher percentage of natives both in total, and relative to exotic species, which dominated in hotter and drier areas. This has important implications in the context of climate change, which predicts the latter conditions to increase in the future, and also for restoration site selection by land managers. These results suggest that when revegetation is dependent upon passive means, wetter sites are more likely to recover with native species than drier ones.

It is imperative that ecologists and land managers work together to develop a management and restoration plan for E. angustifolia, given the expectation that this species is likely to dominate in the future. We hope this research will promote further study regarding E. angustifolia so we can develop a sound management plan for this species.