Occurrence and Spread of Native and Non-Native Plant Species at Idaho National Laboratories. (2011)

Survey, Monitoring and Predicting the Occurrence and Spread of Native and Non-Native Plant Species at Idaho National Laboratories. (2011)

 

Investigators and Affiliations


  • Lisa Rew, Ph.D., Department of Land Resources and Environmental Sciences, Montana State University, Bozeman, Montana
  • Bruce Maxwell, Ph.D., Department of Land Resources and Environmental Sciences, Montana State University, Bozeman, Montana
  • Matt Lavin, Ph.D., Department of Plant Sciences and Plant Pathology, Montana State University, Bozeman, Montana
  • Tyler Brummer, Department of Land Resources and Environmental Sciences, Montana State University, Bozeman, Montana
  • Kimberley Taylor, Department of Land Resources and Environmental Sciences, Montana State University, Bozeman, Montana

 

Funding Source: U. S. Department of Energy-Idaho Operations Office


Background: Management of both non-indigenous plant species (NIS) and rare plants species (RPS) is a high priority in many managed forests, wildlands and rangeland areas. However, rarely do either public or private agencies have sufficient resources to manage all NIS or conserve all RPS. Neither do agencies have sufficient information on the potential impacts of future anthropogenic development. Therefore, a better understanding of the temporal and spatial processes which drive both NIS and RPS population distributions and dynamics is required to improve management effectiveness and efficiency. The difficulty in increasing our knowledge of NIS and RPS population dynamics in the sagebrush-steppe plant community is that they occur with low frequency on the landscape and can be difficult to detect because they are similar in morphology to the co-occurring species. By using knowledge of probable routes of introduction for the NIS, and particular habitat requirements for the RPS, appropriate survey methods can be developed. Repeated sampling can then help to elucidate the spatiotemporal dynamics of select populations. From such data, predictive occurrence maps can be generated for the current landscape, but also for a range of future scenarios including anthropogenic development. Incorporating the information into a decision support management prioritization framework can help resource managers prioritize populations to manage and help evaluate the potential impacts of different disturbance scenarios to minimize the negative (RPS) or positive (NIS) impacts on plant population dynamics.


Objectives:

The goal of this study was to determine the current distribution of NIS and RPS on the INL Site and predict the potential spatial and temporal metapopulation dynamics of these species to help inform management and future development decisions.



Accomplishments through 2011:

Survey detection error and metapopulation dynamics: A total of 45 10-m wide belt transects that originated on roads or facility margins and traveled 2-km away from the road or facility were completed in 2011. Transects were repeats from previous years and were selected according to stratification on fire chronology and proximity to facilities. Presence and absence of eight targeted NIS were recorded along these transects. In addition to repeating the whole transects, two 200-m sections of each transect were repeated again, to evaluate within season detection error. The 200-m sections were randomly located, one within 0 - 1000-m and the other 1001 - 2000-m from the road.

The repeat NIS data were used to evaluate detection error, and also to assess metapopulation dynamics. All analysis was performed in R using the ‘unmarked’ package. To evaluate detection error and effect on predictions we evaluated five Models: Model 1 used just the first visit in a logistic regression, Model 2 used the second visit only in a logistic regression, Model 3 used the maximum of the detection history in a logistic regression, Model 4 used both visits in a hierarchical occupancy model, with the detection process as intercept only, and Model 5 was the same as Model 4 but allowed detection to vary as a function of covariates. Depending on the objectives of each analysis, covariates were either excluded or included for the occupancy process in these five different groups of models.

To assess metapopulation dynamics covariates were carefully constructed based on hypothesized drivers that were determined a priori, which included habitat suitability (HS) (determined via logistic regression on the species P/A data using only environmental predictors), occupied neighborhood density (OND, calculated as the number of neighboring occupied cells divided by the total number of neighboring cells within a given radius) and disturbance variables of paved (continuous variable) and two-track (binary variable) roads, wildfire (binary variable), time since fire for areas that had burned (categorical variable), and Agropyron cristatum “Green Strips” (binary variable).

Biodiversity analysis: Using a subset of the entire 2009-2010 transects, the frequency of all plant species was recorded along 76 transects in two 1000 m sections (0-1000 m and 1001- 2000 m from the road), as well as along a 1000 m transect that followed the road and that was bisected at the mid distance (500 m) by the transect running perpendicular to the road.

The biodiversity data was analyzed to determine the influence of NIS and other ecological parameters (e.g., disturbance and fire) on the abundance and diversity of native plant species including any RPS found during the course of field work. Data were analyzed in R.


Bromus tectorum response to fire: The percent cover of Bromus tectorum, the four most dominant plant species, litter and bareground were recorded in 10-m * 10-m plots. The plots were randomly placed along transects that were stratified by time since fire. Three transects were assessed in areas burned 15 years ago, two transects in areas that burned 11 years ago, two transects in areas that burned three to four years ago, 10 transects in areas that burned one year ago (Jefferson Fire), and seven transects in areas that had not burned recently.

A logistic regression of the binomial family was used to model B. tectorum cover as a function of presence or absence of fire, time since fire, percent litter, percent native vegetation cover, percent NIS vegetation (excluding B. tectorum) cover, and dominant species in the plot. Data were analyzed in R.


Results:
Survey detection error and metapopulation dynamics: Empirical detection rates for the eight targeted NIS – Agropyron cristatum, Alyssum desertorum, Bromus tectorum, Carduus nutans, Descurainia sophia, Halogeton glomeratus, Lepidium perfoliatum, and Sisymbrium altissimum - ranged from 0.24 to 0.94 although the majority of detection rates were greater than 0.75. There was a significant trend that as a species occupancy on the landscape increased, so too did our detection rate (Figure1), albeit that there was a lot of variability at low occupancy rates. Overall detection error did not change species rank occurrence substantially suggesting that if species are managed according to their relative frequency on the landscape detection error is unlikely to change the species order much.

Figure 1

Figure 1. Plot of Estimated Detection Rate as a Function of Occupancy Rate. Detection rate increased as occupancy rate increased (p = 0.01) with no evident effect of year (p = 0.11). The fitted line from a simple linear regression of detection rate as a function of occupancy rate is shown.

Of the eight species evaluated only three of them showed a seasonal change in detection of 10 percent and this did not sufficiently bias regression coefficients to make practical changes to predicted probability of occurrence, nor did observer differences. Evaluation of hierarchical models (McKenzie’s) showed that uncertainty was greater when the hierarchical model was used for three of eight species but was unchanged for four of eight species. Complementarily, predictions of probability of occupancy using general linear regression were robust to choice of model method for four of eight species while predictions changed substantially for three of eight species. Predictions for species with detection higher than 87 percent were robust to each modeling method. In conclusion, we generally found detection rates to be high and logistic regression to be more reliable and interpretable for the development of predictive maps.

Models of spatial variation in colonization and extinction ranged from very simple (six species) to very complex (A. desertorum and S. altissimum). The effect of occupied neighborhood density was the only consistent predictor with greater colonization associated with greater occupied neighborhood density, coupled with lower extinction. Seven of the eight species responded to fire with the four annual Brassicaceae (A. desertorum, D. sophia,, L. perfoliatum, and S. altissimum) having higher colonization and lower extinction in the fire, and colonization decreased and extinction increased as time since fire increased.

Temporal information (2009-2010, 2010-2011) was only available for three species but suggested strong temporal dynamics for S. altissimum and C. nutans that corresponded with yearly climate differences, while A. cristatum followed the same general (increasing metapopulation size) trajectory for both years.

Biodiversity analysis: The ecological factors that have the greatest influence on vascular plant biodiversity are first and foremost the degree of disturbance, followed by the degree of seasonality - whether measured in terms of precipitation, temperature, or water deficit. The least explanatory variables included the abundance and diversity of NIS, as well as the historical legacy of fire. In all fire categories (Figure 2), introduced species (red vertical lines) are distributed evenly along the rank abundance of native species (black open circles), according to a Kolmogorov-Smirnov test, suggesting they add to, rather than replace, native plant diversity. A general trend along this fire chronology was that total plant diversity and the proportion of native plant species to NIS remained approximately the same. At the least, fire by itself did not decrease plant diversity or increase the proportion of NIS.

Figure 2
Figure 2. Rank Abundance (log10 frequency) for the Vascular Plant Species Sampled Along the 76 Transects and Stratified by Fire Interval, Within the INL Development Area.

Principle coordinates (PCO) analysis arrayed the 76 INL Site transects in two dimensions (Figure 3) using community phylogenetic distances (roughly a composite measure of how different each transect is with respect to shared species, genera, families, etc.; results using traditional community metrics such as Bray-Curtis or Euclidean distances rendered the same results as described below but with lower measure of fit). In the left panel (Figure 9-4), the contour lines represent Simpson’s diversity index of native plant species whereas the color of the transect represents the disturbance category. The diversity of native species and disturbance category were determined to be the most explanatory variables of phylogenetic community composition using General Additive Model, Canonical Correspondence Analysis, and distance modeling and model ranking approaches (e.g., AIC). In the right panel, the contour lines represent Simpson’s diversity index of introduced plant species whereas the color of the transect represents the fire category. These last two variables were determined to be some of the least explanatory using the same modeling approaches. The results here suggest that the diversity and abundance of native species is a good indicator of the degree of disturbance that has occurred at a site, whereas the diversity and abundance of NIS is about the same everywhere regardless of ecological variation within the INL sagebrush steppe.

The conclusion is that plant biodiversity within the INL Site boundaries is in good condition because disturbance is minimal in this “accidental wilderness area.” Fire also doesn’t constitute much of a disturbance at least from the perspective of plant biodiversity at this scale (1000 m * 10 m). A logical extension of this would be that the most effective means of reestablishing sagebrush steppe after fire within an infrequently disturbed setting such as the INL Site would be to minimize physical disturbances such as those caused by the creation of fire lines, the establishment of green strips, or by reseeding or revegetation efforts.

With respect to RPS, no federally or state listed rare plants were found during three summers of intensive searching.

Figure 3
Figure 3. Principal Coordinates Analysis of the 76 Transects, Using the Three 1000 m Sections (adjacent to the road, 0-1000 m and 1001-2000 m).

Bromus tectorum response to fire: B. tectorum cover was best explained by the presence or absence of fire in the past 15 years, percent native vegetation cover, percent NIS vegetation cover (excluding B. tectorum), and presence or absence of Elymus lanceolatus, as determined by logistic regression. Presence of fire, presence of E. lanceolatus and percent native vegetation cover all had a negative effect on B. tectorum cover, while percent NIS cover had a positive effect. The odds of finding B. tectorum cover in a burned area were only 48.37 percent of the odds of finding it in an unburned area. The fire chronology results suggested that B. tectorum did not always increase in cover after fire. While other studies have concluded that B. tectorum populations increase after fire, it is possible that in highly undisturbed sites such as the INL, fire may not cause B. tectorum cover to consistently increase. This hypothesis is supported by the fact that higher B. tectorum cover was positively correlated with percent NIS cover, which was higher in more disturbed sites.

The odds of finding B. tectorum cover in a plot with Elymus lanceolatus were only 34.55 percent of the odds of finding B. tectorum in a plot without E. lanceolatus. These results suggest that areas with higher B. tectorum cover occur where E. lanceolatus is not present. This result could be due to differences in habitat preferences between the two species. It could also be a result of either high B. tectorum cover causing reduced E. lanceolatus cover or vice versa. More research is necessary to determine if E. lanceolatus could be used to restore highly disturbed areas infested with B. tectorum if a restoration approach was considered desirable, although the biodiversity results suggest this may not be necessary.

Plans for Continuation: Data analysis, interpretation and finalization of manuscripts are on going but should be completed shortly. Tyler Brummer will be presenting his findings and defending his Master of Science candidacy April 3, 2012.

Publications, Theses, Reports:

  • Brummer TB, Maxwell BD, Higgs M, and Rew LJ. Surveying non-native species occurrence and modeling realized distributions at local and landscape scales. Submitted to Biological Invasions March 2012.
  • Brummer TB, Maxwell BD, Lele S, and Rew LJ. Detection error in plant surveys: to correct or not to correct. In Preparation.
  • Brummer TB, Maxwell BD, Lavin M, and Rew LJ. Regional population dynamics of non-native plant species. In Preparation.
  • Lavin M, Brummer T, Seipel T, Maxwell B, and Rew L (in press) The intermountain flora sets the stage for a community phylogenetic analysis of plant biodiversity in the sagebrush steppe of western North America. Brittonia. (Accepted Dec. 2012).