Temporal invasion regime attributes influence community synchrony and stability in an arid land system

Invasive species have become a major threat to ecosystems across the globe, causing significant ecological and economic damage. To anticipate how communities may respond to future invasions, it is crucial to refine how invader impacts are evaluated, especially in historically uninvaded and highly variable systems such as arid lands. While invader abundance is typically used to predict invader impacts, it may not effectively capture the dynamics that occur over time for established invaders that experience cyclical dynamics (i.e., boom-bust patterns), making it more challenging to track invader impacts. To address this issue, we leveraged a long-term vegetation dataset to develop a novel invasion regime framework for a dominant annual invader in North American deserts, Brassica tournefortii. Using abundance data over time, we evaluated how attributes of this invader’s boom-bust dynamics (i.e., invasion level, boom frequency and magnitude) influence the long-term synchrony and stability of invaded Aeolian sand dunes communities. We found that attributes that captured the temporal effects of the invader were  strong indicators of the impacts of an invader on long-term attributes of communities. Specifically, the mean magnitude of invader booms led to a decrease in species asynchrony and community stability. Increasing boom frequency also decreased community stability but this was more muted. Mean magnitude of invader booms also mediated shifts in the relationship between synchrony and stability, with this relationship becoming more shallow with increasing boom magnitudes. Our research emphasizes the significance of using community metrics that capture temporal dynamics to document invasion impacts within dynamic arid land systems. The invasion regime framework can additionally offer insights into the mechanisms that may enable the persistence of the invader over time. Together this knowledge can be helpful in guiding decision-making and land management strategies aimed at effectively controlling and mitigating the impact of invasive species.

Methods
Study Design & Sampling – we leveraged long-term observational transects that were established in 2002 as part of the Coachella Valley Multiple Species Habitat Conservation Plan Monitoring Program (Allen et al. 2005). Nineteen belt transects were arranged as 100 m x 10 m (0.1ha) rectangles (Figure 2A). Within each belt transect, four 1 m2 quadrats plots are spatially grouped within a cluster, with clusters spaced 44 meters apart (3 clusters per transect: A,B,C; n=12 plots, Figure 2B). Plots were alternated along the center line of the transect for a total of 228 plots. Due to some plots missing vegetation data, we reduced our dataset to 226 plots. Vegetation within these plots was surveyed annually from 2003-2019 using visual estimates of cover for each species present within the aboveground vegetation during peak biomass in the spring, (March-April, dependent on phenology).

To assess traditional metrics of community-level responses to invasion, we created linear-mixed effect models using the ‘lme4’ package in r (Bates et al. 2015). For the first model, we fit Shannon-Weiner index as the response variable, with invasion level (low, moderate, high) as a fixed factor, with cluster nested within transect as a random factor. We used the same model structure as above for the second model but replaced the Shannon-Weiner index with species richness as the response variable. We then compared individual treatment levels in our models using Tukey's HSD post-hoc tests.

To assess how attributes of Brassica tournefortii’s invasion regime attributes influence community synchrony and stability across an invasion gradient, we created two separate mixed effect models, one with community synchrony as the response variable, and the second with community stability as the response variable. Within these models, the following were included as fixed factors: boom frequency, mean boom magnitude and their interactions, with cluster nested within transect as random factors. To make sure our models met the assumptions and to test for collinearity among our predictor variables, we used the performance package in r. VIF values were kept below 10. We included cluster nested within transect as random factors. We used the same model structure for our second model, replacing community stability as the new response variable. Lastly, to test if attributes of an invasion regime mediate the relationship between community synchrony and community stability, we created a third model where we used community stability as a response variable, with community synchrony, mean boom magnitude, and their interactions as fixed factors and cluster nested within transect as a random factor. We focus on the mean boom magnitude invader attribute, as this measure provides the greatest potential to capture carryover effects via large propagule pulse events.  Finally, we used the ‘sjPlot’ package in r to visualize our results and plot the marginal effects of our regression models that had significant interactions (Lüdecke, 2024).