Extreme precipitation promotes invasion in managed grasslands

Abstract

Climate change is increasing the frequency and intensity of extreme events like drought and flooding, which threaten to amplify other global change drivers such as species invasion. We investigate the effect of wet and dry extreme precipitation regimes on invasive species' abundances in Northern tallgrass prairies. Because soil moisture is a key determinant of prairie composition, theory and evidence suggest drought conditions will hinder invasion, whereas wetter conditions will enhance invasion. To test this hypothesis, we explored the effect of precipitation on invasive plant species abundance from 2010–2019 in 25 managed prairies using observations from 267 transects comprising 6675 plots throughout western Minnesota, USA. We estimated how increases in the number of extremely wet or dry months in a year altered overall invasive species abundance and the abundance of the highly invasive grasses Poa pratensis and Bromus inermis. We found that a greater occurrence of abnormally wet months increased invasive species abundance but found mixed evidence that abnormally dry conditions hindered invasion. Further, more moderately wet and dry months reduced native grass abundance. Together, these results suggest that more frequent extremely wet months may intensify invasive dominance and that dry months may not counterbalance these trends. Given the considerable uncertainty still surrounding the interactive effects of climate change and invasion on native plant communities, this research represents an important step towards quantifying the complex influence of precipitation extremes on invasion dynamics in managed ecosystems of critical conservation concern.This article is protected by copyright. All rights reserved.

Causal inference with observational data and unobserved confounding variables

“…Thus, this design allows us to relax the strong assumption that all confounding variables are observed and measured so that we can interpret β 1 as causal, provided other assumptions are met (see Discussion). For ecological examples using this design, see Larsen (2013), Dee et al (2016), Dudney et al (2021), Ratcliffe et al (2023), and Dee et al (2023). We note that accounting for serial correlation, heteroskedasticity, and clustering of the error, such as through cluster robust standard errors, are likely important for both approaches for inference (Abadie et al .…”

“…While these two versions of the fixed effect design look different, they are equivalent(Angrist & Pischke 2008;Wooldridge 2010).Fixed effect designs allow us to relax the strong assumption that all confounding variables are observed, measured, and included as covariates in models for a causal interpretation of β $ when other assumptions are met (see Discussion). For ecological examples, see Larsen (2013),Dee et al (2016),Dudney et al (2021),Ratcliffe et al (2023), andDee et al (2023).…”