Experimental Warming Changes Phenology and Shortens Growing Season of the Dominant Invasive Plant Bromus tectorum (Cheatgrass)

Abstract

Bromus tectorum (cheatgrass) has successfully invaded and established throughout the western United States. Bromus tectorum grows early in the season and this early growth allows B. tectorum to outcompete native species, which has led to dramatic shifts in ecosystem function and plant community composition after B. tectorum invades. If the phenology of native species is unable to track changing climate as effectively as B. tectorum's phenology then climate change may facilitate further invasion. To better understand how B. tectorum phenology will respond to future climate, we tracked the timing of B. tectorum germination, flowering, and senescence over a decade in three in situ climate manipulation experiments with treatments that increased temperatures (2°C and 4°C above ambient), altered precipitation regimes, or applied a combination of each. Linear mixed-effects models were used to analyze treatment effects on the timing of germination, flowering, senescence, and on the length of the vegetative growing season (time from germination to flowering) in each experiment. Altered precipitation treatments were only applied in early years of the study and neither precipitation treatments nor the treatments' legacies significantly affected B. tectorum phenology. The timing of germination did not significantly vary between any warming treatments and their respective ambient plots. However, plots that were warmed had advances in the timing of B. tectorum flowering and senescence, as well as shorter vegetative growing seasons. The phenological advances caused by warming increased with increasing degrees of experimental warming. The greatest differences between warmed and ambient plots were seen in the length of the vegetative growing season, which was shortened by approximately 12 and 7 days in the +4°C and +2°C warming levels, respectively. The effects of experimental warming were small compared to the effects of interannual climate variation, suggesting that interactive controls and the timing of multiple climatic factors are important in determining B. tectorum phenology. Taken together, these results help elucidate how B. tectorum phenology may respond to future climate, increasing our predictive capacity for estimating when to time B. tectorum control efforts and how to more effectively manage this exotic annual grass.

Keywords: Bromus tectorum; climate change; dryland; invasive plants; phenology; phenophase; soil moisture; soil temperature.

FIGURE 1

Daily mean soil volumetric water content estimated from marginal means of linear mixed-effects models from 2009–2019 in the CV4, CV2, and M2 experiments. Experiments are represented with different shaped polygons: circles for CV2, triangles for CV4, and squares for M2. Error bars are 95% confidence intervals calculated on marginal means. There were no significant differences in mean soil volumetric water content between warmed and ambient treatments in any of the three experiments.

FIGURE 2

Empirical data for B. tectorum senescence (A–C), flowering (D–F), and germination (G–I) timing measured weekly in 2010, 2011, and 2015–2019. All phenophases are shown as the average day of year that each phenophase was first observed in each treatment. A value of 0 for day of year corresponds to January 1. A lower value on the y-axis represents earlier timing and a higher value represents later timing. Negative values for day of year indicate that the phenological event took place that number of days prior to January 1. For example, most germination events took place in the Fall and they are shown as negative days for that year (i.e., germination in 2011 took place in the Fall of 2010). Warming treatments are depicted as red triangles and are connected with solid red lines. Ambient treatments are shown with black circles and are connected with dashed black lines. Standard errors of the means are shown with vertical bars associated with the appropriate polygon. Significance of treatment effects is shown in the bottom left of each panel, with NS denoting non-significant effects. Significant differences between the treatment means were determined by Type-II ANOVAs.

FIGURE 3

Model decompositions for vegetative growing season (A–C), flowering (D–F), and germination (G–I) of B. tectorum in each of the three climate manipulation experiments. The estimated mean number of days each phenological stage was delayed or advanced are represented by the horizontal lines. A lower value on the y-axis represents earlier timing and a higher value represents later timing. The estimated means for the ambient plots are represented by the dashed black horizontal lines and are set to zero as all data are shown relative to the ambient plots for each site and each phenological stage. Solid red horizontal lines represent the mean difference in timing (in Julian days) of the warmed plots relative to the ambient plots (black dashed line). When the solid red line is below the dashed black, line there is an estimated warming-induced advance of that phenophase. When the solid red line is above the dashed black line, there is an estimated warming-induced delay of that phenophase. Significance of treatment effects is shown in the bottom left of each panel, with NS denoting non-significant effects. Significant differences between the treatment means were determined by Type-II ANOVAs. The conditional modes of the random effects for the ambient and warming treatments in each year are shown with black circles or red triangles, respectively. Estimated confidence intervals for these conditional modes are shown with vertical bars within the circles or triangles. Interannual variability of these conditional modes are estimated by their difference each year from the horizontal lines of the means.

FIGURE 4

Plot level relationship of date of B. tectorum senescence and mean annual soil temperature at 10 cm depth in the CV4, CV2, and M2 experiments (A–C, respectively). Values from warmed and ambient plots are shown with red triangles and black circles respectively. The linear fits are represented by blue lines and were determined by the geom_smooth function in ggplot2. The linear fits incorporate data from all plots (Ambient and Warming) in each site. R² values for each linear fit is shown in the bottom left of each panel.