Developing an integrated management decision framework for cheatgrass control in the northeastern region of the sagebrush steppe

Objective 1: Cheatgrass was effectively controlled by a fall herbicide application (2019) such that the low cover combined with a drought meant a second herbicide application was not applied in 2020. All other treatments including targeted grazing (0.4 AUM/ac) with cattle or simulated with a weed strimmer, and a split plot of seeding were imposed on all sites across the valley in 2021. Sites were assessed for functional plant cover at peak season in 2022 and 2023.

A single herbicide application reduced the abundance (cover) of cheatgrass significantly one-year post application, and cover remained lower than the controls in the third (p=0.048) but not fourth season. One year after the other treatments were imposed (2022) the herbicide only treatment still had lower cheatgrass abundance but the sprayed and target grazed treatment had higher abundance which is undesirable; none of the other treatments differed from the control. Native grass abundance increased in integrated herbicide with targeted or simulated grazing treatments, but not herbicide, targeted or simulated grazing alone. Seeding did not alter overall cover and few seedlings were observed in either sampling year.  We did observe one general trend that has been observed by others, and which may be helpful for management: as native grass cover increased cheatgrass cover decreased, with the reverse being true for litter which increased as cheatgrass increased. This suggests that areas with higher cover of perennial grasses and lower litter will be more resilient to cheatgrass.  If cover is above 15-20% then sites are less likely to be dominated by the annual grass, cheatgrass.

Objective 2: Cheatgrass was effectively controlled by herbicide application in 2019 and had low abundance in fall 2020. Data on all species were recorded each summer (2021-2023) for the different seeding treatments (no seed, fall broadcast, spring broadcast and spring seed pellet).  Abundance of cheatgrass remained very low (see Objective 3a results) along the native grass gradient. Unfortunately, low numbers of seedlings were recorded in all seed treatments including the control, such that there was no statistical difference in any of the three seasons assessed. Across all treatments the mean seedling density was 2 seedling tillers/plot in one region (Antelope Peak) and 16 seedling tillers/plot in the other (Eastern Sandhills) by the third season of the study.  One interesting finding was not many seed pellets had not broken down in the first season. Seed pellets are made from clay:organic matter:seed at a recommended ratio of 5:3:1. We therefore conducted the second study to improve the ratios for the cold and dry Montana climate. 

In the second experiment there was no difference in seedling establishment between the non-seeded control and the low and high clay ratio seed pellets, across two additional sites in southwest Montana. However, as native grass cover increased mean seedling count decreased (p = 0.013). When native grass cover was split into two categorical variables (<20 and ≥20 percent cover of native grasses) plots with ≥20% cover of native grasses saw lower establishment <1 seedling/plot compared with ~3 seedlings in the low native grass cover plots (p = 0.053; Figure 1).

Objective 3:

3 a) The 12 sites across the Centennial valley represent high-elevation sagebrush steppe for the north-eastern part of the biome. We found that two fall applications of the herbicide, imazapic, effectively reduced cheatgrass for three-four years (Figure 2 and Figure 3), with the greatest reduction in the first season post applications. Within the sprayed sites we evaluated areas that were sprayed or not-sprayed, by placing a tarpaulin on the ground during application. In both regions (Antelope Peak and Eastern Sandhills) cheatgrass was significantly reduced (p<0.01). At the Antelope Peak sites - which has a clay/loam soil – cheatgrass decreased to close to zero for three years but then increased in the fourth year, becoming similar in cover (~15%) to the non-sprayed plots (which naturally declined from ~20 to 15% over the study). At the Eastern Sandhills sites, soils are sandy as the name suggests, cover of cheatgrass remained lower (<1-1.4%) in the sprayed plots than the non-sprayed (3-6%) for the four years (Figure 2).

At the landscape scale we compared sprayed hillslopes with non-sprayed and low infestion “reference” sites.  At this scale cheatgrass was reduced to similar levels by the herbicide in both regions. Specifically, cheatgrass cover was lower in the sprayed than reference sites in the first and third year after herbicide applications; by the fourth year the sprayed area was higher than the reference (Figure 3).  This essentially three-year reduction of cheatgrass from two applications of herbicide is longer than has often been observed in other studies. The longer control may have been due to several factors, including weather, careful timing of herbicide application, and initial lower infestation levels than observed in many areas. The timing of our herbicide was excellent; we had no precipitation until later in the fall, after which cheatgrass emerged and was sprayed at the 2-3 leaf stage (October), with snow and cold temperatures following within a few weeks, thwarting further germination. This represents a best-case scenario because there was only one or two germination cohorts of cheatgrass and we controlled them, with scant germination after due to the change in temperature.  We also had a drier growing season in 2020, the second year after herbicide application ceased, which likely minimized cheatgrass emergence and thus slowed recovery.   

An important finding is that in the paired reference communities we did not observe an increase in cheatgrass throughout the entire study, with values remaining below 1% cover from 2017-2022. This suggests these areas are resilient to cheatgrass, generally.  However, the infested sites were on the same slopes and aspects and as such demonstrate that where additional disturbance occurs cheatgrass can invade, but, at least during the six years of our study cheatgrass did not appear to be increasing in the non-sprayed plots on these slopes.  This does present rather a quandary on when to manage.  One important aim of controlling weeds like cheatgrass is that we obtain a positive response from the remaining vegetation, unfortunately we did not observe that. The herbicide control was not associated with an increase in perennial grass cover or biomass, species richness, nor alpha diversity. This was disappointing. A community analysis found similar results with the sprayed communities being resilient to cheatgrass management (that is no non-target effect), but the plant composition did not more much toward the reference community state. Both sprayed and reference community assemblages shifted over the study, with slight increases native perennial forb cover and a reduction in native perennial grass cover. These changes were correlated with warmer and drier seasons, which could impact the future resiliency of this system but could also demonstrate a natural fluctuations.

3b) Our alternative cheatgrass management strategies were evaluated over three summers (2021-2023). In the first year, post treatment cheatgrass abundance (cover and biomass) was reduced in the medium and high rates of soil amendment, high mustard seed meal and herbicide (indaziflam) (Figure 4). In the second year cheatgrass abundance remained low in the medium soil amendment and the herbicide plots, but by the third season cheatgrass was the same in all treatments (Figure 4). Desired and native perennial grasses abundance was higher in the herbicide treatment than all others, for all years (Figure 5).  Forb abundance did not differ between the treatments but increased each year post treatments.

Objective 4. Develop a decision framework to help livestock producers effectively control cheatgrass on their property.

The decision framework has not yet been fully developed because more analysis is needed and preferably at least one more year of data collection to consolidate our findings.  Our integrated and alternative approaches (Objective 1, 2, 3b) sometimes but performed as well as herbicide but not generally, and herbicide only did not always differ from the no action control. Therefore, we would like to collect at least one more year of data for Objectives 2 and 3b, which would increase our confidence and capacity to develop a decision framework that will be based on thresholds of both annual and perennial grasses. 

At present our preliminary framework conclusions are as follows. Two applications of imazapic reduced cheatgrass but ultimately there were no differences between the 2017 pre-herbicide abundances of cheatgrass and four years after herbicide application for either region (Eastern Sandhills to Antelope Peak) nor sprayed and non-sprayed or reference treatment plots in the Centennial valley, MT. Our herbicide applications were excellent; we applied very late in the fall after rain and emergence of cheatgrass, in both 2017 and 2018. This represents a best-case scenario where there was only one or two germination cohorts of cheatgrass and we controlled them, and there was not much germination after due to a cold change in temperature.  We also had a drier year soon after treatment (2020). We treated cheatgrass at infestation levels that are low – as recommended by other studies – and these levels are of concern to producers and resource managers in our region.  On the infested slopes the cheatgrass cover was around 20% in the denser patches and 9% on average. Our herbicide treatments did reduce cheatgrass for 3 or more years.  However, it must be stated cheatgrass cover was also low and remained that way reference areas (<1%) and declined over time in the non-sprayed plots within the sprayed hillslopes. This suggests that the climate during these six years was not particularly suitable for expansion of cheatgrass.

Evaluation of different management strategies showed that the herbicide indaziflam and the recommended rate of the soil amendment (NutraFix) was effective at controlling cheatgrass for two seasons but not the third season post treatment. And, in our integrated weed management (IWM) study we found that, cheatgrass was not reduced by imazapic and targeted grazing (instead it initially increased) nor targeted grazing alone, though it was by herbicide alone. 

An objective of managing annual grasses is that desired and perennial grass will increase – we did not observe this response at most of our sites.  Perennial grass cover was similar in our sprayed and reference sites in the Eastern Sandhills (10-15%) but higher in the reference area than sprayed area in the Antelope Peak (>20 versus 10-15%).  In our IWM study we did observe an increase in desired perennial grasses in the integrated herbicide with targeted grazing treatments, but not herbicide, targeted or simulated grazing alone. In our other studies we found that the herbicide, indaziflam, was effective at controlling cheatgrass, and perennial grasses responded favorably over three seasons.  In this same study the soil amendments and mustard seed meal did not result in a change to the desired forbs. Finally, our seeding restoration practices did not differ from the non-seeded control but we did observe more seedlings in all treatments where perennial grass cover was less than 20%.