Full methodological details can be found in Lynn et al. (accepted) Journal of Ecology)
Warming experiment
In the fall of 1990 at the Rocky Mountain Biological Laboratory (RMBL) in Gothic, CO, USA, ten permanent 10 x 3m plots were established oriented lengthwise along the ridge of a moraine. Five treatment plots were warmed year-round by overhead electric heaters (~22 W/m2 infrared radiation). The treatment warmed the soil surface (top 15cm) by ~2°C, dried the soil by ~10-20% gravimetric, and increased growing season by ~12 days through warming effects on snowmelt (Harte et al., 1995; Harte & Shaw, 1995; Saleska et al., 2002). Treatments were designed to reflect a warming scenario of doubled atmospheric CO2 based on conditions at the onset of the experiment (Harte & Shaw, 1995). Control plots without mock-heaters were alternated between warmed plots. Each plot was split into three blocks from top to bottom of the ridge, and warming effects on the abiotic environment were greatest in the top (driest) block, where we concentrated our sampling effort. Additional detail on experiment design and upkeep can be found in Harte et al., 2015, Harte & Shaw, 1995, and Rudgers et al., 2014. Other effects of long-term experimental warming included declines in soil carbon (Harte et al., 2015), increased dominance of the shrub sagebrush (Perfors et al., 2003), and increased representation of sedges over grasses without much change in total graminoid cover (Rudgers et al., 2014). On average across plots and years of this study (2015-2017), warmed plots melted out 38 days before the control plots (Julian day of 86 versus 124).
Focal plant species
We studied three dominant perennial grasses (Poaceae): Achnatherum lettermanii, Festuca thurberi, and Poa pratensis. Both A. lettermanii and F. thurberi are bunchgrasses, while P. pratensis is rhizomatous. Prior work on herbivory in the warming meadow had not included grasses (Roy et al., 2004), leaving a large portion of the plant community unstudied. Poa pratensis has significantly declined in the warmed plots compared to controls, and both A. lettermanii and F. thurberi trended towards declines in warmed plots (Rudgers et al., 2014). We focused on herbivore and disease damage to these three most common grass species to gain further insight into a potential indirect mechanism of their decline in response to warming.
Herbivory and disease measurements
In the beginning of each growing season (early to mid-June, depending on snowmelt), we used plastic zip ties to mark three randomly chosen individual tillers on each of six individuals per species per plot. Six individuals per species per the 3 X 3m upper zone made up between 60-100% of individuals of the focal species in the plots, anecdotally. This enabled us to track accumulation of damage on tillers throughout the growing season. If zip ties fell off the tillers, we randomly selected another tiller to track (~5% of observations). We sampled individuals at the top of the ridge in the first block, but for F. thurberi, we could not find the desired six individuals per plant species in plots 7-10. In this case we included individuals in the lower blocks until the desired number was achieved (see Figure S1 and S2 - no differences in the plots where this occurred).
Herbivory and, likely pathogen caused, disease damage were visually estimated by an observer, to the nearest 1% of leaf area damaged. Visual estimates of percentage damage were calibrated between two observers (one student and one expert – J.S. Lynn) by consensus. After one to two days of training, one-student observer carried out the rest of the observations for the season. In total, four observers recorded herbivory (N. Abo-Sido - 2015, I. McCowen -2016, S. Villanueva -2017, and J.S. Lynn -across years). These and similar methods of plant-enemy damage estimation are standard practice for the field (e.g., protocols from herbvar.org; Baskett & Schemske, 2018; Roy et al., 2004). Each grass tiller contained two-three leaves. Additionally, we noted the type of damage among the following classes: cell sucking damage (aphids, leaf hoppers, etc.), chewing damage (caterpillar, grasshopper), or leaf-miner damage (flies, moths). Disease was classified by symptom rather than species due to lack of funds for sequencing. We used the following classes with descriptions: powdery mildew (white powdery damage), rust (reddish-colored dusty damage), black (bulbous black pocks), brown (brown, oozy lesions), or yellow (yellowy discolored lesions) disease. We only counted disease if we could identify pathogen caused disease symptoms (e.g., spores, hyphae, molds, ooze) and no other factors (e.g., discoloration from abiotic stress or mechanical damage).
Climate data and manipulation
We used climate data from the National Atmospheric Deposition Program (http://nadp.slh.wisc.edu/siteOps/ppt/default.aspx) at the CO-10 site ID, which was approximately 50 m from the warming meadow. We investigated how fine-scale weather patterns correlated to the presence and amount of damage. We focused on temperature (°C) and precipitation (mm) as the key climatic variables. We calculated average daily temperature as the midpoint between daily minimum and maximum temperatures. Then, because we observed damage every two weeks, we averaged daily temperature values over the two weeks prior to each sampling date to obtain average conditions leading up to sampling. We summed the amount of precipitation over the two- and four-weeks prior to sampling. The two-week weather windows were chosen to cover weather variation between samplings. The four-week precipitation window was added because of the long “moisture memory” of the soil in the experiment (takes ~ four-weeks of no precipitation to reach below plant wilting point; Harte et al., 1995).
