Restore into the future: post-fire rangeland restoration in the Great Basin

Objective 1. Assess the effects of current post-fire seeding practice on local plant species and genetic diversity in burned areas. We predict that native grass seeding will increase local species and genetic diversity in burned-seeded areas relative to burned-unseeded areas, because burned-unseeded areas are dominated by cheatgrass. We focus on plant diversity at species and genetic level here because we are concerned about the adaptive capacity of plant communities after post-fire seeding. Based on the resilience framework, species diversity buffers against multiple stressors such as drought (Ives and Carpenter 2007, Hautier et al. 2014), and genetic diversity ensures the populations’ potential to evolve when conditions change in the future (Hoffmann and Hercus 2000). Our hypothesis will be supported if species diversity and genetic diversity in burned-seeded areas are higher than those in burned-unseeded areas. Alternatively, seeded areas may have lower diversity than unseeded areas, due to low diversity in seed mixes which contain only 4-5 species from 1-2 source populations (Leger and Baughman 2015). The effect of these low diversity seed mixes on plant diversity of restored communities is not well tested but is likely to reduce the resilience of native perennial grasslands.

Objective 2. Evaluate where on the landscape to prioritize post-fire seeding for native vegetation recovery in burned areas. We predict that dispersal will increase recovery of native species and genetic diversity closer to fire edges, making post-fire seeding more important farther away from the fire edges. Wind dispersal of seeds from unburned areas is the natural process of post-fire recovery in burned areas (Damschen et al. 2014). Seed dispersal exponentially decreases with distance to the nearest seed source or fire edge (Nathan and Muller-Landau 2000). Therefore, in large wildfires, seeds may not reach some burned areas because the distance between unburned areas may be too great for the seeds to disperse. Post-fire seeding attempts to overcome this seed limitation (Foster and Tilman 2003). The effect of seeding on plant diversity may depend on distance from the nearest fire edge. We expect post-fire seeding in the core of these large burns will be most helpful for maintaining a resilient landscape. Our hypothesis will be supported if the effect size of seeding (difference in species and genetic diversity between seeded and unseeded areas) increases with distance from the nearest fire edge.

Objective 3. Identify native grass seed provenance that improves the drought tolerance of restored sites. We predict that the southern populations of native grass species will exhibit higher drought tolerance than the northern populations. This drought tolerance will help provide reliable forage in a hotter/drier future. Our preliminary results indicate that seeds sourced from southern populations in drier sites have higher seedling success rate when grown in a common garden than seeds from local or northern populations in wetter sites (Fig. 1). We expect that this effect is associated with more conservative traits (such as higher water use efficiency) in these populations. A plant trait-based approach using simple-to-measure aboveground traits has been successful in explaining trait-by-environment associations across species (e.g., Baughman et al. 2019, Basey et al. 2015). We will test the association between seed provenance, plant traits, and drought survival. Our hypothesis will be confirmed if southern populations have higher water use efficiency, lower water stress, and higher survival rates than northern populations under drought.

Objective 4. Understand the relationship between the drought tolerance and cheatgrass resistance traits of native grasses. Keeping cheatgrass at bay is important because we expect a tradeoff between drought-tolerant and cheatgrass-resistant traits will decrease survivorship of native grasses under combined drought and invasion. Drought-tolerant individuals have resource-conservative traits (e.g., high water use efficiency (Farquhar et al. 1989) and low growth rates (Craine et al. 2012)) to save energy. In contrast, cheatgrass-resistant individuals are known to have resource-acquisitive traits (e.g., early phenology and high growth rates (Leger et al. 2019)) to compete with annual grasses that emerge in winter and senesce by summer. However, the competitive outcomes for restored natives vs. cheatgrass under contrasting precipitation conditions are still unknown. Understanding the trait association with seedling performance under drought and invasion would inform practitioners which native species and population to use for pre-emptive seeding practices depending on the extent of cheatgrass invasion or the likelihood of future drought in the area. Our hypothesis will be supported if native seedling survival rate increases with drought-tolerance traits under drought but decreases under drought and invasion, and vice versa.

RESEARCH METHODS AND MATERIALS. We tested our hypotheses using two approaches: (1) a Survey where we measure plant diversity at a community and population level in a previously burned area; and 2) a Common Garden Experiment where we measure intraspecific variability in native grass response to drought and cheatgrass treatments.

Objective 1 & 2: Survey of post-fire plant communities.

Sampling design. Our survey took place in Soda Fire, where over 280,000 acres of rangelands burned at the border of Oregon and Idaho in 2015 (MTBS; Fig. 2). Taking advantage of existing USGS monitoring effort, we identified 40 plots stratified by burned-unseeded (n = 15), burned-seeded (n = 15), and unburned-unseeded (n = 10) areas. Using distance from the nearest fire edge as a proxy for landscape dispersal, I further stratified burned-unseeded and burned-seeded plots by five distance classes: <1000 m, 1000-2000 m, 2000-3000 m, 3000-5000 m, >5000 m. Burned-seeded plots distributed across three seeded areas (see Fig 2 for area C: northwest; area D: southwest; area E: east) where they were previously drill seeded with bluebunch wheatgrass Pseudoroegneria spicata) in fall 2015-2017. Bluebunch wheatgrass is a native perennial grass species widely used for post-fire restoration for its high fire tolerance (Miller et al. 2013) and relatively high establishment rate after fire (Hudson 2016). 

Methods. We collected plant composition data and leaf samples of bluebunch wheatgrass at all 40 plots in June 2021. Each plot was delineated by a 10 x 10 m relevé, where number of species is recorded for species richness. Occurrence of each plant taxon was measured every 0.5 m on four 25-m transects radiating from each corner of the relevé using the point-intercept method, and evenness was calculated. Leaf samples were collected from 20 individuals at each plot (n = 800), then air-dried and stored at -80F. We isolated DNA from plant tissues using a modified cetyltrimethy ammonium bromide (CTAB)-based method (Aboul-Maaty and Oraby 2019). We amplifed small quantity of DNA with polymerase chain reaction (PCR) assay using ten known primers for bluebunch wheatgrass (Bushman et al. 2008). To measure genetic diversity of bluebunch wheatgrass populations, we used a genotyping service at the Center for Genome Research and Biocomputing at the Oregon State University to obtain microsatellite data in Fall 2021. With the microsatellite data, allelic richness was calculated as a metric of genetic diversity at the population level.

Analysis. We used ANOVA to test the combined effect seeding on overall species diversity (species richness) and the genetic diversity (allelic richness) of bluebunch wheatgrass, with unburned-unseeded, burned-seeded, and burned-unseeded as explanatory variables. To test how seeding effects change with landscape dispersal, we regressed species and genetic diversity in burned-seeded/burned-unseeded plots against distance from the nearest fire edge in linear models.

Objective 3 & 4: Common Garden Experiment. Experimental design. In October 2019, we established a sagebrush steppe common garden experiment at the Northern Great Basin Experimental Range in Burns, OR. We chose a flat site with low cheatgrass (Bromus tectorum) abundance, a common invasive annual grass in Great Basin rangelands. We set up 1.5 x 3 m plots in a random block design with three rainfall treatments (n = 24 plots; 8 blocks x 3 rainfall treatments): 0% reduction for ambient rainfall, 50% reduction for moderate drought, 80% reduction for severe drought. Each plot was subdivided into twelve 25 x 25 cm subplots for nested seeding treatments (6 populations x 2 replicates). We initially nested cheatgrass competition treatments within rainfall treatments; the average densities of cheatgrass were 145.0 ± 4.0 individuals/m² in competition plots and 7.8 ± 1.0 individuals/m² in non-competition plots. We combined those plots as replicates of seeding treatments because there was no effect of competition on first or second year seedling survival. To reduce microsite variability from shrubs, we first manually removed aboveground biomass of shrubs to prepare the site. We constructed fixed rainout shelters that passively reduce precipitation by a constant percentage (Gherardi & Sala, 2013). Aluminum flashings were inserted 35 cm deep around the shelters to reduce surface and lateral soil flow on to the plots as well as to limit plant roots from accessing water outside the treatment. Plots were fenced with 4-ft high wire mesh that excluded cattle but not native herbivores (e.g. rabbits, gophers). In October 2019, we burned each plot with a propane blow torch to mimic post-fire conditions. To account for microclimates across rainfall treatments, we deployed soil moisture and temperature sensors (TEROS 10 and ZL6 Data Logger, Meter Group) and air temperature sensors (iButton, Maxim Integrated) in six plots. Soil moisture and temperature sensors were installed at 5 cm below the soil surface; air temperature sensors protected by radiation shields (following design by Holden et al., 2013) were installed at 10 cm above the soil surface. These sensors logged data every 2 hours year-round. In September 2020, we raked the annual plants and dug out the perennial forbs and grasses with a shovel; after removing the vegetation, we broadcasted 300 E. elymoides seeds in each subplot (one seed source per subplot). We reseeded the same density in September 2021.

Methods. In July 2021 and 2022, we measured survival rates of seedlings. Germination temperature optima of each seed source were determined by conducting germination trials in a lab at University of Oregon in November 2021. We stratified seeds in a refrigerator at a constant 4 °C for 4 weeks prior to growing them in a growth chamber. We sowed each population (25 seeds per population, 5 replicates) separately on a wet filter paper in a petri dish for 2 weeks in a constant temperature of 8, 10, 12, 15, and 20 °C. We added 2 mL of water to each petri dish at the start, and watered every 2 days to keep the filter moist. Lights were on from 6 am to 6 pm daily. Numbers of germinated seeds were counted on the 14^th day. Mean proportion of germination was calculated by the dividing number of seeds germinated by the number of seeds sown. In July of 2022, we measured leaf ¹³C isotopic composition (δ ¹³C, ‰), which reflects rates of photosynthesis relative to stomatal conductance integrated over the lifespan of the leaf, as a metric of water use efficiency (dehydration avoidance; Cernusak et al., 2013). Previous work demonstrated clinal variation in carbon isotopic composition across populations of E. elymoides (Blumenthal et al., 2021). We sampled fresh leaves from two individuals of second year seedlings per subplot, and oven dried them at 60 ⁰C for 72 hours. Dried leaves were ground to powder in a ball mill, rolled in aluminum tins, and analyzed for carbon isotope composition at the Stable Isotope Lab at the University of Oregon.

Analysis. To compare the effects of seed origin and precipitation treatment (ambient, 50 % and 80 % reduction) on seedling recruitment, we built mixed models with source population and precipitation as fixed factors and block as a random factor. To compare stress tolerance strategies across seed sources, we built one-way ANOVA models with germination temperature optima and carbon isotope signature (δ ¹³C) of second year seedlings as dependent variables and seed sources as an explanatory variable.