Fitness differences override variation-dependent coexistence mechanisms in California grasslands

While most studies of species coexistence focus on the mechanisms that maintain coexistence, it is equally important to understand the mechanisms that structure failed coexistence. For example, California annual grasslands are heavily invaded ecosystems, where non-native species have largely dominated and replaced native communities. These systems are also highly variable, with a high degree of rainfall seasonality and interannual rainfall variability – a quality implicated in the coexistence of functionally distinct species. Yet, despite the apparent strength of this variation, coexistence between native and non-native annuals in this system has faltered. It is therefore uncertain to what degree rainfall variation can offset average fitness differences between native and non-native annual plants in California grasslands to promote coexistence, nor what coexistence mechanisms are most relevant. To test these dynamics, we implemented a competition experiment between five annual species of native forbs and non-native grasses. We grew individuals from each species under varying densities of all other species as competitors, under either wetter or drier early-season rainfall treatments. Using subsequent seed production, we parameterized competition models, assessed the potential for coexistence among species pairs, and quantified the relative influence of variation-dependent coexistence mechanisms. Overall, we found little potential for coexistence. Competition was dominated by the non-native grass Avena fatua, while native forbs were unable to invade non-native grasses. Mutual competitive exclusion was common across almost all species and often contingent on rainfall, suggesting rainfall-mediated priority effects. Among variation-dependent mechanisms, the temporal storage effect had a moderate stabilizing effect for most species, while relative nonlinearity in competition was largely destabilizing, except for the most conservative non-native grass, which benefited from a competitive release under dry conditions. Our findings suggest that rainfall variability does little to mitigate the fitness differences that underlie widespread annual grass invasion in California, but that it influences coexistence dynamics amongst the now-dominant non-native grasses.

Methods
Datasets contain plant phytometer and background competitor stem counts, biomass, and allometric fecundity estimates from a 6-species surface-response competition experiment in a California annual grassland. The six species were: Avena fatua (introduced annual grass), Bromus hordeaceus (introduced annual grass), Vulpia myuros (introduced annual grass), Trifolium hirtum (introduced annual, nitrogen-fixing forb), Lasthenia californica (native annual forb), and Eschscholzia californica (native annual forb). The competition experiment took place in year 3 (Fall 2016 - Spring 2017) of a 3-year (Fall 2014 - Spring 2017) seasonal drought manipulation study at Sierra Foothill Research and Extension Center (SFREC), Browns Valley, California (field coordinates: 39.25117507590813, -121.31116918773333). See Shaw et al. 2022 for details on the overarching 3-year seasonal drought experiment. Competition experiment and data processing abbreviated methods are described next. See Muehleisen et al. 2024 for detailed methods.

Methods for data collection:

Species were seeded October 2016 before germinating rains, and field data collected April-May 2017 during peak production of the growing season. Species were seeded in 50 cm x 50 cm sub-subplots nested within one of two 2 meter x 1 meter subplots, which were nested in one of 16 6.4 meter x 5.2 meter experimental plots. Experimental plots were installed in a single row, west to east, in a southwest-facing field, and blocked in sets of 4 sequentially (e.g., plots 1-4 were in block A, plots 5-8 in block B, etc.). Experimental plots were either droughted in the fall germination window (October to January) using a retractable clear polyethylene roof (hereafter 'shelter'), or were not sheltered and received ambient rainfall in the fall (control plot). There were two fall dry plots and two control plots per block for a total of 8 fall dry plots and 8 control plots. Drought treatment within block was randomized.

In each experimental plot, each of the 6 species were seeded once at low- and high-densities as a background competitor in a 50 x 50 cm sub-subplot (one background species and density per sub-subplot, e.g., Bromus hordeaceus at low density). We selected a seeding rate of 4g per meter^2 for the low density treatment and 16g per meter^2 for the high density treatment based on previous work in a similar California annual grassland (Kraft et al. 2015). Whichever 5 species were not seeded as the background competitor were seeded in the sub-subplot as a phytometer in staggered positions 1-6 (one species per phytometer position, with one position empty when background competitor present). Location of phytometer positions 1-6 were standardized across sub-subplots. Assignment of phytometer seeding position within the sub-subplot, and assignment of background competitor and density within the 2x1m subplots were randomized. To achieve at least 3 individuals per species phytometer, phytometers were seeded based on seed weight, with more seeds sown for species with relatively less seed mass (e.g., Lasthenia californica), and fewer seeds sown for species with larger seed mass (e.g., Avena fatua). This amounted to approximately 8-10 seeds sown for A. fatua and T. hirtum (heavier seed masses), 12-13 for V. myuros and B. hordeaceus, and 15 for L. californica and E. californica. To estimate species biomass and fecundity in the absence of competition, within each experimental plot all 6 species were seeded as phytometers in one sub-subplot that had no competitor (control). All weeds were removed from plots prior prior to seeding, and plots were weeded once in December to mitigate encroachment by non-experimental vegetation. Heavy rains occurred Nov 2016 - Mar 2017. The field site had a slight slope and some plots were disturbed by overland flow as water ran downhill through plots (including in drought-sheltered plots). Some plots were additionally disturbed by gophers or deer. Disturbance is noted in the dataset 'Competition_combined_clean.csv'.

In April 2017, we surveyed the density of background competitors, flowering for forb species, and number of phytometers that matured. We censused competitor density at the entire 50 cm x 50 cm sub-subplot scale when feasible, and otherwise subsampled at 10 cm x 10 cm scale for A. fatua or 5 cm x 5 cm for all other species. When growth of competitors was patchy we recorded competitor percent cover in the sub-subplot. In this grassland, forb species tend to peak in their biomass in April and grasses in May. For each forb species in each sub-subplot, as applicable, we harvested at least 3 phytometer individuals (specimens) or fewer if 3 were not present, or 3 background competitor specimens per density treatment. In May we repeated the harvest for grass species. When available, we also harvested additional forb phytometers present in May for quality control comparison with April biomass. To estimate an allometric fecundity to biomass relationship per species, we harvested at least 20 specimens per species in fall dry and 20 specimens per species in control plots. We harvested these allometric specimens from individuals that remained as background competitors once specimens for phytometers and background competitors were collected. We selected the 20 specimens on a size gradient of small to large, choosing from both low density and high density seeding treatments. Due to poor growth of T. hirtum and E. californica during the experiment (not enough mature individuals for allometric specimens after harvesting phytometers and background competitors), we collected allometric specimens for these species in similar field sites later. These field sites did not have comparable drought experiments and so we collected specimens in ambient conditions (20 each on a size gradient of small to large). All specimens (phytometers, background competitors, and allometric specimens) were dried for 72 hours at 120 degrees F, then weighed on a top-loading balance with ±0.01g accuracy. Any specimen weighing 0.02g or less on the top-loading balance was weighed on a Mettler Toledo AB54-S analytic balance with accuracy of ±0.0001g for a more precise weight. We also counted seeds on all allometric specimens, including seeds that would have been present in the case some already released (e.g., for grass species).

Methods used for data processing:

All data quality control, preparation, and analysis performed in R. We modeled allometric fecundity-biomass relationships using ordinary least squares linear regression, one regression model per species, with the intercept fixed at 0. Specifically, for each species we used the form: 'lm(seeds ~ 0 + biomass)'. We initially modeled relationships separately for data collected in fall dry vs. fall control plots, found no strong differences in relationships based on fall rainfall condition, and so pooled data collected in both conditions for final species allometric models. We rescaled background competitor density from the scale collected in the field to 1 meter^2 scale, adjusting for patchy cover as needed using percent background cover values. We averaged each set of phytometer specimens and background competitor specimens collected per sub-subplot to achieve a single phytometer and competitor biomass per species per sub-subplot. In about 25% of cases (269 of 1056), more phytometers for a species grew in a sub-subplot than were harvested for ANPP. We therefore also projected total biomass of all species phytometers present in a sub-subplot using the mean phytometer weight and count of total phytometers in the field. Finally, we estimated fecundity of phytometers using the allometric relationships derived for each species and biomass of phytometers. Specifically, we used the base R function 'predict()' with the species allometric linear regression model and phytometer biomass data as new data to predict from. Seven phytometers were accidentally not harvested during field data collection and so do not have estimated fecundity.

We generated two csv file datasets, 'Competition_allometric_clean.csv' and 'Competition_combined_clean.csv', using the base R function 'write.csv'. Attribute type descriptors for dataset variables described below follow Ecological Metadata Language ([https://eml.ecoinformatics.org/](https://eml.ecoinformatics.org/)) as closely as possible.

References:

Kraft, N., Godoy, O., & Levine, J. (2015). Plant functional traits and the multidimensional nature of species coexistence. Proceedings of the National Academy of Sciences. 10.1073/pnas.1413650112.
Muehleisen, A. J., White, C. T., Shoemaker, L.G., Suding, K. N., Shaw, E. A., & Hallett, L. M. (2024). Fitness differences override variation-dependent coexistence mechanisms in California grasslands. Journal of Ecology. In press.
Shaw, E. A., White, C. T., Silver, W. L., Suding, K. N., & Hallett, L. M. (2022). Intra-annual precipitation effects on annual grassland productivity and phenology are moderated by community responses. Journal of Ecology, 110, 162–172. [https://doi.org/10.1111/1365-2745.13792](https://doi.org/10.1111/1365-2745.13792)