Model fitting is less sensitive to smaller n, allows for unbalanced designs , and is easier to calculate variances . Our experimental design for many performance measures requires subsampling multiple individuals within plots, and model fitting with random effects allowed for the incorporation of the variance among subsamples, which would be lost if we took the average to calculate an FEV. The performance measures of biomass, cover, and height were log transformed and fit with linear mixed effect models . We performed model selection using backwards stepwise Akaike Information Criterion and the ran ANOVAs followed by post-hoc multiple comparison tests. Percent cover was modeled at a) the community level to capture net community feedbacks that may be hidden when comparing smaller differences among species and b) the individual species level to see if a particular species was driving the feedback. Biomass was assessed only on the community level. Plant height was modelled individually for each species, as height inherently differs due to natural history, and we are primarily interested in how soil conditioning effects within-species height variation. Seed production measures were fitted with generalized linear mixed-effects model for the negative binomial family with the “glmmTMB” package . Likelihood ratio tests were used to determine significance of fixed effects. Native flowering individual was modeled on both the individual species level and community level, as done previously for percent cover. Kruskal-Wallis tests were used on A. triuncialis and S. pulchra seed production, vertical growing racks as the data violated model assumptions due to few surviving individuals. Germination of all seeded species was analyzed with a time-to-event model using the “drc” and “drmSeedGerm” packages .

This method models the cumulative germination curve with the interaction of soil conditioning and community treatment, assuming a log-logistic distribution of germination time and accounting for ungerminated seeds. The model parameters that correspond to time to reach 50% germination and maximum germination were compared via the compParm function to evaluate the drivers of differences in germination patterns between conditioning and community treatments.This long-term field study demonstrated that native and exotic communities both experienced negative feedbacks– performing better in the other community’s soil. The eleven year conditioning phase provided a unique opportunity to assess feedbacks in response to soil changes that may take years to develop, such as build-up of soil organic matter. Setting both experimental phases in the field was necessary to understand the overall strength and role of feedbacks in structuring plant communities in a more realistic setting, as both phases experienced high environmental variability. Any long-term soil changes due to vegetation composition were strong enough to be detected over the observed variability in soil texture, ground squirrel disturbance, and annual precipitation . The two-year feedback phase also experienced strong variation in annual precipitation, as it included an extremely wet year followed by an extremely dry year . The detection of feedbacks required measuring multiple plant responses, as the natives and the exotics did not exhibit feedbacks within the same trait, nor did either show a feedback in above ground biomass, the most commonly measured indicator of plant fitness. Overall, our results suggest the need to measure multiple traits and life stages to capture potential responses to soil conditioning . We detected negative feedbacks on both native and exotic grasses, suggesting California grasslands are structured by negative feedbacks, similar to many other systems , and that positive feedbacks are not a main factor in exotic grass dominance .

The feedbacks were stronger for the native community, which had improved performance in exotic soil, as measured by cover, height, and reproductive productivity. These results are similar to a number of other studies, indicating that negative plant-soil feedbacks are stronger for native than exotic species . Previous greenhouse studies of California grassland species showed that in the short term , native species experienced positive feedbacks and performed worse in exotic conditioned soil, the opposite of our results . This may be due to different feedback mechanisms at play in a shorter conditioning phase or the difference in life stages tested. The exotic grass community had greater below ground biomass in native-conditioned soils. Another feedback experiment with similar species linked greater deep-root exotic biomass to the increased soil organic matter and water holding capacity in deep soils cultured for 10 years by the native perennial Stipa pulchra . Our study did not detect these physical and chemical soil differences but that may be because our site was drier and had higher clay content. It is also possible that these soil changes did occur and were ecologically but not statistically significant in our study. It is also possible that other driving mechanisms increased exotic deep root-growth in native soils at our site, such as potentially enhanced soil aggregation from the deeper perennial roots with more fine root hairs which would increase plant suitability due to more pore space and water flow .The native vs. exotic communities did not differ in their effects on soil water holding capacity, percent carbon or nitrogen, nor soil organic matter. Differences did arise in their effects on net nitrogen mineralization and nitrification, and fungal and bacterial community composition. The exotic plant communities cultivated significantly different fungal and bacterial community compositions from those cultivated by natives, similar to many other studies .

The fungal communities were distinct across soil conditioning treatments at all soil depths, but the bacterial communities differed by conditioning treatment only in the shallowest and deepest soil zones, supporting our hypothesis that differences in rooting depths are a factor in soil conditioning. Changes in microbial composition may directly impact plant performance , or may indirectly alter plant performance by changing resource availability . The native vs exotic communities also differed in effects on the abundances of nitrifying bacteria, AMF, fungal pathogens and saprotrophs, and soil nitrogen cycling, and below we’ll discuss how similar changes have been linked to plant performance in other feedback studies. However, determining the causal mechanisms behind the observed feedbacks is beyond the scope of this experiment, as the observed net feedbacks might be due to these differences in soil properties, properties not measured in this study, or from their interactions.Fungi in the plant pathogen and plant pathogen-saprotrophs guilds were greater in abundance in native-conditioned soil compared to exotic soil. This difference may be due to the life histories of the two grass groups, as the perennial native roots provide a constant food source for pathogens whereas the annual exotic roots die off every spring. Alternatively, fungal pathogenic richness is positively correlated with specific root tip number, which is likely higher in native roots as they are deeper with more fine root hairs . Greater pathogen abundance in native-conditioned soil plots could also be attributed to being largely dominated by S. pulchra in the conditioning phase, and evidence suggests that greater soil pathogen accumulation occurs under monocultures . Negative plant-soil feedbacks are common and thought to maintain plant diversity, and are often attributed to the buildup of localized pathogens . Thus, the strong negative feedback observed in the native grass community may be influenced by lower exposure to pathogenic attack in exoticconditioned soils. In a study with similar species, native grasses also experienced a strong negative feedback, but doubled in growth when native soils were sterilized , supporting the suggestion that pathogens are driving the feedback. Further, pathogens accumulate over time, and the longer conditioning phase of our experiment may explain the difference in feedback direction compared to shorter-term studies . The lower fungal abundance in exotic-cultured soils did not benefit exotic grasses, however, as they performed worse in their own cultivated soils. They may culture their own specific pathogens and also be less susceptible to generalist pathogens than native grasses . Saprotrophs were also greater in native-conditioned soils and could potentially have become parasitic which can occur when soils are dominated by a single species, growing racks and the native soils were largely conditioned by S. pulchra . However, while these could potentially contribute to the natives’ negative PSF, these effects might be short-lived, as saprotroph composition can change quickly in response to the current plant community .We did not find any difference in total N between the two conditioned soils, but net mineralization and nitrification rates were lower in soils conditioned by exotic grasses compared to the native perennials. These results are similar to those found by Parker et al. 2012 and Carey et al. 2017. Lower net rates can occur because decomposition and N release are slower , and/or when there is higher microbial immobilization, resulting in high competition between plants and microbes for N .

If the difference in rates of N cycling influenced native performance, we would expect greater native above ground biomass on native-conditioned soil, which we did not see. Corbin and D’Antonio suggest that changes in mineralization and nitrification rates are easily reversed under a new plant community, and so would not lead to feedbacks. We also detected changes in the nitrifying bacteria; ammonia oxidizing and nitrite oxidizing bacteria were both greater in the deeper exotic-conditioned soils than the native-conditioned soils. As there are fewer deep roots in exotic plots, bacteria in those soils have greater access to ammonia and nitrite. Similarly, Hawkes et al. looked at soils under the native S. pulchra and exotics B. hordeaceus and A. barbata, species used in our study, and also found greater AOB abundance in exotic soils.The presence of a competitor can eclipse, neutralize, or change the strength of a feedback in a species . Assessing plants grown in a community with both intra and inter-specific competition can thus help us understand the proportional role of feedbacks in community structure, particularly if we are interested in communities comprised of species commonly found together in nature, such as our native mix and our exotic mix. Our community treatments do not allow us to tease apart individual species’ contributions to soil conditioning nor touch on whether the direction and magnitude of a feedback is dependent on the specific identity of the conditioning plant and the feedback plant , but did show feedbacks were still observed overall in our two communities. In California grasslands, certain groups can remain dominant over time, but the stability of the group is due to variations in which species dominate within the groups, as environmental conditions change . Our results thus are relevant to the diverse and shifting communities found in California grasslands, especially as feedbacks can be non-additive in a community compared to monoculture . Community type played a major role in native grass abundance, and only slightly affected the exotic grasses. Native establishment was so poor in the mixed native and exotic community that we were not able to analyze most performance measures. This is not surprising, as exotic annual grasses in this system both germinate and grow much faster than native perennials, which allows them to outcompete the native seedlings for light and soil moisture . The few native individuals growing in competition with the exotics, however, still experienced a negative feedback in percent cover. Thus, feedbacks influence natives more subtly in full competition with exotics but can be a stronger control in restoration settings where competition by exotics is actively minimized. As the exotic grasses are the better competitor at the seedling stage, competition with natives did not alter exotic feedback overall. Our results show that competition among native and exotic grasses clearly outweighs the role of plant-soil feedbacks in community structure, with exotic dominance resulting regardless of soil provenance.Our results do not support our hypothesis that exotic annual grass invasion negatively impacts native restoration through the soil, suggesting that soil amelioration may not be necessary to improve restoration success. Studies comparing remnant, restored, and invaded grasslands found that soil biotic communities take years to recover , although another found mycorrhizal communities specific to S. pulchra returned quickly after restoration . Fortunately, it appears that exotic soil conditioning at our site does not majorly hinder native establishment and restoration success but rather benefits natives, even though exotic soils still have been found to decrease native performance when compared to sterilized soil , highlighting native grass susceptibility to pathogens. When establishing native cover is the main goal of a restoration project, reaching 30% native cover is considered a success. Higher cover is very important for the long-term success of a project, as once established, native perennial grasses become more competitive against future exotic annual seed pressure . Thus, the 15% difference in percent cover observed in our study is substantial.