Data Package Metadata   View Summary

Experimental Design. The complete BioCON (Biodiversity, CO2, and N) experiment includes 371 2 x 2 m plots in six circular 20-meter diameter rings, located at the Cedar Creek Ecosystem Science Reserve in Minnesota, USA. Plots were established on secondary successional grassland on a sandy outwash soil after removing the prior vegetation. The BioCON project includes several overlapping and nested experiments. The main biodiversity x CO2 x N experiment design (n=296 plots) consisted of a split-plot arrangement of treatments in a completely randomized design. CO2 treatment (ambient or +180 µmol CO2 mol-1) is the whole-plot factor (ring scale) and is replicated three times among the six rings. N treatment (ambient or enriched with 4 g N m-2 yr-1) was a subplot factor (plot scale) assigned randomly and replicated in half of the individual plots among the six rings (28, 42, 43). Planted richness (1, 4, 9, or 16 species) was a subplot factor (plot scale) assigned randomly among plots in the six rings (28,42,43). All 16 species were planted individually in 8 monoculture plots (2 per unique CO2 and N treatment) and all together in 12 plots per unique CO2 and N treatment. There were 15 plots per unique CO2 and N treatment planted with either 4 or 9 species, with the individual species assignment in each plot drawn at random from the full pool of 16 species.

Plots: The present study includes 108 plots drawn from the main biodiversity x CO2 x N experiment; all of those originally planted with 9 or 16 species and experimentally treated with the complete factorial combination of CO2 and N levels. Thus, within each ring, there were 5 and 4 plots planted with 9 and 16 species, respectively, with ambient N treatment and another 5 and 4 plots planted with 9 and 16 species, respectively, with enriched N treatment. Across rings, there were 15 and 12 plots planted with 9 or 16 species, respectively, at each unique combination of CO2 and N treatment. Beginning in 2007, 2 of the 5 nine species plots at each N treatment level in each ring began to be treated annually with rainfall reduction, and in 2012, 2 of the 5 nine species plots at each N treatment level in each ring (one of which had rainfall reduction treatments) began to be treated annually with warming treatments (30). As described later in the analyses section, we conducted statistical analyses both removing plots that eventually had altered rainfall and/or temperature treatments from the entire analyses and retaining all plots in the analyses. Results were similar with analyses done both ways. In addition, we tested whether rainfall and/or temperature treatments influenced the CO2 x N x year interaction (i.e. testing for those four or five way interactions), using data from 2007-2021 on for rainfall treatments and 2012-2021 on for temperature. None of those four- or five-way interactions were significant (P>0.05). Hence, responses of species richness to CO2 x N and how that changed over time were not influenced by treatment induced variation in rainfall or temperature (that was in any case balanced across CO2 x N treatments). Given similar results whether including or removing plots treated with rainfall and temperature, we used all plots in analyses presented herein.

Species: The 16 species used in this study were all native or naturalized to the Cedar Creek Ecosystem Science Reserve. They include four C4 grasses (Andropogon gerardii, Bouteloua gracilis, Schizachyrium scoparium, Sorghastrum nutans), four C3 grasses (Agropyron repens, Bromus inermis, Koeleria cristata, Poa pratensis), four N-fixing legumes (Amorpha canescens, Lespedeza capitata, Lupinus perennis, Petalostemum villosum) and four non-N-fixing herbaceous species (Achillea millefolium, Anemone cylindrica, Asclepias tuberosa, Solidago rigida). Since the experiment began Agropyron repens has been renamed Elymus repens and Koeleria cristata has been renamed Koeleria macrantha. For consistency with prior publications from this experiment we continue herein to use the prior name. Each 16-species plot was planted in 1997 with 12 g m-2 of seed partitioned equally among the 16 species. Each 9-species plot was a random draw from all 16 species, with 12 g m-2 of seed partitioned equally among the 9 species. All BioCON plots were weeded annually to remove species not in the initial planting; however, the 9 and 16 species plots resist invasion and experienced modest weeding. Enriched N treatments on unweeded grassland plots elsewhere at Cedar Creek had similar effects on species richness as we found in this study, suggesting the overall patterns observed herein are likely representative of unmanipulated as well as manipulated assemblages.

Treatments: Beginning in 1998, the equivalent of 4 g N m-2 yr-1 (NH4NO3) was added to all plots assigned to the enriched N treatment, in three doses during the growing season (in May, June, and July). This N addition is comparable or slightly larger than the average annual net N mineralization rate in similar secondary grasslands on these soils. Beginning in 1998, a free-air CO2 enrichment system was used during each growing season to maintain the CO2 concentration at an average of +180 µmol mol-1 in elevated treatments (three rings) during all daylight hours from spring (early April) to fall (late October to mid-November) each year. The three ambient CO2 rings were treated identically but without additional CO2.

Species composition and richness, biomass sampling and biogeochemistry measurements: In each year (unless otherwise noted), plant species composition and richness, above- and below-ground biomass, % soil moisture, % light transmission, plant C and N, and soil solution N concentration were assessed in every plot (22, 29, 30). Soil solution N concentrations (total) were measured in each plot every year with four 2.5 cm diameter cores taken from 0-20 cm depth during early to midsummer (typically late June). The cores were composited, sieved (2 mm), and extracted with 1 M KCl. Extracts were analyzed for NO3- and NH4+ on an Alpkem auto-analyzer (OI Analytical, USA). Percent soil moisture and light transmission were measured repeatedly throughout each growing season in every plot. Light transmission was measured using an 80-sensor linear array, the AccuPAR LP-80 (Decagon Devices, Pullman, WA, USA). Each sensor measures photosynthetic photon flux density in the 400-700 nm range. For each measurement, the sensors were arrayed above (1 measure) and below (average of 3 measures) the live vegetation in each plot, with the latter divided by the former (x 100) reflecting % light transmission, a proxy for light availability. Soil moisture was measured using time domain reflectometry at the 0-20 cm depth. Average light transmission and percent soil moisture data taken between May 1 and July 31 each year were used to assess treatment effects on these environmental variables, as well as their relations with species richness. Presence and percent cover estimates were made visually in July for each of the 16 species in a permanent 0.50 m2 zone (50 x 100 cm) of each plot that throughout the experiment was neither sampled for biomass nor had soil cores removed. Aboveground biomass was harvested elsewhere in every plot in early August by clipping a 10 x 100 cm strip just above the soil surface; these locations rotated year by year among 10 such locations in each plot. All biomass was collected, sorted to live material and senesced litter, dried and weighed. Live material was considered as current plant biomass, was sorted to species, and used to assess species richness and relative abundance of each species (defined as fraction of total aboveground biomass). The two independent estimates of species richness for each plot (from sorting of clipped biomass and from visible estimates of presence and percent cover, done in different areas of the plot) were averaged for each plot and year. The average was used because (a) there is no a priori reason to consider one measure more reliable than the other (they are well correlated in any case); (b) as they were done in different parts of the plot they double the sampling intensity; and (c) as each was done by different researchers within and among years, use of both together helps smooth out “observer bias”. In cases where clipped and sorted biomass was missing (e.g., in 2005 and 2006 for all nine species plots, in 2020 for low rainfall 9-species plots), we used only the percent cover-based data. In 2019, in a separate study, species richness was assessed in 70 of the 9 and 16 species plots using the identical percent cover method in 324 10 x 10 cm grid cells in the central 1.8 x 1.8 zone of each 2 x 2m plot. The total aggregate number of observed species among all 324 sampled cells in each plot was significantly related for the 70 plots to the number of species originally planted (9 or 16) and the observed neighborhood richness (P less than 0.001, R2=0.73). We applied the coefficients of this model to all years neighborhood richness data in ambient plots to obtain an estimate of total richness at near whole-plot scale. We used this to compare changes over time in richness at plot scale with other published studies of changing grassland richness over time, that tended to be at a similar scale. This is relevant, because as observed in some grasslands (34-36), but not those recovering from disturbance (37,38) the diversity of our experimentally assembled communities declined with time, including under ambient conditions at both neighborhood and plot scales. Richness measured at the neighborhood scale was one-fourth to one-half less than the estimated available species pool at near whole plot scale (3.2 m2), showing that neighborhoods did not contain the full available species pool. Moreover, the fraction of that pool observed at the neighborhood scale declined over the experiment, suggesting increased control of realized neighborhood richness by species interactions over time. Gains in all treatments in neighborhood species richness (SR) from new species recruits would dampen the degree of reduction in SR over the 24 years, and if different combinations of CO2 and N availability led to different magnitude of gains in SR, this could alter the contrasts in their interactive effects. Data on new colonizers were acquired in 10 of the first 11 years of the study in each plot by removing, drying and weighing all individuals of all species not originally planted there. Because we seeded at a relatively high density and had successful establishment of most species, 9 and 16 species mixtures were fully stocked and dense, and difficult to colonize from the beginning of the experiment through the end. For example, in monocultures of all 16 species, new recruit biomass (of any of the other 15 species or of species not included in the experiment) averaged 14.2% (median) of total plot biomass. By contrast, these values were much smaller in plots planted with 9 (1.2% median) and 16 species (0.15%, median), respectively. Thus, 9- and 16-species plantings were on average much less invasible than plots planted with the same species in monocultures and generally resistant to invasion. More germane to this issue, in 9- and 16-species planted plots, neither main effects of CO2 or N nor their interaction had significant effects on non-target species biomass. There was a CO2 x year interaction (P less than 0.001); plots under eCO2 had decreasing proportions of non-focal biomass over time, at both N treatment levels. If numbers of new species gains are associated with the magnitude of new recruit biomass, the lack of CO2 x N interactions suggest that even modest gains in SR from recruitment that would have occurred without species removals were probably unlikely to influence the observed effects and interactions of CO2 and N. It is possible that the aboveground and belowground biomass in these diverse plots was sufficiently dense - regardless of CO2 or N treatment - to prevent these treatments from having a major impact on colonization. Additionally, across the 24 years of this study, total biomass in 9- and 16-species plots grew larger with time, suggesting that resistance to colonization was unlikely to have weakened (and may even have strengthened) for the second half of the study during which no data on removed recruitment are available. Moreover, the eCO2, +N treatment which via light pre-emption reduced species richness the most in the second half of the experiment compared to all other treatment combinations also tended to reduce light the most; thus if new recruits had been permitted, this treatment would have been the least likely to be successfully colonized. In all, these data suggest that allowance of new recruitment would not likely have confounded our interpretation of changing diversity due to CO2 and N over time in this system.