Climate change and primary production: Forty years in a bunchgrass prairie

ANPP. ANPP was measured from 1978–2017. Belowground production was not mea- sured, as it is difficult and destructive [44,45]. Two ANPP methods were employed: 1. For the period 1978–1993, live (green) vegetation was clipped on 10–12 randomly selected 0.1 m2 plots at each site every two weeks from late-May–mid-Oct. Clipped vegetation was separated between grass and dicot (forbs or leaves from woody vegetation), dried for 48 hours at 60˚C, and then weighed. This also provided the relative abundances of grass and dicot. Clipping is time consuming and destructive, so this method was replaced. 2. Since 1994, 10–12 permanent points were randomly selected at each site, and the abun- dance of live (green) vegetation was measured every two weeks by radiometer [46,47] from late-May–mid-Oct. This is non-destructive and requires less time. The radiometer was held at a height to measure a 0.10 m2 area. Live plant biomass is computed from radiometer readings using a site and date specific regression. Each regression was based on radiometer readings of 3–5 0.1 m2 plots that were adjacent to the study location; these reference plots were selected to range from very low to very high plant abundances. Green vegetation in each plot was clipped, separated between grass and dicot, dried for 48 hours at 60˚C, and weighed. Regressions averaged an r of 0.93 (±0.03 SE; range: 0.45–0.99). Relative abun- dances of grass and dicot were based on clippings from reference plots. For either method, ANPP was the sum of increases in live (“green”) plant biomass between

consecutive periods from late-May through mid-Oct. ANPP was allocated to grass and dicot using proportional composition based on clipping at each site and date (see above). To examine whether the two different methods affected ANPP measures, clipping and radi- ometer measures of ANPP were both made in 1994 and found to be highly correlated (r = 0.90, n = 4) with no tendency to over- or under-estimate ANPP (slope = 0.95 ± 0.17). Fur- thermore, all subsequent statistical analyses were conducted with the method as an indepen- dent categorical variable and it was never significant (P always > 0.30). Plant phenology. Two phenology measures were made using biweekly ANPP measure- ments (above): time when peak “green” biomass is attained and degree of summer senescence. Time of peak “green” biomass was determined for each site and year as the Julian day when the greatest “green” biomass was observed. Plant senescence (%) occurs over the summer (July–Aug) as “green” biomass dies or becomes dormant (“brown”) after peak “green” biomass occurs. Senescence was measured as 100 X (1 –lowest observed “green” biomass/peak observed “green” biomass). Plant species composition. Plant species composition was obtained for 1978, 1980–1983 and for all years since 1997. In June 1978, vegetation was clipped and sorted to species for 26– 1 m2 plots randomly located throughout NBR. In June 1980–1983, vegetation was clipped and sorted to species for 30–0.1 m2 plots randomly located at each of two sites (Hill & Triangle,

which were sampled in 1978). Since June 1997, point frame sampling [43,48] was employed at the same two sites (Hill & Triangle), and all analyses were limited to these data, as earlier clip- ping samples were not comparable. At 18 randomly-placed permanent locations (1 m2) at one site and 9 at the other, 100 points in June and Sept were identified to bare ground, litter, moss, lichen, or plant species. Plant species were identified using a number of keys [49–57]. Plant nomenclature, family and life history (e.g., annual, biennial, drought resistance, fire tolerance and invasive) followed the U.S.D.A. Plant Database (http://plants.usda.gov/java/) [58]. U.S.D.A defines drought resistance and fire resistance as none, low, medium, and high. Drought resistance is based on where species grow relative to soil moisture, and fire resistance is based on regrowth and reestablishment after fire (https://plants.usda.gov › charinfo) [58]. We assigned values of 3, 2, 1 or 0 to the U.S.D.A indices (respectively, none, low, medium, and PLOS ONE Bunchgrass prairie and climate change PLOS ONE | https://doi.org/10.1371/journal.pone.0243496 December 23, 2020 4 / 21