Vegetation change over seven years in the largest protected Pacific Northwest Bunchgrass Prairie remnant

Temperate grasslands are one of the most altered ecosystems on Earth. Consequently, conservation of important characteristics of such ecosystems (e.g., biodiversity) is uncertain even within grasslands that have been protected. Invasion by non-native plants is considered a primary threat to intact grasslands. Here, we evaluated native and non-native vegetation composition change over seven years in the largest Pacific Northwest Bunchgrass remnant. We sampled 124 permanent plots across the Zumwalt Prairie Preserve (northeastern Oregon, USA) twice, seven years apart. With data collected from three grassland community types (xeric prairie, mesic prairie, old fields), we asked: (1) how has species composition changed over time; (2) which species showed the greatest changes in abundance; and (3) how did abundance of Ventenata dubia (the most abundant non-native species) relate to patterns of native and non-native plant abundance? Vegetation composition changed in all three plant communities. Ventenata dubia, an annual non-native grass: (1) became the third most dominant species across the study area; (2) was the only non-native that increased in abundance substantially in all three communities; and (3) was negatively related to native perennial forb cover. Relative cover of non-native species decreased in old fields concomitant with increases in native bunchgrass (Festuca idahoensis) and V. dubia cover. Increased cover of native perennial grasses and non-native annual grasses in old fields were associated with loss of bare ground, but not with reductions in non-native perennial grass cover. Native species dominated in the mesic prairie; however, non-native cover (particularly V. dubia) increased (mean cover increased from 3 to 10%) while mean native perennial forb cover decreased (from 30 to 25%) over time. Continued shifts towards non-native annual grass dominance coupled with potentially declining native perennial forbs, may challenge conservation efforts in one of the last large tracts of Pacific Northwest Bunchgrass Prairie.

Field sampling for this study was non-destructive, and did not harm endangered or protected species. If the data are held or will be held in a public repository, include URLs, accession numbers or DOIs. If this information will only be available after acceptance, indicate this by ticking the box below. For example: All XXX files are available from the XXX database (accession number(s) XXX, XXX. prairie, mesic prairie, old fields), we asked: 1) how has species composition changed over time;

30
(2) which species showed the greatest changes in abundance; and (3) how did abundance of 31 Ventenata dubia (the most abundant non-native species) relate to patterns of native and non-32 native plant abundance? Vegetation composition changed in all three plant communities.

33
Ventenata dubia, an annual non-native grass: 1) became the third most dominant species across 34 the study area; 2) was the only non-native that increased in abundance substantially in all three 35 communities; and 3) was negatively related to native perennial forb cover. Relative cover of 36 non-native species decreased in old fields concomitant with increases in native bunchgrass 37 (Festuca idahoensis) and V. dubia cover. Increased cover of native perennial grasses and non-38 native annual grasses in old fields were associated with loss of bare ground, but not with 39 reductions in non-native perennial grass cover. Native species dominated in the mesic prairie; 40 however, non-native cover (particularly V. dubia) increased (mean cover increased from 3 to Temperate grasslands are among the most threatened biomes on Earth and are the most 49 altered ecosystems in North America [1]. Only about 1% of North America's temperate 50 grasslands are in a relatively natural state, and even those that remain have been subjected to 51 significant anthropogenic stressors, e.g., widespread livestock grazing, intentional and accidental 52 introduction of non-native species, and localized settlement/cultivation and subsequent 53 abandonment [1][2]. Because of widespread alteration to temperate grasslands in North America, 54 the ability of these systems to maintain the ecological services, such as biodiversity, that they 55 have historically provided is uncertain [1,3]. Invasion by non-native plant species has been 56 identified as one of the most serious threats to ecosystem function and native biodiversity in anthropogenic stressors that are considered threats to conservation of native temperate grasslands 79 (e.g., legacy effects of past cultivation, livestock grazing, invasion by non-native plants; [1,4]).  grasses (e.g. Poa pratensis and Thinopyrum intermedium) were dominant in or near sites that had 104 been previously cultivated (e.g., old fields). The detection of less invaded mesic communities 105 coexisting alongside annual grass invaded xeric prairie and non-native perennial grass dominated 106 old fields may suggest that never-plowed mesic prairie communities are more resistant to 107 invasion or, alternatively, that invasion by non-native grasses from either the drier or wetter 108 communities is inevitable [22]. The differences in non-native plant composition and abundance 109 across community types in the ZPP suggest that some communities may be more susceptible to

122
Field sampling took place on private lands within the ZPP. Permission for sampling was granted 123 by and in collaboration with the landowner, The Nature Conservancy. Field sampling for this 124 study was non-destructive, and did not harm endangered or protected species.       Non-native relative cover increased across our study area from 27 to 30% over time (two 264 sample paired t-test, two-tailed, t = 2.6, p-value = 0.01; Table 2). This change was primarily 265 focused within the mesic prairie, where non-native relative cover increased from 16 to 22% 266 (paired t = 4.6, p < 0.001; Table 2). There was weak evidence for a decrease (from 68 to 61%; 267 paired t = -1.5, p < 0.14) in non-native relative cover in the old field community, and no 268 evidence of change in the xeric prairie (paired t = 0.45, p = 0.66; Table 2). The two most  (Table 1). Cover and frequency of V. dubia, increased substantially across 273 our study area; V. dubia moved up a rank from fourth to third when evaluated based on cover 274 (mean cover almost doubled over time from 7.6 to 14.9%, Table 1) and from eleventh to third for 275 frequency, making it the third most abundant species in terms of cover and frequency in our 276 study area (Table 1). Rank abundances for annual Bromus spp. and P. pratensis -the second and 277 third most abundant non-natives-were relatively consistent over time. T. intermedium 278 (frequency increased by 10%), and the annual forb Veronica arvensis (mean cover more than 279 doubled from 0.8 to 1.8% and frequency increased by 27%) were two other non-native species 280 that showed signs of increased dominance; however their mean cover and frequency were much 281 lower than V. dubia (Table 1 and [35]).

282
Non-native annual grass cover increased from 13 to 20%, and native perennial forb cover 283 decreased from 23 to 20% over time (Fig.1). There was weak evidence that native perennial 284 grass cover increased from 55 to 58% (change = 3.1% 90% CI = 0.6, 6.0) since the initial 285 sampling effort (Fig. 1). Increases in non-native annual grass mean cover appeared to be focused 286 14 primarily in the old field (increased from 4 to 16%) and mesic prairie communities (increased 287 from 8 to 16% ; Fig 1). Mean native perennial forb cover decreased (from 30 to 25%) in the 288 mesic prairie (Fig. 1). Mean native perennial grass cover increased substantially (from 12 to 289 27%) in old fields (Fig. 1), and was consistent with an increase in total vascular plant cover 290 (11.6%, 95% CI = 2.4, 20.7). There was no evidence for change in cover for any functional 291 group within the xeric prairie (Fig 1). We also calculated the total change in native perennial 292 forb cover (not limited to the dominant species) to make sure that trends in dominant forbs 293 tracked the entire community. Consistent with the dominant forbs, all native perennial forbs 294 decreased by 5% (95% CI = -7.6, -2.9) in our sample area, and the greatest decrease for all native 295 perennial forb foliar cover (-6.6, 95% CI = -9.9, -3.2) was observed in the mesic prairie.  Table 3). The greatest changes were detected in old fields (A = 0.07) followed by the mesic 298 prairie (A = 0.05), and lastly by the xeric prairie (A = 0.03; Table 3). Temporal vectors extracted 299 from the NMS ordination indicated that most old field plots shifted to lower Axis 1 (explaining 300 64% of variation in the distance matrix) scores over time which was associated with lower 301 composition of non-native perennial grass species (Fig. 2). The mesic prairie plots tended to 302 shift down along Axis 2 (explaining 23% of variation in the distance matrix), consistent with 303 increasing non-native relative abundance, and no clear direction of change was observed for the 304 temporal vectors generated from xeric prairie plots (Fig. 2).

305
Species that changed the most in each community 306 Ventenata dubia was among the strongest indicator species for the most recent sampling 307 effort in all community types, and the only non-native that increased abundance across our entire 308 study area, particularly within the mesic prairie and old field communities (Table 4) (Table 4). Ventenata dubia dominance increased substantially (rank 323 abundance in the mesic prairie changed from 11 th to 3 rd over time; [35]). Native perennial forbs 324 showed a slight tendency to decrease in plots with higher initial non-native annual grass cover as 325 well as where non-native annual grass cover increased over time (Fig 3). 326 Notable changes in species abundance within the xeric prairie included increased 327 abundances of two non-native grass species (V. dubia, P. compressa). Other species that showed 328 significant changes in abundance in the xeric prairie were those that tended to be associated with  Table 4).

333
Relationship between V. dubia and native perennial cover 334 Ventenata dubia cover was negatively related to native perennial cover and native 335 perennial forb cover for both sampling periods (Fig. 4 & Table 5). The greatest negative slopes 336 were detected for the 0.9 quantile followed by the 0.5 quantile (Fig 4 & Table 5). The slopes for  Table   338 5). Relationships between V. dubia cover and native perennial grass cover were inconsistent 339 between years and across quantiles. V. dubia cover was negatively related to native perennial 340 grass cover for the 0.9 quantile during the first sampling period but not the 0.5 quantile and for 341 the 0.5 quantile during the second sampling period, but not the 0.9 quantile (Table 5). Because 342 vegetation trends in the old field community were much different than those in the uncultivated 343 prairie, analyses were also performed for the uncultivated prairie alone; however, this did not 344 change our findings or interpretation of the data. Inconsistent relationships between V. dubia and 345 native perennial grasses over time and quantiles suggest complex interactions between these two 346 variables and may be related to factors such as lag time effects between species interactions, 347 variation in native species responses, differential responses by community type, or may reflect a 348 weak relationship between these variables that cannot be distinguished from noise in the data. V. 349 dubia was negatively related to non-native perennial grass cover during both sampling periods.

350
The greatest negative slope (coefficient = -1.3, 95% CI = -1.4, -1.1) was observed for the 0.9 351 quantile during the first sampling period. Slopes for both quantiles decreased substantially and 352 were similar for the second sampling period where it was estimated that a one percent increase in 353 V. dubia cover corresponded to ~ 0.1% decrease in non-native perennial grass cover for the 0.5 354 and 0.9 quantiles (Table 5). V. dubia was positively related to cover of other non-native annual 355 grasses for the 0.5 quantiles during both sampling periods. It was estimated that a 1% increase in 356 V. dubia cover corresponded to approximately a 0.1% increase in the median cover of other non-357 native annual grasses (Table 5).

359
Our study revealed three major findings. First, we found evidence of increased relative of the ecological significance of this result for a couple of reasons. First, our study is based on 432 change between two time periods, and a 5% change in foliar cover is a relatively small change 433 that may be attributable to some unmeasured natural variation in the system; future research will 434 be needed to determine long-term trajectories of perennial forb abundance as well as species 435 specific responses. Second, we do not expect that an overall ~5% decline in perennial forb cover 436 will have significant impacts to ecosystem function or services in the ZPP. We recommend that 437 trends in native perennial forb abundance be closely monitored at the ZPP because continued