How Competition and Wildfire Affect Tree Range Shifts in the American West

13 Due to climate change, plant populations experience environmental conditions to which 14 they are not adapted. Our understanding of the next century’s vegetation geography depends on 15 the distance, direction, and rate at which plants redistribute in response to a changing climate. 16 Although plant redistribution in response to contemporary climate change is widely observed, 17 our understanding of its mechanics is nascent. In this study we test the response of plant range 18 shift rates to wildfire occurrence using 33,838 Forest Inventory Analysis plots across five states 19 in the western United States. Wildfire increased the rate of observed range shifts for 6/8 tree 20 species by more than 22% on average, suggesting that incumbent vegetation can act as a barrier 21 to plant range shifts and that fire management may play an important role in facilitating 22 transitions between vegetation types in response to climate change. 23

3 6 on RS rate. Based on DCB/DCN, the climatic niche difference between source populations and 117 leading-edge populations was greater in areas that burned than in those that did not for six of the 118 eight species. D ! /DCN was, on average, 2.10. By this metric, fire most increased the rate of RS 119 for Lithocarpus densiflorus, which had the greatest ratio--6.35 (though sample size was 120 smallest for this species). Abies grandis was the only species where D ! was appreciably less 121 than D " : D ! /DCN = 0.70. The second measure of RS rate, Schoener's D, was closer to 0 122 (indicating less niche equivalency) for burned plots across each of the eight species, by 22% on 123 average (Table 2). Schoener's D corroborated DCB/DCN for Lithocarpus densiflorus, which had 124 a burned-plot Schoener's D 36% closer to 0 than in unburned plots (the greatest difference in 125 Schoener's D across species). 126 With the RS broken down into each climate variable we found that RS with respect to 127 each of the three components of the climate space was greater in plots that burned than in plots 128 that did not, on average across the eight species (Figure 3). For example, as each species in 129 unburned areas migrated towards plots with either higher or lower Mean Summer Precipitation 130 (MSP), species in burned plots migrated in the same direction (either higher or lower MSP) at a 131 rate that was 100% greater on average (Appendix 1, Fig. A1.1-A1.8). The difference between RS 132 rates was greatest for Mean Winter Precipitation (MWP), where the average RS rate was 208% 133 greater in burned plots. Mean Temperature of the Warmest Month (MTWM) had the least 134 difference between RS rates, but burned plots still had an average RS rate 55% greater than 135 unburned plots for this variable. The range of RS rates across species was greater in burned plots 136 ( Figure 3), probably reflecting the smaller sample size of burned plots. 137

Contextualizing the Direction of RS 138
7 When compared to recent climate change in the study area (Figure 4), the eight species, 139 on average, shifted in the direction that kept them closer to their historic climatic niche. For 140 example, as plots across this study received less MSP, the species shifted towards plots with 141 more summer precipitation--with burned areas exhibiting faster RS than unburned areas ( Figure  142 3a where evolution is unlikely to have made populations tolerant to observed climate changes, we 172 observed range expansions where environmental variables that limit species ranges were 173 lessened/removed. As a possible range-limiting factor for many species, interspecific The observed RS rates, when roughly converted from climatic distance to spatial 246 distance, are also consistent with general expectations. Combining the observed average shifts in 247 MTWM with the normal lapse rate of 6.5ºC/km, we estimate that the non-subalpine trees shifted 248 towards areas of lower MTWM by roughly +48m elevation in unburned plots, and +105m in 249

12
The sources of potential uncertainty in the interpretation of our results include (i) 254 potential sink populations where seedlings occur but have not, and may not, reach maturity (ii) 255 fire succession and fire-regime adaptations and (iii) the relationship between climatic distance 256 and spatial distance (Equation 1). 257 An underlying assumption of the approach was that seedlings eventually mature into 258 trees. This is, of course, not necessarily the case. We attempted to minimize the impact of sink 259 populations on our results by excluding species where the direction of RS was not consistent 260 among different life stages. We used diameter at breast height to delineate seedlings (dbh < 261 2.5cm), saplings ( 2.5cm < dbh < 12.7cm), trees (dbh > 2.5cm), and large trees (dbh > 12.7cm) 262 and compared the RS vectors of seedlings and trees to saplings and large trees. We confirmed 263 that all eight trees used in the primary analysis had consistent RS vector direction between these 264 groups. However, while the sapling group likely contained fewer sink populations than the 265 seedling group, it is still possible that some saplings were members of sink populations and 266 would never reach reproductive maturity. 267 This study provides empirical evidence that wildfire increases the RS rate of tree species 295 that are moving in response to recent climate change. Furthermore, we have argued that the 296 increase in RS rate is likely a product of the reduction of the population size and density of the 297 competitors due to wildfire. The findings of this study bolster previous work suggesting that 298 competition is yet another barrier (in addition to dispersal limitations and geographic barriers) 299 14 that affects the ability of plant species to track their optimal climatic conditions as they move 300 across landscapes. no adult tree (hereafter referred to as "tree") of the species were considered the leading-edge of 320 the migrating population. Subsequently, the FIA plots of each species were separated into 4 321 independent groups: (1) Plots that did not burn with only seedlings present (NS) (2) Plots that did 322 15 not burn with seedlings and trees present (NT) (3) Plots that burned with only seedlings present 323 (BS) (4) Plots that burned with seedlings and trees present (BT). Species were removed from 324 analysis if the number of plots was less than 5 in any of these four categories (NS, NT, BS, BT). 325 Because of the spatial uniformity of FIA sampling, spatial sampling bias is unlikely to be 326 a major source of error in our analysis. However, seedlings were sampled in only a subset of 327 each FIA plot, and it is possible that seedling species present in the full sampling plots were 328 absent in the seedling microplots, potentially reducing the number of NS and BS plots.

Species Selection 349
The climatic difference between seedling-only plots and tree-and-seedling plots may not 350 represent the actual RS direction of a species. For example, seedling establishment may represent 351 sink populations that occur in areas that will not support a reproducing population in the future. 352 To ensure that we included only species for which we could confidently estimate the direction of 353 RS , we compared the climatic RS vector of NS and NT plots with the RS vector of unburned 354 sapling plots (NJ), with dbh between 2.5cm and 12.7cm, and unburned large tree (NG) plots, 355 with dbh greater than 12.7cm 9 . We determined the consistency of the RS direction between the . Figure 5 provides a graphical explanation of this 359 process. We used the cutoff of 50% vector agreement to exclude species where the RS direction 360 was not consistent across life stages. 361 The direction of the RS vectors may differ significantly between the unburned and 362 burned plots, possibly because of species-specific adaptations that affect 363 colonization/regeneration after fire. To avoid the difficulty of comparing and interpreting RS 364 rates of groups that are migrating in different directions, we also excluded species for which the 365 cosine of the angle between RS vectors in burned and unburned plots was less than 0.5 (Figure  366 5). 367

Calculating Climatic Niche Difference and RS Rate 368
The RS rate was proxied by the climatic distance between source populations and 369 leading-edge populations, i.e. the difference between the climatic niche of seedling-only plots 370 and the plots with both trees and seedlings of a species (eq. 1). 371 Where ) is the RS rate and i is either burned (B) or unburned (N), DG is the geographic 373 RS distance per time (t), DC is the climatic RS distance, per time unit defined by the mean age 374 difference between mature trees ( 2 ) and seedlings ( -). The Euclidean distance between the 375 centroids for NS and NT and between BS and BT (derived from the magnitude of the climate RS 376 vectors described above) give D " and D ! , respectively. We do not have information on 2 377 and -, but we have no reason to expect them to be different between burned and unburned 378 plots. Therefore, we use the ratio of D ! to D " as our primary measure of whether RS distances 379 are greater in burned plots. To ensure that the threshold minimum sample size did not greatly impact the results, we 399 repeated our analysis of RS rates for minimum sample size of presence in 5, 10, and 15 FIA 400 plots. The threshold minimum sample size affects the number of species analyzed but not the 401 conclusion that fire facilitates RS (Appendix 5). We focus on the analysis using a minimum 402 sample size of 5, because the number of species meeting the threshold decreased sharply as the 403 threshold increased. Furthermore, two species with NBS = 5 passed the RS vector direction 404 vetting, indicating that this sample size was sufficient to estimate an RS vector that was 405 consistent with populations of larger sample sizes. We also verified that other decisions 406 regarding species exclusion, such as species vetting due to RS vector agreement between life 407 stages and burned-unburned plots, did not produce results that contradicted our conclusions 408