Evidence of a historical frequent, low-severity fire regime in western Washington, USA

Fire is a common disturbance in many forests. We conducted a fire history study on 40 Douglas-fir (Pseudotsuga menziesii (Mirb.) Franco) trees from two sites, Kellett Bluff and Turn Point, in the S...

D r a f t D r a f t Introduction 34 The structure and function of ecosystems can be shaped by individual disturbances and 35 by the disturbance regimes that reflect differences in the type and spatial and temporal patterning 36 of disturbances (Johnson and Miyanishi 2007). Detailed studies of individual sites are necessary 37 to understand the effects of disturbances and/or disturbance regimes at small scales, but can also D r a f t Page 3 38 contribute to large-scale assessments that require synthesizing multiple studies under diverse 39 abiotic conditions (Falk et al. 2011). 40 Fire is a common disturbance in many forests, and understanding historical fire regimes 41 provides a valuable context for understanding current and future fire regimes. In northwestern D r a f t Page 4 61 fires, evaluated the extent and spatial patterns of fires, and examined the seasonality of fires. To 62 explore larger-scale patterns in fire history, we compared the fire histories of these sites with 63 those of other sites in the San Juan Islands. 64

66
Study sites 67 Field work occurred at two sites, Kellett Bluff and Turn Point, in the San Juan Islands. 68 The San Juan Islands are an archipelago between the mainland of North America and Vancouver 69 Island, in the rain shadow of the Olympic Mountains to their southwest. Average annual 70 precipitation at these sites is ~ 790 mm, with a pronounced summer drought: less than 25% of 71 precipitation occurs from April to September (PRISM 2018). The sites are about 11 km apart on 72 separate islands, and are managed by the Bureau of Land Management. 73 Kellett Bluff is a south-facing rocky headland at the southwest tip of Henry Island (N48 74 58' 86", W123 20' 6"). It is 25 ha (63 acres) in size. The site was reserved in the 1890s as a 75 potential lighthouse location, though one was never established -a navigation light was 76 established in the late 1800s and a fog warning bell installed after 1907 (McDonald 1990). Old 77 marine charts of the area show a Coast Salish settlement in Open Bay, just east of Kellett Bluff, 78 and the area was a popular reef netting site until the 1950s. The earliest settler land patents on 79 Henry Island date to 1886, but permanent Euro-American settlement did not occur until the early 80 1900s, when settlers began commercial fishing, sheep farming, and mining limestone. 81 Turn Point is mostly forested with exposed rocky headlands, and is at the northwest tip of 82 Stuart Island (N 48 68' 88", W 123 23' 73"). It is 28 ha (70 acres) in size. A light station was 83 built at the site in 1893 (BLM 2018). When the island was surveyed in 1874, a Native American D r a f t Page 5 84 settlement was noted in Reid Harbor. Sandstone cliffs along the west shore were also noted; 85 these were quarried around 1904. The first land patent application was in 1876, but most 86 settlement occurred in the 1890s, when twelve families received land patents and moved to the 87 island. Cattle-raising, limestone quarrying and fishing provided income for the early settlers of 88 Stuart Island (McDonald 1990). 89 Currently, most of the area at each site is forested, though there are also substantial 90 grassland areas. Douglas-fir (Pseudotsuga menziesii) is the dominant tree, with western red 91 cedar (Thuja plicata), pacific madrone (Arbutus menziesii), grand fir (Abies grandis), and Garry 92 oak (Quercus garryana) in areas. 94 Reconnaissance work began in Autumn 2010. Each site was thoroughly examined to 95 identify fire-scarred live trees, snags, down logs, and intact stumps. Each of these structures was 96 inspected for number of visible scars, soundness of wood, and distance to roads and paths. 97 Selected structures were not clearly visible from roads or paths, and at least one tree length from 98 them. Locations of structures suitable for sampling were recorded using a portable GPS device. 99 A partial cross-section was removed from the boles of live trees and snags. Cross-section 100 width was minimized to ensure the tree's structural integrity but was deep enough to include the 101 pith. Full cross-sections were taken from downed logs and stumps to provide a more complete 102 inventory of fire scars and a higher potential for intact wood for crossdating. Useable samples 103 could not be obtained from all assessed structures. 104 In total, we secured 40 samples, 19 at Kellett Bluff and 21 at Turn Point ( Figure 1; Table  D  r  a  f  t   Page 6 107 Turn Point were obtained during Winter and Spring 2011, and were primarily from live trees (n 108 = 13 of 21). All samples were allowed to dry at room temperature until Spring 2012. Samples 109 were glued onto plywood to prevent breakage and loss, and were then planed and sanded with 110 progressively finer grits (50 to 400) to obtain a smooth, polished surface. Samples were sanded 111 to the point where cell walls were clearly visible under 10x magnification so that false or missing 112 growth rings could be distinguished during crossdating and details of fire scars were evident 113 (Stokes and Smiley 1968;Speer 2010). 114 During sample collection, increment cores were also taken from 16 live Douglas-fir trees 115 at Kellett Bluff that were free from apparent injury and defect. Selected trees were located away 116 from localized ground water sources and depressions and thus were more likely to be 117 climatically sensitive. Increment cores were mounted and sanded using standard 118 dendrochronological procedures (Stokes and Smiley 1968;Fritts and Swetnam 1989). Given the 119 proximity of the sites and their similar climates, we assumed that a chronology built at Kellett 120 Bluff would also be applicable at Turn Point.

121
Dendrochronological analysis 122 A master ring-width chronology was built from the increment cores taken from live trees. 123 Cores were scanned with a high-resolution digital scanner and the images imported into 124 WinDENDRO 2012a software (http://www.regent.qc.ca/products/dendro/DENDRO.html), 125 where ring widths were measured to the nearest 0.01 mm. The ring widths were analyzed using 126 the Dendrochronology Program Library in R (dplR) (Bunn et al. 2015) package in R (R Core 127 Team 2015). The mean interseries correlation was 0.62. Standardized ring width indices were 128 calculated for each tree and averaged to compile a master chronology (Stokes and Smiley 1968;129 Fritts and Swetnam 1989) covering the period from 1650 to 2011 ( Figure S2). In total, we D r a f t identified 84 narrow rings and 36 wide rings during this period (Table S2). We used this pattern 131 of marker years to crossdate fire scar samples to ensure accurate tree ages and dates for past fires 132 (Brown and Swetnam 1994). 133 Cross-section samples were analyzed using a boom-mounted dissecting scope. Fire scar 134 samples from live trees were visually crossdated following the methods outlined in Yamaguchi 135 (1991) and by comparing ring width patterns to those in the master chronology. Samples from 136 dead trees were crossdated by numbering the rings from the outer to the inner ring, noting which 137 rings were narrower or wider than expected relative to surrounding rings, and then evaluating the 138 fit between this set of narrow and wide rings and the master list of marker years (Table S2). 139 Each potential outer ring date from 2011 to 1650 was tested. For each outer ring date, the 140 putative date of each narrow and wide ring was determined. The correspondence between these 141 dates and the master list of marker years was then calculated to determine which outer ring date 142 had the strongest support. We identified the pith and death dates where possible, or the dates of 143 the innermost and outermost extant rings. 144 Fire scars were identified by the characteristic band of previously killed cambium tissue 145 and the subsequent pattern of radial growth healing (McBride 1983;Speer 2010). Scars were 146 carefully examined to distinguish them from other forms of tree injury (Smith et al. 2016). Other 147 types of damage were noted as they may have made the tree susceptible to scarring, but are not 148 reported here. 149 The position of the dead cambium in the growth ring was recorded using six categories: 150 early earlywood, middle earlywood, late earlywood, latewood, dormant, and undetermined 151 (Sutherland et al. 2015). The number of fires associated with each category was tallied and interval after the last fire event was included as an interval in these analyses. 172 Fire synchronicity at these two sites were not compared statistically as the data did not   (Table 1) (Table 3).
D r a f t Individual trees were scarred by fire on average every 40-50 years, but this ranged from 1 199 to 270 years (Table 2). Many fire scars healed over before the next fire was recorded ( Figure 3). 200 The fire return interval was more than three times longer in the post-settlement period than in the 201 historical period. Considering the site as a whole, the mean fire return interval was 13 years, but 202 ranged from 1 to 85 years. The composite fire return interval was 4.5 times longer in the post-203 settlement period than the historical period (27 vs. 6 years).  (Table 3). 210 Individual trees were scarred by fire on average every 35-48 years, though this ranged 211 from 2 to 180 years (Table 4). The fire return interval was 25 years in the historical period but 212 64 years in the post-settlement period. Considering the site as a whole, the mean fire return 213 interval was 7-11 years, but ranged from 1 to 88 years. The mean fire interval rose from 6 years 214 in the historical period to 14 years in the post-settlement period. only happened in 7 years. Even in years where fires occurred at both sites, it is highly probable 240 that they reflect separate ignitions given that the sites are 11 km apart on different islands. A fire 241 that was hot enough to produce embers that could be carried across the water would be unlikely 242 to also be of sufficiently low severity to scar trees without killing them. Lightning from the same D r a f t Page 12 243 storm could conceivably strike both sites though, as explained below, this region has among the 244 lowest lightning strike densities in the continental United States. 245 Widespread fires -those scarring multiple trees within a site -occurred infrequently: in 246 only four years (1868,1870,1895,1902) were fires recorded by two or more trees at a site. 247 Interestingly, these years are in a narrow period of time around the time of Euro-American 248 settlement; the later dates may indicate fires that spread onto these sites from nearby land 249 clearing efforts of settlers (McDadi and Hebda 2008). 250 Although the local fire history studies differ in sampling intensity and temporal depth, 251 when all seven sites are considered together, the longest historical fire-free interval was 29 years 252 (1663 to 1692). Intriguingly, an assessment of the Pacific Northwest region also noted that the 253 period from 1650-1800 experienced reduced area burned (Weisberg and Swanson 2003) 254 suggesting that this might have been a widespread phenomenon. Weisberg and Swanson (2003) 255 also suggested that there was regional synchronicity in the fire regimes of western Oregon and 256 Washington, but our results do not support this conclusion, perhaps because these sites are on 257 islands and/or because of the finer temporal resolution of our study (annual vs. 25-year intervals 258 in Weisberg and Swanson (2003)). history. Low elevation south-facing slopes, the edaphic combination most similar to conditions 267 at our study sites, had a mean fire interval was 85 years. 268 Although these mainland studies concur with our island studies in documenting an 269 abundance of fire on the landscape, methodological differences make it difficult to assess the 270 consistency of these studies with prior research showing that islands have more but smaller fires 271 than on the mainland (Bergeron 1991; Hellberg et al. 2004). Wetzel and Fonda (2000) did not 272 crossdate their samples and used increment cores rather than stem cross-sections. Furthermore, 273 they identified germination dates and fire release markers from increment cores, so their 274 approach is more likely to detect fires that are of high enough severity to cause a release in tree 275 growth (but not so high as to have killed the sampled trees). Wendel and Zabowski (2010) 276 required at least 3 fire scars to indicate a fire, so their estimate is an underestimate of the actual 277 fire frequency. 278 Another difference is that the mainland studies were conducted over much larger areas. 279 At these scales, the authors were able to conduct landscape-scale analyses such as looking at 280 aerial photographs for evidence of even-aged stands that suggest the occurrence of high-severity 281 stand-replacing fires. Given the relatively small areas of our sites, we were not able to conduct 282 these types of analyses. We did observe areas that appeared to be even-aged but, as described 283 below, we feel these are more likely a result of ingrowth in formerly open stands than the 284 replacement of one forest stand with another. 285 Fire histories can be studied indirectly by analyses of charcoal and pollen (Sugimura et 286 al. 2008;Walsh et al. 2015). In this region, studies that used these approaches have estimated 287 fire return intervals that are broadly similar to the mainland studies described above. would expect synchronicity among sites because conditions would be suitable simultaneously at 336 multiple sites. Furthermore, we would expect that synchronicity to be linked to climate. For 337 example, a dry year with a rare lightning storm might serve as the ignition source for fires on 338 separate islands. This does not appear to be the case, however, as Palmer Drought Severity 339 Index (PDSI) during years when fires burned at 2 or more sites did not differ significantly from 340 PDSI during other years (Table S4). 341 Although a few fires occurred soon after Euro-American settlement, fires have not 342 occurred at these sites in several decades. For example, considering all 7 sites together, the last 343 recorded fire was in 1964. This is consistent with patterns in recent decades throughout western 344 North America, and with the notion that historical fires were largely of anthropogenic rather than 345 natural origin. Douglas-fir mixed with prairie (7 stems/ha) but in ca. 2000 they were primarily closed canopy 358 forest (426 stems/ha; Peterson and Hammer 2001). 359 The presence of live, old Douglas-fir trees at both Kellett Bluff and Turn Point indicates 360 that these sites have supported trees for numerous centuries. Many of the fire-scarred trees were 361 rapidly growing, suggesting that they were in relatively open stands with low densities, and 362 recovered rapidly after fire, with scars healing over in less than a decade. Aerial images ( Figure   363 S1) also suggest that in-filling and expansion have occurred at both sites. However, additional 364 research would be required to assess the spatial patterning of stands historically, and the size and 365 extent of grasslands that have experienced in-filling in recent decades.  (Wetzel and Fonda 2000). Further south in Puget Sound, it has been suggested that 396 bear grass (Xerophyllum tenax) savannas in the southeastern Olympic Peninsula were maintained 397 by frequent fire (Peter and Shebitz 2006), though a detailed fire history analysis was not 398 conducted. were dominated by prairies historically (Dunwiddie and Bakker 2011) and it is assumed that 403 these grasslands were also fire-maintained. To our knowledge, no formal fire histories have been 404 conducted in these areas, though soil charcoal evidence suggests that the prairie-forest ecotone 405 was historically stable (Hegarty et al. 2011). Further research to determine the prevalence of 406 frequent, low-severity fires in the Puget Lowlands would be beneficial, as would research into 407 stand structural development patterns. 409 This study did not assess current stand conditions or stand development. Nonetheless, 410 the absence of fires from Kellett Bluff and Turn Point for much of the last century has likely 411 increased the risk that future fires will be more intense due to the accumulation of fuel on the 412 ground, the presence of smaller trees and shrubs that can act as ladder fuels, and the continuity of 413 the canopy. If exposed to fire in their current condition, these stands may experience significant 414 mortality. Wildfire is also a concern on these remote islands because of the logistical difficulties 415 associated with mobilizing a suppression response. 416 Increasing the resilience of these stands would require active management, creating 417 conditions that support low-severity fires and then permitting these fires to occur. Stand Finally, we reiterate that even though both of these sites experienced frequent fires, they 426 supported a tree component for centuries -the oldest tree in this study was more than 500 years 427 old. A program of frequent, low-severity fire may be critical for the long-term persistence of 428 these old-growth trees.          Table 1.

Future options
D r a f t Page 34